Datasets:

---

JetBrains-Research

/

commit-chronicle

like 9

Follow

JetBrains Research 71

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -18,10 +18,11 @@ SYSTEMDS-20 New Data Model\n* 21 Finalize dense tensor blocks OK\n* 22 Sparse double/float tensor blocks\n* 23 Sparse int/bool tensor blocks\n- * 24 Initial tensor dml integration\n+ * 24 Initial tensor dml integration OK\n* 25 Initial data tensor implementation\n* 26 Non-zero default value for sparse (row/col)\n- *\n+ * 27 Tensor buffer pool integration\n+ * 28 Tensor local readers and writers\nSYSTEMDS-30 Builtin and Packaging\n* 31 Shell script for local runs\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/caching/TensorObject.java", "diff": "+/*\n+ * Licensed to the Apache Software Foundation (ASF) under one\n+ * or more contributor license agreements. See the NOTICE file\n+ * distributed with this work for additional information\n+ * regarding copyright ownership. The ASF licenses this file\n+ * to you under the Apache License, Version 2.0 (the\n+ * \"License\"); you may not use this file except in compliance\n+ * with the License. You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing,\n+ * software distributed under the License is distributed on an\n+ * \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+ * KIND, either express or implied. See the License for the\n+ * specific language governing permissions and limitations\n+ * under the License.\n+ */\n+\n+package org.tugraz.sysds.runtime.controlprogram.caching;\n+\n+\n+import java.io.IOException;\n+\n+import org.apache.commons.lang.mutable.MutableBoolean;\n+import org.tugraz.sysds.common.Types.DataType;\n+import org.tugraz.sysds.common.Types.ValueType;\n+import org.tugraz.sysds.runtime.DMLRuntimeException;\n+import org.tugraz.sysds.runtime.data.TensorBlock;\n+import org.tugraz.sysds.runtime.instructions.spark.data.RDDObject;\n+import org.tugraz.sysds.runtime.io.FileFormatProperties;\n+import org.tugraz.sysds.runtime.meta.MatrixCharacteristics;\n+import org.tugraz.sysds.runtime.meta.MetaData;\n+\n+public class TensorObject extends CacheableData<TensorBlock>\n+{\n+ private static final long serialVersionUID = -2843358400200380775L;\n+\n+ protected TensorObject() {\n+ super(DataType.TENSOR, ValueType.STRING);\n+ }\n+\n+ public TensorObject(String fname) {\n+ this();\n+ setFileName(fname);\n+ }\n+\n+ public TensorObject(ValueType vt, String fname) {\n+ super(DataType.TENSOR, vt);\n+ setFileName(fname);\n+ }\n+\n+ public TensorObject(String fname, MetaData meta) {\n+ this();\n+ setFileName(fname);\n+ setMetaData(meta);\n+ }\n+\n+\n+ /**\n+ * Copy constructor that copies meta data but NO data.\n+ *\n+ * @param fo frame object\n+ */\n+ public TensorObject(TensorObject fo) {\n+ super(fo);\n+ }\n+\n+ @Override\n+ public void refreshMetaData() {\n+ if ( _data == null || _metaData ==null ) //refresh only for existing data\n+ throw new DMLRuntimeException(\"Cannot refresh meta data because there is no data or meta data. \");\n+\n+ //update matrix characteristics\n+ MatrixCharacteristics mc = _metaData.getMatrixCharacteristics();\n+ mc.setDimension( _data.getNumRows(),_data.getNumColumns() );\n+ mc.setNonZeros(_data.getNumRows()*_data.getNumColumns());\n+ }\n+\n+ public long getNumRows() {\n+ MatrixCharacteristics mc = getMatrixCharacteristics();\n+ return mc.getRows();\n+ }\n+\n+ public long getNumColumns() {\n+ MatrixCharacteristics mc = getMatrixCharacteristics();\n+ return mc.getCols();\n+ }\n+\n+ @Override\n+ protected TensorBlock readBlobFromCache(String fname) throws IOException {\n+ return (TensorBlock)LazyWriteBuffer.readBlock(fname, false);\n+ }\n+\n+ @Override\n+ protected TensorBlock readBlobFromHDFS(String fname, long rlen, long clen)\n+ throws IOException\n+ {\n+ //TODO read from HDFS\n+ return null;\n+ }\n+\n+ @Override\n+ protected TensorBlock readBlobFromRDD(RDDObject rdd, MutableBoolean status)\n+ throws IOException\n+ {\n+ //TODO read from RDD\n+ return null;\n+ }\n+\n+ @Override\n+ protected void writeBlobToHDFS(String fname, String ofmt, int rep, FileFormatProperties fprop)\n+ throws IOException, DMLRuntimeException\n+ {\n+ //TODO write\n+ }\n+\n+ @Override\n+ protected void writeBlobFromRDDtoHDFS(RDDObject rdd, String fname, String ofmt)\n+ throws IOException, DMLRuntimeException\n+ {\n+ //TODO rdd write\n+ }\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/context/ExecutionContext.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/context/ExecutionContext.java", "diff": "@@ -37,8 +37,8 @@ import org.tugraz.sysds.runtime.controlprogram.caching.CacheableData;\nimport org.tugraz.sysds.runtime.controlprogram.caching.FrameObject;\nimport org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject;\nimport org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject.UpdateType;\n+import org.tugraz.sysds.runtime.controlprogram.caching.TensorObject;\nimport org.tugraz.sysds.runtime.data.TensorBlock;\n-import org.tugraz.sysds.runtime.data.TensorBlockData;\nimport org.tugraz.sysds.runtime.instructions.cp.CPInstruction;\nimport org.tugraz.sysds.runtime.instructions.cp.CPOperand;\nimport org.tugraz.sysds.runtime.instructions.cp.Data;\n@@ -207,17 +207,16 @@ public class ExecutionContext {\nreturn (MatrixObject) dat;\n}\n- public TensorBlockData getTensorObject(String varname) {\n- // TODO use cacheable TensorObject once implemented\n+ public TensorObject getTensorObject(String varname) {\nData dat = getVariable(varname);\n//error handling if non existing or no matrix\nif( dat == null )\nthrow new DMLRuntimeException(\"Variable '\"+varname+\"' does not exist in the symbol table.\");\n- if( !(dat instanceof TensorBlockData) )\n+ if( !(dat instanceof TensorObject) )\nthrow new DMLRuntimeException(\"Variable '\"+varname+\"' is not a tensor.\");\n- return (TensorBlockData) dat;\n+ return (TensorObject) dat;\n}\npublic boolean isFrameObject(String varname) {\n@@ -254,8 +253,7 @@ public class ExecutionContext {\n}\npublic void releaseCacheableData(String varname) {\n- CacheableData<?> dat = getCacheableData(varname);\n- dat.release();\n+ getCacheableData(varname).release();\n}\npublic MatrixCharacteristics getMatrixCharacteristics( String varname ) {\n@@ -289,8 +287,7 @@ public class ExecutionContext {\n* @return matrix block\n*/\npublic MatrixBlock getMatrixInput(String varName) {\n- MatrixObject mo = getMatrixObject(varName);\n- return mo.acquireRead();\n+ return getMatrixObject(varName).acquireRead();\n}\n/**\n@@ -300,9 +297,7 @@ public class ExecutionContext {\n* @return matrix block\n*/\npublic TensorBlock getTensorInput(String varName) {\n- // TODO Cachable tensorBlock\n- TensorBlockData to = getTensorObject(varName);\n- return to.getTensorBlock();\n+ return getTensorObject(varName).acquireRead();\n}\npublic void setMetaData(String varName, long nrows, long ncols) {\n@@ -439,8 +434,7 @@ public class ExecutionContext {\n}\npublic void releaseMatrixInputForGPUInstruction(String varName) {\n- MatrixObject mo = getMatrixObject(varName);\n- mo.getGPUObject(getGPUContext(0)).releaseInput();\n+ getMatrixObject(varName).getGPUObject(getGPUContext(0)).releaseInput();\n}\n/**\n@@ -450,8 +444,7 @@ public class ExecutionContext {\n* @return frame block\n*/\npublic FrameBlock getFrameInput(String varName) {\n- FrameObject fo = getFrameObject(varName);\n- return fo.acquireRead();\n+ return getFrameObject(varName).acquireRead();\n}\n/**\n@@ -460,8 +453,11 @@ public class ExecutionContext {\n* @param varName variable name\n*/\npublic void releaseFrameInput(String varName) {\n- FrameObject fo = getFrameObject(varName);\n- fo.release();\n+ getFrameObject(varName).release();\n+ }\n+\n+ public void releaseTensorInput(String varName) {\n+ getTensorObject(varName).release();\n}\npublic ScalarObject getScalarInput(CPOperand input) {\n@@ -529,8 +525,9 @@ public class ExecutionContext {\n}\npublic void setTensorOutput(String varName, TensorBlock outputData) {\n- TensorBlockData to = getTensorObject(varName);\n- to.setTensorBlock(outputData);\n+ TensorObject to = getTensorObject(varName);\n+ to.acquireModify(outputData);\n+ to.release();\nsetVariable(varName, to);\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DataTensorBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DataTensorBlock.java", "diff": "@@ -18,7 +18,6 @@ package org.tugraz.sysds.runtime.data;\npublic class DataTensorBlock extends TensorBlock\n{\n- private static final long serialVersionUID = -8818191561516565919L;\n//TODO handle of schema via multiple dense or sparse blocks of homogeneous types and col mapping\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/TensorBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/TensorBlock.java", "diff": "package org.tugraz.sysds.runtime.data;\n-import java.io.Serializable;\n+import java.io.DataInput;\n+import java.io.DataOutput;\n+import java.io.IOException;\nimport org.apache.commons.lang.NotImplementedException;\nimport org.tugraz.sysds.common.Types.ValueType;\nimport org.tugraz.sysds.runtime.DMLRuntimeException;\n+import org.tugraz.sysds.runtime.controlprogram.caching.CacheBlock;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport org.tugraz.sysds.runtime.util.UtilFunctions;\n-public class TensorBlock implements Serializable\n+public class TensorBlock implements CacheBlock\n{\n- private static final long serialVersionUID = -4205257127878517048L;\n-\npublic static final double SPARSITY_TURN_POINT = 0.4;\npublic static final ValueType DEFAULT_VTYPE = ValueType.FP64;\npublic static final int[] DEFAULT_DIMS = new int[]{0, 0};\n@@ -171,7 +172,7 @@ public class TensorBlock implements Serializable\npublic boolean allocateDenseBlock(boolean clearNNZ) {\n//allocate block if non-existing or too small (guaranteed to be 0-initialized),\n- // ToDo: use reset instead, since LDRB need to check dimensions for actually available space\n+ // TODO: use reset instead, since LDRB need to check dimensions for actually available space\nlong limit = getLength();\nboolean reset = (_denseBlock == null || _denseBlock.capacity() < limit);\nif( _denseBlock == null )\n@@ -420,4 +421,68 @@ public class TensorBlock implements Serializable\nreturn false;\n}\n+ @Override\n+ public void write(DataOutput out) throws IOException {\n+ // TODO Auto-generated method stub\n+\n+ }\n+\n+ @Override\n+ public void readFields(DataInput in) throws IOException {\n+ // TODO Auto-generated method stub\n+\n+ }\n+\n+ @Override\n+ public int getNumColumns() {\n+ // TODO Auto-generated method stub\n+ return 0;\n+ }\n+\n+ @Override\n+ public long getInMemorySize() {\n+ // TODO Auto-generated method stub\n+ return 0;\n+ }\n+\n+ @Override\n+ public long getExactSerializedSize() {\n+ // TODO Auto-generated method stub\n+ return 0;\n+ }\n+\n+ @Override\n+ public boolean isShallowSerialize() {\n+ // TODO Auto-generated method stub\n+ return false;\n+ }\n+\n+ @Override\n+ public boolean isShallowSerialize(boolean inclConvert) {\n+ return !isSparse();\n+ }\n+\n+ @Override\n+ public void toShallowSerializeBlock() {\n+ // TODO Auto-generated method stub\n+\n+ }\n+\n+ @Override\n+ public void compactEmptyBlock() {\n+ // TODO Auto-generated method stub\n+\n+ }\n+\n+ @Override\n+ public CacheBlock slice(int rl, int ru, int cl, int cu, CacheBlock block) {\n+ // TODO Auto-generated method stub\n+ return null;\n+ }\n+\n+ @Override\n+ public void merge(CacheBlock that, boolean appendOnly) {\n+ // TODO Auto-generated method stub\n+\n+ }\n}\n" }, { "change_type": "DELETE", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/TensorBlockData.java", "new_path": null, "diff": "-/*\n- * Copyright 2019 Graz University of Technology\n- *\n- * Licensed under the Apache License, Version 2.0 (the \"License\");\n- * you may not use this file except in compliance with the License.\n- * You may obtain a copy of the License at\n- *\n- * http://www.apache.org/licenses/LICENSE-2.0\n- *\n- * Unless required by applicable law or agreed to in writing, software\n- * distributed under the License is distributed on an \"AS IS\" BASIS,\n- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n- * See the License for the specific language governing permissions and\n- * limitations under the License.\n- */\n-\n-package org.tugraz.sysds.runtime.data;\n-\n-import org.tugraz.sysds.common.Types;\n-import org.tugraz.sysds.runtime.instructions.cp.Data;\n-\n-/**\n- * Temporary implementation of a `Data`-TensorBlock so the Tensor can be tested without implementing CacheableData for it.\n- */\n-public class TensorBlockData extends Data {\n- private static final long serialVersionUID = -3858118069498977569L;\n-\n- private TensorBlock _tb;\n-\n- public TensorBlockData(Types.ValueType vt) {\n- super(Types.DataType.TENSOR, vt);\n- // TODO handle different parameters\n- _tb = new TensorBlock();\n- }\n-\n- public TensorBlock getTensorBlock() {\n- return _tb;\n- }\n-\n- public void setTensorBlock(TensorBlock tb) {\n- _tb = tb;\n- }\n-\n- @Override\n- public String getDebugName() {\n- return \"TensorBlockData\";\n- }\n-}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/AggregateUnaryCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/AggregateUnaryCPInstruction.java", "diff": "@@ -177,8 +177,7 @@ public class AggregateUnaryCPInstruction extends UnaryCPInstruction\n// TODO use a generalized method on tensorBlock\nDoubleObject out = new DoubleObject(tensorBlock.sum());\n- // TODO once cacheable tensorObjects are used release them\n- //ec.releaseTensorInput(input1.getName());\n+ ec.releaseTensorInput(input1.getName());\nif(output.getDataType() == DataType.SCALAR){\nec.setScalarOutput(output_name, out);\n} else{\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/ParameterizedBuiltinCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/ParameterizedBuiltinCPInstruction.java", "diff": "@@ -37,9 +37,9 @@ import org.tugraz.sysds.runtime.controlprogram.caching.CacheBlock;\nimport org.tugraz.sysds.runtime.controlprogram.caching.CacheableData;\nimport org.tugraz.sysds.runtime.controlprogram.caching.FrameObject;\nimport org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject;\n+import org.tugraz.sysds.runtime.controlprogram.caching.TensorObject;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.data.TensorBlock;\n-import org.tugraz.sysds.runtime.data.TensorBlockData;\nimport org.tugraz.sysds.runtime.functionobjects.ParameterizedBuiltin;\nimport org.tugraz.sysds.runtime.functionobjects.ValueFunction;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\n@@ -322,24 +322,20 @@ public class ParameterizedBuiltinCPInstruction extends ComputationCPInstruction\nString lineseparator = (getParam(\"linesep\") != null) ? getParam(\"linesep\") : TOSTRING_LINESEPARATOR;\n//get input matrix/frame and convert to string\n- // TODO implement cacheableData for tensor so we can simplify this\n- Data dataVariable = ec.getVariable(getParam(\"target\"));\n- String out;\n- if (dataVariable instanceof TensorBlockData) {\n- TensorBlock tensor = ((TensorBlockData) dataVariable).getTensorBlock();\n- // TODO improve truncation to check all dimensions\n- warnOnTrunction(tensor, rows, cols);\n- out = DataConverter.toString(tensor, sparse, separator, lineseparator, \"[\", \"]\",\n- rows, cols, decimal);\n- ec.setScalarOutput(output.getName(), new StringObject(out));\n- return;\n- }\n+ String out = null;\nCacheableData<?> data = ec.getCacheableData(getParam(\"target\"));\nif( data instanceof MatrixObject ) {\nMatrixBlock matrix = (MatrixBlock) data.acquireRead();\nwarnOnTrunction(matrix, rows, cols);\nout = DataConverter.toString(matrix, sparse, separator, lineseparator, rows, cols, decimal);\n}\n+ else if( data instanceof TensorObject ) {\n+ TensorBlock tensor = (TensorBlock) data.acquireRead();\n+ // TODO improve truncation to check all dimensions\n+ warnOnTrunction(tensor, rows, cols);\n+ out = DataConverter.toString(tensor, sparse, separator,\n+ lineseparator, \"[\", \"]\", rows, cols, decimal);\n+ }\nelse if( data instanceof FrameObject ) {\nFrameBlock frame = (FrameBlock) data.acquireRead();\nwarnOnTrunction(frame, rows, cols);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/VariableCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/VariableCPInstruction.java", "diff": "@@ -42,9 +42,9 @@ import org.tugraz.sysds.runtime.controlprogram.caching.CacheableData;\nimport org.tugraz.sysds.runtime.controlprogram.caching.FrameObject;\nimport org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject;\nimport org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject.UpdateType;\n+import org.tugraz.sysds.runtime.controlprogram.caching.TensorObject;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.controlprogram.parfor.util.IDSequence;\n-import org.tugraz.sysds.runtime.data.TensorBlockData;\nimport org.tugraz.sysds.runtime.instructions.Instruction;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\nimport org.tugraz.sysds.runtime.io.FileFormatProperties;\n@@ -491,7 +491,8 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\n{\ncase CreateVariable:\n- if ( getInput1().getDataType() == DataType.MATRIX ) {\n+ if ( getInput1().getDataType() == DataType.MATRIX\n+ || getInput1().getDataType() == DataType.TENSOR ) {\n//create new variable for symbol table and cache\n//(existing objects gets cleared through rmvar instructions)\nString fname = getInput2().getName();\n@@ -500,18 +501,23 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\nfname = new StringBuilder(fname.length()+16).append(fname)\n.append('_').append(_uniqueVarID.getNextID()).toString();\n}\n- MatrixObject mobj = new MatrixObject(getInput1().getValueType(), fname );\n+ CacheableData<?> obj = getInput1().getDataType().isMatrix() ?\n+ new MatrixObject(getInput1().getValueType(), fname) :\n+ new TensorObject(getInput1().getValueType(), fname);\n//clone meta data because it is updated on copy-on-write, otherwise there\n//is potential for hidden side effects between variables.\n- mobj.setMetaData((MetaData)metadata.clone());\n- mobj.setFileFormatProperties(_formatProperties);\n- mobj.setUpdateType(_updateType);\n- mobj.enableCleanup(!getInput1().getName()\n+ obj.setMetaData((MetaData)metadata.clone());\n+ obj.setFileFormatProperties(_formatProperties);\n+ obj.enableCleanup(!getInput1().getName()\n.startsWith(org.tugraz.sysds.lops.Data.PREAD_PREFIX));\n- ec.setVariable(getInput1().getName(), mobj);\n+ ec.setVariable(getInput1().getName(), obj);\n+\n+ if( obj instanceof MatrixObject ) {\n+ ((MatrixObject)obj).setUpdateType(_updateType);\nif(DMLScript.STATISTICS && _updateType.isInPlace())\nStatistics.incrementTotalUIPVar();\n}\n+ }\nelse if( getInput1().getDataType() == DataType.FRAME ) {\nString fname = getInput2().getName();\nFrameObject fobj = new FrameObject(fname);\n@@ -527,29 +533,6 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\n//created variable not called for scalars\nec.setScalarOutput(getInput1().getName(), null);\n}\n- else if ( getInput1().getDataType() == DataType.TENSOR ) {\n- //create new variable for symbol table and cache\n- //(existing objects gets cleared through rmvar instructions)\n- //String fname = getInput2().getName();\n- // check if unique filename needs to be generated\n- //if( Boolean.parseBoolean(getInput3().getName()) ) {\n- // fname = new StringBuilder(fname.length()+16).append(fname)\n- // .append('_').append(_uniqueVarID.getNextID()).toString();\n- //}\n- // TODO Cacheable tensor block\n- TensorBlockData tensor = new TensorBlockData(getInput1().getValueType());\n- //MatrixObject mobj = new MatrixObject(getInput1().getValueType(), fname );\n- //clone meta data because it is updated on copy-on-write, otherwise there\n- //is potential for hidden side effects between variables.\n- //mobj.setMetaData((MetaData)metadata.clone());\n- //mobj.setFileFormatProperties(_formatProperties);\n- //mobj.setUpdateType(_updateType);\n- //mobj.enableCleanup(!getInput1().getName()\n- // .startsWith(org.tugraz.sysds.lops.Data.PREAD_PREFIX));\n- ec.setVariable(getInput1().getName(), tensor);\n- //if(DMLScript.STATISTICS && _updateType.isInPlace())\n- // Statistics.incrementTotalUIPVar();\n- }\nelse {\nthrow new DMLRuntimeException(\"Unexpected data type: \" + getInput1().getDataType());\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/meta/MatrixCharacteristics.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/meta/MatrixCharacteristics.java", "diff": "@@ -56,7 +56,7 @@ public class MatrixCharacteristics implements Serializable\n}\npublic MatrixCharacteristics(MatrixCharacteristics that) {\n- set(that.numRows, that.numColumns, that.numRowsPerBlock, that.numColumnsPerBlock, that.nonZero);\n+ set(that);\n}\npublic MatrixCharacteristics set(long nr, long nc, int bnr, int bnc) {\n@@ -68,21 +68,15 @@ public class MatrixCharacteristics implements Serializable\n}\npublic MatrixCharacteristics set(long nr, long nc, int bnr, int bnc, long nnz) {\n- numRows = nr;\n- numColumns = nc;\n- numRowsPerBlock = bnr;\n- numColumnsPerBlock = bnc;\n+ set(nr, nc, bnr, bnc);\nnonZero = nnz;\nubNnz = false;\nreturn this;\n}\npublic MatrixCharacteristics set(MatrixCharacteristics that) {\n- numRows = that.numRows;\n- numColumns = that.numColumns;\n- numRowsPerBlock = that.numRowsPerBlock;\n- numColumnsPerBlock = that.numColumnsPerBlock;\n- nonZero = that.nonZero;\n+ set(that.numRows, that.numColumns, that.numRowsPerBlock,\n+ that.numColumnsPerBlock, that.nonZero);\nubNnz = that.ubNnz;\nreturn this;\n}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/main/java/org/tugraz/sysds/runtime/meta/TensorCharacteristics.java", "diff": "+/*\n+ * Licensed to the Apache Software Foundation (ASF) under one\n+ * or more contributor license agreements. See the NOTICE file\n+ * distributed with this work for additional information\n+ * regarding copyright ownership. The ASF licenses this file\n+ * to you under the Apache License, Version 2.0 (the\n+ * \"License\"); you may not use this file except in compliance\n+ * with the License. You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing,\n+ * software distributed under the License is distributed on an\n+ * \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+ * KIND, either express or implied. See the License for the\n+ * specific language governing permissions and limitations\n+ * under the License.\n+ */\n+\n+\n+package org.tugraz.sysds.runtime.meta;\n+\n+import java.io.Serializable;\n+import java.util.Arrays;\n+\n+import org.tugraz.sysds.runtime.util.UtilFunctions;\n+\n+\n+public class TensorCharacteristics implements Serializable\n+{\n+ private static final long serialVersionUID = 8300479822915546000L;\n+\n+ private long[] _dims;\n+ private int[] _blkSizes;\n+ private long _nnz = -1;\n+\n+ public TensorCharacteristics() {}\n+\n+ public TensorCharacteristics(long[] dims, long nnz) {\n+ int[] blkSizes = new int[dims.length];\n+ Arrays.fill(blkSizes, -1);\n+ set(dims, blkSizes, nnz);\n+ }\n+\n+ public TensorCharacteristics(long[] dims, int[] blkSizes) {\n+ set(dims, blkSizes, -1);\n+ }\n+\n+ public TensorCharacteristics(long[] dims, int[] blkSizes, long nnz) {\n+ set(dims, blkSizes, nnz);\n+ }\n+\n+ public TensorCharacteristics(TensorCharacteristics that) {\n+ set(that);\n+ }\n+\n+ public TensorCharacteristics set(long[] dims, int[] blkSizes) {\n+ set(dims, blkSizes, -1);\n+ return this;\n+ }\n+\n+ public TensorCharacteristics set(long[] dims, int[] blkSizes, long nnz) {\n+ _dims = dims;\n+ _blkSizes = blkSizes;\n+ return this;\n+ }\n+\n+ public TensorCharacteristics set(TensorCharacteristics that) {\n+ set(that._dims, that._blkSizes, that._nnz);\n+ return this;\n+ }\n+\n+ public int getNumDims() {\n+ return _dims.length;\n+ }\n+\n+ public long getDim(int i) {\n+ return _dims[i];\n+ }\n+\n+ public TensorCharacteristics setDim(int i, long dim) {\n+ _dims[i] = dim;\n+ return this;\n+ }\n+\n+ public TensorCharacteristics setDims(long[] dims) {\n+ _dims = dims;\n+ return this;\n+ }\n+\n+ public long getBlockSize(int i) {\n+ return _blkSizes[i];\n+ }\n+\n+ public TensorCharacteristics setBlockSize(int i, int blksize) {\n+ _blkSizes[i] = blksize;\n+ return this;\n+ }\n+\n+ public TensorCharacteristics setBlockSizes(int[] blkSizes) {\n+ _blkSizes = blkSizes;\n+ return this;\n+ }\n+\n+ public long getLength() {\n+ return UtilFunctions.prod(_dims);\n+ }\n+\n+ public long getNumBlocks() {\n+ long ret = 1;\n+ for( int i=0; i<getNumDims(); i++ )\n+ ret *= getNumBlocks(i);\n+ return ret;\n+ }\n+\n+ public long getNumBlocks(int i) {\n+ return Math.max((long) Math.ceil((double)getDim(i) / getBlockSize(i)), 1);\n+ }\n+\n+ @Override\n+ public String toString() {\n+ return \"[\"+Arrays.toString(_dims)+\", nnz=\"+_nnz\n+ + \", blocks \"+Arrays.toString(_blkSizes)+\"]\";\n+ }\n+\n+ @Override\n+ public boolean equals (Object anObject) {\n+ if( !(anObject instanceof TensorCharacteristics) )\n+ return false;\n+ TensorCharacteristics tc = (TensorCharacteristics) anObject;\n+ return Arrays.equals(_dims, tc._dims)\n+ && Arrays.equals(_blkSizes, tc._blkSizes)\n+ && _nnz == tc._nnz;\n+ }\n+\n+ @Override\n+ public int hashCode() {\n+ return UtilFunctions.intHashCode(UtilFunctions.intHashCode(\n+ Arrays.hashCode(_dims), Arrays.hashCode(_blkSizes)),\n+ Long.hashCode(_nnz));\n+ }\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/util/UtilFunctions.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/util/UtilFunctions.java", "diff": "@@ -698,6 +698,13 @@ public class UtilFunctions\nreturn StreamSupport.stream(iterable.spliterator(), false);\n}\n+ public static long prod(long[] arr) {\n+ long ret = 1;\n+ for(int i=0; i<arr.length; i++)\n+ ret *= arr[i];\n+ return ret;\n+ }\n+\npublic static long prod(int[] arr) {\nlong ret = 1;\nfor(int i=0; i<arr.length; i++)\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Types.java", "new_path": "src/main/java/org/tugraz/sysds/common/Types.java", "diff": "@@ -55,6 +55,9 @@ public class Types\npublic boolean isList() {\nreturn this == LIST;\n}\n+ public boolean isUnknown() {\n+ return this == UNKNOWN;\n+ }\n}\n/**\n@@ -65,6 +68,9 @@ public class Types\npublic boolean isNumeric() {\nreturn this == INT32 || this == INT64 || this == FP32 || this == FP64;\n}\n+ public boolean isUnknown() {\n+ return this == UNKNOWN;\n+ }\npublic boolean isPseudoNumeric() {\nreturn isNumeric() || this == BOOLEAN;\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/DataGenOp.java", "new_path": "src/main/java/org/tugraz/sysds/hops/DataGenOp.java", "diff": "@@ -86,7 +86,9 @@ public class DataGenOp extends MultiThreadedHop\n* @param inputParameters HashMap of the input parameters for Rand Hop\n*/\npublic DataGenOp(DataGenMethod mthd, DataIdentifier id, HashMap<String, Hop> inputParameters) {\n- super(id.getName(), id.getDataType(), id.getValueType());\n+ super(id.getName(),\n+ id.getDataType().isUnknown() ? DataType.MATRIX : id.getDataType(),\n+ id.getValueType().isUnknown() ? ValueType.FP64 : id.getValueType());\n_id = id;\n_op = mthd;\n@@ -171,8 +173,8 @@ public class DataGenOp extends MultiThreadedHop\ninputLops.put(cur.getKey(), getInput().get(cur.getValue()).constructLops());\n}\n- DataGen rnd = new DataGen(_op, _id, inputLops,_baseDir,\n- getDataType(), getValueType(), et);\n+ DataGen rnd = new DataGen(_op, _id, inputLops,\n+ _baseDir, getDataType(), getValueType(), et);\nint k = OptimizerUtils.getConstrainedNumThreads(_maxNumThreads);\nrnd.setNumThreads(k);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/rewrite/HopRewriteUtils.java", "new_path": "src/main/java/org/tugraz/sysds/hops/rewrite/HopRewriteUtils.java", "diff": "@@ -305,6 +305,7 @@ public class HopRewriteUtils\nHashMap<String, Hop> params = new HashMap<>();\nparams.put(DataExpression.RAND_ROWS, rows);\nparams.put(DataExpression.RAND_COLS, cols);\n+ params.put(DataExpression.RAND_DIMS, new LiteralOp(\"-1\")); //TODO\nparams.put(DataExpression.RAND_MIN, val);\nparams.put(DataExpression.RAND_MAX, val);\nparams.put(DataExpression.RAND_PDF, new LiteralOp(DataExpression.RAND_PDF_UNIFORM));\n@@ -338,6 +339,7 @@ public class HopRewriteUtils\nHop cols = inputGen.getInput().get(params.get(DataExpression.RAND_COLS));\nHop min = inputGen.getInput().get(params.get(DataExpression.RAND_MIN));\nHop max = inputGen.getInput().get(params.get(DataExpression.RAND_MAX));\n+ Hop dims = inputGen.getInput().get(params.get(DataExpression.RAND_DIMS));\nHop pdf = inputGen.getInput().get(params.get(DataExpression.RAND_PDF));\nHop mean = inputGen.getInput().get(params.get(DataExpression.RAND_LAMBDA));\nHop sparsity = inputGen.getInput().get(params.get(DataExpression.RAND_SPARSITY));\n@@ -361,6 +363,7 @@ public class HopRewriteUtils\nparams2.put(DataExpression.RAND_COLS, cols);\nparams2.put(DataExpression.RAND_MIN, sminHop);\nparams2.put(DataExpression.RAND_MAX, smaxHop);\n+ params2.put(DataExpression.RAND_DIMS, dims);\nparams2.put(DataExpression.RAND_PDF, pdf);\nparams2.put(DataExpression.RAND_LAMBDA, mean);\nparams2.put(DataExpression.RAND_SPARSITY, sparsity);\n@@ -388,6 +391,7 @@ public class HopRewriteUtils\nHashMap<String, Hop> params = new HashMap<>();\nparams.put(DataExpression.RAND_ROWS, rows);\nparams.put(DataExpression.RAND_COLS, cols);\n+ params.put(DataExpression.RAND_DIMS, new LiteralOp(\"-1\")); //TODO\nparams.put(DataExpression.RAND_MIN, val);\nparams.put(DataExpression.RAND_MAX, val);\nparams.put(DataExpression.RAND_PDF, new LiteralOp(DataExpression.RAND_PDF_UNIFORM));\n@@ -422,6 +426,7 @@ public class HopRewriteUtils\nparams.put(DataExpression.RAND_COLS, cols);\nparams.put(DataExpression.RAND_MIN, val);\nparams.put(DataExpression.RAND_MAX, val);\n+ params.put(DataExpression.RAND_DIMS, new LiteralOp(\"-1\")); //TODO\nparams.put(DataExpression.RAND_PDF, new LiteralOp(DataExpression.RAND_PDF_UNIFORM));\nparams.put(DataExpression.RAND_LAMBDA,new LiteralOp(-1.0));\nparams.put(DataExpression.RAND_SPARSITY, new LiteralOp(1.0));\n@@ -444,8 +449,8 @@ public class HopRewriteUtils\nHashMap<String, Hop> params = new HashMap<>();\nparams.put(DataExpression.RAND_ROWS, rowInput);\nparams.put(DataExpression.RAND_COLS, colInput);\n- params.put(DataExpression.RAND_MIN, val);\nparams.put(DataExpression.RAND_DIMS, dimsInput);\n+ params.put(DataExpression.RAND_MIN, val);\nparams.put(DataExpression.RAND_MAX, val);\nparams.put(DataExpression.RAND_PDF, new LiteralOp(DataExpression.RAND_PDF_UNIFORM));\nparams.put(DataExpression.RAND_LAMBDA, new LiteralOp(-1.0));\n@@ -479,6 +484,7 @@ public class HopRewriteUtils\nHashMap<String, Hop> params = new HashMap<>();\nparams.put(DataExpression.RAND_ROWS, new LiteralOp(rows));\nparams.put(DataExpression.RAND_COLS, new LiteralOp(cols));\n+ params.put(DataExpression.RAND_DIMS, new LiteralOp(\"-1\")); //TODO\nparams.put(DataExpression.RAND_MIN, str);\nparams.put(DataExpression.RAND_MAX, str);\nparams.put(DataExpression.RAND_SEED, new LiteralOp(DataGenOp.UNSPECIFIED_SEED));\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/parser/DataIdentifier.java", "new_path": "src/main/java/org/tugraz/sysds/parser/DataIdentifier.java", "diff": "@@ -34,14 +34,6 @@ public class DataIdentifier extends Identifier\nsetParseInfo(passed);\n}\n- @Override\n- public Expression rewriteExpression(String prefix) {\n- DataIdentifier newId = new DataIdentifier(this);\n- String newIdName = prefix + _name;\n- newId.setName(newIdName);\n- return newId;\n- }\n-\npublic DataIdentifier(String name){\nsuper();\n_name = name;\n@@ -51,6 +43,14 @@ public class DataIdentifier extends Identifier\n_name = null;\n}\n+ @Override\n+ public Expression rewriteExpression(String prefix) {\n+ DataIdentifier newId = new DataIdentifier(this);\n+ String newIdName = prefix + _name;\n+ newId.setName(newIdName);\n+ return newId;\n+ }\n+\npublic String getName(){\nreturn _name;\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/RandSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/RandSPInstruction.java", "diff": "@@ -159,7 +159,7 @@ public class RandSPInstruction extends UnarySPInstruction {\nDataGenMethod method = DataGenMethod.INVALID;\nif ( opcode.equalsIgnoreCase(DataGen.RAND_OPCODE) ) {\nmethod = DataGenMethod.RAND;\n- InstructionUtils.checkNumFields ( str, 12 );\n+ InstructionUtils.checkNumFields ( str, 13 );\n}\nelse if ( opcode.equalsIgnoreCase(DataGen.SEQ_OPCODE) ) {\nmethod = DataGenMethod.SEQ;\n@@ -179,20 +179,21 @@ public class RandSPInstruction extends UnarySPInstruction {\nif ( method == DataGenMethod.RAND ) {\nCPOperand rows = new CPOperand(s[1]);\nCPOperand cols = new CPOperand(s[2]);\n- int rpb = Integer.parseInt(s[3]);\n- int cpb = Integer.parseInt(s[4]);\n- double minValue = !s[5].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n- Double.valueOf(s[5]).doubleValue() : -1;\n- double maxValue = !s[6].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ //TODO: handle dims param s[3]\n+ int rpb = Integer.parseInt(s[4]);\n+ int cpb = Integer.parseInt(s[5]);\n+ double minValue = !s[6].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\nDouble.valueOf(s[6]).doubleValue() : -1;\n- double sparsity = !s[7].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ double maxValue = !s[7].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\nDouble.valueOf(s[7]).doubleValue() : -1;\n- long seed = !s[8].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n- Long.valueOf(s[8]).longValue() : -1;\n- String dir = s[9];\n- String pdf = s[10];\n- String pdfParams = !s[11].contains( Lop.VARIABLE_NAME_PLACEHOLDER) ?\n- s[11] : null;\n+ double sparsity = !s[8].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ Double.valueOf(s[8]).doubleValue() : -1;\n+ long seed = !s[9].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ Long.valueOf(s[9]).longValue() : -1;\n+ String dir = s[10];\n+ String pdf = s[11];\n+ String pdfParams = !s[12].contains( Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ s[12] : null;\nreturn new RandSPInstruction(op, method, null, out, rows, cols, rpb, cpb, minValue, maxValue, sparsity, seed, dir, pdf, pdfParams, opcode, str);\n}\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Types.java", "new_path": "src/main/java/org/tugraz/sysds/common/Types.java", "diff": "@@ -95,4 +95,14 @@ public class Types\n}\n}\n}\n+\n+ /**\n+ * Serialization block types (empty, dense, sparse, ultra-sparse)\n+ */\n+ public enum BlockType{\n+ EMPTY_BLOCK,\n+ ULTRA_SPARSE_BLOCK,\n+ SPARSE_BLOCK,\n+ DENSE_BLOCK,\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/caching/CacheDataInput.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/caching/CacheDataInput.java", "diff": "@@ -106,7 +106,10 @@ public class CacheDataInput implements DataInput, MatrixBlockDataInput\n@Override\npublic float readFloat() throws IOException {\n- throw new IOException(\"Not supported.\");\n+ int tmp = baToInt(_buff, _count);\n+ float tmp2 = Float.intBitsToFloat(tmp);\n+ _count += 4;\n+ return tmp2;\n}\n@Override\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/caching/CacheDataOutput.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/caching/CacheDataOutput.java", "diff": "@@ -102,7 +102,9 @@ public class CacheDataOutput implements DataOutput, MatrixBlockDataOutput\n@Override\npublic void writeFloat(float v) throws IOException {\n- throw new IOException(\"Not supported.\");\n+ int tmp = Float.floatToRawIntBits(v);\n+ intToBa(tmp, _buff, _count);\n+ _count += 4;\n}\n@Override\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/TensorBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/TensorBlock.java", "diff": "@@ -21,6 +21,7 @@ import java.io.DataOutput;\nimport java.io.IOException;\nimport org.apache.commons.lang.NotImplementedException;\n+import org.tugraz.sysds.common.Types.BlockType;\nimport org.tugraz.sysds.common.Types.ValueType;\nimport org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.caching.CacheBlock;\n@@ -35,7 +36,7 @@ public class TensorBlock implements CacheBlock\npublic static final SparseBlock.Type DEFAULT_SPARSEBLOCK = SparseBlock.Type.MCSR;\n//constant value type of tensor block\n- protected final ValueType _vt;\n+ protected ValueType _vt;\n//min 2 dimensions to preserve proper matrix semantics\nprotected int[] _dims; //[2,inf)\n@@ -223,7 +224,8 @@ public class TensorBlock implements CacheBlock\nreturn getDim(0);\n}\n- public int getNumCols() {\n+ @Override\n+ public int getNumColumns() {\nreturn getDim(1);\n}\n@@ -303,6 +305,85 @@ public class TensorBlock implements CacheBlock\nreturn _sparseBlock;\n}\n+ ////////\n+ // Input/Output functions\n+\n+ @Override\n+ @SuppressWarnings(\"incomplete-switch\")\n+ public void readFields(DataInput in)\n+ throws IOException\n+ {\n+ //step 1: read header\n+ _vt = ValueType.values()[in.readByte()];\n+ _dims = new int[in.readInt()];\n+ for(int i=0; i<_dims.length; i++)\n+ _dims[i] = in.readInt();\n+ _nnz = in.readLong();\n+\n+ //step 2: read tensor block data\n+ switch( BlockType.values()[in.readByte()] ) {\n+ case EMPTY_BLOCK:\n+ reset(_dims);\n+ return;\n+ case DENSE_BLOCK: {\n+ allocateDenseBlock(false);\n+ DenseBlock a = getDenseBlock();\n+ int odims = (int) UtilFunctions.prod(_dims, 1);\n+ for( int i=0; i<getNumRows(); i++ )\n+ for( int j=0; j<odims; j++ ) {\n+ switch( _vt ) {\n+ case FP32: a.set(i, j, in.readFloat()); break;\n+ case FP64: a.set(i, j, in.readDouble()); break;\n+ case INT32: a.set(i, j, in.readInt()); break;\n+ case INT64: a.set(i, j, in.readLong()); break;\n+ case BOOLEAN: a.set(i, j, in.readByte()); break;\n+ }\n+ }\n+ break;\n+ }\n+ case SPARSE_BLOCK:\n+ case ULTRA_SPARSE_BLOCK:\n+ throw new NotImplementedException();\n+ }\n+ }\n+\n+ @Override\n+ @SuppressWarnings(\"incomplete-switch\")\n+ public void write(DataOutput out)\n+ throws IOException\n+ {\n+ //step 1: write header\n+ out.writeByte(_vt.ordinal()); // value type\n+ out.writeInt(getNumDims()); // num dims\n+ for(int i=0; i<getNumDims(); i++)\n+ out.writeInt(getDim(i)); // dim\n+ out.writeLong(getNonZeros()); // nnz\n+\n+ //step 2: write tensor block data\n+ if( isEmpty(false) ) {\n+ //empty blocks do not need to materialize row information\n+ out.writeByte(BlockType.EMPTY_BLOCK.ordinal());\n+ }\n+ else if( !isSparse() ) {\n+ out.writeByte(BlockType.DENSE_BLOCK.ordinal());\n+ DenseBlock a = getDenseBlock();\n+ int odims = (int) UtilFunctions.prod(_dims, 1);\n+ for( int i=0; i<getNumRows(); i++ )\n+ for( int j=0; j<odims; j++ ) {\n+ switch( _vt ) {\n+ case FP32: out.writeFloat((float)a.get(i, j)); break;\n+ case FP64: out.writeDouble(a.get(i, j)); break;\n+ case INT32: out.writeInt((int)a.get(i, j)); break;\n+ case INT64: out.writeLong((long)a.get(i, j)); break;\n+ case BOOLEAN: out.writeBoolean(a.get(i, j) != 0); break;\n+ }\n+ }\n+ }\n+ else {\n+ throw new NotImplementedException();\n+ }\n+ }\n+\n////////\n// Basic modification\n@@ -316,6 +397,16 @@ public class TensorBlock implements CacheBlock\n}\n}\n+ public double get(int r, int c) {\n+ if (_sparse) {\n+ // TODO: Implement sparse\n+ throw new NotImplementedException();\n+ //return _sparseBlock.get(ix);\n+ } else {\n+ return _denseBlock.get(r, c);\n+ }\n+ }\n+\npublic String getString(int[] ix) {\nif (_sparse) {\n// TODO: Implement sparse\n@@ -334,6 +425,14 @@ public class TensorBlock implements CacheBlock\n}\n}\n+ public void set(int r, int c, double v) {\n+ if (_sparse) {\n+ throw new NotImplementedException();\n+ } else {\n+ _denseBlock.set(r, c, v);\n+ }\n+ }\n+\npublic void set(int[] ix, String v) {\nif (_sparse) {\nthrow new NotImplementedException();\n@@ -421,24 +520,6 @@ public class TensorBlock implements CacheBlock\nreturn false;\n}\n- @Override\n- public void write(DataOutput out) throws IOException {\n- // TODO Auto-generated method stub\n-\n- }\n-\n- @Override\n- public void readFields(DataInput in) throws IOException {\n- // TODO Auto-generated method stub\n-\n- }\n-\n- @Override\n- public int getNumColumns() {\n- // TODO Auto-generated method stub\n- return 0;\n- }\n-\n@Override\npublic long getInMemorySize() {\n// TODO Auto-generated method stub\n@@ -447,8 +528,30 @@ public class TensorBlock implements CacheBlock\n@Override\npublic long getExactSerializedSize() {\n- // TODO Auto-generated method stub\n- return 0;\n+ //header size (vt, num dims, dims, nnz, type)\n+ long size = 4 * (1+_dims.length) + 8 + 2;\n+ //serialized representation\n+ if( !isSparse() ) {\n+ switch( _vt ) {\n+ case INT32:\n+ case FP32:\n+ size += 4 * getLength(); break;\n+ case INT64:\n+ case FP64:\n+ size += 8 * getLength(); break;\n+ case BOOLEAN:\n+ //TODO perf bits instead of bytes\n+ size += getLength(); break;\n+ //size += Math.ceil((double)getLength() / 64); break;\n+ case STRING:\n+ case UNKNOWN:\n+ throw new NotImplementedException();\n+ }\n+ }\n+ else {\n+ throw new NotImplementedException();\n+ }\n+ return size;\n}\n@Override\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/ParameterizedBuiltinCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/ParameterizedBuiltinCPInstruction.java", "diff": "@@ -378,15 +378,14 @@ public class ParameterizedBuiltinCPInstruction extends ComputationCPInstruction\nprivate void warnOnTrunction(TensorBlock data, int rows, int cols) {\n//warn on truncation because users might not be aware and use toString for verification\nif( (getParam(\"rows\")==null && data.getNumRows()>rows)\n- || (getParam(\"cols\")==null && data.getNumCols()>cols) )\n+ || (getParam(\"cols\")==null && data.getNumColumns()>cols) )\n{\nStringBuilder sb = new StringBuilder();\nIntStream.range(0, data.getNumDims()).forEach((i) -> {\n- if ((i == data.getNumDims() - 1)) {\n+ if ((i == data.getNumDims() - 1))\nsb.append(data.getDim(i));\n- } else {\n+ else\nsb.append(data.getDim(i)).append(\"x\");\n- }\n});\nLOG.warn(\"Truncating \"+data.getClass().getSimpleName()+\" of size \"+sb.toString()+\" to \"+rows+\"x\"+cols+\". \"\n+ \"Use toString(X, rows=..., cols=...) if necessary.\");\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/matrix/data/MatrixBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/matrix/data/MatrixBlock.java", "diff": "@@ -38,6 +38,7 @@ import java.util.stream.IntStream;\nimport org.apache.commons.lang3.concurrent.ConcurrentUtils;\nimport org.apache.commons.math3.random.Well1024a;\nimport org.apache.hadoop.io.DataInputBuffer;\n+import org.tugraz.sysds.common.Types.BlockType;\nimport org.tugraz.sysds.conf.ConfigurationManager;\nimport org.tugraz.sysds.hops.OptimizerUtils;\nimport org.tugraz.sysds.lops.MMTSJ.MMTSJType;\n@@ -129,13 +130,6 @@ public class MatrixBlock extends MatrixValue implements CacheBlock, Externalizab\n//basic header (int rlen, int clen, byte type)\npublic static final int HEADER_SIZE = 9;\n- public enum BlockType{\n- EMPTY_BLOCK,\n- ULTRA_SPARSE_BLOCK, //ultra sparse representation, in-mem same as sparse\n- SPARSE_BLOCK, //sparse representation, see sparseRows\n- DENSE_BLOCK, //dense representation, see denseBlock\n- }\n-\n//matrix meta data\nprotected int rlen = -1;\nprotected int clen = -1;\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/util/DataConverter.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/util/DataConverter.java", "diff": "@@ -729,6 +729,27 @@ public class DataConverter\nreturn frame;\n}\n+ public static TensorBlock convertToTensorBlock(MatrixBlock mb, ValueType vt) {\n+ TensorBlock ret = new TensorBlock(vt, new int[] {mb.getNumRows(), mb.getNumColumns()});\n+ ret.allocateDenseBlock(true);\n+ if( mb.getNonZeros() > 0 ) {\n+ if( mb.isInSparseFormat() ) {\n+ Iterator<IJV> iter = mb.getSparseBlockIterator();\n+ while( iter.hasNext() ) {\n+ IJV cell = iter.next();\n+ ret.set(cell.getI(), cell.getJ(), cell.getV());\n+ }\n+ }\n+ else {\n+ double[] a = mb.getDenseBlockValues();\n+ for( int i=0, ix=0; i<mb.getNumRows(); i++ )\n+ for( int j=0; j<mb.getNumColumns(); j++, ix++ )\n+ ret.set(i, j, a[ix]);\n+ }\n+ }\n+ return ret;\n+ }\n+\npublic static MatrixBlock[] convertToMatrixBlockPartitions( MatrixBlock mb, boolean colwise )\n{\nMatrixBlock[] ret = null;\n@@ -909,7 +930,7 @@ public class DataConverter\n}\npublic static String toString(TensorBlock mb) {\n- return toString(mb, false, \" \", \"\\n\", \"[\", \"]\", mb.getNumRows(), mb.getNumCols(), 3);\n+ return toString(mb, false, \" \", \"\\n\", \"[\", \"]\", mb.getNumRows(), mb.getNumColumns(), 3);\n}\n/**\n@@ -932,7 +953,7 @@ public class DataConverter\n// Setup number of rows and columns to print\nint rlen = tb.getNumRows();\n- int clen = tb.getNumCols();\n+ int clen = tb.getNumColumns();\nint rowLength = rlen;\nint colLength = clen;\nif (rowsToPrint >= 0)\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/TestUtils.java", "new_path": "src/test/java/org/tugraz/sysds/test/TestUtils.java", "diff": "@@ -2043,6 +2043,11 @@ public class TestUtils\nreturn data;\n}\n+ public static MatrixBlock round(MatrixBlock data) {\n+ return DataConverter.convertToMatrixBlock(\n+ round(DataConverter.convertToDoubleMatrix(data)));\n+ }\n+\npublic static double[][] floor(double[][] data) {\nfor(int i=0; i<data.length; i++)\nfor(int j=0; j<data[i].length; j++)\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/component/tensor/TensorConstructionTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/component/tensor/TensorConstructionTest.java", "diff": "@@ -30,7 +30,7 @@ public class TensorConstructionTest\nAssert.assertEquals(ValueType.FP64, tb.getValueType());\nAssert.assertEquals(2, tb.getNumDims());\nAssert.assertEquals(0, tb.getNumRows());\n- Assert.assertEquals(0, tb.getNumCols());\n+ Assert.assertEquals(0, tb.getNumColumns());\nAssert.assertEquals(0, tb.getNonZeros());\nAssert.assertTrue(tb.isSparse());\nAssert.assertFalse(tb.isMatrix());\n@@ -42,7 +42,7 @@ public class TensorConstructionTest\nAssert.assertEquals(ValueType.FP64, tb.getValueType());\nAssert.assertEquals(2, tb.getNumDims());\nAssert.assertEquals(1, tb.getNumRows());\n- Assert.assertEquals(1, tb.getNumCols());\n+ Assert.assertEquals(1, tb.getNumColumns());\nAssert.assertEquals(1, tb.getNonZeros());\nAssert.assertFalse(tb.isSparse());\nAssert.assertFalse(tb.isMatrix());\n@@ -54,7 +54,7 @@ public class TensorConstructionTest\nAssert.assertEquals(ValueType.INT64, tb.getValueType());\nAssert.assertEquals(3, tb.getNumDims());\nAssert.assertEquals(11, tb.getNumRows());\n- Assert.assertEquals(12, tb.getNumCols());\n+ Assert.assertEquals(12, tb.getNumColumns());\nAssert.assertEquals(13, tb.getDim(2));\nAssert.assertEquals(0, tb.getNonZeros());\nAssert.assertTrue(tb.isSparse());\n@@ -67,7 +67,7 @@ public class TensorConstructionTest\nAssert.assertEquals(ValueType.INT64, tb.getValueType());\nAssert.assertEquals(3, tb.getNumDims());\nAssert.assertEquals(11, tb.getNumRows());\n- Assert.assertEquals(12, tb.getNumCols());\n+ Assert.assertEquals(12, tb.getNumColumns());\nAssert.assertEquals(13, tb.getDim(2));\nAssert.assertEquals(0, tb.getNonZeros());\nAssert.assertFalse(tb.isSparse());\n@@ -80,7 +80,7 @@ public class TensorConstructionTest\nAssert.assertEquals(ValueType.BOOLEAN, tb.getValueType());\nAssert.assertEquals(2, tb.getNumDims());\nAssert.assertEquals(11, tb.getNumRows());\n- Assert.assertEquals(12, tb.getNumCols());\n+ Assert.assertEquals(12, tb.getNumColumns());\nAssert.assertEquals(12, tb.getDim(1));\nAssert.assertEquals(0, tb.getNonZeros());\nAssert.assertTrue(tb.isSparse());\n@@ -93,7 +93,7 @@ public class TensorConstructionTest\nAssert.assertEquals(ValueType.FP64, tb.getValueType());\nAssert.assertEquals(2, tb.getNumDims());\nAssert.assertEquals(0, tb.getNumRows());\n- Assert.assertEquals(0, tb.getNumCols());\n+ Assert.assertEquals(0, tb.getNumColumns());\nAssert.assertEquals(0, tb.getNonZeros());\nAssert.assertTrue(tb.isSparse());\nAssert.assertFalse(tb.isMatrix());\n@@ -105,7 +105,7 @@ public class TensorConstructionTest\nAssert.assertEquals(ValueType.FP64, tb.getValueType());\nAssert.assertEquals(2, tb.getNumDims());\nAssert.assertEquals(1, tb.getNumRows());\n- Assert.assertEquals(1, tb.getNumCols());\n+ Assert.assertEquals(1, tb.getNumColumns());\nAssert.assertEquals(1, tb.getNonZeros());\nAssert.assertFalse(tb.isSparse());\nAssert.assertFalse(tb.isMatrix());\n@@ -117,7 +117,7 @@ public class TensorConstructionTest\nAssert.assertEquals(ValueType.INT64, tb.getValueType());\nAssert.assertEquals(3, tb.getNumDims());\nAssert.assertEquals(11, tb.getNumRows());\n- Assert.assertEquals(12, tb.getNumCols());\n+ Assert.assertEquals(12, tb.getNumColumns());\nAssert.assertEquals(13, tb.getDim(2));\nAssert.assertEquals(0, tb.getNonZeros());\nAssert.assertTrue(tb.isSparse());\n@@ -130,7 +130,7 @@ public class TensorConstructionTest\nAssert.assertEquals(ValueType.INT64, tb.getValueType());\nAssert.assertEquals(3, tb.getNumDims());\nAssert.assertEquals(11, tb.getNumRows());\n- Assert.assertEquals(12, tb.getNumCols());\n+ Assert.assertEquals(12, tb.getNumColumns());\nAssert.assertEquals(13, tb.getDim(2));\nAssert.assertEquals(0, tb.getNonZeros());\nAssert.assertFalse(tb.isSparse());\n@@ -143,7 +143,7 @@ public class TensorConstructionTest\nAssert.assertEquals(ValueType.BOOLEAN, tb.getValueType());\nAssert.assertEquals(2, tb.getNumDims());\nAssert.assertEquals(11, tb.getNumRows());\n- Assert.assertEquals(12, tb.getNumCols());\n+ Assert.assertEquals(12, tb.getNumColumns());\nAssert.assertEquals(12, tb.getDim(1));\nAssert.assertEquals(0, tb.getNonZeros());\nAssert.assertTrue(tb.isSparse());\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/java/org/tugraz/sysds/test/component/tensor/TensorSerializationTest.java", "diff": "+/*\n+ * Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed under the Apache License, Version 2.0 (the \"License\");\n+ * you may not use this file except in compliance with the License.\n+ * You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing, software\n+ * distributed under the License is distributed on an \"AS IS\" BASIS,\n+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+ * See the License for the specific language governing permissions and\n+ * limitations under the License.\n+ */\n+\n+package org.tugraz.sysds.test.component.tensor;\n+\n+import java.io.DataInput;\n+import java.io.DataOutput;\n+\n+import org.junit.Assert;\n+import org.junit.Test;\n+import org.tugraz.sysds.common.Types.ValueType;\n+import org.tugraz.sysds.runtime.DMLRuntimeException;\n+import org.tugraz.sysds.runtime.controlprogram.caching.CacheDataInput;\n+import org.tugraz.sysds.runtime.controlprogram.caching.CacheDataOutput;\n+import org.tugraz.sysds.runtime.data.TensorBlock;\n+import org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\n+import org.tugraz.sysds.runtime.util.DataConverter;\n+import org.tugraz.sysds.test.TestUtils;\n+\n+\n+public class TensorSerializationTest\n+{\n+ @Test\n+ public void testSerializeTensorFP32() {\n+ TensorBlock tb1 = createTensorBlock(ValueType.FP32, 70, 30, 0.7);\n+ TensorBlock tb2 = serializeAndDeserialize(tb1);\n+ compareTensorBlocks(tb1, tb2);\n+ }\n+\n+ @Test\n+ public void testSerializeTensorFP64() {\n+ TensorBlock tb1 = createTensorBlock(ValueType.FP64, 70, 30, 0.7);\n+ TensorBlock tb2 = serializeAndDeserialize(tb1);\n+ compareTensorBlocks(tb1, tb2);\n+ }\n+\n+ @Test\n+ public void testSerializeTensorINT32() {\n+ TensorBlock tb1 = createTensorBlock(ValueType.INT32, 70, 30, 0.7);\n+ TensorBlock tb2 = serializeAndDeserialize(tb1);\n+ compareTensorBlocks(tb1, tb2);\n+ }\n+\n+ @Test\n+ public void testSerializeTensorINT64() {\n+ TensorBlock tb1 = createTensorBlock(ValueType.INT64, 70, 30, 0.7);\n+ TensorBlock tb2 = serializeAndDeserialize(tb1);\n+ compareTensorBlocks(tb1, tb2);\n+ }\n+\n+ @Test\n+ public void testSerializeTensorBoolean() {\n+ TensorBlock tb1 = createTensorBlock(ValueType.BOOLEAN, 70, 30, 0.7);\n+ TensorBlock tb2 = serializeAndDeserialize(tb1);\n+ compareTensorBlocks(tb1, tb2);\n+ }\n+\n+ private TensorBlock serializeAndDeserialize(TensorBlock tb1) {\n+ try {\n+ //serialize and deserialize tensor block\n+ byte[] bdata = new byte[(int)tb1.getExactSerializedSize()];\n+ DataOutput dout = new CacheDataOutput(bdata);\n+ tb1.write(dout); //tb1 serialized into bdata\n+ DataInput din = new CacheDataInput(bdata);\n+ TensorBlock tb2 = new TensorBlock();\n+ tb2.readFields(din); //bdata deserialized into tb2\n+ return tb2;\n+ }\n+ catch(Exception ex) {\n+ throw new DMLRuntimeException(ex);\n+ }\n+ }\n+\n+ private TensorBlock createTensorBlock(ValueType vt, int rows, int cols, double sparsity) {\n+ return DataConverter.convertToTensorBlock(TestUtils.round(\n+ MatrixBlock.randOperations(rows, cols, sparsity, 0, 1, \"uniform\", 7)), vt);\n+ }\n+\n+ private void compareTensorBlocks(TensorBlock tb1, TensorBlock tb2) {\n+ Assert.assertEquals(tb1.getValueType(), tb2.getValueType());\n+ Assert.assertEquals(tb1.getNumRows(), tb2.getNumRows());\n+ Assert.assertEquals(tb1.getNumColumns(), tb2.getNumColumns());\n+ for(int i=0; i<tb1.getNumRows(); i++)\n+ for(int j=0; j<tb1.getNumColumns(); j++)\n+ Assert.assertEquals(Double.valueOf(tb1.get(i, j)),\n+ Double.valueOf(tb2.get(i, j)));\n+ }\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/io/binary/SerializeTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/io/binary/SerializeTest.java", "diff": "@@ -34,7 +34,6 @@ import org.tugraz.sysds.test.TestUtils;\npublic class SerializeTest extends AutomatedTestBase\n{\n-\nprivate final static String TEST_NAME = \"SerializeTest\";\nprivate final static String TEST_DIR = \"functions/io/binary/\";\nprivate final static String TEST_CLASS_DIR = TEST_DIR + SerializeTest.class.getSimpleName() + \"/\";\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -52,10 +52,10 @@ SYSTEMDS-60 Update SystemML improvements\nSYSTEMDS-70 Lineage Tracing and Reuse OK\n* 71 Basic collection of lineage traces OK\n- * 72 Serialization and deserialization of lineage traces\n- * 73 Deduplication of lineage traces\n- (loops, nested loops, branchless loops)\n- * 74 Performance features lineage tracing\n+ * 72 Serialization and deserialization of lineage traces OK\n+ * 73 Deduplication of lineage traces OK\n+ (loops, nested loops, branchless loops) OK\n+ * 74 Performance features lineage tracing OK\n* 75 Reuse cache based on lineage traces\n* 76 Generate runtime plan from lineage trace OK\n* 77 New builtin function for obtaining lineage OK\n@@ -75,11 +75,12 @@ SYSTEMDS-110 New Builtin Functions\n* 111 Time builtin function for script-level measurements OK\n* 112 Image data augmentation builtin functions OK\n* 113 Builtin functions for linear regression algorithms OK\n-<<<<<<< HEAD\n-=======\n-\n->>>>>>> Add LargeDenseBlock Representations with new support for Strings for DRB\nSYSTEMDS-120 Performance Features\n* 121 Avoid spark context creation on parfor result merge OK\n* 122 Reduce thread contention on parfor left indexing OK\n+\n+SYSTEMDS-130 IPA and Size Propagation\n+ * 131 New IPA pass for function call forwarding OK\n+ * 132 Hop Size Propagation for Tensors\n+ *\n\\ No newline at end of file\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/FunctionOp.java", "new_path": "src/main/java/org/tugraz/sysds/hops/FunctionOp.java", "diff": "@@ -102,6 +102,14 @@ public class FunctionOp extends Hop\n_fname = fname;\n}\n+ public void setFunctionNamespace( String fnamespace ) {\n+ _fnamespace = fnamespace;\n+ }\n+\n+ public void setInputVariableNames(String[] names) {\n+ _inputNames = names;\n+ }\n+\npublic ArrayList<Hop> getOutputs() {\nreturn _outputHops;\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/Hop.java", "new_path": "src/main/java/org/tugraz/sysds/hops/Hop.java", "diff": "@@ -715,6 +715,11 @@ public abstract class Hop implements ParseInfo\nh._parent.add(this);\n}\n+ public void addAllInputs( ArrayList<Hop> list ) {\n+ for( Hop h : list )\n+ addInput(h);\n+ }\n+\npublic int getRowsInBlock() {\nreturn _rows_in_block;\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/ipa/FunctionCallGraph.java", "new_path": "src/main/java/org/tugraz/sysds/hops/ipa/FunctionCallGraph.java", "diff": "@@ -184,6 +184,22 @@ public class FunctionCallGraph\n_fCallsSB.get(fkey).remove(sb);\n}\n+ /**\n+ * Replaces a function call to fkeyOld with a call to fkey,\n+ * but using the function op and statement block from the old.\n+ *\n+ * @param fkeyOld old function key of called function\n+ * @param fkey new function key of called function\n+ */\n+ public void replaceFunctionCalls(String fkeyOld, String fkey) {\n+ ArrayList<FunctionOp> fopTmp = _fCalls.get(fkeyOld);\n+ ArrayList<StatementBlock> sbTmp =_fCallsSB.get(fkeyOld);\n+ _fCalls.remove(fkeyOld);\n+ _fCallsSB.remove(fkeyOld);\n+ _fCalls.put(fkey, fopTmp);\n+ _fCallsSB.put(fkey, sbTmp);\n+ }\n+\n/**\n* Indicates if the given function is either directly or indirectly recursive.\n* An example of an indirect recursive function is foo2 in the following call\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/main/java/org/tugraz/sysds/hops/ipa/IPAPassForwardFunctionCalls.java", "diff": "+/*\n+ * Licensed to the Apache Software Foundation (ASF) under one\n+ * or more contributor license agreements. See the NOTICE file\n+ * distributed with this work for additional information\n+ * regarding copyright ownership. The ASF licenses this file\n+ * to you under the Apache License, Version 2.0 (the\n+ * \"License\"); you may not use this file except in compliance\n+ * with the License. You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing,\n+ * software distributed under the License is distributed on an\n+ * \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+ * KIND, either express or implied. See the License for the\n+ * specific language governing permissions and limitations\n+ * under the License.\n+ */\n+\n+package org.tugraz.sysds.hops.ipa;\n+\n+import java.util.ArrayList;\n+import java.util.Arrays;\n+import java.util.HashMap;\n+import java.util.HashSet;\n+import java.util.stream.IntStream;\n+\n+import org.tugraz.sysds.hops.FunctionOp;\n+import org.tugraz.sysds.hops.Hop;\n+import org.tugraz.sysds.hops.LiteralOp;\n+import org.tugraz.sysds.hops.Hop.DataOpTypes;\n+import org.tugraz.sysds.hops.rewrite.HopRewriteUtils;\n+import org.tugraz.sysds.parser.DMLProgram;\n+import org.tugraz.sysds.parser.FunctionStatement;\n+import org.tugraz.sysds.parser.FunctionStatementBlock;\n+\n+/**\n+ * This rewrite forwards a function call to a function with a simple\n+ * function call that only consumes function parameters and literals\n+ * into the original call location.\n+ */\n+public class IPAPassForwardFunctionCalls extends IPAPass\n+{\n+ @Override\n+ public boolean isApplicable(FunctionCallGraph fgraph) {\n+ return InterProceduralAnalysis.FORWARD_SIMPLE_FUN_CALLS;\n+ }\n+\n+ @Override\n+ public void rewriteProgram( DMLProgram prog, FunctionCallGraph fgraph, FunctionCallSizeInfo fcallSizes )\n+ {\n+ for( String fkey : fgraph.getReachableFunctions() ) {\n+ FunctionStatementBlock fsb = prog.getFunctionStatementBlock(fkey);\n+ FunctionStatement fstmt = (FunctionStatement)fsb.getStatement(0);\n+\n+ //step 1: basic application filter: simple forwarding call\n+ if( fstmt.getBody().size() > 1 || !containsFunctionOp(fstmt.getBody().get(0).getHops())\n+ || !hasOnlySimplyArguments((FunctionOp)fstmt.getBody().get(0).getHops().get(0)))\n+ continue;\n+ if( LOG.isDebugEnabled() )\n+ LOG.debug(\"IPA: Forward-function-call candidate L1: '\"+fkey+\"'\");\n+\n+ //step 2: check consistent output ordering\n+ FunctionOp call2 = (FunctionOp)fstmt.getBody().get(0).getHops().get(0);\n+ if( !hasConsistentOutputOrdering(fstmt, call2)\n+ || fgraph.getFunctionCalls(fkey).size() > 1)\n+ continue;\n+ if( LOG.isDebugEnabled() )\n+ LOG.debug(\"IPA: Forward-function-call candidate L2: '\"+fkey+\"'\");\n+\n+ //step 3: check and rewire input arguments (single call guaranteed)\n+ FunctionOp call1 = fgraph.getFunctionCalls(fkey).get(0);\n+ if( hasValidVariableNames(call1) && hasValidVariableNames(call2)\n+ && isFirstSubsetOfSecond(call2.getInputVariableNames(), call1.getInputVariableNames())) {\n+ //step 4: rewire input arguments\n+ call1.setFunctionName(call2.getFunctionName());\n+ call1.setFunctionNamespace(call2.getFunctionNamespace());\n+ reconcileFunctionInputsInPlace(call1, call2);\n+ //step 5: update function call graph\n+ fgraph.replaceFunctionCalls(call1.getFunctionKey(), fkey);\n+\n+ if( LOG.isDebugEnabled() )\n+ LOG.debug(\"IPA: Forward-function-call: replaced '\"\n+ + fkey +\"' with '\"+call2.getFunctionKey()+\"'\");\n+ }\n+ }\n+ }\n+\n+ private static boolean containsFunctionOp(ArrayList<Hop> hops) {\n+ if( hops==null || hops.isEmpty() )\n+ return false;\n+ Hop.resetVisitStatus(hops);\n+ boolean ret = HopRewriteUtils.containsOp(hops, FunctionOp.class);\n+ Hop.resetVisitStatus(hops);\n+ return ret;\n+ }\n+\n+ private static boolean hasOnlySimplyArguments(FunctionOp fop) {\n+ return fop.getInput().stream().allMatch(h -> h instanceof LiteralOp\n+ || HopRewriteUtils.isData(h, DataOpTypes.TRANSIENTREAD));\n+ }\n+\n+ private static boolean hasConsistentOutputOrdering(FunctionStatement fstmt, FunctionOp fop2) {\n+ int len = Math.min(fstmt.getOutputParams().size(), fop2.getOutputVariableNames().length);\n+ return IntStream.range(0, len).allMatch(i ->\n+ fstmt.getOutputParams().get(i).getName().equals(fop2.getOutputVariableNames()[i]));\n+ }\n+\n+ private static boolean hasValidVariableNames(FunctionOp fop) {\n+ return fop.getInputVariableNames() != null\n+ && Arrays.stream(fop.getInputVariableNames()).allMatch(s -> s != null);\n+ }\n+\n+ private static boolean isFirstSubsetOfSecond(String[] first, String[] second) {\n+ //build phase: second\n+ HashSet<String> probe = new HashSet<>();\n+ for( String s : second )\n+ probe.add(s);\n+ //probe phase: first\n+ return Arrays.stream(first).allMatch(s -> probe.contains(s));\n+ }\n+\n+ private static void reconcileFunctionInputsInPlace(FunctionOp call1, FunctionOp call2) {\n+ //prepare all input of call2 for probing\n+ HashMap<String,Hop> probe = new HashMap<>();\n+ for( int i=0; i<call2.getInput().size(); i++ )\n+ probe.put(call2.getInputVariableNames()[i], call2.getInput().get(i));\n+\n+ //construct new inputs for call1\n+ ArrayList<Hop> inputs = new ArrayList<>();\n+ for( int i=0; i<call1.getInput().size(); i++ )\n+ if( probe.containsKey(call1.getInputVariableNames()[i]) ) {\n+ inputs.add( (probe.get(call1.getInputVariableNames()[i]) instanceof LiteralOp) ?\n+ probe.get(call1.getInputVariableNames()[i]) : call1.getInput().get(i));\n+ }\n+ HopRewriteUtils.removeAllChildReferences(call1);\n+ call1.addAllInputs(inputs);\n+ call1.setInputVariableNames(call2.getInputVariableNames());\n+ }\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/ipa/InterProceduralAnalysis.java", "new_path": "src/main/java/org/tugraz/sysds/hops/ipa/InterProceduralAnalysis.java", "diff": "@@ -93,6 +93,7 @@ public class InterProceduralAnalysis\nprotected static final boolean APPLY_DYNAMIC_REWRITES = true; //apply dynamic hop dag and statement block rewrites\nprotected static final int INLINING_MAX_NUM_OPS = 10; //inline single-statement functions w/ #ops <= threshold, other than dataops and literals\nprotected static final boolean ELIMINATE_DEAD_CODE = true; //remove dead code (e.g., assigments) not used later on\n+ protected static final boolean FORWARD_SIMPLE_FUN_CALLS = true; //replace a call to a simple forwarding function with the function itself\nstatic {\n// for internal debugging only\n@@ -133,6 +134,7 @@ public class InterProceduralAnalysis\n_passes.add(new IPAPassRemoveConstantBinaryOps());\n_passes.add(new IPAPassPropagateReplaceLiterals());\n_passes.add(new IPAPassInlineFunctions());\n+ _passes.add(new IPAPassForwardFunctionCalls());\n_passes.add(new IPAPassEliminateDeadCode());\n//note: apply rewrites last because statement block rewrites\n//might merge relevant statement blocks in special cases, which\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/MatrixReshapeSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/MatrixReshapeSPInstruction.java", "diff": "@@ -60,20 +60,21 @@ public class MatrixReshapeSPInstruction extends UnarySPInstruction\npublic static MatrixReshapeSPInstruction parseInstruction ( String str ) {\nString[] parts = InstructionUtils.getInstructionPartsWithValueType(str);\n- InstructionUtils.checkNumFields( parts, 6 );\n+ InstructionUtils.checkNumFields( parts, 7 );\nString opcode = parts[0];\nCPOperand in1 = new CPOperand(parts[1]);\n- CPOperand in2 = new CPOperand(parts[2]);\n- CPOperand in3 = new CPOperand(parts[3]);\n- CPOperand in4 = new CPOperand(parts[4]);\n- CPOperand out = new CPOperand(parts[5]);\n- boolean outputEmptyBlocks = Boolean.parseBoolean(parts[6]);\n+ CPOperand rows = new CPOperand(parts[2]);\n+ CPOperand cols = new CPOperand(parts[3]);\n+ //TODO handle dims for tensors parts[4]\n+ CPOperand byRow = new CPOperand(parts[5]);\n+ CPOperand out = new CPOperand(parts[6]);\n+ boolean outputEmptyBlocks = Boolean.parseBoolean(parts[7]);\nif(!opcode.equalsIgnoreCase(\"rshape\"))\nthrow new DMLRuntimeException(\"Unknown opcode while parsing an MatrixReshapeInstruction: \" + str);\nelse\n- return new MatrixReshapeSPInstruction(new Operator(true), in1, in2, in3, in4, out, outputEmptyBlocks, opcode, str);\n+ return new MatrixReshapeSPInstruction(new Operator(true), in1, rows, cols, byRow, out, outputEmptyBlocks, opcode, str);\n}\n@Override\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/ipa/FunctionCallGraph.java", "new_path": "src/main/java/org/tugraz/sysds/hops/ipa/FunctionCallGraph.java", "diff": "@@ -198,6 +198,12 @@ public class FunctionCallGraph\n_fCallsSB.remove(fkeyOld);\n_fCalls.put(fkey, fopTmp);\n_fCallsSB.put(fkey, sbTmp);\n+ //additional cleanups fold no longer reachable\n+ _fRecursive.remove(fkeyOld);\n+ _fSideEffectFree.remove(fkeyOld);\n+ _fGraph.remove(fkeyOld);\n+ for( HashSet<String> hs : _fGraph.values() )\n+ hs.remove(fkeyOld);\n}\n/**\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/ipa/FunctionCallSizeInfo.java", "new_path": "src/main/java/org/tugraz/sysds/hops/ipa/FunctionCallSizeInfo.java", "diff": "@@ -210,7 +210,7 @@ public class FunctionCallSizeInfo\n//step 1: determine function candidates by evaluating all function calls\nfor( String fkey : _fgraph.getReachableFunctions() ) {\nList<FunctionOp> flist = _fgraph.getFunctionCalls(fkey);\n- if( flist.isEmpty() ) //robustness removed functions\n+ if( flist == null || flist.isEmpty() ) //robustness removed functions\ncontinue;\n//condition 1: function called just once\n@@ -249,7 +249,7 @@ public class FunctionCallSizeInfo\n//(considered for valid functions only)\nfor( String fkey : _fcand ) {\nList<FunctionOp> flist = _fgraph.getFunctionCalls(fkey);\n- if( flist.isEmpty() ) //robustness removed functions\n+ if( flist == null || flist.isEmpty() ) //robustness removed functions\ncontinue;\nFunctionOp first = flist.get(0);\nHashSet<Integer> tmp = new HashSet<>();\n@@ -267,7 +267,7 @@ public class FunctionCallSizeInfo\n//(considered for all functions)\nfor( String fkey : _fgraph.getReachableFunctions() ) {\nList<FunctionOp> flist = _fgraph.getFunctionCalls(fkey);\n- if( flist.isEmpty() ) //robustness removed functions\n+ if( flist == null || flist.isEmpty() ) //robustness removed functions\ncontinue;\nFunctionOp first = flist.get(0);\n//initialize w/ all literals of first call\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/ipa/IPAPassForwardFunctionCalls.java", "new_path": "src/main/java/org/tugraz/sysds/hops/ipa/IPAPassForwardFunctionCalls.java", "diff": "@@ -76,8 +76,9 @@ public class IPAPassForwardFunctionCalls extends IPAPass\ncall1.setFunctionName(call2.getFunctionName());\ncall1.setFunctionNamespace(call2.getFunctionNamespace());\nreconcileFunctionInputsInPlace(call1, call2);\n- //step 5: update function call graph\n- fgraph.replaceFunctionCalls(call1.getFunctionKey(), fkey);\n+ //step 5: update function call graph (old, new)\n+ fgraph.replaceFunctionCalls(fkey, call2.getFunctionKey());\n+ prog.removeFunctionStatementBlock(fkey);\nif( LOG.isDebugEnabled() )\nLOG.debug(\"IPA: Forward-function-call: replaced '\"\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/ipa/InterProceduralAnalysis.java", "new_path": "src/main/java/org/tugraz/sysds/hops/ipa/InterProceduralAnalysis.java", "diff": "@@ -72,7 +72,6 @@ import org.tugraz.sysds.runtime.meta.MetaDataFormat;\n* Additionally, IPA also covers the removal of unused functions, the decision on\n* recompile once functions, the removal of unnecessary checkpoints, and the\n* global removal of constant binary operations such as X * ones.\n- *\n*/\npublic class InterProceduralAnalysis\n{\n@@ -134,11 +133,11 @@ public class InterProceduralAnalysis\n_passes.add(new IPAPassRemoveConstantBinaryOps());\n_passes.add(new IPAPassPropagateReplaceLiterals());\n_passes.add(new IPAPassInlineFunctions());\n- _passes.add(new IPAPassForwardFunctionCalls());\n_passes.add(new IPAPassEliminateDeadCode());\n//note: apply rewrites last because statement block rewrites\n//might merge relevant statement blocks in special cases, which\n//would require an update of the function call graph\n+ _passes.add(new IPAPassForwardFunctionCalls());\n_passes.add(new IPAPassApplyStaticAndDynamicHopRewrites());\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/parser/DMLProgram.java", "new_path": "src/main/java/org/tugraz/sysds/parser/DMLProgram.java", "diff": "@@ -70,6 +70,11 @@ public class DMLProgram\nreturn getFunctionStatementBlock(tmp[0], tmp[1]);\n}\n+ public void removeFunctionStatementBlock(String fkey) {\n+ String[] tmp = splitFunctionKey(fkey);\n+ removeFunctionStatementBlock(tmp[0], tmp[1]);\n+ }\n+\npublic FunctionStatementBlock getFunctionStatementBlock(String namespaceKey, String functionName) {\nDMLProgram namespaceProgram = this.getNamespaces().get(namespaceKey);\nif (namespaceProgram == null)\n@@ -80,6 +85,13 @@ public class DMLProgram\nreturn retVal;\n}\n+ public void removeFunctionStatementBlock(String namespaceKey, String functionName) {\n+ DMLProgram namespaceProgram = this.getNamespaces().get(namespaceKey);\n+ // for the namespace DMLProgram, get the specified function (if exists) in its current namespace\n+ if (namespaceProgram != null)\n+ namespaceProgram._functionBlocks.remove(functionName);\n+ }\n+\npublic HashMap<String, FunctionStatementBlock> getFunctionStatementBlocks(String namespaceKey) {\nDMLProgram namespaceProgram = this.getNamespaces().get(namespaceKey);\nif (namespaceProgram == null){\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/DataGenCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/DataGenCPInstruction.java", "diff": "@@ -357,7 +357,7 @@ public class DataGenCPInstruction extends UnaryCPInstruction {\npublic LineageItem[] getLineageItems() {\nString tmpInstStr = instString;\nif (getSeed() == DataGenOp.UNSPECIFIED_SEED) {\n- int position = (method == DataGenMethod.RAND) ? 9 :\n+ int position = (method == DataGenMethod.RAND) ? 10 :\n(method == DataGenMethod.SAMPLE) ? 5 : 0;\ntmpInstStr = InstructionUtils.replaceOperand(\ntmpInstStr, position, String.valueOf(runtimeSeed));\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageItemUtils.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageItemUtils.java", "diff": "@@ -169,6 +169,7 @@ public class LineageItemUtils {\nHashMap<String, Hop> params = new HashMap<>();\nparams.put(DataExpression.RAND_ROWS, new LiteralOp(rand.getRows()));\nparams.put(DataExpression.RAND_COLS, new LiteralOp(rand.getCols()));\n+ params.put(DataExpression.RAND_DIMS, new LiteralOp(\"-1\"));\nparams.put(DataExpression.RAND_MIN, new LiteralOp(rand.getMinValue()));\nparams.put(DataExpression.RAND_MAX, new LiteralOp(rand.getMaxValue()));\nparams.put(DataExpression.RAND_PDF, new LiteralOp(rand.getPdf()));\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/matrix/data/LibMatrixDatagen.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/matrix/data/LibMatrixDatagen.java", "diff": "@@ -202,7 +202,7 @@ public class LibMatrixDatagen\n// Special case shortcuts for efficiency\nif ( rgen._pdf == RandomMatrixGenerator.PDF.UNIFORM) {\n- if ( min == 0.0 && max == 0.0 ) { //all zeros\n+ if ( min == 0.0 && max == 0.0 || sparsity == 0 ) { //all zeros\nout.reset(rows, cols, true);\nreturn;\n}\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/api/DMLOptions.java", "new_path": "src/main/java/org/tugraz/sysds/api/DMLOptions.java", "diff": "@@ -59,6 +59,7 @@ public class DMLOptions {\npublic boolean help = false; // whether to print the usage option\npublic boolean lineage = false; // whether compute lineage trace\npublic boolean lineage_dedup = false; // whether deduplicate lineage items\n+ public boolean lineage_reuse = false; // whether lineage-based reuse of intermediates\npublic final static DMLOptions defaultOptions = new DMLOptions(null);\n@@ -108,6 +109,8 @@ public class DMLOptions {\nif (lineageType != null){\nif (lineageType.equalsIgnoreCase(\"dedup\"))\ndmlOptions.lineage_dedup = lineageType.equalsIgnoreCase(\"dedup\");\n+ else if (lineageType.equalsIgnoreCase(\"reuse\"))\n+ dmlOptions.lineage_reuse = lineageType.equalsIgnoreCase(\"reuse\");\nelse\nthrow new org.apache.commons.cli.ParseException(\"Invalid argument specified for -lineage option\");\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/api/DMLScript.java", "new_path": "src/main/java/org/tugraz/sysds/api/DMLScript.java", "diff": "@@ -97,6 +97,7 @@ public class DMLScript\npublic static String GPU_MEMORY_ALLOCATOR = \"cuda\"; // GPU memory allocator to use\npublic static boolean LINEAGE = DMLOptions.defaultOptions.lineage; // whether compute lineage trace\npublic static boolean LINEAGE_DEDUP = DMLOptions.defaultOptions.lineage_dedup; // whether deduplicate lineage items\n+ public static boolean LINEAGE_REUSE = DMLOptions.defaultOptions.lineage_reuse; // whether lineage-based reuse\npublic static boolean USE_ACCELERATOR = DMLOptions.defaultOptions.gpu;\npublic static boolean FORCE_ACCELERATOR = DMLOptions.defaultOptions.forceGPU;\n@@ -198,6 +199,7 @@ public class DMLScript\nEXEC_MODE = dmlOptions.execMode;\nLINEAGE = dmlOptions.lineage;\nLINEAGE_DEDUP = dmlOptions.lineage_dedup;\n+ LINEAGE_REUSE = dmlOptions.lineage_reuse;\nString fnameOptConfig = dmlOptions.configFile;\nboolean isFile = dmlOptions.filePath != null;\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/MatrixIndexingCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/MatrixIndexingCPInstruction.java", "diff": "@@ -28,10 +28,14 @@ import org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject;\nimport org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject.UpdateType;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\n+import org.tugraz.sysds.runtime.lineage.Lineage;\n+import org.tugraz.sysds.runtime.lineage.LineageItem;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport org.tugraz.sysds.runtime.util.IndexRange;\nimport org.tugraz.sysds.utils.Statistics;\n+import java.util.Arrays;\n+\npublic final class MatrixIndexingCPInstruction extends IndexingCPInstruction {\npublic MatrixIndexingCPInstruction(CPOperand in, CPOperand rl, CPOperand ru, CPOperand cl, CPOperand cu,\n@@ -117,4 +121,11 @@ public final class MatrixIndexingCPInstruction extends IndexingCPInstruction {\nelse\nthrow new DMLRuntimeException(\"Invalid opcode (\" + opcode +\") encountered in MatrixIndexingCPInstruction.\");\n}\n+\n+ @Override\n+ public LineageItem[] getLineageItems() {\n+ LineageItem[] tmp = Arrays.asList(input1,input2,input3,colLower,colUpper,rowLower,rowUpper)\n+ .stream().filter(c -> c!=null).map(c -> Lineage.getOrCreate(c)).toArray(LineageItem[]::new);\n+ return new LineageItem[]{new LineageItem(output.getName(), getOpcode(), tmp)};\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "diff": "package org.tugraz.sysds.runtime.lineage;\nimport org.tugraz.sysds.api.DMLScript;\n+import org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.instructions.Instruction;\nimport org.tugraz.sysds.runtime.instructions.cp.ComputationCPInstruction;\n-import org.tugraz.sysds.runtime.instructions.cp.Data;\n-import org.tugraz.sysds.runtime.instructions.cp.DataGenCPInstruction;\n+import org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport java.util.HashMap;\nimport java.util.Map;\npublic class LineageCache {\n- private static final Map<LineageItem, Data> _cache = new HashMap<>();\n-\n- public static void put(LineageItem key, Data value) {\n- _cache.put(key, value);\n- }\n+ private static final Map<LineageItem, MatrixBlock> _cache = new HashMap<>();\npublic static void put(Instruction inst, ExecutionContext ec) {\n- if (inst instanceof DataGenCPInstruction) {\n- LineageItem[] items = ((LineageTraceable) inst).getLineageItems();\n- for (LineageItem item : items) {\n- Data d = ec.getVariable(((ComputationCPInstruction) inst).output);\n- LineageCache.put(item, d);\n+ if (!DMLScript.LINEAGE_REUSE)\n+ return;\n+\n+ if( inst instanceof ComputationCPInstruction\n+ &&((ComputationCPInstruction) inst).output.getDataType().isMatrix() ) {\n+\n+ for (LineageItem item : ((LineageTraceable) inst).getLineageItems()) {\n+ MatrixObject mo = ec.getMatrixObject(((ComputationCPInstruction) inst).output);\n+ LineageCache._cache.put(item, mo.acquireReadAndRelease());\n}\n}\n}\n@@ -31,7 +31,7 @@ public class LineageCache {\nreturn _cache.containsKey(key);\n}\n- public static Data get(LineageItem key) {\n+ public static MatrixBlock get(LineageItem key) {\nreturn _cache.get(key);\n}\n@@ -40,17 +40,16 @@ public class LineageCache {\n}\npublic static boolean reuse(Instruction inst, ExecutionContext ec) {\n- if (!DMLScript.LINEAGE)\n+ if (!DMLScript.LINEAGE && DMLScript.LINEAGE_REUSE)\nreturn false;\nif (inst instanceof ComputationCPInstruction) {\nboolean reused = true;\nLineageItem[] items = ((ComputationCPInstruction) inst).getLineageItems();\nfor (LineageItem item : items) {\n- if (LineageCache.probe(item)) {\n- Data d = LineageCache.get(item);\n- ec.setVariable(((ComputationCPInstruction) inst).output.getName(), d);\n- } else\n+ if (LineageCache.probe(item))\n+ ec.setMatrixOutput(((ComputationCPInstruction) inst).output.getName(), LineageCache.get(item));\n+ else\nreused = false;\n}\nreturn reused && items.length > 0;\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/component/misc/CLIOptionsParserTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/component/misc/CLIOptionsParserTest.java", "diff": "@@ -129,6 +129,7 @@ public class CLIOptionsParserTest {\nDMLOptions o = DMLOptions.parseCLArguments(args);\nAssert.assertEquals(true, o.lineage);\nAssert.assertEquals(false, o.lineage_dedup);\n+ Assert.assertEquals(false, o.lineage_reuse);\n}\n@Test\n@@ -138,15 +139,43 @@ public class CLIOptionsParserTest {\nDMLOptions o = DMLOptions.parseCLArguments(args);\nAssert.assertEquals(true, o.lineage);\nAssert.assertEquals(true, o.lineage_dedup);\n+ Assert.assertEquals(false, o.lineage_reuse);\n+ }\n+\n+ @Test\n+ public void testLineageReuse() throws Exception {\n+ String cl = \"systemml -f test.dml -lineage reuse\";\n+ String[] args = cl.split(\" \");\n+ DMLOptions o = DMLOptions.parseCLArguments(args);\n+ Assert.assertEquals(true, o.lineage);\n+ Assert.assertEquals(true, o.lineage_reuse);\n+ Assert.assertEquals(false, o.lineage_dedup);\n+ }\n+\n+ @Test\n+ public void testLineageDedupAndReuse() throws Exception {\n+ String cl = \"systemml -f test.dml -lineage dedup reuse\";\n+ String[] args = cl.split(\" \");\n+ DMLOptions o = DMLOptions.parseCLArguments(args);\n+ Assert.assertEquals(true, o.lineage);\n+ Assert.assertEquals(true, o.lineage_dedup);\n+ Assert.assertEquals(true, o.lineage_reuse);\n}\n@Test(expected = ParseException.class)\n- public void testBadLineageOption() throws Exception {\n+ public void testBadLineageOptionDedup() throws Exception {\nString cl = \"systemml -f test.dml -lineage ded\";\nString[] args = cl.split(\" \");\nDMLOptions.parseCLArguments(args);\n}\n+ @Test(expected = ParseException.class)\n+ public void testBadLineageOptionReuse() throws Exception {\n+ String cl = \"systemml -f test.dml -lineage rese\";\n+ String[] args = cl.split(\" \");\n+ DMLOptions.parseCLArguments(args);\n+ }\n+\n@Test\npublic void testGPUForce() throws Exception {\nString cl = \"systemml -f test.dml -gpu force\";\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/FullReuseTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/FullReuseTest.java", "diff": "@@ -19,14 +19,13 @@ package org.tugraz.sysds.test.functions.lineage;\nimport org.junit.Test;\nimport org.tugraz.sysds.hops.OptimizerUtils;\nimport org.tugraz.sysds.runtime.lineage.Lineage;\n-import org.tugraz.sysds.runtime.lineage.LineageItem;\n-import org.tugraz.sysds.runtime.lineage.LineageParser;\n+import org.tugraz.sysds.runtime.matrix.data.MatrixValue;\nimport org.tugraz.sysds.test.AutomatedTestBase;\nimport org.tugraz.sysds.test.TestConfiguration;\nimport org.tugraz.sysds.test.TestUtils;\n-import org.tugraz.sysds.utils.Explain;\nimport java.util.ArrayList;\n+import java.util.HashMap;\nimport java.util.List;\npublic class FullReuseTest extends AutomatedTestBase {\n@@ -34,17 +33,15 @@ public class FullReuseTest extends AutomatedTestBase {\nprotected static final String TEST_DIR = \"functions/lineage/\";\nprotected static final String TEST_NAME1 = \"FullReuse1\";\nprotected static final String TEST_NAME2 = \"FullReuse2\";\n+ protected static final String TEST_NAME3 = \"FullReuse3\";\nprotected String TEST_CLASS_DIR = TEST_DIR + FullReuseTest.class.getSimpleName() + \"/\";\n- protected static final int numRecords = 1024;\n- protected static final int numFeatures = 1024;\n-\n-\n@Override\npublic void setUp() {\nTestUtils.clearAssertionInformation();\naddTestConfiguration(TEST_NAME1, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME1));\naddTestConfiguration(TEST_NAME2, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME2));\n+ addTestConfiguration(TEST_NAME3, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME3));\n}\n@Test\n@@ -57,6 +54,11 @@ public class FullReuseTest extends AutomatedTestBase {\ntestLineageTrace(TEST_NAME2);\n}\n+ @Test\n+ public void testLineageTrace3() {\n+ testLineageTrace(TEST_NAME3);\n+ }\n+\npublic void testLineageTrace(String testname) {\nboolean old_simplification = OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION;\nboolean old_sum_product = OptimizerUtils.ALLOW_SUM_PRODUCT_REWRITES;\n@@ -67,34 +69,37 @@ public class FullReuseTest extends AutomatedTestBase {\nOptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = false;\nOptimizerUtils.ALLOW_SUM_PRODUCT_REWRITES = false;\n- int rows = numRecords;\n- int cols = numFeatures;\n-\ngetAndLoadTestConfiguration(testname);\n+ fullDMLScriptName = getScript();\n- List<String> proArgs = new ArrayList<String>();\n-\n+ // Without lineage-based reuse enabled\n+ List<String> proArgs = new ArrayList<>();\nproArgs.add(\"-stats\");\nproArgs.add(\"-lineage\");\n- proArgs.add(\"-explain\");\n+// proArgs.add(\"-explain\");\nproArgs.add(\"-args\");\n- proArgs.add(input(\"X\"));\nproArgs.add(output(\"X\"));\nprogramArgs = proArgs.toArray(new String[proArgs.size()]);\n- fullDMLScriptName = getScript();\n-\n- double[][] X = getRandomMatrix(rows, cols, 0, 1, 0.8, -1);\n- writeInputMatrixWithMTD(\"X\", X, true);\n-\nLineage.resetInternalState();\nrunTest(true, EXCEPTION_NOT_EXPECTED, null, -1);\n+ HashMap<MatrixValue.CellIndex, Double> X_orig = readDMLMatrixFromHDFS(\"X\");\n- String X_lineage = readDMLLineageFromHDFS(\"X\");\n+ // With lineage-based reuse enabled\n+ proArgs.clear();\n+ proArgs.add(\"-stats\");\n+ proArgs.add(\"-lineage\");\n+ proArgs.add(\"reuse\");\n+// proArgs.add(\"-explain\");\n+ proArgs.add(\"-args\");\n+ proArgs.add(output(\"X\"));\n+ programArgs = proArgs.toArray(new String[proArgs.size()]);\n- LineageItem X_li = LineageParser.parseLineageTrace(X_lineage);\n+ Lineage.resetInternalState();\n+ runTest(true, EXCEPTION_NOT_EXPECTED, null, -1);\n+ HashMap<MatrixValue.CellIndex, Double> X_reused = readDMLMatrixFromHDFS(\"X\");\n- TestUtils.compareScalars(X_lineage, Explain.explain(X_li));\n+ TestUtils.compareMatrices(X_orig, X_reused, 1e-6, \"Origin\", \"Reused\");\n} finally {\nOptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = old_simplification;\nOptimizerUtils.ALLOW_SUM_PRODUCT_REWRITES = old_sum_product;\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/lineage/FullReuse1.dml", "new_path": "src/test/scripts/functions/lineage/FullReuse1.dml", "diff": "# limitations under the License.\n#\n#-------------------------------------------------------------\n+# Increase k for better performance gains\n-# How to invoke this dml script LineageTrace.dml?\n-# Assume LR_HOME is set to the home of the dml script\n-# Assume rows = 20 and cols = 20 for X\n-# hadoop jar SystemML.jar -f $LR_HOME/LineageTrace.dml -args \"$INPUT_DIR/X\" \"$OUTPUT_DIR/X\" \"$OUTPUT_DIR/Y\"\n+X = rand(rows=1024, cols=1024, seed=42);\n+k = 10\n-X = read($1);\n-\n-for(i in 1:10){\n+for(i in 1:k){\ntmp = t(X) %*% X;\n}\n-write(tmp, $2, format=\"text\");\n+write(tmp, $1, format=\"text\");\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/lineage/FullReuse2.dml", "new_path": "src/test/scripts/functions/lineage/FullReuse2.dml", "diff": "# limitations under the License.\n#\n#-------------------------------------------------------------\n+# Increase rows and cols for better performance gains\n-# How to invoke this dml script LineageTrace.dml?\n-# Assume LR_HOME is set to the home of the dml script\n-# Assume rows = 20 and cols = 20 for X\n-# hadoop jar SystemML.jar -f $LR_HOME/LineageTrace.dml -args \"$INPUT_DIR/X\" \"$OUTPUT_DIR/X\" \"$OUTPUT_DIR/Y\"\n+r = 100000\n+c = 100\n-# Increase rows and cols for performance comparision\n-X = rand(rows=10000, cols=100);\n-y = rand(rows=10000, cols=1);\n+X = rand(rows=r, cols=c, seed=42);\n+y = rand(rows=r, cols=1, seed=43);\nj = 10\n-R = matrix(0, ncol(X), j);\n+R = matrix(0, c, j);\nfor(i in 1:j) {\nlambda = 10 ^ -i;\n- print(\"lambda: \" + toString(lambda));\n-\nA = t(X) %*% X + diag(matrix(lambda, rows=ncol(X), cols=1));\nb = t(X) %*% y;\nbeta = solve(A, b);\n@@ -40,4 +36,4 @@ for(i in 1:j) {\nR[,i] = beta;\n}\n-write(R, $2, format=\"text\");\n+write(R, $1, format=\"text\");\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/lineage/FullReuse3.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+# Increase rows and cols for better performance gain\n+\n+X = rand(rows=1024, cols=16, seed=42);\n+tmp = X[,1];\n+R = matrix(0, ncol(X), 1);\n+\n+for(i in 2:ncol(X)) {\n+ A = t(tmp) %*% tmp;\n+ R[i,] = sum(A)\n+\n+ tmp = cbind(tmp, X[,i]);\n+ A = t(tmp) %*% tmp;\n+ R[i,] += sum(A)\n+}\n+\n+write(R, $1, format=\"text\");\n\\ No newline at end of file\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -75,6 +75,7 @@ SYSTEMDS-110 New Builtin Functions\n* 111 Time builtin function for script-level measurements OK\n* 112 Image data augmentation builtin functions OK\n* 113 Builtin functions for linear regression algorithms OK\n+ * 114 Builtin function for stepwise regression OK\nSYSTEMDS-120 Performance Features\n* 121 Avoid spark context creation on parfor result merge OK\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "new_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "diff": "@@ -19,6 +19,8 @@ package org.tugraz.sysds.common;\nimport java.util.EnumSet;\nimport java.util.HashMap;\n+import org.tugraz.sysds.common.Types.ReturnType;\n+\n/**\n* Enum to represent all builtin functions in the default name space.\n* Each function is either native or implemented by a DML script. In\n@@ -38,8 +40,8 @@ public enum Builtins {\nATAN(\"atan\", false),\nAVG_POOL(\"avg_pool\", false),\nAVG_POOL_BACKWARD(\"avg_pool_backward\", false),\n- BATCH_NORM2D(\"batch_norm2d\", false),\n- BATCH_NORM2D_BACKWARD(\"batch_norm2d_backward\", false),\n+ BATCH_NORM2D(\"batch_norm2d\", false, ReturnType.MULTI_RETURN),\n+ BATCH_NORM2D_BACKWARD(\"batch_norm2d_backward\", false, ReturnType.MULTI_RETURN),\nBIASADD(\"bias_add\", false),\nBIASMULT(\"bias_multiply\", false),\nBITWAND(\"bitwAnd\", false),\n@@ -75,7 +77,7 @@ public enum Builtins {\nCUMSUM(\"cumsum\", false),\nCUMSUMPROD(\"cumsumprod\", false),\nDIAG(\"diag\", false),\n- EIGEN(\"eigen\", false),\n+ EIGEN(\"eigen\", false, ReturnType.MULTI_RETURN),\nEXISTS(\"exists\", false),\nEXP(\"exp\", false),\nEVAL(\"eval\", false),\n@@ -95,9 +97,9 @@ public enum Builtins {\nLMDS(\"lmDS\", true),\nLMPREDICT(\"lmpredict\", true),\nLOG(\"log\", false),\n- LSTM(\"lstm\", false),\n- LSTM_BACKWARD(\"lstm_backward\", false),\n- LU(\"lu\", false),\n+ LSTM(\"lstm\", false, ReturnType.MULTI_RETURN),\n+ LSTM_BACKWARD(\"lstm_backward\", false, ReturnType.MULTI_RETURN),\n+ LU(\"lu\", false, ReturnType.MULTI_RETURN),\nMEAN(\"mean\", \"avg\", false),\nMIN(\"min\", \"pmin\", false),\nMAX(\"max\", \"pmax\", false),\n@@ -112,7 +114,7 @@ public enum Builtins {\nOUTLIER(\"outlier\", true, false), //TODO parameterize opposite\nPPRED(\"ppred\", false),\nPROD(\"prod\", false),\n- QR(\"qr\", false),\n+ QR(\"qr\", false, ReturnType.MULTI_RETURN),\nQUANTILE(\"quantile\", false),\nRANGE(\"range\", false),\nRBIND(\"rbind\", false),\n@@ -134,11 +136,11 @@ public enum Builtins {\nSIGN(\"sign\", false),\nSIN(\"sin\", false),\nSINH(\"sinh\", false),\n- STEPLM(\"steplm\",true),\n+ STEPLM(\"steplm\",true, ReturnType.MULTI_RETURN),\nSOLVE(\"solve\", false),\nSQRT(\"sqrt\", false),\nSUM(\"sum\", false),\n- SVD(\"svd\", false),\n+ SVD(\"svd\", false, ReturnType.MULTI_RETURN),\nTRANS(\"t\", false),\nTABLE(\"table\", \"ctable\", false),\nTAN(\"tan\", false),\n@@ -179,22 +181,31 @@ public enum Builtins {\nUPPER_TRI(\"upper.tri\", false, true);\nBuiltins(String name, boolean script) {\n- this(name, null, script, false);\n+ this(name, null, script, false, ReturnType.SINGLE_RETURN);\n+ }\n+\n+ Builtins(String name, boolean script, ReturnType retType) {\n+ this(name, null, script, false, retType);\n}\nBuiltins(String name, boolean script, boolean parameterized) {\n- this(name, null, script, parameterized);\n+ this(name, null, script, parameterized, ReturnType.SINGLE_RETURN);\n}\nBuiltins(String name, String alias, boolean script) {\n- this(name, alias, script, false);\n+ this(name, alias, script, false, ReturnType.SINGLE_RETURN);\n}\nBuiltins(String name, String alias, boolean script, boolean parameterized) {\n+ this(name, alias, script, false, ReturnType.SINGLE_RETURN);\n+ }\n+\n+ Builtins(String name, String alias, boolean script, boolean parameterized, ReturnType retType) {\n_name = name;\n_alias = alias;\n_script = script;\n_parameterized = parameterized;\n+ _retType = retType;\n}\nprivate final static String BUILTIN_DIR = \"scripts/builtin/\";\n@@ -213,6 +224,7 @@ public enum Builtins {\nprivate final String _alias;\nprivate final boolean _script;\nprivate final boolean _parameterized;\n+ private final ReturnType _retType;\npublic String getName() {\nreturn _name;\n@@ -230,6 +242,10 @@ public enum Builtins {\nreturn _parameterized;\n}\n+ public boolean isMultiReturn() {\n+ return _retType == ReturnType.MULTI_RETURN;\n+ }\n+\npublic static boolean contains(String name, boolean script, boolean parameterized) {\nBuiltins tmp = get(name);\nreturn tmp != null && script == tmp.isScript()\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Types.java", "new_path": "src/main/java/org/tugraz/sysds/common/Types.java", "diff": "@@ -105,4 +105,14 @@ public class Types\nSPARSE_BLOCK,\nDENSE_BLOCK,\n}\n+\n+ /**\n+ * Type of builtin or user-defined function with regard to its\n+ * number of return variables.\n+ */\n+ public enum ReturnType {\n+ NO_RETURN,\n+ SINGLE_RETURN,\n+ MULTI_RETURN\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/parser/AssignmentStatement.java", "new_path": "src/main/java/org/tugraz/sysds/parser/AssignmentStatement.java", "diff": "@@ -137,10 +137,8 @@ public class AssignmentStatement extends Statement\n@Override\npublic String toString(){\nStringBuilder sb = new StringBuilder();\n- for (int i=0; i< _targetList.size(); i++){\n- DataIdentifier di = _targetList.get(i);\n- sb.append(di);\n- }\n+ for (int i=0; i< _targetList.size(); i++)\n+ sb.append(_targetList.get(i));\nsb.append(_isAccum ? \" += \" : \" = \");\nif (_source instanceof StringIdentifier) {\nsb.append(\"\\\"\");\n@@ -150,7 +148,6 @@ public class AssignmentStatement extends Statement\nsb.append(_source.toString());\n}\nsb.append(\";\");\n-\nreturn sb.toString();\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/parser/BuiltinFunctionExpression.java", "new_path": "src/main/java/org/tugraz/sysds/parser/BuiltinFunctionExpression.java", "diff": "@@ -1555,19 +1555,7 @@ public class BuiltinFunctionExpression extends DataIdentifier\n@Override\npublic boolean multipleReturns() {\n- switch(_opcode) {\n- case QR:\n- case LU:\n- case EIGEN:\n- case LSTM:\n- case LSTM_BACKWARD:\n- case BATCH_NORM2D:\n- case BATCH_NORM2D_BACKWARD:\n- case SVD:\n- return true;\n- default:\n- return false;\n- }\n+ return _opcode.isMultiReturn();\n}\nprivate static boolean isConstant(Expression expr) {\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/parser/MultiAssignmentStatement.java", "new_path": "src/main/java/org/tugraz/sysds/parser/MultiAssignmentStatement.java", "diff": "package org.tugraz.sysds.parser;\nimport java.util.ArrayList;\n+import java.util.Arrays;\nimport java.util.List;\npublic class MultiAssignmentStatement extends Statement\n@@ -106,26 +107,16 @@ public class MultiAssignmentStatement extends Statement\n}\n@Override\n- public String toString()\n- {\n+ public String toString() {\nStringBuilder sb = new StringBuilder();\n- sb.append(\"[\");\n-\n- for( int i=0; i< _targetList.size(); i++ )\n- {\n- sb.append(_targetList.get(i).toString());\n- if (i < _targetList.size() - 1)\n- sb.append(\",\");\n- }\n- sb.append(\"] = \");\n+ sb.append(Arrays.toString(_targetList.toArray()));\n+ sb.append(\" = \");\nsb.append(_source.toString());\nsb.append(\";\");\n-\nreturn sb.toString();\n}\npublic void setSource(FunctionCallIdentifier s) {\n_source = s;\n-\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/parser/dml/DmlSyntacticValidator.java", "new_path": "src/main/java/org/tugraz/sysds/parser/dml/DmlSyntacticValidator.java", "diff": "@@ -589,6 +589,7 @@ public class DmlSyntacticValidator implements DmlListener {\nsetMultiAssignmentStatement(targetList, e, ctx, ctx.info);\nreturn;\n}\n+ handleDMLBodiedBuiltinFunction(functionName, namespace, ctx);\n}\n// Override default namespace for imported non-built-in function\n@@ -598,6 +599,16 @@ public class DmlSyntacticValidator implements DmlListener {\nsetMultiAssignmentStatement(targetList, functCall, ctx, ctx.info);\n}\n+ private void handleDMLBodiedBuiltinFunction(String functionName, String namespace, ParserRuleContext ctx) {\n+ if( Builtins.contains(functionName, true, false) ) {\n+ //load and add builtin DML-bodied functions\n+ String filePath = Builtins.getFilePath(functionName);\n+ DMLProgram prog = parseAndAddImportedFunctions(namespace, filePath, ctx);\n+ for( Entry<String,FunctionStatementBlock> f : prog.getNamedFunctionStatementBlocks().entrySet() )\n+ builtinFuns.addFunctionStatementBlock(f.getKey(), f.getValue());\n+ }\n+ }\n+\n// -----------------------------------------------------------------\n// Control Statements - Guards & Loops\n@@ -1610,14 +1621,7 @@ public class DmlSyntacticValidator implements DmlListener {\nsetAssignmentStatement(ctx, info, target, e);\nreturn;\n}\n-\n- if( Builtins.contains(functionName, true, false) ) {\n- //load and add builtin DML-bodied functions\n- String filePath = Builtins.getFilePath(functionName);\n- DMLProgram prog = parseAndAddImportedFunctions(namespace, filePath, ctx);\n- for( Entry<String,FunctionStatementBlock> f : prog.getNamedFunctionStatementBlocks().entrySet() )\n- builtinFuns.addFunctionStatementBlock(f.getKey(), f.getValue());\n- }\n+ handleDMLBodiedBuiltinFunction(functionName, namespace, ctx);\n}\n// handle user-defined functions\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinSTEPLmTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinSTEPLmTest.java", "diff": "@@ -52,6 +52,11 @@ public class BuiltinSTEPLmTest extends AutomatedTestBase {\nrunSTEPLmTest(false, ExecType.CP, BuiltinSTEPLmTest.LinregType.AUTO);\n}\n+ @Test\n+ public void testLmMatrixSparseCPlm() {\n+ runSTEPLmTest(false, ExecType.CP, BuiltinSTEPLmTest.LinregType.AUTO);\n+ }\n+\n@Test\npublic void testLmMatrixDenseSPlm() {\nrunSTEPLmTest(false, ExecType.SPARK, BuiltinSTEPLmTest.LinregType.AUTO);\n@@ -62,7 +67,6 @@ public class BuiltinSTEPLmTest extends AutomatedTestBase {\nrunSTEPLmTest(true, ExecType.SPARK, BuiltinSTEPLmTest.LinregType.AUTO);\n}\n-\nprivate void runSTEPLmTest(boolean sparse, ExecType instType, BuiltinSTEPLmTest.LinregType linregAlgo) {\nExecMode platformOld = setExecMode(instType);\n@@ -76,7 +80,6 @@ public class BuiltinSTEPLmTest extends AutomatedTestBase {\nString HOME = SCRIPT_DIR + TEST_DIR;\n-\nfullDMLScriptName = HOME + dml_test_name + \".dml\";\nprogramArgs = new String[]{\"-explain\", \"-args\", input(\"A\"), input(\"B\"), output(\"C\"), output(\"S\")};\nfullRScriptName = HOME + TEST_NAME + \".R\";\n@@ -92,18 +95,15 @@ public class BuiltinSTEPLmTest extends AutomatedTestBase {\nrunRScript(true);\n//compare matrices\n-\nHashMap<CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(\"C\");\nHashMap<CellIndex, Double> dmfile1 = readDMLMatrixFromHDFS(\"S\");\nHashMap<CellIndex, Double> rfile = readRMatrixFromFS(\"C\");\nHashMap<CellIndex, Double> rfile1 = readRMatrixFromFS(\"S\");\nTestUtils.compareMatrices(dmlfile, rfile, eps, \"Stat-DML\", \"Stat-R\");\nTestUtils.compareMatrices(dmfile1, rfile1, eps, \"Stat-DML\", \"Stat-R\");\n-\n- } finally {\n+ }\n+ finally {\nrtplatform = platformOld;\n}\n}\n-\n}\n-\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/CPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/CPInstruction.java", "diff": "package org.tugraz.sysds.runtime.instructions.cp;\n+import org.tugraz.sysds.api.DMLScript;\nimport org.tugraz.sysds.common.Types;\nimport org.tugraz.sysds.lops.Lop;\nimport org.tugraz.sysds.common.Types.DataType;\n@@ -31,6 +32,7 @@ import org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.data.TensorBlock;\nimport org.tugraz.sysds.runtime.instructions.CPInstructionParser;\nimport org.tugraz.sysds.runtime.instructions.Instruction;\n+import org.tugraz.sysds.runtime.lineage.Lineage;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport org.tugraz.sysds.runtime.matrix.operators.Operator;\nimport org.tugraz.sysds.runtime.util.UtilFunctions;\n@@ -107,15 +109,19 @@ public abstract class CPInstruction extends Instruction\n@Override\npublic Instruction preprocessInstruction(ExecutionContext ec) {\n- Instruction tmp = this;\n+ //default preprocess behavior (e.g., debug state)\n+ Instruction tmp = super.preprocessInstruction(ec);\n+\n//instruction patching\nif( tmp.requiresLabelUpdate() ) { //update labels only if required\n//note: no exchange of updated instruction as labels might change in the general case\nString updInst = updateLabels(tmp.toString(), ec.getVariables());\ntmp = CPInstructionParser.parseSingleInstruction(updInst);\n+ // Corrected lineage trace for patched instructions\n+ if (DMLScript.LINEAGE)\n+ Lineage.trace(tmp, ec);\n}\n- //default preprocess behavior (e.g., lineage tracing)\n- return super.preprocessInstruction(ec);\n+ return tmp;\n}\n@Override\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/lops/Transform.java", "new_path": "src/main/java/org/tugraz/sysds/lops/Transform.java", "diff": "@@ -32,7 +32,6 @@ import org.tugraz.sysds.common.Types.ValueType;\n* Lop to perform transpose/vector to diag operations\n* This lop can change the keys and hence break alignment.\n*/\n-\npublic class Transform extends Lop\n{\npublic enum OperationTypes {\n@@ -133,7 +132,6 @@ public class Transform extends Lop\ndefault:\nthrow new UnsupportedOperationException(this.printErrorLocation() + \"Instruction is not defined for Transform operation \" + operation);\n-\n}\n}\n@@ -141,38 +139,32 @@ public class Transform extends Lop\n@Override\npublic String getInstructions(String input1, String output) {\n- StringBuilder sb = new StringBuilder();\n- sb.append( getExecType() );\n- sb.append( OPERAND_DELIMITOR );\n- sb.append( getOpcode() );\n- sb.append( OPERAND_DELIMITOR );\n- sb.append( getInputs().get(0).prepInputOperand(input1));\n- sb.append( OPERAND_DELIMITOR );\n- sb.append( this.prepOutputOperand(output));\n-\n- if( getExecType()==ExecType.CP && operation == OperationTypes.Transpose ) {\n- sb.append( OPERAND_DELIMITOR );\n- sb.append( _numThreads );\n+ //opcodes: r', rev, rdiag\n+ return getInstructions(input1, 1, output);\n}\n- return sb.toString();\n+ @Override\n+ public String getInstructions(String input1, String input2, String input3, String input4, String output) {\n+ //opcodes: rsort\n+ return getInstructions(input1, 4, output);\n}\n@Override\npublic String getInstructions(String input1, String input2, String input3, String input4, String input5, String output) {\n- //only used for reshape\n+ //opcodes: rshape\n+ return getInstructions(input1, 5, output);\n+ }\n+ private String getInstructions(String input1, int numInputs, String output) {\nStringBuilder sb = new StringBuilder();\nsb.append( getExecType() );\n-\nsb.append( OPERAND_DELIMITOR );\nsb.append( getOpcode() );\nsb.append( OPERAND_DELIMITOR );\nsb.append( getInputs().get(0).prepInputOperand(input1));\n//rows, cols, byrow\n- String[] inputX = new String[]{input2,input3,input4,input5};\n- for( int i=1; i<=(inputX.length); i++ ) {\n+ for( int i = 1; i < numInputs; i++ ) {\nLop ltmp = getInputs().get(i);\nsb.append( OPERAND_DELIMITOR );\nsb.append( ltmp.prepScalarInputOperand(getExecType()));\n@@ -182,11 +174,14 @@ public class Transform extends Lop\nsb.append( OPERAND_DELIMITOR );\nsb.append( this.prepOutputOperand(output));\n+ if( getExecType()==ExecType.CP && operation == OperationTypes.Transpose ) {\n+ sb.append( OPERAND_DELIMITOR );\n+ sb.append( _numThreads );\n+ }\nif( getExecType()==ExecType.SPARK && operation == OperationTypes.Reshape ) {\nsb.append( OPERAND_DELIMITOR );\nsb.append( _outputEmptyBlock );\n}\n-\nif( getExecType()==ExecType.SPARK && operation == OperationTypes.Sort ){\nsb.append( OPERAND_DELIMITOR );\nsb.append( _bSortIndInMem );\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/ipa/IPAPassForwardFunctionCalls.java", "new_path": "src/main/java/org/tugraz/sysds/hops/ipa/IPAPassForwardFunctionCalls.java", "diff": "@@ -54,7 +54,7 @@ public class IPAPassForwardFunctionCalls extends IPAPass\nFunctionStatement fstmt = (FunctionStatement)fsb.getStatement(0);\n//step 1: basic application filter: simple forwarding call\n- if( fstmt.getBody().size() > 1 || !containsFunctionOp(fstmt.getBody().get(0).getHops())\n+ if( fstmt.getBody().size() != 1 || !containsFunctionOp(fstmt.getBody().get(0).getHops())\n|| !hasOnlySimplyArguments((FunctionOp)fstmt.getBody().get(0).getHops().get(0)))\ncontinue;\nif( LOG.isDebugEnabled() )\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/RandSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/RandSPInstruction.java", "diff": "@@ -45,6 +45,7 @@ import org.tugraz.sysds.lops.DataGen;\nimport org.tugraz.sysds.lops.Lop;\nimport org.tugraz.sysds.common.Types.DataType;\nimport org.tugraz.sysds.common.Types.ValueType;\n+import org.tugraz.sysds.conf.ConfigurationManager;\nimport org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.controlprogram.context.SparkExecutionContext;\n@@ -253,7 +254,8 @@ public class RandSPInstruction extends UnarySPInstruction {\nLOG.trace(\"Process RandSPInstruction rand with seed = \"+lSeed+\".\");\n//step 2: potential in-memory rand operations if applicable\n- if( isMemAvail(lrows, lcols, sparsity, minValue, maxValue)\n+ if( ConfigurationManager.isDynamicRecompilation()\n+ && isMemAvail(lrows, lcols, sparsity, minValue, maxValue)\n&& DMLScript.getGlobalExecMode() != ExecMode.SPARK )\n{\nRandomMatrixGenerator rgen = LibMatrixDatagen.createRandomMatrixGenerator(\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/PredicateRecompileTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/PredicateRecompileTest.java", "diff": "@@ -283,22 +283,21 @@ public class PredicateRecompileTest extends AutomatedTestBase\nelse\n{\nif( IPA ) {\n- //old expected numbers before IPA\n- if( testname.equals(TEST_NAME1) )\n- Assert.assertEquals(\"Unexpected number of executed Spark instructions.\",\n- 4 - ((evalExpr||constFold)?4:0), Statistics.getNoOfExecutedSPInst()); //rand, 2xgmr while pred, 1x gmr while body\n- else //if( testname.equals(TEST_NAME2) )\n+ int expected = (testname.equals(TEST_NAME1) ?\n+ 4 - ((evalExpr||constFold)?4:0) :\n+ 3 - ((evalExpr||constFold)?3:0))\n+ + ((!testname.equals(TEST_NAME2)&&!(evalExpr||constFold))?1:0); //loop checkpoint\nAssert.assertEquals(\"Unexpected number of executed Spark instructions.\",\n- 3 - ((evalExpr||constFold)?3:0), Statistics.getNoOfExecutedSPInst()); //rand, 1xgmr if pred, 1x gmr if body\n+ expected, Statistics.getNoOfExecutedSPInst());\n}\nelse {\n//old expected numbers before IPA\n- if( testname.equals(TEST_NAME1) )\n- Assert.assertEquals(\"Unexpected number of executed Spark instructions.\",\n- 5 - ((evalExpr||constFold)?1:0), Statistics.getNoOfExecutedSPInst()); //rand, 2xgmr while pred, 1x gmr while body\n- else //if( testname.equals(TEST_NAME2) )\n+ int expected = (testname.equals(TEST_NAME1) ?\n+ 4 - ((evalExpr||constFold)?1:0) :\n+ 3 - ((evalExpr||constFold)?1:0))\n+ + (!testname.equals(TEST_NAME2)?1:0); //loop checkpoint\nAssert.assertEquals(\"Unexpected number of executed Spark instructions.\",\n- 3 - ((evalExpr||constFold)?1:0), Statistics.getNoOfExecutedSPInst()); //rand, 1xgmr if pred, 1x gmr if body\n+ expected, Statistics.getNoOfExecutedSPInst());\n}\n}\n@@ -306,13 +305,11 @@ public class PredicateRecompileTest extends AutomatedTestBase\nHashMap<CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(\"R\");\nAssert.assertEquals(Double.valueOf((double)val), dmlfile.get(new CellIndex(1,1)));\n}\n- finally\n- {\n+ finally {\nCompilerConfig.FLAG_DYN_RECOMPILE = oldFlagRecompile;\nOptimizerUtils.ALLOW_SIZE_EXPRESSION_EVALUATION = oldFlagEval;\nOptimizerUtils.ALLOW_CONSTANT_FOLDING = oldFlagFold;\nOptimizerUtils.ALLOW_INTER_PROCEDURAL_ANALYSIS = oldFlagIPA;\n-\nOptimizerUtils.ALLOW_RAND_JOB_RECOMPILE = oldFlagRand1;\nOptimizerUtils.ALLOW_BRANCH_REMOVAL = oldFlagRand2;\nOptimizerUtils.ALLOW_WORSTCASE_SIZE_EXPRESSION_EVALUATION = oldFlagRand3;\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/ReblockRecompileTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/ReblockRecompileTest.java", "diff": "@@ -37,7 +37,6 @@ import org.tugraz.sysds.utils.Statistics;\npublic class ReblockRecompileTest extends AutomatedTestBase\n{\n-\nprivate final static String TEST_NAME1 = \"rblk_recompile1\";\nprivate final static String TEST_NAME2 = \"rblk_recompile2\";\nprivate final static String TEST_NAME3 = \"rblk_recompile3\";\n@@ -75,21 +74,6 @@ public class ReblockRecompileTest extends AutomatedTestBase\nrunReblockTest(2, 296304710250949L);\n}\n- @Test\n- public void testReblockGroupedAggregateRand() {\n- runReblockTest(3, System.nanoTime());\n- }\n-\n- @Test //failed before for this particular seed\n- public void testReblockGroupedAggregateFixed() {\n- runReblockTest(3, 296304710250949L);\n- }\n-\n- /**\n- *\n- * @param scriptNum\n- * @param seed\n- */\nprivate void runReblockTest(int scriptNum, long seed)\n{\nString TEST_NAME = null;\n@@ -124,13 +108,14 @@ public class ReblockRecompileTest extends AutomatedTestBase\nrunTest(true, exceptionExpected, null, -1); //0 due to recompile\nrunRScript(true);\n- Assert.assertEquals(\"Unexpected number of executed MR jobs.\",\n+ Assert.assertEquals(\"Unexpected number of executed Spark instructions.\",\n0, Statistics.getNoOfExecutedSPInst());\n//compare matrices\ntry\n{\n- MatrixBlock mo = DataConverter.readMatrixFromHDFS(output(\"R\"), InputInfo.BinaryBlockInputInfo, rows, 1, OptimizerUtils.DEFAULT_BLOCKSIZE, OptimizerUtils.DEFAULT_BLOCKSIZE);\n+ MatrixBlock mo = DataConverter.readMatrixFromHDFS(output(\"R\"), InputInfo.BinaryBlockInputInfo,\n+ rows, 1, OptimizerUtils.DEFAULT_BLOCKSIZE, OptimizerUtils.DEFAULT_BLOCKSIZE);\nHashMap<CellIndex, Double> dmlfile = new HashMap<CellIndex,Double>();\nfor( int i=0; i<mo.getNumRows(); i++ )\nfor( int j=0; j<mo.getNumColumns(); j++ )\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/RemoveEmptyRecompileTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/RemoveEmptyRecompileTest.java", "diff": "@@ -42,7 +42,6 @@ import org.tugraz.sysds.utils.Statistics;\n*/\npublic class RemoveEmptyRecompileTest extends AutomatedTestBase\n{\n-\nprivate final static String TEST_NAME = \"remove_empty_recompile\";\nprivate final static String TEST_DIR = \"functions/recompile/\";\n@@ -71,175 +70,142 @@ public class RemoveEmptyRecompileTest extends AutomatedTestBase\n@Override\n- public void setUp()\n- {\n+ public void setUp() {\nTestUtils.clearAssertionInformation();\naddTestConfiguration(TEST_NAME, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME, new String[] { \"R\" }));\n}\n@Test\n- public void testRemoveEmptySumNonEmpty()\n- {\n+ public void testRemoveEmptySumNonEmpty() {\nrunRemoveEmptyTest(OpType.SUM, false);\n}\n@Test\n- public void testRemoveEmptyRoundNonEmpty()\n- {\n+ public void testRemoveEmptyRoundNonEmpty() {\nrunRemoveEmptyTest(OpType.ROUND, false);\n}\n@Test\n- public void testRemoveEmptyTransposeNonEmpty()\n- {\n+ public void testRemoveEmptyTransposeNonEmpty() {\nrunRemoveEmptyTest(OpType.TRANSPOSE, false);\n}\n@Test\n- public void testRemoveEmptyMultLeftNonEmpty()\n- {\n+ public void testRemoveEmptyMultLeftNonEmpty() {\nrunRemoveEmptyTest(OpType.MULT_LEFT, false);\n}\n@Test\n- public void testRemoveEmptyMultRightNonEmpty()\n- {\n+ public void testRemoveEmptyMultRightNonEmpty() {\nrunRemoveEmptyTest(OpType.MULT_RIGHT, false);\n}\n@Test\n- public void testRemoveEmptyPlusLeftNonEmpty()\n- {\n+ public void testRemoveEmptyPlusLeftNonEmpty() {\nrunRemoveEmptyTest(OpType.PLUS_LEFT, false);\n}\n@Test\n- public void testRemoveEmptyPlusRightNonEmpty()\n- {\n+ public void testRemoveEmptyPlusRightNonEmpty() {\nrunRemoveEmptyTest(OpType.PLUS_RIGHT, false);\n}\n@Test\n- public void testRemoveEmptyMinusLeftNonEmpty()\n- {\n+ public void testRemoveEmptyMinusLeftNonEmpty() {\nrunRemoveEmptyTest(OpType.MINUS_LEFT, false);\n}\n@Test\n- public void testRemoveEmptyMinusRightNonEmpty()\n- {\n+ public void testRemoveEmptyMinusRightNonEmpty() {\nrunRemoveEmptyTest(OpType.MINUS_RIGHT, false);\n}\n@Test\n- public void testRemoveEmptyMatMultLeftNonEmpty()\n- {\n+ public void testRemoveEmptyMatMultLeftNonEmpty() {\nrunRemoveEmptyTest(OpType.MM_LEFT, false);\n}\n@Test\n- public void testRemoveEmptyMatMultRightNonEmpty()\n- {\n+ public void testRemoveEmptyMatMultRightNonEmpty() {\nrunRemoveEmptyTest(OpType.MM_RIGHT, false);\n}\n@Test\n- public void testRemoveEmptyRIXNonEmpty()\n- {\n+ public void testRemoveEmptyRIXNonEmpty() {\nrunRemoveEmptyTest(OpType.RIX, false);\n}\n@Test\n- public void testRemoveEmptyLIXNonEmpty()\n- {\n+ public void testRemoveEmptyLIXNonEmpty() {\nrunRemoveEmptyTest(OpType.LIX, false);\n}\n@Test\n- public void testRemoveEmptySumEmpty()\n- {\n+ public void testRemoveEmptySumEmpty() {\nrunRemoveEmptyTest(OpType.SUM, true);\n}\n@Test\n- public void testRemoveEmptyRoundEmpty()\n- {\n+ public void testRemoveEmptyRoundEmpty() {\nrunRemoveEmptyTest(OpType.ROUND, true);\n}\n@Test\n- public void testRemoveEmptyTransposeEmpty()\n- {\n+ public void testRemoveEmptyTransposeEmpty() {\nrunRemoveEmptyTest(OpType.TRANSPOSE, true);\n}\n@Test\n- public void testRemoveEmptyMultLeftEmpty()\n- {\n+ public void testRemoveEmptyMultLeftEmpty() {\nrunRemoveEmptyTest(OpType.MULT_LEFT, true);\n}\n@Test\n- public void testRemoveEmptyMultRightEmpty()\n- {\n+ public void testRemoveEmptyMultRightEmpty() {\nrunRemoveEmptyTest(OpType.MULT_RIGHT, true);\n}\n@Test\n- public void testRemoveEmptyPlusLeftEmpty()\n- {\n+ public void testRemoveEmptyPlusLeftEmpty() {\nrunRemoveEmptyTest(OpType.PLUS_LEFT, true);\n}\n@Test\n- public void testRemoveEmptyPlusRightEmpty()\n- {\n+ public void testRemoveEmptyPlusRightEmpty() {\nrunRemoveEmptyTest(OpType.PLUS_RIGHT, true);\n}\n@Test\n- public void testRemoveEmptyMinusLeftEmpty()\n- {\n+ public void testRemoveEmptyMinusLeftEmpty() {\nrunRemoveEmptyTest(OpType.MINUS_LEFT, true);\n}\n@Test\n- public void testRemoveEmptyMinusRightEmpty()\n- {\n+ public void testRemoveEmptyMinusRightEmpty() {\nrunRemoveEmptyTest(OpType.MINUS_RIGHT, true);\n}\n@Test\n- public void testRemoveEmptyMatMultLeftEmpty()\n- {\n+ public void testRemoveEmptyMatMultLeftEmpty() {\nrunRemoveEmptyTest(OpType.MM_LEFT, true);\n}\n@Test\n- public void testRemoveEmptyMatMultRightEmpty()\n- {\n+ public void testRemoveEmptyMatMultRightEmpty() {\nrunRemoveEmptyTest(OpType.MM_RIGHT, true);\n}\n@Test\n- public void testRemoveEmptyRIXEmpty()\n- {\n+ public void testRemoveEmptyRIXEmpty() {\nrunRemoveEmptyTest(OpType.RIX, true);\n}\n@Test\n- public void testRemoveEmptyLIXEmpty()\n- {\n+ public void testRemoveEmptyLIXEmpty() {\nrunRemoveEmptyTest(OpType.LIX, true);\n}\n-\n- /**\n- *\n- * @param type\n- * @param empty\n- */\nprivate void runRemoveEmptyTest( OpType type, boolean empty )\n{\nboolean oldFlagIPA = OptimizerUtils.ALLOW_INTER_PROCEDURAL_ANALYSIS;\n@@ -272,13 +238,13 @@ public class RemoveEmptyRecompileTest extends AutomatedTestBase\nrunTest(true, false, null, -1);\nrunRScript(true);\n- //CHECK compiled MR jobs\n- int expectNumCompiled = 21; //reblock, 10xGMR, 2x(MMCJ+GMR), 2xGMR(LIX), write\n- Assert.assertEquals(\"Unexpected number of compiled MR jobs.\",\n+ //CHECK compiled Spark jobs\n+ int expectNumCompiled = 24; //reblock, 4x1, 9x2, write\n+ Assert.assertEquals(\"Unexpected number of compiled Spark jobs.\",\nexpectNumCompiled, Statistics.getNoOfCompiledSPInst());\n- //CHECK executed MR jobs\n+ //CHECK executed Spark jobs\nint expectNumExecuted = 0;\n- Assert.assertEquals(\"Unexpected number of executed MR jobs.\",\n+ Assert.assertEquals(\"Unexpected number of executed Spark jobs.\",\nexpectNumExecuted, Statistics.getNoOfExecutedSPInst());\n//CHECK rewrite application\n@@ -295,19 +261,15 @@ public class RemoveEmptyRecompileTest extends AutomatedTestBase\nHashMap<CellIndex, Double> rfile = readRMatrixFromFS(\"R\");\nTestUtils.compareMatrices(dmlfile, rfile, eps, \"DML\", \"R\");\n}\n- finally\n- {\n+ finally {\nOptimizerUtils.ALLOW_INTER_PROCEDURAL_ANALYSIS = oldFlagIPA;\n}\n}\n-\n- private static String getOpcode( OpType type )\n- {\n+ private static String getOpcode( OpType type ) {\nswitch(type){\n//for sum, literal replacement of unary aggregates applies\ncase SUM: return \"rlit\";//return \"uak+\";\n-\ncase ROUND: return \"round\";\ncase TRANSPOSE: return \"r'\";\ncase MULT_LEFT:\n@@ -321,7 +283,6 @@ public class RemoveEmptyRecompileTest extends AutomatedTestBase\ncase RIX: return RightIndex.OPCODE;\ncase LIX: return LeftIndex.OPCODE;\n}\n-\nreturn null;\n}\n}\n\\ No newline at end of file\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "new_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "diff": "@@ -197,7 +197,7 @@ public enum Builtins {\n}\nBuiltins(String name, String alias, boolean script, boolean parameterized) {\n- this(name, alias, script, false, ReturnType.SINGLE_RETURN);\n+ this(name, alias, script, parameterized, ReturnType.SINGLE_RETURN);\n}\nBuiltins(String name, String alias, boolean script, boolean parameterized, ReturnType retType) {\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/ProgramBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/ProgramBlock.java", "diff": "@@ -370,5 +370,4 @@ public abstract class ProgramBlock implements ParseInfo\n_text = parseInfo.getText();\n_filename = parseInfo.getFilename();\n}\n-\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/WhileProgramBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/WhileProgramBlock.java", "diff": "@@ -35,13 +35,11 @@ import org.tugraz.sysds.runtime.instructions.cp.BooleanObject;\npublic class WhileProgramBlock extends ProgramBlock\n{\nprivate ArrayList<Instruction> _predicate;\n- private ArrayList <Instruction> _exitInstructions ;\nprivate ArrayList<ProgramBlock> _childBlocks;\npublic WhileProgramBlock(Program prog, ArrayList<Instruction> predicate) {\nsuper(prog);\n_predicate = predicate;\n- _exitInstructions = new ArrayList<>();\n_childBlocks = new ArrayList<>();\n}\n@@ -49,14 +47,6 @@ public class WhileProgramBlock extends ProgramBlock\n_childBlocks.add(childBlock);\n}\n- public void setExitInstructions2(ArrayList<Instruction> exitInstructions) {\n- _exitInstructions = exitInstructions;\n- }\n-\n- public void setExitInstructions1(ArrayList<Instruction> predicate) {\n- _predicate = predicate;\n- }\n-\npublic ArrayList<Instruction> getPredicate() {\nreturn _predicate;\n}\n@@ -125,14 +115,6 @@ public class WhileProgramBlock extends ProgramBlock\ncatch (Exception e) {\nthrow new DMLRuntimeException(printBlockErrorLocation() + \"Error evaluating while program block\", e);\n}\n-\n- //execute exit instructions\n- try {\n- executeInstructions(_exitInstructions, ec);\n- }\n- catch(Exception e) {\n- throw new DMLRuntimeException(printBlockErrorLocation() + \"Error executing while exit instructions.\", e);\n- }\n}\npublic void setChildBlocks(ArrayList<ProgramBlock> childs) {\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/util/ProgramConverter.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/util/ProgramConverter.java", "diff": "@@ -1332,16 +1332,11 @@ public class ProgramConverter\n//predicate instructions\nArrayList<Instruction> inst = parseInstructions(st.nextToken(),id);\n- //exit instructions\n- ArrayList<Instruction> exit = parseInstructions(st.nextToken(),id);\n-\n//program blocks\nArrayList<ProgramBlock> pbs = rParseProgramBlocks(st.nextToken(), prog, id);\nWhileProgramBlock wpb = new WhileProgramBlock(prog,inst);\n- wpb.setExitInstructions2(exit);\nwpb.setChildBlocks(pbs);\n-\nreturn wpb;\n}\n" } ]

[ { "change_type": "MODIFY", "old_path": "scripts/builtin/steplm.dml", "new_path": "scripts/builtin/steplm.dml", "diff": "# AVG_RES_Y Average of the residual Y - pred(Y|X), i.e. residual bias\nm_steplm = function(Matrix[Double] X, Matrix[Double] y, Integer icpt = 0, Double reg = 1e-7, Double tol = 1e-7, Integer maxi = 0, Boolean verbose = TRUE)\n- return(Matrix[Double] C, Matrix[Double] S)\n-{\n+return(Matrix[Double] C, Matrix[Double] S) {\n+\n# currently only the forward selection strategy in supported: start\n# from one feature and iteratively add features until AIC improves\ndir = \"forward\";\n@@ -103,7 +103,7 @@ m_steplm = function(Matrix[Double] X, Matrix[Double] y, Integer icpt = 0, Double\nfor (i in 1:m_orig, check = 0) {\ncolumns_fixed_ordered_1 = matrix(i, rows = 1, cols = 1);\n[AIC_1, beta_out_i] = linear_regression(X_orig[, i], y, icpt);\n- print(\" mibeta \" + toString(beta_out_i))\n+\nAICs[1, i] = AIC_1;\nAIC_cur = as.scalar(AICs[1, i]);\nif ((AIC_cur < AIC_best) & ((AIC_best - AIC_cur) > abs(thr * AIC_best))) {\n@@ -128,8 +128,8 @@ m_steplm = function(Matrix[Double] X, Matrix[Double] y, Integer icpt = 0, Double\nbeta_out = B;\nS = Selected;\nC = beta_out;\n- stop(\"\");\n- }\n+ } else {\n+\nprint(\"Best AIC \" + AIC_best + \" achieved with feature: \" + column_best);\ncolumns_fixed[1, column_best] = 1;\n@@ -178,7 +178,6 @@ m_steplm = function(Matrix[Double] X, Matrix[Double] y, Integer icpt = 0, Double\ncontinue = FALSE;\n}\n}\n-\n# run linear regression with selected set of features\nprint(\"Running linear regression with selected features...\");\n[AIC, beta_out] = linear_regression(X_global, y, icpt);\n@@ -189,11 +188,11 @@ m_steplm = function(Matrix[Double] X, Matrix[Double] y, Integer icpt = 0, Double\nbeta_out = reorder_matrix(boa_ncol, beta_out, Selected);\nS = Selected;\nC = beta_out;\n+ }\n} else {\nstop(\"Currently only forward selection strategy is supported!\");\n}\n}\n-\n# Computes linear regression using lm and outputs AIC.\nlinear_regression = function(Matrix[Double] X, Matrix[Double] y, Integer icpt = 0)\nreturn(Double AIC, Matrix[Double] beta) {\n@@ -207,6 +206,7 @@ linear_regression = function(Matrix[Double] X, Matrix[Double] y, Integer icpt =\nX = cbind(X, ones_n);\nm = m - 1;\n}\n+\nm_ext = ncol(X);\n# BEGIN THE DIRECT SOLVE ALGORITHM (EXTERNAL CALL)\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinSTEPLmTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinSTEPLmTest.java", "diff": "@@ -26,8 +26,8 @@ import org.tugraz.sysds.test.TestUtils;\nimport java.util.HashMap;\n-public class BuiltinSTEPLmTest extends AutomatedTestBase {\n-\n+public class BuiltinSTEPLmTest extends AutomatedTestBase\n+{\nprivate final static String TEST_NAME = \"steplm\";\nprivate final static String TEST_DIR = \"functions/builtin/\";\nprivate static final String TEST_CLASS_DIR = TEST_DIR + BuiltinSTEPLmTest.class.getSimpleName() + \"/\";\n@@ -38,10 +38,6 @@ public class BuiltinSTEPLmTest extends AutomatedTestBase {\nprivate final static double spSparse = 0.3;\nprivate final static double spDense = 0.7;\n- public enum LinregType {\n- CG, DS, AUTO\n- }\n-\n@Override\npublic void setUp() {\naddTestConfiguration(TEST_NAME, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME, new String[]{\"B\"}));\n@@ -49,32 +45,29 @@ public class BuiltinSTEPLmTest extends AutomatedTestBase {\n@Test\npublic void testLmMatrixDenseCPlm() {\n- runSTEPLmTest(false, ExecType.CP, BuiltinSTEPLmTest.LinregType.AUTO);\n+ runSTEPLmTest(false, ExecType.CP);\n}\n@Test\npublic void testLmMatrixSparseCPlm() {\n- runSTEPLmTest(false, ExecType.CP, BuiltinSTEPLmTest.LinregType.AUTO);\n+ runSTEPLmTest(true, ExecType.CP);\n}\n@Test\npublic void testLmMatrixDenseSPlm() {\n- runSTEPLmTest(false, ExecType.SPARK, BuiltinSTEPLmTest.LinregType.AUTO);\n+ runSTEPLmTest(false, ExecType.SPARK);\n}\n@Test\npublic void testLmMatrixSparseSPlm() {\n- runSTEPLmTest(true, ExecType.SPARK, BuiltinSTEPLmTest.LinregType.AUTO);\n+ runSTEPLmTest(true, ExecType.SPARK);\n}\n- private void runSTEPLmTest(boolean sparse, ExecType instType, BuiltinSTEPLmTest.LinregType linregAlgo) {\n+ private void runSTEPLmTest(boolean sparse, ExecType instType) {\nExecMode platformOld = setExecMode(instType);\n-\nString dml_test_name = TEST_NAME;\ntry {\n- //disableOutAndExpectedDeletion();\n-\nloadTestConfiguration(getTestConfiguration(TEST_NAME));\ndouble sparsity = sparse ? spSparse : spDense;\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/builtin/steplm.dml", "new_path": "src/test/scripts/functions/builtin/steplm.dml", "diff": "#\n#-------------------------------------------------------------\n-X = read($1)\n-y = read($2)\n-[S, C] = steplm(X = X, y = y, icpt = 1);\n-write(S, \"S\");\n-write(C, \"C\");\n+X = read($1);\n+y = read($2);\n+\n+[C, S] = steplm(X = X, y = y, icpt = 1);\n+\n+write(C, $3);\n+write(S, $4);\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlock.java", "diff": "@@ -501,6 +501,15 @@ public abstract class DenseBlock implements Serializable\n*/\npublic abstract DenseBlock set(int[] ix, double v);\n+ /**\n+ * Set the specified cell to the given value.\n+ *\n+ * @param ix cell indexes\n+ * @param v value\n+ * @return self\n+ */\n+ public abstract DenseBlock set(int[] ix, long v);\n+\n/**\n* Set the specified cell to the given value.\n*\n@@ -549,6 +558,14 @@ public abstract class DenseBlock implements Serializable\n*/\npublic abstract String getString(int[] ix);\n+ /**\n+ * Get the value of a given cell as long\n+ *\n+ * @param ix cell indexes\n+ * @return value as long\n+ */\n+ public abstract long getLong(int[] ix);\n+\n@Override\npublic String toString() {\nStringBuilder sb = new StringBuilder();\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockBool.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockBool.java", "diff": "@@ -187,6 +187,12 @@ public class DenseBlockBool extends DenseBlockDRB\nreturn this;\n}\n+ @Override\n+ public DenseBlock set(int[] ix, long v) {\n+ _data.set(pos(ix), v != 0);\n+ return this;\n+ }\n+\n@Override\npublic DenseBlock set(int[] ix, String v) {\n_data.set(pos(ix), Boolean.parseBoolean(v));\n@@ -207,4 +213,9 @@ public class DenseBlockBool extends DenseBlockDRB\npublic String getString(int[] ix) {\nreturn String.valueOf(_data.get(pos(ix)));\n}\n+\n+ @Override\n+ public long getLong(int[] ix) {\n+ return _data.get(pos(ix)) ? 1 : 0;\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockFP32.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockFP32.java", "diff": "@@ -140,6 +140,12 @@ public class DenseBlockFP32 extends DenseBlockDRB\nreturn this;\n}\n+ @Override\n+ public DenseBlock set(int[] ix, long v) {\n+ _data[pos(ix)] = v;\n+ return this;\n+ }\n+\n@Override\npublic DenseBlock set(int[] ix, String v) {\n_data[pos(ix)] = Float.parseFloat(v);\n@@ -160,4 +166,9 @@ public class DenseBlockFP32 extends DenseBlockDRB\npublic String getString(int[] ix) {\nreturn String.valueOf(_data[pos(ix)]);\n}\n+\n+ @Override\n+ public long getLong(int[] ix) {\n+ return UtilFunctions.toLong(_data[pos(ix)]);\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockFP64.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockFP64.java", "diff": "@@ -160,6 +160,12 @@ public class DenseBlockFP64 extends DenseBlockDRB\nreturn this;\n}\n+ @Override\n+ public DenseBlock set(int[] ix, long v) {\n+ _data[pos(ix)] = v;\n+ return this;\n+ }\n+\n@Override\npublic DenseBlock set(int[] ix, String v) {\n_data[pos(ix)] = Double.parseDouble(v);\n@@ -180,4 +186,9 @@ public class DenseBlockFP64 extends DenseBlockDRB\npublic String getString(int[] ix) {\nreturn String.valueOf(_data[pos(ix)]);\n}\n+\n+ @Override\n+ public long getLong(int[] ix) {\n+ return UtilFunctions.toLong(_data[pos(ix)]);\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockInt32.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockInt32.java", "diff": "@@ -135,6 +135,12 @@ public class DenseBlockInt32 extends DenseBlockDRB\nreturn this;\n}\n+ @Override\n+ public DenseBlock set(int[] ix, long v) {\n+ _data[pos(ix)] = (int) v;\n+ return this;\n+ }\n+\n@Override\npublic DenseBlock set(int[] ix, String v) {\n_data[pos(ix)] = Integer.parseInt(v);\n@@ -155,4 +161,9 @@ public class DenseBlockInt32 extends DenseBlockDRB\npublic String getString(int[] ix) {\nreturn String.valueOf(_data[pos(ix)]);\n}\n+\n+ @Override\n+ public long getLong(int[] ix) {\n+ return _data[pos(ix)];\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockInt64.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockInt64.java", "diff": "@@ -133,6 +133,12 @@ public class DenseBlockInt64 extends DenseBlockDRB\nreturn this;\n}\n+ @Override\n+ public DenseBlock set(int[] ix, long v) {\n+ _data[pos(ix)] = v;\n+ return this;\n+ }\n+\n@Override\npublic DenseBlock set(int[] ix, String v) {\n_data[pos(ix)] = Long.parseLong(v);\n@@ -153,4 +159,9 @@ public class DenseBlockInt64 extends DenseBlockDRB\npublic String getString(int[] ix) {\nreturn String.valueOf(_data[pos(ix)]);\n}\n+\n+ @Override\n+ public long getLong(int[] ix) {\n+ return _data[pos(ix)];\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockLBool.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockLBool.java", "diff": "@@ -74,7 +74,7 @@ public class DenseBlockLBool extends DenseBlockLDRB\nint numBlocks = UtilFunctions.toInt(Math.ceil((double) rlen / newBlockSize));\nif (_blen == newBlockSize && dataLength <= capacity()) {\nfor (int i = 0; i < numBlocks; i++) {\n- int toIndex = (int)Math.min((long)newBlockSize, dataLength - i * newBlockSize) * _odims[0];\n+ int toIndex = (int)Math.min(newBlockSize, dataLength - i * newBlockSize) * _odims[0];\n_blocks[i].set(0, toIndex, bv);\n// Clear old data so we can use cardinality for computeNnz\n_blocks[i].set(toIndex, _blocks[i].size(), false);\n@@ -167,6 +167,11 @@ public class DenseBlockLBool extends DenseBlockLDRB\nreturn this;\n}\n+ @Override\n+ public DenseBlock set(int[] ix, long v) {\n+ _blocks[index(ix[0])].set(pos(ix), v != 0);\n+ return this;\n+ }\n@Override\npublic DenseBlock set(int[] ix, String v) {\n_blocks[index(ix[0])].set(pos(ix), Boolean.parseBoolean(v));\n@@ -187,4 +192,9 @@ public class DenseBlockLBool extends DenseBlockLDRB\npublic String getString(int[] ix) {\nreturn String.valueOf(_blocks[index(ix[0])].get(pos(ix)));\n}\n+\n+ @Override\n+ public long getLong(int[] ix) {\n+ return _blocks[index(ix[0])].get(pos(ix)) ? 1 : 0;\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockLFP32.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockLFP32.java", "diff": "@@ -116,6 +116,12 @@ public class DenseBlockLFP32 extends DenseBlockLDRB\nreturn this;\n}\n+ @Override\n+ public DenseBlock set(int[] ix, long v) {\n+ _blocks[index(ix[0])][pos(ix)] = v;\n+ return this;\n+ }\n+\n@Override\npublic DenseBlock set(int[] ix, String v) {\n_blocks[index(ix[0])][pos(ix)] = Float.parseFloat(v);\n@@ -136,4 +142,9 @@ public class DenseBlockLFP32 extends DenseBlockLDRB\npublic String getString(int[] ix) {\nreturn String.valueOf(_blocks[index(ix[0])][pos(ix)]);\n}\n+\n+ @Override\n+ public long getLong(int[] ix) {\n+ return UtilFunctions.toLong(_blocks[index(ix[0])][pos(ix)]);\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockLFP64.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockLFP64.java", "diff": "@@ -114,6 +114,12 @@ public class DenseBlockLFP64 extends DenseBlockLDRB\nreturn this;\n}\n+ @Override\n+ public DenseBlock set(int[] ix, long v) {\n+ _blocks[index(ix[0])][pos(ix)] = v;\n+ return this;\n+ }\n+\n@Override\npublic DenseBlock set(int[] ix, String v) {\n_blocks[index(ix[0])][pos(ix)] = Double.parseDouble(v);\n@@ -134,4 +140,9 @@ public class DenseBlockLFP64 extends DenseBlockLDRB\npublic String getString(int[] ix) {\nreturn String.valueOf(_blocks[index(ix[0])][pos(ix)]);\n}\n+\n+ @Override\n+ public long getLong(int[] ix) {\n+ return UtilFunctions.toLong(_blocks[index(ix[0])][pos(ix)]);\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockLInt32.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockLInt32.java", "diff": "@@ -117,6 +117,12 @@ public class DenseBlockLInt32 extends DenseBlockLDRB\nreturn this;\n}\n+ @Override\n+ public DenseBlock set(int[] ix, long v) {\n+ _blocks[index(ix[0])][pos(ix)] = (int) v;\n+ return this;\n+ }\n+\n@Override\npublic DenseBlock set(int[] ix, String v) {\n_blocks[index(ix[0])][pos(ix)] = Integer.parseInt(v);\n@@ -137,4 +143,9 @@ public class DenseBlockLInt32 extends DenseBlockLDRB\npublic String getString(int[] ix) {\nreturn String.valueOf(_blocks[ix[0]][pos(ix)]);\n}\n+\n+ @Override\n+ public long getLong(int[] ix) {\n+ return _blocks[index(ix[0])][pos(ix)];\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockLInt64.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockLInt64.java", "diff": "@@ -117,6 +117,12 @@ public class DenseBlockLInt64 extends DenseBlockLDRB\nreturn this;\n}\n+ @Override\n+ public DenseBlock set(int[] ix, long v) {\n+ _blocks[index(ix[0])][pos(ix)] = v;\n+ return this;\n+ }\n+\n@Override\npublic DenseBlock set(int[] ix, String v) {\n_blocks[index(ix[0])][pos(ix)] = Long.parseLong(v);\n@@ -137,4 +143,9 @@ public class DenseBlockLInt64 extends DenseBlockLDRB\npublic String getString(int[] ix) {\nreturn String.valueOf(_blocks[index(ix[0])][pos(ix)]);\n}\n+\n+ @Override\n+ public long getLong(int[] ix) {\n+ return _blocks[index(ix[0])][pos(ix)];\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockLString.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockLString.java", "diff": "@@ -126,6 +126,12 @@ public class DenseBlockLString extends DenseBlockLDRB\nreturn this;\n}\n+ @Override\n+ public DenseBlock set(int[] ix, long v) {\n+ _blocks[index(ix[0])][pos(ix)] = String.valueOf(v);\n+ return this;\n+ }\n+\n@Override\npublic DenseBlock set(int[] ix, String v) {\n_blocks[index(ix[0])][pos(ix)] = v;\n@@ -146,4 +152,9 @@ public class DenseBlockLString extends DenseBlockLDRB\npublic String getString(int[] ix) {\nreturn _blocks[index(ix[0])][pos(ix)];\n}\n+\n+ @Override\n+ public long getLong(int[] ix) {\n+ return Long.parseLong(_blocks[index(ix[0])][pos(ix)]);\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockString.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DenseBlockString.java", "diff": "@@ -147,6 +147,12 @@ public class DenseBlockString extends DenseBlockDRB {\nreturn this;\n}\n+ @Override\n+ public DenseBlock set(int[] ix, long v) {\n+ _data[pos(ix)] = String.valueOf(v);\n+ return this;\n+ }\n+\n@Override\npublic double get(int r, int c) {\nreturn Double.parseDouble(_data[pos(r, c)]);\n@@ -167,4 +173,9 @@ public class DenseBlockString extends DenseBlockDRB {\npublic String getString(int[] ix) {\nreturn _data[pos(ix)];\n}\n+\n+ @Override\n+ public long getLong(int[] ix) {\n+ return Long.parseLong(_data[pos(ix)]);\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/TensorBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/TensorBlock.java", "diff": "@@ -412,6 +412,15 @@ public class TensorBlock implements CacheBlock\n}\n}\n+ public long getLong(int[] ix) {\n+ if (_sparse) {\n+ // TODO: Implement sparse\n+ throw new NotImplementedException();\n+ } else {\n+ return _denseBlock.getLong(ix);\n+ }\n+ }\n+\npublic String getString(int[] ix) {\nif (_sparse) {\n// TODO: Implement sparse\n@@ -438,6 +447,14 @@ public class TensorBlock implements CacheBlock\n}\n}\n+ public void set(int[] ix, long v) {\n+ if (_sparse) {\n+ throw new NotImplementedException();\n+ } else {\n+ _denseBlock.set(ix, v);\n+ }\n+ }\n+\npublic void set(int[] ix, String v) {\nif (_sparse) {\nthrow new NotImplementedException();\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -59,6 +59,7 @@ SYSTEMDS-70 Lineage Tracing and Reuse OK\n* 75 Reuse cache based on lineage traces OK\n* 76 Generate runtime plan from lineage trace OK\n* 77 New builtin function for obtaining lineage OK\n+ * 78 Extended lineage tracing (parfor, funs, codegen)\nSYSTEMDS-80 Improved distributed operations\n* 81 Avoid unnecessary replication on rmm\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/Instruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/Instruction.java", "diff": "@@ -191,8 +191,7 @@ public abstract class Instruction\nreturn extendedOpcode;\n}\n- public boolean requiresLabelUpdate()\n- {\n+ public boolean requiresLabelUpdate() {\nreturn instString.contains( Lop.VARIABLE_NAME_PLACEHOLDER );\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/ComputationCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/ComputationCPInstruction.java", "diff": "@@ -23,12 +23,11 @@ import org.tugraz.sysds.api.DMLScript;\nimport org.tugraz.sysds.common.Types.ExecMode;\nimport org.tugraz.sysds.hops.OptimizerUtils;\nimport org.tugraz.sysds.runtime.controlprogram.caching.CacheableData;\n-import org.tugraz.sysds.runtime.lineage.Lineage;\nimport org.tugraz.sysds.runtime.lineage.LineageTraceable;\nimport org.tugraz.sysds.runtime.lineage.LineageItem;\n+import org.tugraz.sysds.runtime.lineage.LineageItemUtils;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport org.tugraz.sysds.runtime.matrix.operators.Operator;\n-import java.util.ArrayList;\npublic abstract class ComputationCPInstruction extends CPInstruction implements LineageTraceable {\n@@ -77,14 +76,7 @@ public abstract class ComputationCPInstruction extends CPInstruction implements\n@Override\npublic LineageItem[] getLineageItems() {\n- ArrayList<LineageItem> lineages = new ArrayList<>();\n- if (input1 != null)\n- lineages.add(Lineage.getOrCreate(input1));\n- if (input2 != null)\n- lineages.add(Lineage.getOrCreate(input2));\n- if (input3 != null)\n- lineages.add(Lineage.getOrCreate(input3));\nreturn new LineageItem[]{new LineageItem(output.getName(),\n- getOpcode(), lineages.toArray(new LineageItem[0]))};\n+ getOpcode(), LineageItemUtils.getLineage(input1,input2,input3))};\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/MatrixIndexingCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/MatrixIndexingCPInstruction.java", "diff": "@@ -28,14 +28,12 @@ import org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject;\nimport org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject.UpdateType;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\n-import org.tugraz.sysds.runtime.lineage.Lineage;\nimport org.tugraz.sysds.runtime.lineage.LineageItem;\n+import org.tugraz.sysds.runtime.lineage.LineageItemUtils;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport org.tugraz.sysds.runtime.util.IndexRange;\nimport org.tugraz.sysds.utils.Statistics;\n-import java.util.Arrays;\n-\npublic final class MatrixIndexingCPInstruction extends IndexingCPInstruction {\npublic MatrixIndexingCPInstruction(CPOperand in, CPOperand rl, CPOperand ru, CPOperand cl, CPOperand cu,\n@@ -124,8 +122,7 @@ public final class MatrixIndexingCPInstruction extends IndexingCPInstruction {\n@Override\npublic LineageItem[] getLineageItems() {\n- LineageItem[] tmp = Arrays.asList(input1,input2,input3,colLower,colUpper,rowLower,rowUpper)\n- .stream().filter(c -> c!=null).map(c -> Lineage.getOrCreate(c)).toArray(LineageItem[]::new);\n- return new LineageItem[]{new LineageItem(output.getName(), getOpcode(), tmp)};\n+ return new LineageItem[]{new LineageItem(output.getName(), getOpcode(),\n+ LineageItemUtils.getLineage(input1,input2,input3,colLower,colUpper,rowLower,rowUpper))};\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/MultiReturnBuiltinCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/MultiReturnBuiltinCPInstruction.java", "diff": "@@ -26,8 +26,8 @@ import org.tugraz.sysds.common.Types.ValueType;\nimport org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\n-import org.tugraz.sysds.runtime.lineage.Lineage;\nimport org.tugraz.sysds.runtime.lineage.LineageItem;\n+import org.tugraz.sysds.runtime.lineage.LineageItemUtils;\nimport org.tugraz.sysds.runtime.matrix.data.LibCommonsMath;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport org.tugraz.sysds.runtime.matrix.operators.Operator;\n@@ -112,19 +112,10 @@ public class MultiReturnBuiltinCPInstruction extends ComputationCPInstruction {\n@Override\npublic LineageItem[] getLineageItems() {\n- ArrayList<LineageItem> lineages = new ArrayList<>();\n- if (input1 != null)\n- lineages.add(Lineage.getOrCreate(input1));\n- if (input2 != null)\n- lineages.add(Lineage.getOrCreate(input2));\n- if (input3 != null)\n- lineages.add(Lineage.getOrCreate(input3));\n-\n+ LineageItem[] inputLineage = LineageItemUtils.getLineage(input1,input2,input3);\nArrayList<LineageItem> items = new ArrayList<>();\n- for (CPOperand out : _outputs) {\n- items.add(new LineageItem(out.getName(),\n- getOpcode(), lineages.toArray(new LineageItem[0])));\n- }\n+ for (CPOperand out : _outputs)\n+ items.add(new LineageItem(out.getName(), getOpcode(), inputLineage));\nreturn items.toArray(new LineageItem[items.size()]);\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/SpoofCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/SpoofCPInstruction.java", "diff": "@@ -26,6 +26,8 @@ import org.tugraz.sysds.runtime.codegen.CodegenUtils;\nimport org.tugraz.sysds.runtime.codegen.SpoofOperator;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\n+import org.tugraz.sysds.runtime.lineage.LineageItem;\n+import org.tugraz.sysds.runtime.lineage.LineageItemUtils;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\npublic class SpoofCPInstruction extends ComputationCPInstruction {\n@@ -92,4 +94,10 @@ public class SpoofCPInstruction extends ComputationCPInstruction {\nif(input.getDataType()==DataType.MATRIX)\nec.releaseMatrixInput(input.getName());\n}\n+\n+ @Override\n+ public LineageItem[] getLineageItems() {\n+ return new LineageItem[]{new LineageItem(output.getName(),\n+ getOpcode(), LineageItemUtils.getLineage(_in))};\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/VariableCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/VariableCPInstruction.java", "diff": "@@ -54,6 +54,7 @@ import org.tugraz.sysds.runtime.io.WriterMatrixMarket;\nimport org.tugraz.sysds.runtime.io.WriterTextCSV;\nimport org.tugraz.sysds.runtime.lineage.Lineage;\nimport org.tugraz.sysds.runtime.lineage.LineageItem;\n+import org.tugraz.sysds.runtime.lineage.LineageItemUtils;\nimport org.tugraz.sysds.runtime.lineage.LineageTraceable;\nimport org.tugraz.sysds.runtime.matrix.data.FrameBlock;\nimport org.tugraz.sysds.runtime.matrix.data.InputInfo;\n@@ -1190,6 +1191,16 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\ngetOpcode(), lineages.toArray(new LineageItem[0]));\nbreak;\n}\n+ case CastAsBooleanVariable:\n+ case CastAsDoubleVariable:\n+ case CastAsIntegerVariable:\n+ case CastAsScalarVariable:\n+ case CastAsMatrixVariable:\n+ case CastAsFrameVariable:{\n+ li = new LineageItem(getOutputVariableName(),\n+ getOpcode(), LineageItemUtils.getLineage(getInput1()));\n+ break;\n+ }\ncase RemoveVariable:\ndefault:\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageItemUtils.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageItemUtils.java", "diff": "@@ -150,6 +150,11 @@ public class LineageItemUtils {\nreturn ec.getVariable(varname);\n}\n+ public static LineageItem[] getLineage(CPOperand... operands) {\n+ return Arrays.stream(operands).filter(c -> c!=null)\n+ .map(c -> Lineage.getOrCreate(c)).toArray(LineageItem[]::new);\n+ }\n+\nprivate static void rConstructHops(LineageItem item, HashMap<Long, Hop> operands) {\nif (item.isVisited())\nreturn;\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageMap.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageMap.java", "diff": "@@ -114,6 +114,15 @@ public class LineageMap {\nprocessMoveLI(li);\nbreak;\n}\n+ case CastAsBooleanVariable:\n+ case CastAsDoubleVariable:\n+ case CastAsIntegerVariable:\n+ case CastAsScalarVariable:\n+ case CastAsMatrixVariable:\n+ case CastAsFrameVariable: {\n+ addLineageItem(li);\n+ break;\n+ }\ndefault:\nthrow new DMLRuntimeException(\"Unknown VariableCPInstruction (\" + inst.getOpcode() + \") traced.\");\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/codegenalg/AlgorithmMSVM.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/codegenalg/AlgorithmMSVM.java", "diff": "@@ -22,6 +22,7 @@ package org.tugraz.sysds.test.functions.codegenalg;\nimport java.io.File;\nimport java.util.HashMap;\n+import org.apache.commons.lang.ArrayUtils;\nimport org.junit.Assert;\nimport org.junit.Test;\nimport org.tugraz.sysds.api.DMLScript;\n@@ -148,7 +149,31 @@ public class AlgorithmMSVM extends AutomatedTestBase\nrunMSVMTest(TEST_NAME1, true, true, 4, ExecType.CP, TestType.FUSE_NO_REDUNDANCY);\n}\n- private void runMSVMTest( String testname, boolean rewrites, boolean sparse, int numClasses, ExecType instType, TestType testType)\n+ private void runMSVMTest( String testname, boolean rewrites, boolean sparse, int numClasses, ExecType instType, TestType testType) {\n+ runMSVMTest(testname, rewrites, sparse, false, numClasses, instType, testType);\n+ }\n+\n+ @Test\n+ public void testMSVMDenseMulRewritesCPLineage() {\n+ runMSVMTest(TEST_NAME1, true, false, true, 4, ExecType.CP, TestType.DEFAULT);\n+ }\n+\n+ @Test\n+ public void testMSVMSparseMulRewritesCPLineage() {\n+ runMSVMTest(TEST_NAME1, true, true, true, 4, ExecType.CP, TestType.DEFAULT);\n+ }\n+\n+ @Test\n+ public void testMSVMDenseMulCPLineage() {\n+ runMSVMTest(TEST_NAME1, false, false, true, 4, ExecType.CP, TestType.DEFAULT);\n+ }\n+\n+ @Test\n+ public void testMSVMSparseMulCPLineage() {\n+ runMSVMTest(TEST_NAME1, false, true, true, 4, ExecType.CP, TestType.DEFAULT);\n+ }\n+\n+ private void runMSVMTest( String testname, boolean rewrites, boolean sparse, boolean lineage, int numClasses, ExecType instType, TestType testType)\n{\nboolean oldFlag = OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION;\nExecMode platformOld = rtplatform;\n@@ -171,6 +196,8 @@ public class AlgorithmMSVM extends AutomatedTestBase\nprogramArgs = new String[]{ \"-explain\", \"-stats\", \"-nvargs\", \"X=\"+input(\"X\"), \"Y=\"+input(\"Y\"),\n\"icpt=\"+String.valueOf(intercept), \"tol=\"+String.valueOf(epsilon), \"reg=0.001\",\n\"maxiter=\"+String.valueOf(maxiter), \"model=\"+output(\"w\"), \"Log= \"};\n+ if( lineage )\n+ programArgs = (String[])ArrayUtils.addAll(new String[]{\"-lineage\"}, programArgs);\nrCmd = getRCmd(inputDir(), String.valueOf(intercept),String.valueOf(epsilon),\nString.valueOf(maxiter), expectedDir());\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/FunctionCallCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/FunctionCallCPInstruction.java", "diff": "@@ -37,6 +37,7 @@ import org.tugraz.sysds.runtime.controlprogram.context.ExecutionContextFactory;\nimport org.tugraz.sysds.runtime.instructions.Instruction;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\nimport org.tugraz.sysds.runtime.io.IOUtilFunctions;\n+import org.tugraz.sysds.runtime.lineage.Lineage;\npublic class FunctionCallCPInstruction extends CPInstruction {\nprivate final String _functionName;\n@@ -111,6 +112,7 @@ public class FunctionCallCPInstruction extends CPInstruction {\n// create bindings to formal parameters for given function call\n// These are the bindings passed to the FunctionProgramBlock for function execution\nLocalVariableMap functionVariables = new LocalVariableMap();\n+ Lineage lineage = DMLScript.LINEAGE ? new Lineage() : null;\nfor( int i=0; i<_boundInputs.length; i++) {\n//error handling non-existing variables\nCPOperand input = _boundInputs[i];\n@@ -138,6 +140,10 @@ public class FunctionCallCPInstruction extends CPInstruction {\n//set input parameter\nfunctionVariables.put(currFormalParam.getName(), value);\n+\n+ //map lineage to function arguments\n+ if( lineage != null )\n+ lineage.set(currFormalParam.getName(), ec.getLineageItem(input));\n}\n// Pin the input variables so that they do not get deleted\n@@ -146,12 +152,13 @@ public class FunctionCallCPInstruction extends CPInstruction {\n// Create a symbol table under a new execution context for the function invocation,\n// and copy the function arguments into the created table.\n- ExecutionContext fn_ec = ExecutionContextFactory.createContext(false, ec.getProgram());\n+ ExecutionContext fn_ec = ExecutionContextFactory.createContext(false, false, ec.getProgram());\nif (DMLScript.USE_ACCELERATOR) {\nfn_ec.setGPUContexts(ec.getGPUContexts());\nfn_ec.getGPUContext(0).initializeThread();\n}\nfn_ec.setVariables(functionVariables);\n+ fn_ec.setLineage(lineage);\n// execute the function block\ntry {\nfpb._functionName = this._functionName;\n@@ -187,7 +194,8 @@ public class FunctionCallCPInstruction extends CPInstruction {\nint numOutputs = Math.min(_boundOutputNames.size(), fpb.getOutputParams().size());\nfor (int i=0; i< numOutputs; i++) {\nString boundVarName = _boundOutputNames.get(i);\n- Data boundValue = retVars.get(fpb.getOutputParams().get(i).getName());\n+ String retVarName = fpb.getOutputParams().get(i).getName();\n+ Data boundValue = retVars.get(retVarName);\nif (boundValue == null)\nthrow new DMLRuntimeException(boundVarName + \" was not assigned a return value\");\n@@ -198,6 +206,10 @@ public class FunctionCallCPInstruction extends CPInstruction {\n//add/replace data in symbol table\nec.setVariable(boundVarName, boundValue);\n+\n+ //map lineage of function returns back to calling site\n+ if( lineage != null ) //unchanged ref\n+ ec.getLineage().set(boundVarName, lineage.get(retVarName));\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageMap.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageMap.java", "diff": "@@ -6,6 +6,7 @@ import org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.instructions.Instruction;\nimport org.tugraz.sysds.runtime.instructions.cp.CPOperand;\n+import org.tugraz.sysds.runtime.instructions.cp.FunctionCallCPInstruction;\nimport org.tugraz.sysds.runtime.instructions.cp.VariableCPInstruction;\nimport org.tugraz.sysds.runtime.lineage.LineageItem.LineageItemType;\nimport org.tugraz.sysds.utils.Explain;\n@@ -28,6 +29,8 @@ public class LineageMap {\n}\npublic void trace(Instruction inst, ExecutionContext ec) {\n+ if( inst instanceof FunctionCallCPInstruction )\n+ return; // no need for lineage tracing\nif (!(inst instanceof LineageTraceable))\nthrow new DMLRuntimeException(\"Unknown Instruction (\" + inst.getOpcode() + \") traced.\");\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/FullReuseTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/FullReuseTest.java", "diff": "@@ -18,6 +18,7 @@ package org.tugraz.sysds.test.functions.lineage;\nimport org.junit.Test;\nimport org.tugraz.sysds.hops.OptimizerUtils;\n+import org.tugraz.sysds.hops.recompile.Recompiler;\nimport org.tugraz.sysds.runtime.lineage.Lineage;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixValue;\nimport org.tugraz.sysds.test.AutomatedTestBase;\n@@ -76,7 +77,6 @@ public class FullReuseTest extends AutomatedTestBase {\nList<String> proArgs = new ArrayList<>();\nproArgs.add(\"-stats\");\nproArgs.add(\"-lineage\");\n-// proArgs.add(\"-explain\");\nproArgs.add(\"-args\");\nproArgs.add(output(\"X\"));\nprogramArgs = proArgs.toArray(new String[proArgs.size()]);\n@@ -90,7 +90,6 @@ public class FullReuseTest extends AutomatedTestBase {\nproArgs.add(\"-stats\");\nproArgs.add(\"-lineage\");\nproArgs.add(\"reuse\");\n-// proArgs.add(\"-explain\");\nproArgs.add(\"-args\");\nproArgs.add(output(\"X\"));\nprogramArgs = proArgs.toArray(new String[proArgs.size()]);\n@@ -100,9 +99,11 @@ public class FullReuseTest extends AutomatedTestBase {\nHashMap<MatrixValue.CellIndex, Double> X_reused = readDMLMatrixFromHDFS(\"X\");\nTestUtils.compareMatrices(X_orig, X_reused, 1e-6, \"Origin\", \"Reused\");\n- } finally {\n+ }\n+ finally {\nOptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = old_simplification;\nOptimizerUtils.ALLOW_SUM_PRODUCT_REWRITES = old_sum_product;\n+ Recompiler.reinitRecompiler();\n}\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageReadTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageReadTest.java", "diff": "@@ -18,6 +18,7 @@ package org.tugraz.sysds.test.functions.lineage;\nimport org.junit.Test;\nimport org.tugraz.sysds.hops.OptimizerUtils;\n+import org.tugraz.sysds.hops.recompile.Recompiler;\nimport org.tugraz.sysds.runtime.lineage.LineageItem;\nimport org.tugraz.sysds.runtime.lineage.LineageParser;\nimport org.tugraz.sysds.test.AutomatedTestBase;\n@@ -37,9 +38,12 @@ public class LineageReadTest extends AutomatedTestBase {\n@Test\npublic void testLineageRead() {\n+ boolean oldRewrites = OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION;\n+ boolean oldRewrites2 = OptimizerUtils.ALLOW_SUM_PRODUCT_REWRITES;\n+\n+ try {\nOptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = false;\nOptimizerUtils.ALLOW_SUM_PRODUCT_REWRITES = false;\n-\ngetAndLoadTestConfiguration(TEST_NAME);\nString lineage =\n@@ -55,4 +59,10 @@ public class LineageReadTest extends AutomatedTestBase {\nLineageItem li = LineageParser.parseLineageTrace(lineage);\nTestUtils.compareScalars(lineage, Explain.explain(li));\n}\n+ finally {\n+ OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = oldRewrites;\n+ OptimizerUtils.ALLOW_SUM_PRODUCT_REWRITES = oldRewrites2;\n+ Recompiler.reinitRecompiler();\n+ }\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageTraceDedupTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageTraceDedupTest.java", "diff": "@@ -19,6 +19,7 @@ package org.tugraz.sysds.test.functions.lineage;\nimport org.junit.Test;\nimport org.tugraz.sysds.common.Types;\nimport org.tugraz.sysds.hops.OptimizerUtils;\n+import org.tugraz.sysds.hops.recompile.Recompiler;\nimport org.tugraz.sysds.runtime.lineage.Lineage;\nimport org.tugraz.sysds.runtime.lineage.LineageItem;\nimport org.tugraz.sysds.runtime.lineage.LineageParser;\n@@ -142,10 +143,12 @@ public class LineageTraceDedupTest extends AutomatedTestBase {\nString dedup_trace = readDMLLineageFromHDFS(\"R\");\nLineageItem dedup_li = LineageParser.parseLineageTrace(dedup_trace);\nassertEquals(dedup_li, li);\n- } finally {\n+ }\n+ finally {\nOptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = old_simplification;\nOptimizerUtils.ALLOW_SUM_PRODUCT_REWRITES = old_sum_product;\nAutomatedTestBase.rtplatform = old_rtplatform;\n+ Recompiler.reinitRecompiler();\n}\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageTraceEqualsTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageTraceEqualsTest.java", "diff": "@@ -21,6 +21,7 @@ import java.util.List;\nimport org.junit.Test;\nimport org.tugraz.sysds.hops.OptimizerUtils;\n+import org.tugraz.sysds.hops.recompile.Recompiler;\nimport org.tugraz.sysds.runtime.lineage.Lineage;\nimport org.tugraz.sysds.runtime.lineage.LineageItem;\nimport org.tugraz.sysds.runtime.lineage.LineageParser;\n@@ -98,9 +99,11 @@ public class LineageTraceEqualsTest extends AutomatedTestBase {\nassertTrue(X_li.hashCode() == Z_li.hashCode());\nassertTrue(X_li.equals(Z_li));\n- } finally {\n+ }\n+ finally {\nOptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = old_simplification;\nOptimizerUtils.ALLOW_SUM_PRODUCT_REWRITES = old_sum_product;\n+ Recompiler.reinitRecompiler();\n}\n}\n}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageTraceFunctionTest.java", "diff": "+/*\n+ * Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed under the Apache License, Version 2.0 (the \"License\");\n+ * you may not use this file except in compliance with the License.\n+ * You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing, software\n+ * distributed under the License is distributed on an \"AS IS\" BASIS,\n+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+ * See the License for the specific language governing permissions and\n+ * limitations under the License.\n+ */\n+\n+package org.tugraz.sysds.test.functions.lineage;\n+\n+import java.util.ArrayList;\n+import java.util.HashMap;\n+import java.util.List;\n+\n+import org.junit.Test;\n+import org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject;\n+import org.tugraz.sysds.runtime.instructions.cp.Data;\n+import org.tugraz.sysds.runtime.lineage.Lineage;\n+import org.tugraz.sysds.runtime.lineage.LineageItem;\n+import org.tugraz.sysds.runtime.lineage.LineageItemUtils;\n+import org.tugraz.sysds.runtime.lineage.LineageParser;\n+import org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\n+import org.tugraz.sysds.runtime.matrix.data.MatrixValue.CellIndex;\n+import org.tugraz.sysds.test.AutomatedTestBase;\n+import org.tugraz.sysds.test.TestConfiguration;\n+import org.tugraz.sysds.test.TestUtils;\n+\n+public class LineageTraceFunctionTest extends AutomatedTestBase\n+{\n+ protected static final String TEST_DIR = \"functions/lineage/\";\n+ protected static final String TEST_NAME1 = \"LineageTraceFun1\"; //rand - matrix result\n+ protected static final String TEST_NAME2 = \"LineageTraceFun2\"; //rand - matrix result\n+\n+ protected String TEST_CLASS_DIR = TEST_DIR + LineageTraceFunctionTest.class.getSimpleName() + \"/\";\n+\n+ protected static final int numRecords = 50;\n+ protected static final int numFeatures = 10;\n+\n+ public LineageTraceFunctionTest() {\n+\n+ }\n+\n+ @Override\n+ public void setUp() {\n+ TestUtils.clearAssertionInformation();\n+ addTestConfiguration( TEST_NAME1, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME1, new String[] {\"R\"}) );\n+ addTestConfiguration( TEST_NAME2, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME2, new String[] {\"R\"}) );\n+ }\n+\n+ @Test\n+ public void testLineageTraceFunction1() {\n+ testLineageTraceFunction(TEST_NAME1);\n+ }\n+\n+ @Test\n+ public void testLineageTraceFunction2() {\n+ testLineageTraceFunction(TEST_NAME2);\n+ }\n+\n+ private void testLineageTraceFunction(String testname) {\n+ System.out.println(\"------------ BEGIN \" + testname + \"------------\");\n+\n+ getAndLoadTestConfiguration(testname);\n+ List<String> proArgs = new ArrayList<String>();\n+\n+ proArgs.add(\"-explain\");\n+ proArgs.add(\"-lineage\");\n+ proArgs.add(\"-args\");\n+ proArgs.add(input(\"X\"));\n+ proArgs.add(output(\"R\"));\n+ proArgs.add(String.valueOf(numRecords));\n+ proArgs.add(String.valueOf(numFeatures));\n+ programArgs = proArgs.toArray(new String[proArgs.size()]);\n+ fullDMLScriptName = getScript();\n+\n+ //run the test\n+ Lineage.resetInternalState();\n+ runTest(true, EXCEPTION_NOT_EXPECTED, null, -1);\n+\n+ //get lineage and generate program\n+ String Rtrace = readDMLLineageFromHDFS(\"R\");\n+ LineageItem R = LineageParser.parseLineageTrace(Rtrace);\n+ Data ret = LineageItemUtils.computeByLineage(R);\n+\n+ HashMap<CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(\"R\");\n+ MatrixBlock tmp = ((MatrixObject)ret).acquireReadAndRelease();\n+ TestUtils.compareMatrices(dmlfile, tmp, 1e-6);\n+ }\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageTraceTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageTraceTest.java", "diff": "@@ -21,6 +21,7 @@ import java.util.List;\nimport org.junit.Test;\nimport org.tugraz.sysds.hops.OptimizerUtils;\n+import org.tugraz.sysds.hops.recompile.Recompiler;\nimport org.tugraz.sysds.runtime.lineage.Lineage;\nimport org.tugraz.sysds.runtime.lineage.LineageItem;\nimport org.tugraz.sysds.runtime.lineage.LineageParser;\n@@ -112,9 +113,11 @@ public class LineageTraceTest extends AutomatedTestBase {\nTestUtils.compareScalars(X_lineage, Explain.explain(X_li));\nTestUtils.compareScalars(Y_lineage, Explain.explain(Y_li));\n- } finally {\n+ }\n+ finally {\nOptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = old_simplification;\nOptimizerUtils.ALLOW_SUM_PRODUCT_REWRITES = old_sum_product;\n+ Recompiler.reinitRecompiler();\n}\n}\n}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/lineage/LineageTraceFun1.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+foo = function(Matrix[Double] X, Matrix[Double] Y)\n+ return (Matrix[Double] R, Matrix[Double] S)\n+{\n+ if( sum(X) > 0 ) {\n+ R = X + X/2 + X/3 + Y;\n+ S = Y + Y/2 + Y/3 + X;\n+ }\n+ else {\n+ R = X - 1;\n+ S = Y - 1;\n+ }\n+}\n+\n+X = rand(rows=$3, cols=$4, min=-1, max=10, seed=7);\n+Y = X^2;\n+[R, S] = foo(Y, X);\n+R = R - S;\n+\n+print(lineage(R));\n+write(R, $2);\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/lineage/LineageTraceFun2.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+foo = function(Matrix[Double] X, Matrix[Double] Y)\n+ return (Matrix[Double] R, Matrix[Double] S)\n+{\n+ if( sum(X) > 0 ) {\n+ [R, S] = foo( -abs(X), -abs(Y) );\n+ }\n+ else {\n+ R = X - 1;\n+ S = Y - 1;\n+ }\n+}\n+\n+X = rand(rows=$3, cols=$4, min=-1, max=10, seed=7);\n+Y = X^2;\n+[R, S] = foo(Y, X);\n+R = R - S;\n+\n+print(lineage(R));\n+write(R, $2);\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/util/ProgramConverter.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/util/ProgramConverter.java", "diff": "@@ -176,6 +176,7 @@ public class ProgramConverter\n{\nExecutionContext cpec = ExecutionContextFactory.createContext(false, ec.getProgram());\ncpec.setVariables((LocalVariableMap) ec.getVariables().clone());\n+ if( ec.getLineage() != null )\ncpec.setLineage(new Lineage(ec.getLineage()));\n//handle result variables with in-place update flag\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/jmlc/JMLCParfor2ForCompileTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/jmlc/JMLCParfor2ForCompileTest.java", "diff": "@@ -68,6 +68,7 @@ public class JMLCParfor2ForCompileTest extends AutomatedTestBase\nconn.close();\n}\ncatch(Exception ex) {\n+ ex.printStackTrace();\nAssert.fail(\"JMLC parfor test failed: \"+ex.getMessage());\n}\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/parfor/ParForParallelRemoteResultMergeTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/parfor/ParForParallelRemoteResultMergeTest.java", "diff": "@@ -45,8 +45,7 @@ public class ParForParallelRemoteResultMergeTest extends AutomatedTestBase\n@Override\n- public void setUp()\n- {\n+ public void setUp() {\naddTestConfiguration(TEST_NAME1,\nnew TestConfiguration(TEST_CLASS_DIR, TEST_NAME1, new String[] { \"R\" }) );\naddTestConfiguration(TEST_NAME2,\n@@ -54,36 +53,25 @@ public class ParForParallelRemoteResultMergeTest extends AutomatedTestBase\n}\n@Test\n- public void testMultipleResultMergeFewDense()\n- {\n+ public void testMultipleResultMergeFewDense() {\nrunParallelRemoteResultMerge(TEST_NAME1, false);\n}\n@Test\n- public void testMultipleResultMergeFewSparse()\n- {\n+ public void testMultipleResultMergeFewSparse() {\nrunParallelRemoteResultMerge(TEST_NAME1, true);\n}\n@Test\n- public void testMultipleResultMergeManyDense()\n- {\n+ public void testMultipleResultMergeManyDense() {\nrunParallelRemoteResultMerge(TEST_NAME2, false);\n}\n@Test\n- public void testMultipleResultMergeManySparse()\n- {\n+ public void testMultipleResultMergeManySparse() {\nrunParallelRemoteResultMerge(TEST_NAME2, true);\n}\n-\n- /**\n- *\n- * @param outer execution mode of outer parfor loop\n- * @param inner execution mode of inner parfor loop\n- * @param instType execution mode of instructions\n- */\nprivate void runParallelRemoteResultMerge( String test_name, boolean sparse )\n{\n//inst exec type, influenced via rows\n@@ -95,7 +83,6 @@ public class ParForParallelRemoteResultMergeTest extends AutomatedTestBase\nconfig.addVariable(\"cols\", cols);\nloadTestConfiguration(config);\n- /* This is for running the junit test the new way, i.e., construct the arguments directly */\nString HOME = SCRIPT_DIR + TEST_DIR;\nfullDMLScriptName = HOME + TEST_NAME + \".dml\";\nprogramArgs = new String[]{\"-args\", input(\"V\"),\n@@ -119,10 +106,10 @@ public class ParForParallelRemoteResultMergeTest extends AutomatedTestBase\n//compare num MR jobs\nif( TEST_NAME.equals(TEST_NAME1) ) //2 results\n- Assert.assertEquals(\"Unexpected number of executed MR jobs.\",\n+ Assert.assertEquals(\"Unexpected number of executed Spark jobs.\",\n3, Statistics.getNoOfExecutedSPInst());\nelse if ( TEST_NAME.equals(TEST_NAME2) ) //32 results\n- Assert.assertEquals(\"Unexpected number of executed MR jobs.\",\n+ Assert.assertEquals(\"Unexpected number of executed Spark jobs.\",\n33, Statistics.getNoOfExecutedSPInst());\n//compare matrices\nHashMap<CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(\"R\");\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/parfor/parfor_pr_resultmerge1a.dml", "new_path": "src/test/scripts/functions/parfor/parfor_pr_resultmerge1a.dml", "diff": "#-------------------------------------------------------------\n#\n+# Modifications Copyright 2019 Graz University of Technology\n+#\n# Licensed to the Apache Software Foundation (ASF) under one\n# or more contributor license agreements. See the NOTICE file\n# distributed with this work for additional information\n@@ -25,8 +27,7 @@ m = $2;\nn = $3;\nR1 = matrix(0,rows=m,cols=n);\n-parfor( i in 1:(n-7), par=8, mode=REMOTE_MR, resultmerge=REMOTE_MR, taskpartitioner=FACTORING, opt=NONE )\n-{\n+parfor( i in 1:(n-7), par=8, mode=REMOTE_SPARK, resultmerge=REMOTE_SPARK, taskpartitioner=FACTORING, opt=NONE ) {\nX = V[,i];\nR1[,i] = X;\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/parfor/parfor_pr_resultmerge1b.dml", "new_path": "src/test/scripts/functions/parfor/parfor_pr_resultmerge1b.dml", "diff": "#-------------------------------------------------------------\n#\n+# Modifications Copyright 2019 Graz University of Technology\n+#\n# Licensed to the Apache Software Foundation (ASF) under one\n# or more contributor license agreements. See the NOTICE file\n# distributed with this work for additional information\n@@ -25,7 +27,7 @@ m = $2;\nn = $3;\nR1 = matrix(1,rows=m,cols=n);\n-parfor( i in 1:(n-7), par=8, mode=REMOTE_MR, resultmerge=REMOTE_MR, taskpartitioner=FACTORING, opt=NONE )\n+parfor( i in 1:(n-7), par=8, mode=REMOTE_SPARK, resultmerge=REMOTE_SPARK, taskpartitioner=FACTORING, opt=NONE )\n{\nX = V[,i];\nR1[,i] = X;\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/parfor/parfor_pr_resultmerge2.dml", "new_path": "src/test/scripts/functions/parfor/parfor_pr_resultmerge2.dml", "diff": "#-------------------------------------------------------------\n#\n+# Modifications Copyright 2019 Graz University of Technology\n+#\n# Licensed to the Apache Software Foundation (ASF) under one\n# or more contributor license agreements. See the NOTICE file\n# distributed with this work for additional information\n@@ -26,7 +28,7 @@ n = $3;\nR1 = matrix(0,rows=m,cols=n);\nR2 = matrix(0,rows=m,cols=n);\n-parfor( i in 1:n, par=8, mode=REMOTE_MR, resultmerge=REMOTE_MR, taskpartitioner=FACTORING, opt=NONE )\n+parfor( i in 1:n, par=8, mode=REMOTE_SPARK, resultmerge=REMOTE_SPARK, taskpartitioner=FACTORING, opt=NONE )\n{\nX = V[,i];\nR1[,i] = X;\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/parfor/parfor_pr_resultmerge32.dml", "new_path": "src/test/scripts/functions/parfor/parfor_pr_resultmerge32.dml", "diff": "#-------------------------------------------------------------\n#\n+# Modifications Copyright 2019 Graz University of Technology\n+#\n# Licensed to the Apache Software Foundation (ASF) under one\n# or more contributor license agreements. See the NOTICE file\n# distributed with this work for additional information\n@@ -56,7 +58,7 @@ R29 = matrix(0,rows=m,cols=n);\nR30 = matrix(0,rows=m,cols=n);\nR31 = matrix(0,rows=m,cols=n);\nR32 = matrix(0,rows=m,cols=n);\n-parfor( i in 1:n, par=8, mode=REMOTE_MR, resultmerge=REMOTE_MR, taskpartitioner=FACTORING, opt=NONE )\n+parfor( i in 1:n, par=8, mode=REMOTE_SPARK, resultmerge=REMOTE_SPARK, taskpartitioner=FACTORING, opt=NONE )\n{\nX = V[,i];\nR1[,i] = X;\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/utils/Statistics.java", "new_path": "src/main/java/org/tugraz/sysds/utils/Statistics.java", "diff": "@@ -703,18 +703,17 @@ public class Statistics\nlong count = tmp[len - 1 - i].getValue().count.longValue();\nint numLines = wrappedInstruction.length;\n- String miscFormatString = \"%s\";\nfor(int wrapIter = 0; wrapIter < numLines; wrapIter++) {\nString instStr = (wrapIter < wrappedInstruction.length) ? wrappedInstruction[wrapIter] : \"\";\nif(wrapIter == 0) {\n// Display instruction count\nsb.append(String.format(\n- \" %\" + maxNumLen + \"d %-\" + maxInstLen + \"s %\" + maxTimeSLen + \"s %\" + maxCountLen + \"d\" + miscFormatString,\n+ \" %\" + maxNumLen + \"d %-\" + maxInstLen + \"s %\" + maxTimeSLen + \"s %\" + maxCountLen + \"d\",\n(i + 1), instStr, timeSString, count));\n}\nelse {\nsb.append(String.format(\n- \" %\" + maxNumLen + \"s %-\" + maxInstLen + \"s %\" + maxTimeSLen + \"s %\" + maxCountLen + \"s\" + miscFormatString,\n+ \" %\" + maxNumLen + \"s %-\" + maxInstLen + \"s %\" + maxTimeSLen + \"s %\" + maxCountLen + \"s\",\n\"\", instStr, \"\", \"\"));\n}\nsb.append(\"\\n\");\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/codegen/OuterProdTmplTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/codegen/OuterProdTmplTest.java", "diff": "@@ -44,6 +44,7 @@ public class OuterProdTmplTest extends AutomatedTestBase\nprivate static final String TEST_NAME7 = \"wdivmmTransposeOut\";\nprivate static final String TEST_NAME8 = \"wSparseUnsafeOuterProduct\";\nprivate static final String TEST_NAME9 = \"wdivmmNeq\";\n+ private static final String TEST_NAME10 = \"rmseDist\";\nprivate static final String TEST_DIR = \"functions/codegen/\";\nprivate static final String TEST_CLASS_DIR = TEST_DIR + OuterProdTmplTest.class.getSimpleName() + \"/\";\n@@ -64,6 +65,7 @@ public class OuterProdTmplTest extends AutomatedTestBase\naddTestConfiguration( TEST_NAME7, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME7, new String[] { \"7\" }) );\naddTestConfiguration( TEST_NAME8, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME8, new String[] { \"8\" }) );\naddTestConfiguration( TEST_NAME9, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME9, new String[] { \"9\" }) );\n+ addTestConfiguration( TEST_NAME10, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME10, new String[] { \"10\" }) );\n}\n@Test\n@@ -186,6 +188,16 @@ public class OuterProdTmplTest extends AutomatedTestBase\ntestCodegenIntegrationWithInput( TEST_NAME9, true, ExecType.SPARK );\n}\n+ @Test\n+ public void testCodegenOuterProd10NoRewrite() {\n+ testCodegenIntegration( TEST_NAME10, false, ExecType.CP );\n+ }\n+\n+ @Test\n+ public void testCodegenOuterProd10NoRewriteSP() {\n+ testCodegenIntegrationWithInput( TEST_NAME10, false, ExecType.SPARK );\n+ }\n+\nprivate void testCodegenIntegration( String testname, boolean rewrites, ExecType instType )\n{\nboolean oldFlag = OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION;\n@@ -238,7 +250,6 @@ public class OuterProdTmplTest extends AutomatedTestBase\nOptimizerUtils.ALLOW_AUTO_VECTORIZATION = true;\nOptimizerUtils.ALLOW_OPERATOR_FUSION = true;\n}\n-\n}\nprivate void testCodegenIntegrationWithInput( String testname, boolean rewrites, ExecType instType )\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/codegen/rmseDist.R", "diff": "+#-------------------------------------------------------------\n+#\n+# Licensed to the Apache Software Foundation (ASF) under one\n+# or more contributor license agreements. See the NOTICE file\n+# distributed with this work for additional information\n+# regarding copyright ownership. The ASF licenses this file\n+# to you under the Apache License, Version 2.0 (the\n+# \"License\"); you may not use this file except in compliance\n+# with the License. You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing,\n+# software distributed under the License is distributed on an\n+# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+# KIND, either express or implied. See the License for the\n+# specific language governing permissions and limitations\n+# under the License.\n+#\n+#-------------------------------------------------------------\n+\n+args<-commandArgs(TRUE)\n+options(digits=22)\n+library(\"Matrix\")\n+\n+U = matrix( 4, 4000, 10)\n+V = matrix( 5, 10, 2000)\n+X = U %*% V\n+X[1:3500,] = matrix(0,3500,2000);\n+\n+T1 = rowSums(U^2)%*%matrix(1,1,ncol(V));\n+T2 = matrix(1,nrow(U),1)%*%t(as.matrix(colSums(V^2)))\n+D = sqrt(-2 * U %*% V + T1 + T2);\n+\n+s = sum(rowSums((X != 0) * (X - D))^2)\n+S = as.matrix(s);\n+\n+writeMM(as(S, \"CsparseMatrix\"), paste(args[2], \"S\", sep=\"\"));\n+\n\\ No newline at end of file\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/codegen/rmseDist.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Licensed to the Apache Software Foundation (ASF) under one\n+# or more contributor license agreements. See the NOTICE file\n+# distributed with this work for additional information\n+# regarding copyright ownership. The ASF licenses this file\n+# to you under the Apache License, Version 2.0 (the\n+# \"License\"); you may not use this file except in compliance\n+# with the License. You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing,\n+# software distributed under the License is distributed on an\n+# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+# KIND, either express or implied. See the License for the\n+# specific language governing permissions and limitations\n+# under the License.\n+#\n+#-------------------------------------------------------------\n+\n+U = matrix( 4, rows=4000, cols=10)\n+V = matrix( 5, rows=10, cols=2000)\n+X = U %*% V\n+X[1:3500,] = matrix(0,3500,2000);\n+\n+while(FALSE){}\n+\n+#TODO + rowSums(U^2) + colSums(V^2)\n+# requires broadcasting support in outer\n+\n+T1 = rowSums(U^2)%*%matrix(1,1,ncol(V));\n+T2 = matrix(1,nrow(U),1)%*%colSums(V^2)\n+\n+D = sqrt(-2 * U %*% V + T1 + T2);\n+S = as.matrix(sum(rowSums((X != 0) * (X - D))^2));\n+\n+write(S,$1)\n+\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/caching/TensorObject.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/caching/TensorObject.java", "diff": "@@ -86,10 +86,6 @@ public class TensorObject extends CacheableData<HomogTensor>\ntc.setNonZeros(_data.getNonZeros());\n}\n- public int[] getDims() {\n- return _data.getDims();\n- }\n-\npublic long getNumRows() {\nDataCharacteristics dc = getDataCharacteristics();\nreturn dc.getRows();\n@@ -126,7 +122,7 @@ public class TensorObject extends CacheableData<HomogTensor>\nTensorCharacteristics tc = (TensorCharacteristics) _metaData.getDataCharacteristics();\n// TODO correct blocksize;\n// TODO read from RDD\n- return SparkExecutionContext.toTensorBlock((JavaPairRDD<TensorIndexes, TensorBlock>)rdd.getRDD(), tc);\n+ return SparkExecutionContext.toTensorBlock((JavaPairRDD<TensorIndexes, HomogTensor>)rdd.getRDD(), tc);\n}\n@Override\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/context/SparkExecutionContext.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/context/SparkExecutionContext.java", "diff": "@@ -51,8 +51,8 @@ import org.tugraz.sysds.runtime.controlprogram.caching.FrameObject;\nimport org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject;\nimport org.tugraz.sysds.runtime.controlprogram.caching.TensorObject;\nimport org.tugraz.sysds.runtime.controlprogram.parfor.stat.InfrastructureAnalyzer;\n+import org.tugraz.sysds.runtime.data.HomogTensor;\nimport org.tugraz.sysds.runtime.data.SparseBlock;\n-import org.tugraz.sysds.runtime.data.TensorBlock;\nimport org.tugraz.sysds.runtime.data.TensorIndexes;\nimport org.tugraz.sysds.runtime.instructions.cp.Data;\nimport org.tugraz.sysds.runtime.instructions.spark.data.BroadcastObject;\n@@ -317,19 +317,19 @@ public class SparkExecutionContext extends ExecutionContext\n* variables.\n*\n* @param varname variable name\n- * @return JavaPairRDD of TensorIndexes-TensorBlocks\n+ * @return JavaPairRDD of TensorIndexes-HomogTensors\n*/\n@SuppressWarnings(\"unchecked\")\n- public JavaPairRDD<TensorIndexes, TensorBlock> getBinaryTensorBlockRDDHandleForVariable(String varname ) {\n+ public JavaPairRDD<TensorIndexes, HomogTensor> getBinaryTensorBlockRDDHandleForVariable(String varname ) {\nTensorObject to = getTensorObject(varname);\n- return (JavaPairRDD<TensorIndexes, TensorBlock>)\n+ return (JavaPairRDD<TensorIndexes, HomogTensor>)\ngetRDDHandleForTensorObject(to, InputInfo.BinaryTensorBlockInputInfo, -1, true);\n}\n@SuppressWarnings(\"unchecked\")\n- public JavaPairRDD<TensorIndexes, TensorBlock> getBinaryTensorBlockRDDHandleForVariable(String varname, int numParts, boolean inclEmpty ) {\n+ public JavaPairRDD<TensorIndexes, HomogTensor> getBinaryTensorBlockRDDHandleForVariable(String varname, int numParts, boolean inclEmpty ) {\nTensorObject to = getTensorObject(varname);\n- return (JavaPairRDD<TensorIndexes, TensorBlock>)\n+ return (JavaPairRDD<TensorIndexes, HomogTensor>)\ngetRDDHandleForTensorObject(to, InputInfo.BinaryTensorBlockInputInfo, numParts, inclEmpty);\n}\n@@ -473,10 +473,10 @@ public class SparkExecutionContext extends ExecutionContext\n// TODO implement hadoop read write for tensor\nthrow new DMLRuntimeException(\"Tensor can not yet be written or read to hadoopFile\");\n/*rdd = sc.hadoopFile(to.getFileName(), inputInfo.inputFormatClass, inputInfo.inputKeyClass, inputInfo.inputValueClass);\n- rdd = SparkUtils.copyBinaryBlockTensor((JavaPairRDD<TensorIndexes, TensorBlock>) rdd); //cp is workaround for read bug\n+ rdd = SparkUtils.copyBinaryBlockTensor((JavaPairRDD<TensorIndexes, HomogTensor>) rdd); //cp is workaround for read bug\nfromFile = true;*/\n} else { //default case\n- TensorBlock mb = to.acquireRead(); //pin matrix in memory\n+ HomogTensor mb = to.acquireRead(); //pin matrix in memory\nint[] blen = new int[dc.getNumDims()];\nfor (int i = 0; i < blen.length; i++) {\nblen[i] = (int) dc.getBlockSize(i);\n@@ -503,7 +503,7 @@ public class SparkExecutionContext extends ExecutionContext\n//rdd = sc.hadoopFile(to.getFileName(), inputInfo.inputFormatClass, inputInfo.inputKeyClass, inputInfo.inputValueClass);\n//note: this copy is still required in Spark 1.4 because spark hands out whatever the inputformat\n//recordreader returns; the javadoc explicitly recommend to copy all key/value pairs\n- //rdd = SparkUtils.copyBinaryBlockTensor((JavaPairRDD<TensorIndexes, TensorBlock>) rdd); //cp is workaround for read bug\n+ //rdd = SparkUtils.copyBinaryBlockTensor((JavaPairRDD<TensorIndexes, HomogTensor>) rdd); //cp is workaround for read bug\n// TODO: TensorMarket?\n} else {\n// TODO support other Input formats\n@@ -834,14 +834,14 @@ public class SparkExecutionContext extends ExecutionContext\nreturn result;\n}\n- public static JavaPairRDD<TensorIndexes, TensorBlock> toTensorJavaPairRDD(JavaSparkContext sc, TensorBlock src, int[] blen) {\n+ public static JavaPairRDD<TensorIndexes, HomogTensor> toTensorJavaPairRDD(JavaSparkContext sc, HomogTensor src, int[] blen) {\nreturn toTensorJavaPairRDD(sc, src, blen, -1, true);\n}\n- public static JavaPairRDD<TensorIndexes, TensorBlock> toTensorJavaPairRDD(JavaSparkContext sc, TensorBlock src,\n+ public static JavaPairRDD<TensorIndexes, HomogTensor> toTensorJavaPairRDD(JavaSparkContext sc, HomogTensor src,\nint[] blen, int numParts, boolean inclEmpty) {\nlong t0 = DMLScript.STATISTICS ? System.nanoTime() : 0;\n- List<Tuple2<TensorIndexes, TensorBlock>> list;\n+ List<Tuple2<TensorIndexes, HomogTensor>> list;\nboolean singleBlock = true;\nfor (int i = 0; i < blen.length; i++) {\n@@ -866,7 +866,7 @@ public class SparkExecutionContext extends ExecutionContext\n.collect(Collectors.toList());\n}\n- JavaPairRDD<TensorIndexes, TensorBlock> result = (numParts > 1) ?\n+ JavaPairRDD<TensorIndexes, HomogTensor> result = (numParts > 1) ?\nsc.parallelizePairs(list, numParts) : sc.parallelizePairs(list);\nif (DMLScript.STATISTICS) {\n@@ -899,7 +899,7 @@ public class SparkExecutionContext extends ExecutionContext\n}\n}\n- private static Tuple2<TensorIndexes,TensorBlock> createIndexedTensorBlock(TensorBlock mb, TensorCharacteristics tc, long ix) {\n+ private static Tuple2<TensorIndexes,HomogTensor> createIndexedTensorBlock(HomogTensor mb, TensorCharacteristics tc, long ix) {\ntry {\n//compute block indexes\nlong[] blockIx = new long[tc.getNumDims()];\n@@ -913,7 +913,7 @@ public class SparkExecutionContext extends ExecutionContext\nix /= tc.getNumBlocks(i);\n}\n// TODO: sparse\n- TensorBlock outBlock = new TensorBlock(mb.getValueType(), outDims, false);\n+ HomogTensor outBlock = new HomogTensor(mb.getValueType(), outDims, false);\noutBlock = mb.slice(offset, outBlock);\n//create key-value pair\nfor (int i = 0; i < blockIx.length; i++) {\n@@ -1127,7 +1127,7 @@ public class SparkExecutionContext extends ExecutionContext\nreturn out;\n}\n- public static TensorBlock toTensorBlock(JavaPairRDD<TensorIndexes, TensorBlock> rdd, DataCharacteristics dc) {\n+ public static HomogTensor toTensorBlock(JavaPairRDD<TensorIndexes, HomogTensor> rdd, DataCharacteristics dc) {\nlong t0 = DMLScript.STATISTICS ? System.nanoTime() : 0;\n// TODO special case single block\n@@ -1136,17 +1136,17 @@ public class SparkExecutionContext extends ExecutionContext\nidims[i] = (int)dc.getDim(i);\n}\n// TODO asynchronous allocation\n- List<Tuple2<TensorIndexes, TensorBlock>> list = rdd.collect();\n+ List<Tuple2<TensorIndexes, HomogTensor>> list = rdd.collect();\nValueType vt = list.get(0)._2.getValueType();\n- TensorBlock out = new TensorBlock(vt, idims);\n+ HomogTensor out = new HomogTensor(vt, idims);\nout.allocateDenseBlock();\n//copy blocks one-at-a-time into output matrix block\n- for( Tuple2<TensorIndexes, TensorBlock> keyval : list )\n+ for( Tuple2<TensorIndexes, HomogTensor> keyval : list )\n{\n//unpack index-block pair\nTensorIndexes ix = keyval._1();\n- TensorBlock block = keyval._2();\n+ HomogTensor block = keyval._2();\n//compute row/column block offsets\nint[] lower = new int[ix.getNumDims()];\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/HomogTensor.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/HomogTensor.java", "diff": "@@ -37,7 +37,7 @@ import java.io.IOException;\nimport java.io.ObjectInput;\nimport java.io.ObjectOutput;\n-public class HomogTensor extends Tensor implements Externalizable\n+public class HomogTensor extends TensorBlock implements Externalizable\n{\nprivate static final long serialVersionUID = -1887367304030494999L;\n@@ -227,12 +227,6 @@ public class HomogTensor extends Tensor implements Externalizable\nreturn _vt;\n}\n- public int[] getDims() {\n- int[] dims = new int[_dims.length];\n- Array.copy(_dims, 0, dims, 0, _dims.length);\n- return dims;\n- }\n-\npublic long getNonZeros() {\nreturn _nnz;\n}\n@@ -290,7 +284,7 @@ public class HomogTensor extends Tensor implements Externalizable\nallocateDenseBlock(false);\nDenseBlock a = getDenseBlock();\nint odims = (int) UtilFunctions.prod(_dims, 1);\n- int ix = new int[getNumDims()];\n+ int[] ix = new int[getNumDims()];\nfor( int i=0; i<getNumRows(); i++ ) {\nix[0] = i;\nfor (int j = 0; j < odims; j++) {\n@@ -298,7 +292,7 @@ public class HomogTensor extends Tensor implements Externalizable\nswitch (_vt) {\ncase FP32: a.set(i, j, in.readFloat()); break;\ncase FP64: a.set(i, j, in.readDouble()); break;\n- case INT32: a.set(ix, (long)in.readInt()); break;\n+ case INT32: a.set(ix, in.readInt()); break;\ncase INT64: a.set(ix, in.readLong()); break;\ncase BOOLEAN: a.set(i, j, in.readByte()); break;\ncase STRING: a.set(ix, in.readUTF()); break;\n@@ -712,7 +706,7 @@ public class HomogTensor extends Tensor implements Externalizable\n}\n@Override\n- public void readExternal(ObjectInput in) throws IOException, ClassNotFoundException {\n+ public void readExternal(ObjectInput in) throws IOException {\nreadFields(in);\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/LibTensorAgg.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/LibTensorAgg.java", "diff": "@@ -212,7 +212,7 @@ public class LibTensorAgg {\n//out.binaryOperationsInPlace(laop.increOp, partout);\n}\n- private static void sum(HomogTensor in, HomogTensor out, Plus kplus, int rl, int ru) {\n+ private static void sum(HomogTensor in, HomogTensor out, Plus plus, int rl, int ru) {\n// TODO: SparseBlock\nif (in.isSparse()) {\nthrow new DMLRuntimeException(\"Sparse aggregation not implemented for Tensor\");\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/VariableCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/VariableCPInstruction.java", "diff": "@@ -41,7 +41,7 @@ import org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject.UpdateType;\nimport org.tugraz.sysds.runtime.controlprogram.caching.TensorObject;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.controlprogram.parfor.util.IDSequence;\n-import org.tugraz.sysds.runtime.data.TensorBlock;\n+import org.tugraz.sysds.runtime.data.HomogTensor;\nimport org.tugraz.sysds.runtime.instructions.Instruction;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\nimport org.tugraz.sysds.runtime.io.FileFormatProperties;\n@@ -638,7 +638,7 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\nec.setScalarOutput(output.getName(), new DoubleObject(value));\n}\nelse if( getInput1().getDataType().isTensor() ) {\n- TensorBlock tBlock = ec.getTensorInput(getInput1().getName());\n+ HomogTensor tBlock = ec.getTensorInput(getInput1().getName());\nif( tBlock.getNumDims() != 2 || tBlock.getNumRows() != 1 || tBlock.getNumColumns() != 1 )\nthrow new DMLRuntimeException(\"Dimension mismatch - unable to cast tensor '\"+getInput1().getName()+\"' to scalar.\");\nswitch (tBlock.getValueType()) {\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/AggregateUnarySPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/AggregateUnarySPInstruction.java", "diff": "@@ -31,7 +31,7 @@ import org.tugraz.sysds.lops.PartialAggregate.CorrectionLocationType;\nimport org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.controlprogram.context.SparkExecutionContext;\n-import org.tugraz.sysds.runtime.data.TensorBlock;\n+import org.tugraz.sysds.runtime.data.HomogTensor;\nimport org.tugraz.sysds.runtime.data.TensorIndexes;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\nimport org.tugraz.sysds.runtime.instructions.cp.CPOperand;\n@@ -147,8 +147,8 @@ public class AggregateUnarySPInstruction extends UnarySPInstruction {\nSparkExecutionContext sec = (SparkExecutionContext)ec;\n//get input\n- JavaPairRDD<TensorIndexes, TensorBlock> in = sec.getBinaryTensorBlockRDDHandleForVariable( input1.getName() );\n- JavaPairRDD<TensorIndexes, TensorBlock> out = in;\n+ JavaPairRDD<TensorIndexes, HomogTensor> in = sec.getBinaryTensorBlockRDDHandleForVariable( input1.getName() );\n+ JavaPairRDD<TensorIndexes, HomogTensor> out = in;\n// TODO: filter input blocks for trace\n//execute unary aggregate operation\n@@ -160,13 +160,13 @@ public class AggregateUnarySPInstruction extends UnarySPInstruction {\nif( _aggtype == SparkAggType.SINGLE_BLOCK )\n{\n// TODO filter non empty blocks if sparse safe\n- JavaRDD<TensorBlock> out2 = out.map(new RDDUTensorAggFunction2(auop));\n- TensorBlock out3 = RDDAggregateUtils.aggStableTensor(out2, aggop);\n+ JavaRDD<HomogTensor> out2 = out.map(new RDDUTensorAggFunction2(auop));\n+ HomogTensor out3 = RDDAggregateUtils.aggStableTensor(out2, aggop);\n//put output block into symbol table (no lineage because single block)\n//this also includes implicit maintenance of data characteristics\n// TODO generalize to drop depending on location of correction\n- TensorBlock out4 = new TensorBlock(out3.getValueType(), new int[]{1, 1}, false);\n+ HomogTensor out4 = new HomogTensor(out3.getValueType(), new int[]{1, 1}, false);\nout4.set(0, 0, out3.get(0, 0));\nsec.setTensorOutput(output.getName(), out4);\n}\n@@ -263,7 +263,7 @@ public class AggregateUnarySPInstruction extends UnarySPInstruction {\n/**\n* Similar to RDDUAggFunction but single output block.\n*/\n- public static class RDDUTensorAggFunction2 implements Function<Tuple2<TensorIndexes, TensorBlock>, TensorBlock>\n+ public static class RDDUTensorAggFunction2 implements Function<Tuple2<TensorIndexes, HomogTensor>, HomogTensor>\n{\nprivate static final long serialVersionUID = -6258769067791011763L;\n@@ -274,11 +274,11 @@ public class AggregateUnarySPInstruction extends UnarySPInstruction {\n}\n@Override\n- public TensorBlock call( Tuple2<TensorIndexes, TensorBlock> arg0 )\n+ public HomogTensor call( Tuple2<TensorIndexes, HomogTensor> arg0 )\nthrows Exception\n{\n//unary aggregate operation (always keep the correction)\n- return arg0._2.aggregateUnaryOperations(_op, new TensorBlock());\n+ return arg0._2.aggregateUnaryOperations(_op, new HomogTensor());\n}\n}\n@@ -317,7 +317,7 @@ public class AggregateUnarySPInstruction extends UnarySPInstruction {\n}\n}\n- private static class RDDUTensorAggValueFunction implements Function<TensorBlock, TensorBlock>\n+ private static class RDDUTensorAggValueFunction implements Function<HomogTensor, HomogTensor>\n{\nprivate static final long serialVersionUID = -968274963539513423L;\n@@ -329,17 +329,17 @@ public class AggregateUnarySPInstruction extends UnarySPInstruction {\n}\n@Override\n- public TensorBlock call( TensorBlock arg0 )\n+ public HomogTensor call( HomogTensor arg0 )\nthrows Exception\n{\n- TensorBlock blkOut = new TensorBlock();\n+ HomogTensor blkOut = new HomogTensor();\n//unary aggregate operation\narg0.aggregateUnaryOperations(_op, blkOut);\n//always drop correction since no aggregation\n// TODO generalize to drop depending on location of correction\n- TensorBlock out = new TensorBlock(blkOut.getValueType(), new int[]{1, 1}, false);\n+ HomogTensor out = new HomogTensor(blkOut.getValueType(), new int[]{1, 1}, false);\nout.set(0, 0, blkOut.get(0, 0));\n//output new tuple\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/MatrixReshapeSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/MatrixReshapeSPInstruction.java", "diff": "@@ -29,7 +29,7 @@ import org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.controlprogram.context.SparkExecutionContext;\nimport org.tugraz.sysds.runtime.data.IndexedTensorBlock;\n-import org.tugraz.sysds.runtime.data.TensorBlock;\n+import org.tugraz.sysds.runtime.data.HomogTensor;\nimport org.tugraz.sysds.runtime.data.TensorIndexes;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\nimport org.tugraz.sysds.runtime.instructions.cp.CPOperand;\n@@ -119,9 +119,9 @@ public class MatrixReshapeSPInstruction extends UnarySPInstruction\nsec.addLineageRDD(output.getName(), input1.getName());\n} else {\n// TODO Tensor reshape\n- JavaPairRDD<TensorIndexes, TensorBlock> in1 = sec.getBinaryTensorBlockRDDHandleForVariable(input1.getName(),\n+ JavaPairRDD<TensorIndexes, HomogTensor> in1 = sec.getBinaryTensorBlockRDDHandleForVariable(input1.getName(),\n-1, _outputEmptyBlocks);\n- JavaPairRDD<TensorIndexes, TensorBlock> out = in1.flatMapToPair(\n+ JavaPairRDD<TensorIndexes, HomogTensor> out = in1.flatMapToPair(\nnew RDDTensorReshapeFunction(mcIn, mcOut, byRow, _outputEmptyBlocks));\n// TODO merge by key\n//out = RDDAggregateUtils.mergeByKey(out);\n@@ -163,8 +163,8 @@ public class MatrixReshapeSPInstruction extends UnarySPInstruction\n}\n@SuppressWarnings(\"unused\")\n- private static class RDDTensorReshapeFunction implements PairFlatMapFunction<Tuple2<TensorIndexes, TensorBlock>,\n- TensorIndexes, TensorBlock> {\n+ private static class RDDTensorReshapeFunction implements PairFlatMapFunction<Tuple2<TensorIndexes, HomogTensor>,\n+ TensorIndexes, HomogTensor> {\nprivate static final long serialVersionUID = 8030648988828223639L;\nprivate final DataCharacteristics _mcIn;\n@@ -180,7 +180,7 @@ public class MatrixReshapeSPInstruction extends UnarySPInstruction\n}\n@Override\n- public Iterator<Tuple2<TensorIndexes, TensorBlock>> call(Tuple2<TensorIndexes, TensorBlock> arg0)\n+ public Iterator<Tuple2<TensorIndexes, HomogTensor>> call(Tuple2<TensorIndexes, HomogTensor> arg0)\nthrows Exception {\n//input conversion (for libmatrixreorg compatibility)\nIndexedTensorBlock in = SparkUtils.toIndexedTensorBlock(arg0);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/RandSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/RandSPInstruction.java", "diff": "@@ -45,7 +45,7 @@ import org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.controlprogram.context.SparkExecutionContext;\nimport org.tugraz.sysds.runtime.controlprogram.parfor.stat.InfrastructureAnalyzer;\n-import org.tugraz.sysds.runtime.data.TensorBlock;\n+import org.tugraz.sysds.runtime.data.HomogTensor;\nimport org.tugraz.sysds.runtime.data.TensorIndexes;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\nimport org.tugraz.sysds.runtime.instructions.cp.CPOperand;\n@@ -475,7 +475,7 @@ public class RandSPInstruction extends UnarySPInstruction {\n//step 4: execute rand instruction over seed input\n// TODO getDimLengthPerBlock accurate for each dimension\n- JavaPairRDD<TensorIndexes, TensorBlock> out = seedsRDD\n+ JavaPairRDD<TensorIndexes, HomogTensor> out = seedsRDD\n.mapToPair(new GenerateRandomTensorBlock(output.getValueType(), tDims, blockSizes,\nsparsity, minValueStr, maxValueStr, pdf, pdfParams));\n@@ -854,7 +854,7 @@ public class RandSPInstruction extends UnarySPInstruction {\n}\n}\n- private static class GenerateRandomTensorBlock implements PairFunction<Tuple2<TensorIndexes, Long>, TensorIndexes, TensorBlock>\n+ private static class GenerateRandomTensorBlock implements PairFunction<Tuple2<TensorIndexes, Long>, TensorIndexes, HomogTensor>\n{\nprivate static final long serialVersionUID = -512119897654170462L;\n@@ -881,7 +881,7 @@ public class RandSPInstruction extends UnarySPInstruction {\n}\n@Override\n- public Tuple2<TensorIndexes, TensorBlock> call(Tuple2<TensorIndexes, Long> kv)\n+ public Tuple2<TensorIndexes, HomogTensor> call(Tuple2<TensorIndexes, Long> kv)\nthrows Exception\n{\n//compute local block size:\n@@ -895,7 +895,7 @@ public class RandSPInstruction extends UnarySPInstruction {\nint clen = (int) UtilFunctions.prod(blockDims, 1);\nlong seed = kv._2;\n- TensorBlock tb = new TensorBlock(_vt, blockDims);\n+ HomogTensor tb = new HomogTensor(_vt, blockDims);\n// TODO implement sparse support\ntb.allocateDenseBlock();\nif (!_min.equals(_max)) {\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/functions/CopyTensorBlockFunction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/functions/CopyTensorBlockFunction.java", "diff": "package org.tugraz.sysds.runtime.instructions.spark.functions;\nimport org.apache.spark.api.java.function.Function;\n-import org.tugraz.sysds.runtime.data.TensorBlock;\n+import org.tugraz.sysds.runtime.data.HomogTensor;\n/**\n* General purpose copy function for binary block rdds. This function can be used in\n* mapValues (copy tensor blocks). It supports both deep and shallow copies of values.\n*/\n-public class CopyTensorBlockFunction implements Function<TensorBlock, TensorBlock> {\n+public class CopyTensorBlockFunction implements Function<HomogTensor, HomogTensor> {\nprivate static final long serialVersionUID = 707987326466592670L;\nprivate boolean _deepCopy;\n@@ -40,10 +40,10 @@ public class CopyTensorBlockFunction implements Function<TensorBlock, TensorBloc\n}\n@Override\n- public TensorBlock call(TensorBlock arg0)\n+ public HomogTensor call(HomogTensor arg0)\nthrows Exception {\nif (_deepCopy)\n- return new TensorBlock(arg0);\n+ return new HomogTensor(arg0);\nelse\nreturn arg0;\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/functions/CopyTensorBlockPairFunction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/functions/CopyTensorBlockPairFunction.java", "diff": "package org.tugraz.sysds.runtime.instructions.spark.functions;\nimport org.apache.spark.api.java.function.PairFlatMapFunction;\n-import org.tugraz.sysds.runtime.data.TensorBlock;\n+import org.tugraz.sysds.runtime.data.HomogTensor;\nimport org.tugraz.sysds.runtime.data.TensorIndexes;\nimport org.tugraz.sysds.runtime.instructions.spark.data.LazyIterableIterator;\nimport scala.Tuple2;\n@@ -33,10 +33,10 @@ import java.util.Iterator;\n* mapToPair (copy tensor indexes and blocks). It supports both deep and shallow copies\n* of key/value pairs.\n*/\n-public class CopyTensorBlockPairFunction implements PairFlatMapFunction<Iterator<Tuple2<TensorIndexes, TensorBlock>>, TensorIndexes, TensorBlock> {\n+public class CopyTensorBlockPairFunction implements PairFlatMapFunction<Iterator<Tuple2<TensorIndexes, HomogTensor>>, TensorIndexes, HomogTensor> {\nprivate static final long serialVersionUID = 605514365345997070L;\n- private boolean _deepCopy = true;\n+ private boolean _deepCopy;\npublic CopyTensorBlockPairFunction() {\nthis(true);\n@@ -47,25 +47,24 @@ public class CopyTensorBlockPairFunction implements PairFlatMapFunction<Iterator\n}\n@Override\n- public LazyIterableIterator<Tuple2<TensorIndexes, TensorBlock>> call(Iterator<Tuple2<TensorIndexes, TensorBlock>> arg0)\n+ public LazyIterableIterator<Tuple2<TensorIndexes, HomogTensor>> call(Iterator<Tuple2<TensorIndexes, HomogTensor>> arg0)\nthrows Exception {\nreturn new CopyBlockPairIterator(arg0);\n}\n- private class CopyBlockPairIterator extends LazyIterableIterator<Tuple2<TensorIndexes, TensorBlock>> {\n- public CopyBlockPairIterator(Iterator<Tuple2<TensorIndexes, TensorBlock>> iter) {\n+ private class CopyBlockPairIterator extends LazyIterableIterator<Tuple2<TensorIndexes, HomogTensor>> {\n+ public CopyBlockPairIterator(Iterator<Tuple2<TensorIndexes, HomogTensor>> iter) {\nsuper(iter);\n}\n@Override\n- protected Tuple2<TensorIndexes, TensorBlock> computeNext(Tuple2<TensorIndexes, TensorBlock> arg)\n- throws Exception {\n+ protected Tuple2<TensorIndexes, HomogTensor> computeNext(Tuple2<TensorIndexes, HomogTensor> arg) {\nif (_deepCopy) {\nTensorIndexes ix = new TensorIndexes(arg._1());\n- TensorBlock block;\n+ HomogTensor block;\n// TODO: always create deep copies in more memory-efficient CSR representation\n// if block is already in sparse format\n- block = new TensorBlock(arg._2());\n+ block = new HomogTensor(arg._2());\nreturn new Tuple2<>(ix, block);\n} else {\nreturn arg;\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/utils/RDDAggregateUtils.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/utils/RDDAggregateUtils.java", "diff": "@@ -25,9 +25,10 @@ import org.apache.spark.api.java.function.Function;\nimport org.apache.spark.api.java.function.Function2;\nimport org.tugraz.sysds.lops.PartialAggregate.CorrectionLocationType;\nimport org.tugraz.sysds.runtime.DMLRuntimeException;\n-import org.tugraz.sysds.runtime.data.TensorBlock;\n+import org.tugraz.sysds.runtime.data.HomogTensor;\nimport org.tugraz.sysds.runtime.data.TensorIndexes;\nimport org.tugraz.sysds.runtime.functionobjects.KahanPlus;\n+import org.tugraz.sysds.runtime.functionobjects.Plus;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\nimport org.tugraz.sysds.runtime.instructions.cp.KahanObject;\nimport org.tugraz.sysds.runtime.instructions.spark.AggregateUnarySPInstruction.RDDUAggFunction2;\n@@ -152,29 +153,29 @@ public class RDDAggregateUtils\n/**\n* Single block aggregation over pair rdds with corrections for numerical stability.\n*\n- * @param in tensor as {@code JavaPairRDD<TensorIndexes, TensorBlock>}\n+ * @param in tensor as {@code JavaPairRDD<TensorIndexes, HomogTensor>}\n* @param aop aggregate operator\n* @return tensor block\n*/\n- public static TensorBlock aggStableTensor(JavaPairRDD<TensorIndexes, TensorBlock> in, AggregateOperator aop) {\n+ public static HomogTensor aggStableTensor(JavaPairRDD<TensorIndexes, HomogTensor> in, AggregateOperator aop) {\nreturn aggStableTensor( in.values(), aop);\n}\n/**\n* Single block aggregation over rdds with corrections for numerical stability.\n*\n- * @param in tensor as {@code JavaRDD<TensorBlock>}\n+ * @param in tensor as {@code JavaRDD<HomogTensor>}\n* @param aop aggregate operator\n* @return tensor block\n*/\n- public static TensorBlock aggStableTensor( JavaRDD<TensorBlock> in, AggregateOperator aop )\n+ public static HomogTensor aggStableTensor( JavaRDD<HomogTensor> in, AggregateOperator aop )\n{\n//stable aggregate of all blocks with correction block per function instance\n//reduce-all aggregate via fold instead of reduce to allow\n//for update in-place w/o deep copy of left-hand-side blocks\nreturn in.fold(\n- new TensorBlock(),\n+ new HomogTensor(),\nnew AggregateSingleTensorBlockFunction(aop) );\n}\npublic static JavaPairRDD<MatrixIndexes, MatrixBlock> aggByKeyStable( JavaPairRDD<MatrixIndexes, MatrixBlock> in,\n@@ -648,19 +649,18 @@ public class RDDAggregateUtils\n* drop them at the end because during aggregation we dont know if we produce an\n* intermediate or the final aggregate.\n*/\n- private static class AggregateSingleTensorBlockFunction implements Function2<TensorBlock, TensorBlock, TensorBlock>\n+ private static class AggregateSingleTensorBlockFunction implements Function2<HomogTensor, HomogTensor, HomogTensor>\n{\nprivate static final long serialVersionUID = 5665180309149919945L;\nprivate AggregateOperator _op = null;\n- //private MatrixBlock _corr = null;\npublic AggregateSingleTensorBlockFunction( AggregateOperator op ) {\n_op = op;\n}\n@Override\n- public TensorBlock call(TensorBlock arg0, TensorBlock arg1)\n+ public HomogTensor call(HomogTensor arg0, HomogTensor arg1)\nthrows Exception\n{\n//prepare combiner block\n@@ -671,7 +671,10 @@ public class RDDAggregateUtils\nreturn arg0;\n}\n- // TODO correction\n+ // TODO remove once KahanPlus is completely replaced by plus\n+ if (_op.increOp.fn instanceof KahanPlus) {\n+ _op = new AggregateOperator(0, Plus.getPlusFnObject());\n+ }\n//aggregate second input (in-place)\narg0.incrementalAggregate(_op, arg1);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/utils/SparkUtils.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/utils/SparkUtils.java", "diff": "@@ -34,7 +34,7 @@ import org.tugraz.sysds.lops.Checkpoint;\nimport org.tugraz.sysds.runtime.controlprogram.context.SparkExecutionContext;\nimport org.tugraz.sysds.runtime.controlprogram.parfor.stat.InfrastructureAnalyzer;\nimport org.tugraz.sysds.runtime.data.IndexedTensorBlock;\n-import org.tugraz.sysds.runtime.data.TensorBlock;\n+import org.tugraz.sysds.runtime.data.HomogTensor;\nimport org.tugraz.sysds.runtime.data.TensorIndexes;\nimport org.tugraz.sysds.runtime.instructions.spark.functions.CopyBinaryCellFunction;\nimport org.tugraz.sysds.runtime.instructions.spark.functions.CopyMatrixBlockFunction;\n@@ -72,11 +72,11 @@ public class SparkUtils\nreturn new IndexedMatrixValue(ix, mb);\n}\n- public static IndexedTensorBlock toIndexedTensorBlock( Tuple2<TensorIndexes,TensorBlock> in ) {\n+ public static IndexedTensorBlock toIndexedTensorBlock( Tuple2<TensorIndexes,HomogTensor> in ) {\nreturn new IndexedTensorBlock(in._1(), in._2());\n}\n- public static IndexedTensorBlock toIndexedTensorBlock(TensorIndexes ix, TensorBlock mb ) {\n+ public static IndexedTensorBlock toIndexedTensorBlock(TensorIndexes ix, HomogTensor mb ) {\nreturn new IndexedTensorBlock(ix, mb);\n}\n@@ -171,11 +171,11 @@ public class SparkUtils\n* Creates a partitioning-preserving deep copy of the input tensor RDD, where\n* the indexes and values are copied.\n*\n- * @param in tensor as {@code JavaPairRDD<TensorIndexes,TensorBlock>}\n- * @return tensor as {@code JavaPairRDD<TensorIndexes,TensorBlock>}\n+ * @param in tensor as {@code JavaPairRDD<TensorIndexes,HomogTensor>}\n+ * @return tensor as {@code JavaPairRDD<TensorIndexes,HomogTensor>}\n*/\n- public static JavaPairRDD<TensorIndexes, TensorBlock> copyBinaryBlockTensor(\n- JavaPairRDD<TensorIndexes, TensorBlock> in) {\n+ public static JavaPairRDD<TensorIndexes, HomogTensor> copyBinaryBlockTensor(\n+ JavaPairRDD<TensorIndexes, HomogTensor> in) {\nreturn copyBinaryBlockTensor(in, true);\n}\n@@ -183,12 +183,12 @@ public class SparkUtils\n* Creates a partitioning-preserving copy of the input tensor RDD. If a deep copy is\n* requested, indexes and values are copied, otherwise they are simply passed through.\n*\n- * @param in tensor as {@code JavaPairRDD<TensorIndexes,TensorBlock>}\n+ * @param in tensor as {@code JavaPairRDD<TensorIndexes,HomogTensor>}\n* @param deep if true, perform deep copy\n- * @return tensor as {@code JavaPairRDD<TensorIndexes,TensorBlock>}\n+ * @return tensor as {@code JavaPairRDD<TensorIndexes,HomogTensor>}\n*/\n- public static JavaPairRDD<TensorIndexes, TensorBlock> copyBinaryBlockTensor(\n- JavaPairRDD<TensorIndexes, TensorBlock> in, boolean deep) {\n+ public static JavaPairRDD<TensorIndexes, HomogTensor> copyBinaryBlockTensor(\n+ JavaPairRDD<TensorIndexes, HomogTensor> in, boolean deep) {\nif (!deep) //pass through of indexes and blocks\nreturn in.mapValues(new CopyTensorBlockFunction(false));\nelse //requires key access, so use mappartitions\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/util/DataConverter.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/util/DataConverter.java", "diff": "@@ -1063,7 +1063,7 @@ public class DataConverter\n/**\n* Concatenates a single tensor value to the `StringBuilder` by converting it to the correct format.\n*\n- * @param tb the TensorBlock\n+ * @param tb the HomogTensor\n* @param sb the StringBuilder to use\n* @param df DecimalFormat with the correct settings for double or float values\n* @param ix the index of the TensorBlock value\n@@ -1193,7 +1193,7 @@ public class DataConverter\nbreak;\ncase TENSOR: {\n// Dimensions given as vector\n- TensorBlock in = ec.getTensorInput(dims.getName());\n+ HomogTensor in = ec.getTensorInput(dims.getName());\nboolean colVec = false;\nif (!in.isVector()) {\nthrow new DMLRuntimeException(\"Dimensions tensor has to be a vector.\");\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/meta/MatrixCharacteristics.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/meta/MatrixCharacteristics.java", "diff": "@@ -76,8 +76,8 @@ public class MatrixCharacteristics extends DataCharacteristics\n@Override\npublic DataCharacteristics set(DataCharacteristics that) {\n- set(that.getRows(), that.getCols(), that.getRowsPerBlock(), that.getColsPerBlock(), getNonZeros());\n- ubNnz = !that.nnzKnown();\n+ set(that.getRows(), that.getCols(), that.getRowsPerBlock(), that.getColsPerBlock(), that.getNonZeros());\n+ ubNnz = (that instanceof MatrixCharacteristics && ((MatrixCharacteristics)that).ubNnz);\nreturn this;\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/meta/MetaDataFormat.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/meta/MetaDataFormat.java", "diff": "@@ -46,10 +46,9 @@ public class MetaDataFormat extends MetaData\n@Override\npublic Object clone() {\n- if (_dc instanceof MatrixCharacteristics) {\n+ if (_dc instanceof MatrixCharacteristics)\nreturn new MetaDataFormat(new MatrixCharacteristics(_dc), oinfo, iinfo);\n- } else {\n+ else\nreturn new MetaDataFormat(new TensorCharacteristics(_dc), oinfo, iinfo);\n}\n}\n-}\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/DataGenOp.java", "new_path": "src/main/java/org/tugraz/sysds/hops/DataGenOp.java", "diff": "@@ -93,6 +93,13 @@ public class DataGenOp extends MultiThreadedHop\n_id = id;\n_op = mthd;\n+ //ensure all parameters existing and consistent with data type\n+ //TODO remove once this unnecessary parameter is cleaned up\n+ if( !inputParameters.containsKey(DataExpression.RAND_TENSOR) )\n+ inputParameters.put(DataExpression.RAND_TENSOR, new LiteralOp(false));\n+ else if (HopRewriteUtils.isLiteralOfValue(inputParameters.get(DataExpression.RAND_TENSOR), true))\n+ setDataType(DataType.TENSOR);\n+\nint index = 0;\nfor( Entry<String, Hop> e: inputParameters.entrySet() ) {\nString s = e.getKey();\n@@ -510,14 +517,13 @@ public class DataGenOp extends MultiThreadedHop\n&& _paramIndexMap!=null && that2._paramIndexMap!=null\n&& _maxNumThreads == that2._maxNumThreads );\n- if( ret )\n- {\n- for( Entry<String,Integer> e : _paramIndexMap.entrySet() )\n- {\n+ if( ret ) {\n+ for( Entry<String,Integer> e : _paramIndexMap.entrySet() ) {\nString key1 = e.getKey();\nint pos1 = e.getValue();\n- int pos2 = that2._paramIndexMap.get(key1);\n- ret &= ( that2.getInput().get(pos2)!=null\n+ int pos2 = that2._paramIndexMap.containsKey(key1) ?\n+ that2._paramIndexMap.get(key1) : -1;\n+ ret &= ( pos2 >=0 && that2.getInput().get(pos2)!=null\n&& getInput().get(pos1) == that2.getInput().get(pos2) );\n}\n@@ -531,8 +537,6 @@ public class DataGenOp extends MultiThreadedHop\nret = false;\n}\n}\n-\nreturn ret;\n}\n-\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/parser/DataExpression.java", "new_path": "src/main/java/org/tugraz/sysds/parser/DataExpression.java", "diff": "@@ -105,7 +105,7 @@ public class DataExpression extends DataIdentifier\npublic static final String[] RAND_VALID_PARAM_NAMES =\n{RAND_ROWS, RAND_COLS, RAND_DIMS, RAND_MIN, RAND_MAX, RAND_SPARSITY, RAND_SEED, RAND_PDF, RAND_LAMBDA,\n- RAND_TENSOR};\n+ RAND_TENSOR}; //FIXME: why is this istensor required at all\npublic static final String[] RESHAPE_VALID_PARAM_NAMES =\n{ RAND_BY_ROW, RAND_DIMNAMES, RAND_DATA, RAND_ROWS, RAND_COLS, RAND_DIMS};\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/apache/sysml/runtime/util/UtilFunctions.java", "new_path": "src/main/java/org/apache/sysml/runtime/util/UtilFunctions.java", "diff": "@@ -323,11 +323,13 @@ public class UtilFunctions\n}\npublic static int toInt( double val ) {\n- return (int) Math.floor( val + DOUBLE_EPS );\n+ return (int) (Math.signum(val)\n+ * Math.floor(Math.abs(val) + DOUBLE_EPS));\n}\npublic static long toLong( double val ) {\n- return (long) Math.floor( val + DOUBLE_EPS );\n+ return (long) (Math.signum(val)\n+ * Math.floor(Math.abs(val) + DOUBLE_EPS));\n}\npublic static int toInt(Object obj) {\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/VariableCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/VariableCPInstruction.java", "diff": "@@ -356,7 +356,7 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\nInputInfo ii = OutputInfo.getMatchingInputInfo(oi);\nMetaDataFormat iimd = null;\n- if (dt == DataType.MATRIX) {\n+ if (dt == DataType.MATRIX || dt == DataType.FRAME) {\nDataCharacteristics mc = new MatrixCharacteristics();\nif (parts.length == 6) {\n// do nothing\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Types.java", "new_path": "src/main/java/org/tugraz/sysds/common/Types.java", "diff": "@@ -85,11 +85,18 @@ public class Types\n}\n}\npublic static ValueType fromExternalString(String value) {\n+ //for now we support both internal and external strings\n+ //until we have completely changed the external types\nString lvalue = (value != null) ? value.toUpperCase() : null;\nswitch(lvalue) {\n+ case \"FP32\": return FP32;\n+ case \"FP64\":\ncase \"DOUBLE\": return FP64;\n+ case \"INT32\": return INT32;\n+ case \"INT64\":\ncase \"INT\": return INT64;\ncase \"BOOLEAN\": return BOOLEAN;\n+ case \"STRING\": return STRING;\ndefault:\nthrow new DMLRuntimeException(\"Unknown value type: \"+value);\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/caching/FrameObject.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/caching/FrameObject.java", "diff": "@@ -125,7 +125,7 @@ public class FrameObject extends CacheableData<FrameBlock>\nString[] parts = schema.split(DataExpression.DEFAULT_DELIM_DELIMITER);\n_schema = new ValueType[parts.length];\nfor( int i=0; i<parts.length; i++ )\n- _schema[i] = ValueType.valueOf(parts[i].toUpperCase());\n+ _schema[i] = ValueType.fromExternalString(parts[i].toUpperCase());\n}\n}\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/append/AppendChainTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/append/AppendChainTest.java", "diff": "@@ -33,7 +33,6 @@ import org.tugraz.sysds.utils.Statistics;\npublic class AppendChainTest extends AutomatedTestBase\n{\n-\nprivate final static String TEST_NAME = \"AppendChainTest\";\nprivate final static String TEST_DIR = \"functions/append/\";\nprivate final static String TEST_CLASS_DIR = TEST_DIR + AppendChainTest.class.getSimpleName() + \"/\";\n@@ -61,12 +60,12 @@ public class AppendChainTest extends AutomatedTestBase\n@Test\npublic void testAppendChainVectorDenseCP() {\n- commonAppendTest(ExecMode.HYBRID, rows, cols1, cols2a, cols3a, false);\n+ commonAppendTest(ExecMode.SINGLE_NODE, rows, cols1, cols2a, cols3a, false);\n}\n@Test\npublic void testAppendChainMatrixDenseCP() {\n- commonAppendTest(ExecMode.HYBRID, rows, cols1, cols2b, cols3b, false);\n+ commonAppendTest(ExecMode.SINGLE_NODE, rows, cols1, cols2b, cols3b, false);\n}\n// ------------------------------------------------------\n@@ -94,18 +93,17 @@ public class AppendChainTest extends AutomatedTestBase\n@Test\npublic void testAppendChainVectorSparseCP() {\n- commonAppendTest(ExecMode.HYBRID, rows, cols1, cols2a, cols3a, true);\n+ commonAppendTest(ExecMode.SINGLE_NODE, rows, cols1, cols2a, cols3a, true);\n}\n@Test\npublic void testAppendChainMatrixSparseCP() {\n- commonAppendTest(ExecMode.HYBRID, rows, cols1, cols2b, cols3b, true);\n+ commonAppendTest(ExecMode.SINGLE_NODE, rows, cols1, cols2b, cols3b, true);\n}\npublic void commonAppendTest(ExecMode platform, int rows, int cols1, int cols2, int cols3, boolean sparse)\n{\nTestConfiguration config = getAndLoadTestConfiguration(TEST_NAME);\n-\nExecMode prevPlfm=rtplatform;\nboolean sparkConfigOld = DMLScript.USE_LOCAL_SPARK_CONFIG;\n@@ -118,24 +116,18 @@ public class AppendChainTest extends AutomatedTestBase\nconfig.addVariable(\"rows\", rows);\nconfig.addVariable(\"cols\", cols1);\n- //This is for running the junit test the new way, i.e., construct the arguments directly\nString RI_HOME = SCRIPT_DIR + TEST_DIR;\nfullDMLScriptName = RI_HOME + TEST_NAME + \".dml\";\nprogramArgs = new String[]{\"-args\", input(\"A\"),\n- Long.toString(rows),\n- Long.toString(cols1),\n- input(\"B1\"),\n- Long.toString(cols2),\n- input(\"B2\"),\n- Long.toString(cols3),\n- output(\"C\") };\n+ Long.toString(rows), Long.toString(cols1),\n+ input(\"B1\"), Long.toString(cols2), input(\"B2\"),\n+ Long.toString(cols3), output(\"C\") };\nfullRScriptName = RI_HOME + TEST_NAME + \".R\";\nrCmd = \"Rscript\" + \" \" + fullRScriptName + \" \" +\ninputDir() + \" \"+ expectedDir();\ndouble sparsity = sparse ? sparsity2 : sparsity1;\ndouble sparsity2 = 1-sparsity;\n-\ndouble[][] A = getRandomMatrix(rows, cols1, min, max, sparsity, 11);\nwriteInputMatrix(\"A\", A, true);\ndouble[][] B1= getRandomMatrix(rows, cols2, min, max, sparsity2, 21);\n@@ -143,15 +135,13 @@ public class AppendChainTest extends AutomatedTestBase\ndouble[][] B2= getRandomMatrix(rows, cols2, min, max, sparsity, 31);\nwriteInputMatrix(\"B2\", B2, true);\n- boolean exceptionExpected = false;\n- int expectedCompiledMRJobs = 1;\n- int expectedExecutedMRJobs = 0;\n- runTest(true, exceptionExpected, null, expectedCompiledMRJobs);\n+ int expectedCompiled = platform==ExecMode.SINGLE_NODE ?\n+ 0 : 8; //3x(rblk+chkpt), append, write\n+ runTest(true, false, null, expectedCompiled);\nrunRScript(true);\n- Assert.assertEquals(\"Wrong number of executed MR jobs.\",\n- expectedExecutedMRJobs, Statistics.getNoOfExecutedSPInst());\n- //compare result data\n+ Assert.assertEquals(\"Wrong number of executed Spark jobs.\",\n+ expectedCompiled, Statistics.getNoOfExecutedSPInst());\nfor(String file: config.getOutputFiles()) {\nHashMap<CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(file);\nHashMap<CellIndex, Double> rfile = readRMatrixFromFS(file);\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/append/AppendMatrixTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/append/AppendMatrixTest.java", "diff": "@@ -37,7 +37,6 @@ import org.tugraz.sysds.utils.Statistics;\npublic class AppendMatrixTest extends AutomatedTestBase\n{\n-\nprivate final static String TEST_NAME = \"AppendMatrixTest\";\nprivate final static String TEST_DIR = \"functions/append/\";\nprivate final static String TEST_CLASS_DIR = TEST_DIR + AppendMatrixTest.class.getSimpleName() + \"/\";\n@@ -61,7 +60,6 @@ public class AppendMatrixTest extends AutomatedTestBase\nprivate final static int cols2d = 1920;\nprivate final static int cols3d = 990;\n-\nprivate final static double sparsity1 = 0.5;\nprivate final static double sparsity2 = 0.01;\n@@ -145,35 +143,14 @@ public class AppendMatrixTest extends AutomatedTestBase\ncommonAppendTest(ExecMode.SPARK, rows, cols1d, cols2d, true, AppendMethod.MR_GAPPEND);\n}\n- // -----------------------------------------------------------------\n-\n- //NOTE: different dimension use cases only relvant for MR\n- /*\n- @Test\n- public void testAppendInBlock2CP() {\n- commonAppendTest(ExecMode.SINGLE_NODE, rows, cols1b, cols2b);\n- }\n-\n- @Test\n- public void testAppendOutBlock1CP() {\n- commonAppendTest(ExecMode.SINGLE_NODE, rows, cols1c, cols2c);\n- }\n-\n- @Test\n- public void testAppendOutBlock2CP() {\n- commonAppendTest(ExecMode.SINGLE_NODE, rows, cols1d, cols2d);\n- }*/\npublic void commonAppendTest(ExecMode platform, int rows, int cols1, int cols2, boolean sparse, AppendMethod forcedAppendMethod)\n{\nTestConfiguration config = getAndLoadTestConfiguration(TEST_NAME);\n-\nExecMode prevPlfm=rtplatform;\n-\ndouble sparsity = (sparse) ? sparsity2 : sparsity1;\nboolean sparkConfigOld = DMLScript.USE_LOCAL_SPARK_CONFIG;\n- try\n- {\n+ try {\nif(forcedAppendMethod != null) {\nBinaryOp.FORCED_APPEND_METHOD = forcedAppendMethod;\n}\n@@ -184,15 +161,11 @@ public class AppendMatrixTest extends AutomatedTestBase\nconfig.addVariable(\"rows\", rows);\nconfig.addVariable(\"cols\", cols1);\n- /* This is for running the junit test the new way, i.e., construct the arguments directly */\nString RI_HOME = SCRIPT_DIR + TEST_DIR;\nfullDMLScriptName = RI_HOME + TEST_NAME + \".dml\";\nprogramArgs = new String[]{\"-args\", input(\"A\"),\n- Long.toString(rows),\n- Long.toString(cols1),\n- input(\"B\"),\n- Long.toString(cols2),\n- output(\"C\") };\n+ Long.toString(rows), Long.toString(cols1),\n+ input(\"B\"), Long.toString(cols2), output(\"C\") };\nfullRScriptName = RI_HOME + TEST_NAME + \".R\";\nrCmd = \"Rscript\" + \" \" + fullRScriptName + \" \" +\ninputDir() + \" \" + expectedDir();\n@@ -204,28 +177,24 @@ public class AppendMatrixTest extends AutomatedTestBase\ndouble[][] B= getRandomMatrix(rows, cols2, min, max, sparsity, System.currentTimeMillis());\nwriteInputMatrix(\"B\", B, true);\n- boolean exceptionExpected = false;\n- int expectedCompiledMRJobs = 1;\n- int expectedExecutedMRJobs = 0;\n- runTest(true, exceptionExpected, null, expectedCompiledMRJobs);\n+ int expectedCompiled = platform==ExecMode.SINGLE_NODE ?\n+ 0 : 6; //2x(rblk+chkpt), append, write\n+ runTest(true, false, null, expectedCompiled);\nrunRScript(true);\n- Assert.assertEquals(\"Wrong number of executed MR jobs.\",\n- expectedExecutedMRJobs, Statistics.getNoOfExecutedSPInst());\n- for(String file: config.getOutputFiles())\n- {\n+ Assert.assertEquals(\"Wrong number of executed Spark jobs.\",\n+ expectedCompiled, Statistics.getNoOfExecutedSPInst());\n+ for(String file: config.getOutputFiles()) {\nHashMap<CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(file);\nHashMap<CellIndex, Double> rfile = readRMatrixFromFS(file);\nTestUtils.compareMatrices(dmlfile, rfile, epsilon, file+\"-DML\", file+\"-R\");\n}\n}\n- finally\n- {\n+ finally {\n//reset execution platform\nrtplatform = prevPlfm;\nDMLScript.USE_LOCAL_SPARK_CONFIG = sparkConfigOld;\nBinaryOp.FORCED_APPEND_METHOD = null;\n}\n}\n-\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/BranchRemovalTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/BranchRemovalTest.java", "diff": "@@ -37,15 +37,12 @@ public class BranchRemovalTest extends AutomatedTestBase\nprivate final static int rows = 10;\nprivate final static int cols = 15;\n-\n@Override\n- public void setUp()\n- {\n+ public void setUp() {\naddTestConfiguration(TEST_NAME,\nnew TestConfiguration(TEST_CLASS_DIR, TEST_NAME, new String[] { \"X\" }) );\n}\n-\n@Test\npublic void testTrueConditionNoBranchRemovalNoIPA() {\nrunBranchRemovalTest(true, false, false);\n@@ -123,7 +120,7 @@ public class BranchRemovalTest extends AutomatedTestBase\nTestUtils.compareMatrices(dmlfile, rfile, 0, \"Stat-DML\", \"Stat-R\");\n//check expected number of compiled and executed MR jobs\n- int expectedNumCompiled = 5; //reblock, 3xGMR (append), write\n+ int expectedNumCompiled = 4; //reblock, 2 append, write\nint expectedNumExecuted = 0;\nif( branchRemoval )\nexpectedNumCompiled = 1; //reblock\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/IPAConstantPropagationTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/IPAConstantPropagationTest.java", "diff": "@@ -116,8 +116,8 @@ public class IPAConstantPropagationTest extends AutomatedTestBase\nHashMap<CellIndex, Double> rfile = readRMatrixFromFS(\"X\");\nTestUtils.compareMatrices(dmlfile, rfile, 0, \"Stat-DML\", \"Stat-R\");\n- //check expected number of compiled and executed MR jobs\n- int expectedNumCompiled = ( branchRemoval && !update ) ? 0 : 1; //rand\n+ //check expected number of compiled and executed Spark jobs\n+ int expectedNumCompiled = ( branchRemoval && !update ) ? 0 : 2; //rand/write\nint expectedNumExecuted = 0;\ncheckNumCompiledSparkInst(expectedNumCompiled);\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/IPAPropagationSizeMultipleFunctionsTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/IPAPropagationSizeMultipleFunctionsTest.java", "diff": "@@ -45,8 +45,7 @@ public class IPAPropagationSizeMultipleFunctionsTest extends AutomatedTestBase\nprivate final static double sparsity = 0.7;\n@Override\n- public void setUp()\n- {\n+ public void setUp() {\nTestUtils.clearAssertionInformation();\naddTestConfiguration( TEST_NAME1, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME1, new String[] { \"R\" }) );\naddTestConfiguration( TEST_NAME2, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME2, new String[] { \"R\" }) );\n@@ -55,7 +54,6 @@ public class IPAPropagationSizeMultipleFunctionsTest extends AutomatedTestBase\naddTestConfiguration( TEST_NAME5, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME5, new String[] { \"R\" }) );\n}\n-\n@Test\npublic void testFunctionSizePropagationSameInput() {\nrunIPASizePropagationMultipleFunctionsTest(TEST_NAME1, false);\n@@ -137,9 +135,9 @@ public class IPAPropagationSizeMultipleFunctionsTest extends AutomatedTestBase\nHashMap<CellIndex, Double> rfile = readRMatrixFromFS(\"R\");\nTestUtils.compareMatrices(dmlfile, rfile, 0, \"Stat-DML\", \"Stat-R\");\n- //check expected number of compiled and executed MR jobs\n+ //check expected number of compiled and executed Spark jobs\nint expectedNumCompiled = (IPA) ? ((TEST_NAME.equals(TEST_NAME5))?2:1) :\n- (TEST_NAME.equals(TEST_NAME5)?5:4); //reblock, 2xGMR foo, GMR\n+ (TEST_NAME.equals(TEST_NAME5)?6:5); //reblock, rix, +, write + 1 per fun\nint expectedNumExecuted = 0;\ncheckNumCompiledSparkInst(expectedNumCompiled);\n@@ -149,5 +147,4 @@ public class IPAPropagationSizeMultipleFunctionsTest extends AutomatedTestBase\nOptimizerUtils.ALLOW_INTER_PROCEDURAL_ANALYSIS = oldFlagIPA;\n}\n}\n-\n}\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/ipa/IPAPassForwardFunctionCalls.java", "new_path": "src/main/java/org/tugraz/sysds/hops/ipa/IPAPassForwardFunctionCalls.java", "diff": "@@ -54,7 +54,8 @@ public class IPAPassForwardFunctionCalls extends IPAPass\nFunctionStatement fstmt = (FunctionStatement)fsb.getStatement(0);\n//step 1: basic application filter: simple forwarding call\n- if( fstmt.getBody().size() != 1 || !containsFunctionOp(fstmt.getBody().get(0).getHops())\n+ if( fstmt.getBody().size() != 1 || fstmt.getBody().get(0).getHops().size() != 1\n+ || !containsFunctionOp(fstmt.getBody().get(0).getHops())\n|| !hasOnlySimplyArguments((FunctionOp)fstmt.getBody().get(0).getHops().get(0)))\ncontinue;\nif( LOG.isDebugEnabled() )\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/applications/NNTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/applications/NNTest.java", "diff": "@@ -39,5 +39,4 @@ public class NNTest extends MLContextTestBase {\nsetUnexpectedStdOut(ERROR_STRING);\nml.execute(script);\n}\n-\n}\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/utils/RDDAggregateUtils.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/utils/RDDAggregateUtils.java", "diff": "@@ -28,6 +28,7 @@ import org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.data.TensorBlock;\nimport org.tugraz.sysds.runtime.data.TensorIndexes;\nimport org.tugraz.sysds.runtime.functionobjects.KahanPlus;\n+import org.tugraz.sysds.runtime.functionobjects.Plus;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\nimport org.tugraz.sysds.runtime.instructions.cp.KahanObject;\nimport org.tugraz.sysds.runtime.instructions.spark.AggregateUnarySPInstruction.RDDUAggFunction2;\n@@ -671,7 +672,10 @@ public class RDDAggregateUtils\nreturn arg0;\n}\n- // TODO correction\n+ // TODO remove once KahanPlus is completely replaced by plus\n+ if (_op.increOp.fn instanceof KahanPlus) {\n+ _op = new AggregateOperator(0, Plus.getPlusFnObject());\n+ }\n//aggregate second input (in-place)\narg0.incrementalAggregate(_op, arg1);\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/api/DMLOptions.java", "new_path": "src/main/java/org/tugraz/sysds/api/DMLOptions.java", "diff": "@@ -105,7 +105,8 @@ public class DMLOptions {\ndmlOptions.help = line.hasOption(\"help\");\nif (line.hasOption(\"lineage\")){\ndmlOptions.lineage = true;\n- String lineageType = line.getOptionValue(\"lineage\");\n+ String lineageTypes[] = line.getOptionValues(\"lineage\");\n+ for (String lineageType : lineageTypes) {\nif (lineageType != null){\nif (lineageType.equalsIgnoreCase(\"dedup\"))\ndmlOptions.lineage_dedup = lineageType.equalsIgnoreCase(\"dedup\");\n@@ -115,6 +116,7 @@ public class DMLOptions {\nthrow new org.apache.commons.cli.ParseException(\"Invalid argument specified for -lineage option\");\n}\n}\n+ }\ndmlOptions.debug = line.hasOption(\"debug\");\ndmlOptions.gpu = line.hasOption(\"gpu\");\nif (dmlOptions.gpu) {\n@@ -247,7 +249,7 @@ public class DMLOptions {\nOption helpOpt = OptionBuilder.withDescription(\"shows usage message\")\n.create(\"help\");\nOption lineageOpt = OptionBuilder.withDescription(\"computes lineage traces\")\n- .hasOptionalArg().create(\"lineage\");\n+ .hasArgs().create(\"lineage\");\noptions.addOption(configOpt);\noptions.addOption(cleanOpt);\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/api/DMLOptions.java", "new_path": "src/main/java/org/tugraz/sysds/api/DMLOptions.java", "diff": "@@ -106,6 +106,7 @@ public class DMLOptions {\nif (line.hasOption(\"lineage\")){\ndmlOptions.lineage = true;\nString lineageTypes[] = line.getOptionValues(\"lineage\");\n+ if (lineageTypes != null) {\nfor (String lineageType : lineageTypes) {\nif (lineageType != null){\nif (lineageType.equalsIgnoreCase(\"dedup\"))\n@@ -117,6 +118,7 @@ public class DMLOptions {\n}\n}\n}\n+ }\ndmlOptions.debug = line.hasOption(\"debug\");\ndmlOptions.gpu = line.hasOption(\"gpu\");\nif (dmlOptions.gpu) {\n@@ -249,7 +251,7 @@ public class DMLOptions {\nOption helpOpt = OptionBuilder.withDescription(\"shows usage message\")\n.create(\"help\");\nOption lineageOpt = OptionBuilder.withDescription(\"computes lineage traces\")\n- .hasArgs().create(\"lineage\");\n+ .hasOptionalArgs().create(\"lineage\");\noptions.addOption(configOpt);\noptions.addOption(cleanOpt);\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DataTensor.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DataTensor.java", "diff": "@@ -99,7 +99,7 @@ public class DataTensor extends TensorBlock {\nif (_colsdata == null) {\nallocateBlock();\n} else {\n- BasicTensor[] newCols = new BasicTensor[getDim(1)];\n+ //BasicTensor[] newCols = new BasicTensor[getDim(1)];\nif (_colsdata.length > getDim(1))\n_colsdata = Arrays.copyOfRange(_colsdata, 0, getDim(1));\nint[] blockDims = toInternalDims(dims);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/MatrixReshapeSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/MatrixReshapeSPInstruction.java", "diff": "@@ -29,7 +29,6 @@ import org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.controlprogram.context.SparkExecutionContext;\nimport org.tugraz.sysds.runtime.data.IndexedTensorBlock;\n-import org.tugraz.sysds.runtime.data.BasicTensor;\nimport org.tugraz.sysds.runtime.data.TensorBlock;\nimport org.tugraz.sysds.runtime.data.TensorIndexes;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/indexing/LeftIndexingUpdateInPlaceTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/indexing/LeftIndexingUpdateInPlaceTest.java", "diff": "@@ -145,7 +145,7 @@ public class LeftIndexingUpdateInPlaceTest extends AutomatedTestBase\nwriteInputMatrixWithMTD(\"B\", B, true);\n//run dml and r script\n- runTest(true, false, null, 1);\n+ runTest(true, false, null, 2); //2xrblk\nrunRScript(true);\nHashMap<CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(\"R\");\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/RewriteComplexMapMultChainTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/RewriteComplexMapMultChainTest.java", "diff": "@@ -32,7 +32,6 @@ import org.tugraz.sysds.test.TestUtils;\npublic class RewriteComplexMapMultChainTest extends AutomatedTestBase\n{\n-\nprivate final static String TEST_NAME1 = \"rewrite_mapmultchain1\";\nprivate final static String TEST_NAME2 = \"rewrite_mapmultchain2\";\nprivate final static String TEST_DIR = \"functions/recompile/\";\n@@ -48,7 +47,6 @@ public class RewriteComplexMapMultChainTest extends AutomatedTestBase\nprivate final static double sparsity = 0.7;\nprivate final static double eps = 0.0000001;\n-\n@Override\npublic void setUp()\n{\n@@ -58,7 +56,6 @@ public class RewriteComplexMapMultChainTest extends AutomatedTestBase\nnew TestConfiguration(TEST_CLASS_DIR, TEST_NAME2, new String[] { \"HV\" }) );\n}\n-\n@Test\npublic void testRewriteExpr1SingleColumnCP() throws IOException {\nrunRewriteMapMultChain(TEST_NAME1, true, ExecType.CP);\n@@ -114,15 +111,14 @@ public class RewriteComplexMapMultChainTest extends AutomatedTestBase\nHashMap<CellIndex, Double> rfile = readRMatrixFromFS(\"HV\");\nTestUtils.compareMatrices(dmlfile, rfile, eps, \"DML\", \"R\");\n- //check expected number of compiled and executed MR jobs\n- int expectedNumCompiled = (et==ExecType.CP)?1:(singleCol?4:6); //mapmultchain if single column\n+ //check expected number of compiled and executed Spark jobs\n+ int expectedNumCompiled = (et==ExecType.CP)?3:(singleCol?4:6); //mapmultchain if single column\nint expectedNumExecuted = (et==ExecType.CP)?0:(singleCol?4:6); //mapmultchain if single column\ncheckNumCompiledSparkInst(expectedNumCompiled);\ncheckNumExecutedSparkInst(expectedNumExecuted);\n}\n- finally\n- {\n+ finally {\nrtplatform = platformOld;\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/SparsityRecompileTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/SparsityRecompileTest.java", "diff": "@@ -49,7 +49,7 @@ public class SparsityRecompileTest extends AutomatedTestBase\nprivate final static String TEST_CLASS_DIR = TEST_DIR + SparsityRecompileTest.class.getSimpleName() + \"/\";\nprivate final static long rows = 1000;\n- private final static long cols = 500000;\n+ private final static long cols = 1000000;\nprivate final static double sparsity = 0.00001d;\nprivate final static double val = 7.0;\n@@ -128,21 +128,20 @@ public class SparsityRecompileTest extends AutomatedTestBase\nDataConverter.writeMatrixToHDFS(mb, input(\"V\"), OutputInfo.TextCellOutputInfo, mc);\nHDFSTool.writeMetaDataFile(input(\"V.mtd\"), ValueType.FP64, mc, OutputInfo.TextCellOutputInfo);\n- boolean exceptionExpected = false;\n- runTest(true, exceptionExpected, null, -1);\n+ runTest(true, false, null, -1);\n- //CHECK compiled MR jobs\n- int expectNumCompiled = (testname.equals(TEST_NAME2)?3:4) //reblock,GMR,GMR,GMR (one GMR less for if)\n- + (testname.equals(TEST_NAME4)?2:0);//(+2 resultmerge)\n- Assert.assertEquals(\"Unexpected number of compiled MR jobs.\",\n+ //CHECK compiled Spark jobs\n+ int expectNumCompiled = (testname.equals(TEST_NAME2)?3:4) //-1 for if\n+ + (testname.equals(TEST_NAME4)?3:0);//(+2 resultmerge, 1 sum)\n+ Assert.assertEquals(\"Unexpected number of compiled Spark jobs.\",\nexpectNumCompiled, Statistics.getNoOfCompiledSPInst());\n- //CHECK executed MR jobs\n- int expectNumExecuted = -1;\n- if( recompile ) expectNumExecuted = 0 + ((testname.equals(TEST_NAME4))?2:0); //(+2 resultmerge)\n- else expectNumExecuted = (testname.equals(TEST_NAME2)?3:4) //reblock,GMR,GMR,GMR (one GMR less for if)\n- + ((testname.equals(TEST_NAME4))?2:0); //(+2 resultmerge)\n- Assert.assertEquals(\"Unexpected number of executed MR jobs.\",\n+ //CHECK executed Spark jobs\n+ int expectNumExecuted = recompile ?\n+ ((testname.equals(TEST_NAME4))?2:0) : //(+2 resultmerge)\n+ (testname.equals(TEST_NAME2)?3:4) //reblock + 3 (-1 for if)\n+ + ((testname.equals(TEST_NAME4))?3:0); //(+2 resultmerge, 1 sum)\n+ Assert.assertEquals(\"Unexpected number of executed Spark jobs.\",\nexpectNumExecuted, Statistics.getNoOfExecutedSPInst());\n//compare matrices\n@@ -151,11 +150,9 @@ public class SparsityRecompileTest extends AutomatedTestBase\n}\ncatch(Exception ex) {\nthrow new RuntimeException(ex);\n- //Assert.fail(\"Failed to run test: \"+ex.getMessage());\n}\nfinally {\nCompilerConfig.FLAG_DYN_RECOMPILE = oldFlagRecompile;\n}\n}\n-\n}\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/OptimizerUtils.java", "new_path": "src/main/java/org/tugraz/sysds/hops/OptimizerUtils.java", "diff": "@@ -325,22 +325,6 @@ public class OptimizerUtils\ncase 4:\ncconf.set(ConfigType.OPT_LEVEL, OptimizationLevel.O4_GLOBAL_TIME_MEMORY.ordinal());\nbreak;\n- // opt level 4: debug mode (no interfering rewrites)\n- case 5:\n- cconf.set(ConfigType.OPT_LEVEL, OptimizationLevel.O5_DEBUG_MODE.ordinal());\n- ALLOW_CONSTANT_FOLDING = false;\n- ALLOW_COMMON_SUBEXPRESSION_ELIMINATION = false;\n- ALLOW_ALGEBRAIC_SIMPLIFICATION = false;\n- ALLOW_INTER_PROCEDURAL_ANALYSIS = false;\n- ALLOW_BRANCH_REMOVAL = false;\n- ALLOW_SIZE_EXPRESSION_EVALUATION = false;\n- ALLOW_WORSTCASE_SIZE_EXPRESSION_EVALUATION = false;\n- ALLOW_RAND_JOB_RECOMPILE = false;\n- ALLOW_SUM_PRODUCT_REWRITES = false;\n- ALLOW_SPLIT_HOP_DAGS = false;\n- cconf.set(ConfigType.ALLOW_DYN_RECOMPILATION, false);\n- cconf.set(ConfigType.ALLOW_INDIVIDUAL_SB_SPECIFIC_OPS, false);\n- break;\n// opt level 6 and7: SPOOF w/o fused operators, otherwise same as O2\n// (hidden optimization levels not documented on purpose, as they will\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/binary/matrix/MapMultChainTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/binary/matrix/MapMultChainTest.java", "diff": "@@ -240,10 +240,11 @@ public class MapMultChainTest extends AutomatedTestBase\n//check compiled/executed jobs\nint numInputs = testname.equals(TEST_NAME1) ? 2 : 3;\n- int expectedNumCompiled = numInputs\n- + ((instType==ExecType.SPARK)?(sumProductRewrites?1:2):0);\n+ int expectedNumCompiled = numInputs + ((instType==ExecType.SPARK) ?\n+ (numInputs + (sumProductRewrites?2:((numInputs==2)?4:5))):0);\ncheckNumCompiledSparkInst(expectedNumCompiled);\n- checkNumExecutedSparkInst(expectedNumCompiled - numInputs);\n+ checkNumExecutedSparkInst(expectedNumCompiled\n+ - ((instType==ExecType.CP)?numInputs:0));\n}\nfinally {\nrtplatform = platformOld;\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/SparsityFunctionRecompileTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/SparsityFunctionRecompileTest.java", "diff": "@@ -50,7 +50,7 @@ public class SparsityFunctionRecompileTest extends AutomatedTestBase\nSparsityFunctionRecompileTest.class.getSimpleName() + \"/\";\nprivate final static long rows = 1000;\n- private final static long cols = 500000;\n+ private final static long cols = 1000000;\nprivate final static double sparsity = 0.00001d;\nprivate final static double val = 7.0;\n@@ -162,7 +162,7 @@ public class SparsityFunctionRecompileTest extends AutomatedTestBase\nString HOME = SCRIPT_DIR + TEST_DIR;\nfullDMLScriptName = HOME + testname + \".dml\";\n- programArgs = new String[]{\"-stats\", \"-args\",\n+ programArgs = new String[]{\"-explain\", \"-args\",\ninput(\"V\"), Double.toString(val), output(\"R\") };\nCompilerConfig.FLAG_DYN_RECOMPILE = recompile;\n@@ -180,7 +180,7 @@ public class SparsityFunctionRecompileTest extends AutomatedTestBase\n//CHECK compiled Spark jobs\nint expectNumCompiled = 1 //rblk\n- + (testname.equals(TEST_NAME2) ? (IPA?0:5) : (IPA?1:4)) //if no write on IPA\n+ + (testname.equals(TEST_NAME2) ? (IPA?2:5) : (IPA?3:4)) //if no write on IPA\n+ (testname.equals(TEST_NAME4)? 2 : 0); //(+2 parfor resultmerge);\nAssert.assertEquals(\"Unexpected number of compiled Spark jobs.\",\nexpectNumCompiled, Statistics.getNoOfCompiledSPInst());\n@@ -188,8 +188,8 @@ public class SparsityFunctionRecompileTest extends AutomatedTestBase\n//CHECK executed Spark jobs\nint expectNumExecuted = recompile ?\n(testname.equals(TEST_NAME4)?2:0) : //(2x resultmerge)\n- (testname.equals(TEST_NAME2) ? (IPA?1:5) :\n- (testname.equals(TEST_NAME4) ? (IPA?4:7) : (IPA?2:5)));\n+ (testname.equals(TEST_NAME2) ? (IPA?3:5) :\n+ (testname.equals(TEST_NAME4) ? (IPA?6:7) : (IPA?4:5)));\nAssert.assertEquals(\"Unexpected number of executed Spark jobs.\",\nexpectNumExecuted, Statistics.getNoOfExecutedSPInst());\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/applications/ID3Test.java", "new_path": "src/test/java/org/tugraz/sysds/test/applications/ID3Test.java", "diff": "@@ -99,7 +99,7 @@ public class ID3Test extends AutomatedTestBase\n//check also num actually executed jobs\nif(AutomatedTestBase.rtplatform != ExecMode.SPARK) {\nlong actualSP = Statistics.getNoOfExecutedSPInst();\n- Assert.assertEquals(\"Wrong number of executed jobs: expected 2 but executed \"+actualSP+\".\", 2, actualSP);\n+ Assert.assertEquals(\"Wrong number of executed jobs: expected 0 but executed \"+actualSP+\".\", 0, actualSP);\n}\n//compare results\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/binary/matrix/BinUaggChainTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/binary/matrix/BinUaggChainTest.java", "diff": "@@ -21,6 +21,7 @@ package org.tugraz.sysds.test.functions.binary.matrix;\nimport java.util.HashMap;\n+import org.junit.Assert;\nimport org.junit.Test;\nimport org.tugraz.sysds.api.DMLScript;\nimport org.tugraz.sysds.common.Types.ExecMode;\n@@ -30,13 +31,8 @@ import org.tugraz.sysds.test.AutomatedTestBase;\nimport org.tugraz.sysds.test.TestConfiguration;\nimport org.tugraz.sysds.test.TestUtils;\n-/**\n- * TODO: extend test by various binary operator - unary aggregate operator combinations.\n- *\n- */\npublic class BinUaggChainTest extends AutomatedTestBase\n{\n-\nprivate final static String TEST_NAME1 = \"BinUaggChain_Col\";\nprivate final static String TEST_DIR = \"functions/binary/matrix/\";\nprivate final static String TEST_CLASS_DIR = TEST_DIR + BinUaggChainTest.class.getSimpleName() + \"/\";\n@@ -50,8 +46,7 @@ public class BinUaggChainTest extends AutomatedTestBase\nprivate final static double sparsity2 = 0.1; //sparse\n@Override\n- public void setUp()\n- {\n+ public void setUp() {\nTestUtils.clearAssertionInformation();\naddTestConfiguration(TEST_NAME1, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME1, new String[] { \"B\" }));\n}\n@@ -59,39 +54,25 @@ public class BinUaggChainTest extends AutomatedTestBase\n// -------------------------\n@Test\n- public void testBinUaggChainColSingleDenseSP()\n- {\n+ public void testBinUaggChainColSingleDenseSP() {\nrunBinUaggTest(TEST_NAME1, true, false, ExecType.SPARK);\n}\n@Test\n- public void testBinUaggChainColSingleSparseSP()\n- {\n+ public void testBinUaggChainColSingleSparseSP() {\nrunBinUaggTest(TEST_NAME1, true, true, ExecType.SPARK);\n}\n@Test\n- public void testBinUaggChainColMultiDenseSP()\n- {\n+ public void testBinUaggChainColMultiDenseSP() {\nrunBinUaggTest(TEST_NAME1, false, false, ExecType.SPARK);\n}\n@Test\n- public void testBinUaggChainColMultiSparseSP()\n- {\n+ public void testBinUaggChainColMultiSparseSP() {\nrunBinUaggTest(TEST_NAME1, false, true, ExecType.SPARK);\n}\n- // ----------------------\n-\n-\n-\n- /**\n- *\n- * @param sparseM1\n- * @param sparseM2\n- * @param instType\n- */\nprivate void runBinUaggTest( String testname, boolean singleBlock, boolean sparse, ExecType instType)\n{\n//rtplatform for MR\n@@ -110,10 +91,9 @@ public class BinUaggChainTest extends AutomatedTestBase\nString TEST_NAME = testname;\ngetAndLoadTestConfiguration(TEST_NAME);\n- /* This is for running the junit test the new way, i.e., construct the arguments directly */\nString HOME = SCRIPT_DIR + TEST_DIR;\nfullDMLScriptName = HOME + TEST_NAME + \".dml\";\n- programArgs = new String[]{\"-args\", input(\"A\"), output(\"B\")};\n+ programArgs = new String[]{\"-stats\",\"-args\", input(\"A\"), output(\"B\")};\nfullRScriptName = HOME + TEST_NAME + \".R\";\nrCmd = \"Rscript\" + \" \" + fullRScriptName + \" \" + inputDir() + \" \" + expectedDir();\n@@ -131,18 +111,16 @@ public class BinUaggChainTest extends AutomatedTestBase\nTestUtils.compareMatrices(dmlfile, rfile, eps, \"Stat-DML\", \"Stat-R\");\n//check compiled/executed jobs\n- if( rtplatform != ExecMode.SINGLE_NODE ) {\n- int expectedNumCompiled = (singleBlock)?1:3;\n- int expectedNumExecuted = (singleBlock)?1:3;\n+ int expectedNumCompiled = (singleBlock)?4:5;\n+ int expectedNumExecuted = (singleBlock)?4:5;\ncheckNumCompiledSparkInst(expectedNumCompiled);\ncheckNumExecutedSparkInst(expectedNumExecuted);\n+ Assert.assertEquals(singleBlock,\n+ heavyHittersContainsSubString(\"binuaggchain\"));\n}\n- }\n- finally\n- {\n+ finally {\nrtplatform = platformOld;\nDMLScript.USE_LOCAL_SPARK_CONFIG = sparkConfigOld;\n}\n}\n-\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/parfor/ParForRepeatedOptimizationTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/parfor/ParForRepeatedOptimizationTest.java", "diff": "@@ -38,7 +38,6 @@ import org.tugraz.sysds.utils.Statistics;\npublic class ParForRepeatedOptimizationTest extends AutomatedTestBase\n{\n-\nprivate final static String TEST_NAME1 = \"parfor_repeatedopt1\";\nprivate final static String TEST_NAME2 = \"parfor_repeatedopt2\";\nprivate final static String TEST_NAME3 = \"parfor_repeatedopt3\";\n@@ -63,58 +62,50 @@ public class ParForRepeatedOptimizationTest extends AutomatedTestBase\n}\n@BeforeClass\n- public static void init()\n- {\n+ public static void init() {\nTestUtils.clearDirectory(TEST_DATA_DIR + TEST_CLASS_DIR);\n}\n@AfterClass\n- public static void cleanUp()\n- {\n+ public static void cleanUp() {\nif (TEST_CACHE_ENABLED) {\nTestUtils.clearDirectory(TEST_DATA_DIR + TEST_CLASS_DIR);\n}\n}\n@Test\n- public void testParForRepeatedOptNoReuseNoUpdateCP()\n- {\n- int numExpectedMRJobs = 1+3; //reblock, 3*partition\n+ public void testParForRepeatedOptNoReuseNoUpdateCP() {\n+ int numExpectedMRJobs = 1+8; //reblock, 3*partition, 4*checkpoints, 1\nrunParForRepeatedOptTest( false, false, false, ExecType.CP, numExpectedMRJobs );\n}\n@Test\n- public void testParForRepeatedOptNoReuseUpdateCP()\n- {\n- int numExpectedMRJobs = 1+3+2; //reblock, 3*partition, 2*GMR (previously 3GMR, now 1GMR removed on V*1)\n+ public void testParForRepeatedOptNoReuseUpdateCP() {\n+ int numExpectedMRJobs = 1+3+6; //reblock, 3*partition, 4*checkpoints, 2\nrunParForRepeatedOptTest( false, true, false, ExecType.CP, numExpectedMRJobs );\n}\n@Test\n- public void testParForRepeatedOptNoReuseChangedDimCP()\n- {\n- int numExpectedMRJobs = 1+3+3; //reblock, 3*partition, 3*GMR\n+ public void testParForRepeatedOptNoReuseChangedDimCP() {\n+ int numExpectedMRJobs = 1+3+7; //reblock, 3*partition, 4*checkpoints, 3\nrunParForRepeatedOptTest( false, false, true, ExecType.CP, numExpectedMRJobs );\n}\n@Test\n- public void testParForRepeatedOptReuseNoUpdateCP()\n- {\n- int numExpectedMRJobs = 1+1; //reblock, partition\n+ public void testParForRepeatedOptReuseNoUpdateCP() {\n+ int numExpectedMRJobs = 1+1 + 5; //reblock, partition, ?\nrunParForRepeatedOptTest( true, false, false, ExecType.CP, numExpectedMRJobs );\n}\n@Test\n- public void testParForRepeatedOptReuseUpdateCP()\n- {\n- int numExpectedMRJobs = 1+3+2; //reblock, 3*partition, 2*GMR (previously 3GMR, now 1GMR removed on V*1)\n+ public void testParForRepeatedOptReuseUpdateCP() {\n+ int numExpectedMRJobs = 1+3+6; //reblock, 3*partition, 4*checkpoint, 2\nrunParForRepeatedOptTest( true, true, false, ExecType.CP, numExpectedMRJobs );\n}\n@Test\n- public void testParForRepeatedOptReuseChangedDimCP()\n- {\n- int numExpectedMRJobs = 1+3+3; //reblock, 3*partition, 3*GMR\n+ public void testParForRepeatedOptReuseChangedDimCP() {\n+ int numExpectedMRJobs = 1+3+7; //reblock, 3*partition, 4*checkpoints, 3\nrunParForRepeatedOptTest( true, false, true, ExecType.CP, numExpectedMRJobs );\n}\n@@ -171,7 +162,8 @@ public class ParForRepeatedOptimizationTest extends AutomatedTestBase\nrunTest(true, false, null, -1);\nrunRScript(true);\n- Assert.assertEquals(\"Unexpected number of executed MR jobs.\", numExpectedMR, Statistics.getNoOfExecutedSPInst());\n+ Assert.assertEquals(\"Unexpected number of executed Spark jobs.\",\n+ numExpectedMR, Statistics.getNoOfExecutedSPInst());\n//compare matrices\nHashMap<CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(\"R\");\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/ternary/CTableSequenceTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/ternary/CTableSequenceTest.java", "diff": "@@ -21,6 +21,7 @@ package org.tugraz.sysds.test.functions.ternary;\nimport java.util.HashMap;\n+import org.junit.Assert;\nimport org.junit.Test;\nimport org.tugraz.sysds.api.DMLScript;\nimport org.tugraz.sysds.common.Types.ExecMode;\n@@ -45,7 +46,6 @@ import org.tugraz.sysds.test.TestUtils;\n*/\npublic class CTableSequenceTest extends AutomatedTestBase\n{\n-\nprivate final static String TEST_NAME1 = \"CTableSequenceLeft\";\nprivate final static String TEST_NAME2 = \"CTableSequenceRight\";\n@@ -56,118 +56,97 @@ public class CTableSequenceTest extends AutomatedTestBase\nprivate final static int rows = 2407;\nprivate final static int maxVal = 7;\n-\n@Override\n- public void setUp()\n- {\n+ public void setUp() {\nTestUtils.clearAssertionInformation();\naddTestConfiguration(TEST_NAME1, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME1, new String[] { \"B\" }) );\naddTestConfiguration(TEST_NAME2, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME2, new String[] { \"B\" }) );\n}\n@Test\n- public void testCTableSequenceLeftNoRewriteSP()\n- {\n+ public void testCTableSequenceLeftNoRewriteSP() {\nrunCTableSequenceTest(false, true, false, ExecType.SPARK);\n}\n@Test\n- public void testCTableSequenceLeftRewriteSP()\n- {\n+ public void testCTableSequenceLeftRewriteSP() {\nrunCTableSequenceTest(true, true, false, ExecType.SPARK);\n}\n@Test\n- public void testCTableSequenceRightNoRewriteSP()\n- {\n+ public void testCTableSequenceRightNoRewriteSP() {\nrunCTableSequenceTest(false, false, false, ExecType.SPARK);\n}\n@Test\n- public void testCTableSequenceRightRewriteSP()\n- {\n+ public void testCTableSequenceRightRewriteSP() {\nrunCTableSequenceTest(true, false, false, ExecType.SPARK);\n}\n@Test\n- public void testCTableSequenceLeftNoRewriteAggSP()\n- {\n+ public void testCTableSequenceLeftNoRewriteAggSP() {\nrunCTableSequenceTest(false, true, true, ExecType.SPARK);\n}\n@Test\n- public void testCTableSequenceLeftRewriteAggSP()\n- {\n+ public void testCTableSequenceLeftRewriteAggSP() {\nrunCTableSequenceTest(true, true, true, ExecType.SPARK);\n}\n@Test\n- public void testCTableSequenceRightNoRewriteAggSP()\n- {\n+ public void testCTableSequenceRightNoRewriteAggSP() {\nrunCTableSequenceTest(false, false, true, ExecType.SPARK);\n}\n@Test\n- public void testCTableSequenceRightRewriteAggSP()\n- {\n+ public void testCTableSequenceRightRewriteAggSP() {\nrunCTableSequenceTest(true, false, true, ExecType.SPARK);\n}\n@Test\n- public void testCTableSequenceLeftNoRewriteCP()\n- {\n+ public void testCTableSequenceLeftNoRewriteCP() {\nrunCTableSequenceTest(false, true, false, ExecType.CP);\n}\n@Test\n- public void testCTableSequenceLeftRewriteCP()\n- {\n+ public void testCTableSequenceLeftRewriteCP() {\nrunCTableSequenceTest(true, true, false, ExecType.CP);\n}\n@Test\n- public void testCTableSequenceRightNoRewriteCP()\n- {\n+ public void testCTableSequenceRightNoRewriteCP() {\nrunCTableSequenceTest(false, false, false, ExecType.CP);\n}\n@Test\n- public void testCTableSequenceRightRewriteCP()\n- {\n+ public void testCTableSequenceRightRewriteCP() {\nrunCTableSequenceTest(true, false, false, ExecType.CP);\n}\n-\n@Test\n- public void testCTableSequenceLeftNoRewriteAggCP()\n- {\n+ public void testCTableSequenceLeftNoRewriteAggCP() {\nrunCTableSequenceTest(false, true, true, ExecType.CP);\n}\n@Test\n- public void testCTableSequenceLeftRewriteAggCP()\n- {\n+ public void testCTableSequenceLeftRewriteAggCP() {\nrunCTableSequenceTest(true, true, true, ExecType.CP);\n}\n@Test\n- public void testCTableSequenceRightNoRewriteAggCP()\n- {\n+ public void testCTableSequenceRightNoRewriteAggCP() {\nrunCTableSequenceTest(false, false, true, ExecType.CP);\n}\n@Test\n- public void testCTableSequenceRightRewriteAggCP()\n- {\n+ public void testCTableSequenceRightRewriteAggCP() {\nrunCTableSequenceTest(true, false, true, ExecType.CP);\n}\nprivate void runCTableSequenceTest(boolean rewrite, boolean left, boolean withAgg, ExecType et)\n{\nString TEST_NAME = left ? TEST_NAME1 : TEST_NAME2;\n-\n- //rtplatform for MR\nExecMode platformOld = rtplatform;\nboolean rewriteOld = TernaryOp.ALLOW_CTABLE_SEQUENCE_REWRITES;\n@@ -187,10 +166,9 @@ public class CTableSequenceTest extends AutomatedTestBase\nTestConfiguration config = getTestConfiguration(TEST_NAME);\nloadTestConfiguration(config);\n- /* This is for running the junit test the new way, i.e., construct the arguments directly */\nString HOME = SCRIPT_DIR + TEST_DIR;\nfullDMLScriptName = HOME + TEST_NAME + \".dml\";\n- programArgs = new String[]{\"-explain\",\"-args\", input(\"A\"),\n+ programArgs = new String[]{\"-stats\",\"-args\", input(\"A\"),\nInteger.toString(rows),\nInteger.toString(1),\nInteger.toString(withAgg?1:0),\n@@ -213,8 +191,10 @@ public class CTableSequenceTest extends AutomatedTestBase\n//w/ rewrite: 4 instead of 6 because seq and aggregation are not required for ctable_expand\n//2 for CP due to reblock jobs for input and table\n- int expectedNumCompiled = ((et==ExecType.CP) ? 2 :(rewrite ? 4 : 6))+(withAgg ? 1 : 0);\n+ int expectedNumCompiled = ((et==ExecType.CP) ? 2 : 5)+(withAgg ? 1 : 0);\ncheckNumCompiledSparkInst(expectedNumCompiled);\n+ Assert.assertEquals(left & rewrite,\n+ heavyHittersContainsSubString(\"ctableexpand\"));\n}\nfinally {\nrtplatform = platformOld;\n" } ]

[ { "change_type": "ADD", "old_path": null, "new_path": "scripts/builtin/slicefinder.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+\n+m_slicefinder = function(Matrix[Double] X0, Matrix[Double] W, Integer k = 1) return(Matrix[Double] result) {\n+\n+ f = 2;\n+ beta = W;\n+ # number of features combined\n+ col = ncol(X0);\n+ row = nrow(X0);\n+ val_matrix = matrix(0, rows = 2, cols = col - 1);\n+ vcol = ncol(val_matrix);\n+ empty_row = matrix(0, rows = 1, cols = col - 1);\n+ print(\"col: \" + col + \" row = \" + row);\n+ #first scan, spot data, make first slices\n+\n+ for (j in 1:col - 1) {\n+ vector = order(target = X0[, j], by = 1, decreasing = FALSE);\n+ val_matrix[2, j] = vector[1, 1];\n+ val_counter = 1;\n+ print(\"Col \" + j);\n+ for (i in 1:row) {\n+ if (as.scalar(val_matrix[val_counter + 1, j]) != as.scalar(vector[i, 1])) {\n+ if (nrow(val_matrix) == val_counter + 1)\n+ val_matrix = rbind(val_matrix, empty_row);\n+ val_counter = val_counter + 1;\n+ val_matrix[val_counter + 1, j] = vector[i, 1];\n+ }\n+ }\n+\n+ val_matrix[1, j] = val_counter;\n+ #here I can add some condition to split the values from each column if val_counter is too big;\n+ ################################################\n+ #this code relates to large datasets\n+ /* if (val_counter > k) {\n+ position = floor(val_counter / k);\n+ for (a in 1:k) {\n+ if (a == k) {\n+ pos = as.scalar(val_matrix[1, j]) + 1;\n+ tresh = val_matrix[pos, j];\n+ val_matrix[a + 1, j] = tresh;\n+ } else {\n+ pos = position * a;\n+ tresh = val_matrix[pos, j];\n+ val_matrix[a + 1, j] = tresh;\n+ }\n+ }\n+ }\n+ */\n+ ##################################################\n+ }\n+\n+ # now val_matrix[1:4,]) is a treshhold matrix that define clear slices\n+ print(toString(val_matrix));\n+\n+ #start selecting slices\n+ vrow = nrow(val_matrix);\n+ vcol = ncol(val_matrix);\n+ totalrows = (vrow - 1) * vcol;\n+ print(\"vrow: \" + vrow);\n+ print(\"vcol: \" + vcol);\n+ print(\"totalrows: \" + totalrows);\n+\n+ #######################################\n+ Y0 = X0[1:nrow(X0), ncol(X0)];\n+ Y = lmpredict(X = X0[1:nrow(X0), 1:col - 1], w = beta, icpt = 0);\n+ [error0, diff0] = standart_error(Y, Y0);\n+ print(\"Error0: \" + error0);\n+ print(\"diff0: \" + diff0);\n+\n+ #####################################################\n+\n+ set_matrix = matrix(0, rows = 1, cols = 2 + (8 * f));\n+ set_row = matrix(0, rows = 1, cols = 2 + (8 * f));\n+\n+ cont = 1;\n+\n+ b0 = 1;\n+ b1 = col - 1;\n+ slice_number = 0;\n+ pointer_col = 1;\n+ pointer_row = 2;\n+\n+ set_matrix = first_slices(val_matrix, set_matrix, X0,set_row, beta);\n+\n+ ress = order(target = set_matrix, by = 1, decreasing = TRUE);\n+ set_matrix = double_features(val_matrix, set_matrix, X0,Y, set_row, beta);\n+\n+\n+ ress = order(target = set_matrix, by = 1, decreasing = TRUE);\n+ set_rows = nrow(set_matrix);\n+ set_cols = ncol(set_matrix);\n+ print(\"Second ress\");\n+ print(toString(ress));\n+ print(set_rows);\n+\n+ result = ress;\n+}\n+\n+standart_error = function(matrix[double] Y, matrix[double] Y0) return(double error, double diff) {\n+ diff = var(Y0 - Y);\n+ error = sqrt(sum((Y0 - Y)^2) / (nrow(Y) - 2));\n+}\n+\n+index = function(matrix[double] X, Integer column, double value, Integer mode) return(Integer pos) {\n+ begin = 1;\n+ e = nrow(X) + 1;\n+ while (begin < e - 1) {\n+ pos = as.integer(floor((begin + e) / 2));\n+ if (mode == 0) {\n+ if (as.scalar(X[pos, column]) < value)\n+ begin = pos;\n+ else\n+ e = pos;\n+ }\n+ else if (mode == 1) {\n+ if (as.scalar(X[pos, column]) <= value)\n+ begin = pos;\n+ else\n+ e = pos;\n+ }\n+ }\n+}\n+\n+first_slices = function(Matrix[Double] val_matrix, Matrix[Double] set_matrix, Matrix[Double] X0, Matrix[Double] set_row, Matrix[Double] beta) return(Matrix[Double] set_matrix) {\n+ col = ncol(X0);\n+ row = nrow(X0);\n+ vrow = nrow(val_matrix);\n+ vcol = ncol(val_matrix);\n+ cont = nrow(set_matrix);\n+ b0 = 1;\n+ b1 = col - 1;\n+\n+ for (j in 1:vcol) {\n+ num_value = as.scalar(val_matrix[1,j]);\n+ x = order(target = X0, by = j, decreasing = FALSE);\n+ print(\"my col: \" + j)\n+ for (i in 2:num_value+1) {\n+ swich = 1;\n+ if (nrow(set_matrix) < cont)\n+ set_matrix = rbind(set_matrix, set_row);\n+\n+ if (swich == 1) {\n+ value = as.scalar(val_matrix[i, j]);\n+ a0 = index(x, j, value, 0);\n+ a1 = index(x, j, value, 1);\n+ slice_matrix = x[a0:a1, b0:b1];\n+ Y0 = x[a0:a1, col];\n+ Y = lmpredict(X = slice_matrix, w = beta, icpt = 0);\n+ [error, diff] = standart_error(Y, Y0);\n+ set_matrix[cont,1:10] = t(as.matrix(list(diff, error, value,\n+ j, nrow(slice_matrix), ncol(slice_matrix), a0, a1, b0, b1)))\n+ cont = cont + 1;\n+ swich = 0;\n+ }\n+ }\n+ }\n+}\n+\n+\n+double_features = function(Matrix[Double] val_matrix, Matrix[Double] set_matrix, Matrix[Double] X0,Matrix[Double] Y, Matrix[Double] set_row, Matrix[Double] beta) return(Matrix[Double] set_matrix) {\n+\n+ vrow = nrow(val_matrix);\n+ vcol = ncol(val_matrix);\n+ cont = nrow(set_matrix);\n+ col = ncol(X0);\n+ row = nrow(X0);\n+ totalrows = (vrow - 1) * vcol;\n+ b0 = 1;\n+ b1 = col - 1;\n+ slice_number = 2;\n+\n+ for (j in 1:vcol) {\n+ num_value = as.scalar(val_matrix[1,j]);\n+ x = order(target = X0, by = j, decreasing = FALSE);\n+ if(j == num_value + 1)\n+ vrow = vrow -1;\n+\n+ for (i in 2:vrow) {\n+ if (i > 2 | j > 1)\n+ slice_number = slice_number + 1;\n+\n+ for (a in slice_number:totalrows) {\n+ num_col = as.scalar(set_matrix[a, 4]);\n+ x_x = order(target = X0, by = num_col, decreasing = FALSE);\n+\n+ value_A = as.scalar(set_matrix[a, 3]);\n+ a00 = as.scalar(set_matrix[a,7]);\n+ a11 = as.scalar(set_matrix[a,8]);\n+ #print(\"a0 y a1: \" + a00 + \" \" + a11)\n+ A = x_x[a00:a11, b0:b1];\n+ Ya = x_x[a00:a11, col];\n+\n+ if (nrow(set_matrix) <= cont)\n+ set_matrix = rbind(set_matrix, set_row);\n+\n+ value_B = as.scalar(val_matrix[i,j]);\n+ a0 = index(x, j, value_B, 0);\n+ a1 = index(x, j, value_B, 1);\n+ B = x[a0:a1, b0:b1];\n+ slice_matrix = rbind(A, B);\n+ Yb = x[a0:a1, col];\n+\n+ Y0 = rbind(Ya, Yb);\n+ Y = lmpredict(X = slice_matrix, w = beta, icpt = 0);\n+\n+ [error, diff] = standart_error(Y, Y0);\n+\n+ set_matrix[cont, ] = t(as.matrix(list(diff, error, value_A, num_col, nrow(A),\n+ ncol(A), a00, a11, b0, b1, value_B, j, nrow(B), ncol(B), a0, a1, b0, b1)));\n+ cont = cont + 1;\n+ }\n+ }\n+ }\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "new_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "diff": "@@ -137,6 +137,7 @@ public enum Builtins {\nSIN(\"sin\", false),\nSINH(\"sinh\", false),\nSTEPLM(\"steplm\",true, ReturnType.MULTI_RETURN),\n+ SLICEFINDER(\"slicefinder\", true),\nSOLVE(\"solve\", false),\nSQRT(\"sqrt\", false),\nSUM(\"sum\", false),\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/TestUtils.java", "new_path": "src/test/java/org/tugraz/sysds/test/TestUtils.java", "diff": "@@ -2060,6 +2060,13 @@ public class TestUtils\nreturn data;\n}\n+ public static double[][] ceil(double[][] data) {\n+ for(int i=0; i<data.length; i++)\n+ for(int j=0; j<data[i].length; j++)\n+ data[i][j]=Math.ceil(data[i][j]);\n+ return data;\n+ }\n+\npublic static double[][] floor(double[][] data, int col) {\nfor(int i=0; i<data.length; i++)\ndata[i][col]=Math.floor(data[i][col]);\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinSliceFinderTest.java", "diff": "+/*\n+ * Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed under the Apache License, Version 2.0 (the \"License\");\n+ * you may not use this file except in compliance with the License.\n+ * You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing, software\n+ * distributed under the License is distributed on an \"AS IS\" BASIS,\n+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+ * See the License for the specific language governing permissions and\n+ * limitations under the License.\n+ */\n+\n+package org.tugraz.sysds.test.functions.builtin;\n+\n+\n+import org.junit.Test;\n+import org.tugraz.sysds.common.Types.ExecMode;\n+import org.tugraz.sysds.lops.LopProperties.ExecType;\n+import org.tugraz.sysds.test.AutomatedTestBase;\n+import org.tugraz.sysds.test.TestConfiguration;\n+import org.tugraz.sysds.test.TestUtils;\n+\n+//package io;\n+import java.util.*;\n+\n+public class BuiltinSliceFinderTest extends AutomatedTestBase {\n+\n+ private final static String TEST_NAME = \"slicefinder\";\n+ private final static String TEST_DIR = \"functions/builtin/\";\n+ private static final String TEST_CLASS_DIR = TEST_DIR + BuiltinSliceFinderTest.class.getSimpleName() + \"/\";\n+\n+ private final static int rows = 32000;\n+ private final static int cols = 10;\n+\n+ @Override\n+ public void setUp() {\n+ addTestConfiguration(TEST_NAME, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME, new String[]{\"B\"}));\n+ }\n+\n+ @Test\n+ public void SingleFreatureTest() {\n+ runslicefindertest(1,true, ExecType.CP, BuiltinLmTest.LinregType.AUTO);\n+ }\n+\n+ @Test\n+ public void MultipleValuesOneFeature() {\n+ runslicefindertest(2,true, ExecType.CP, BuiltinLmTest.LinregType.AUTO);\n+ }\n+\n+ @Test\n+ public void MultipleFeaturesSingleValues() {\n+ runslicefindertest(3,true, ExecType.CP, BuiltinLmTest.LinregType.AUTO);\n+ }\n+\n+ private void runslicefindertest(int test,boolean sparse, ExecType instType, BuiltinLmTest.LinregType linregAlgo) {\n+ ExecMode platformOld = setExecMode(instType);\n+ String dml_test_name = TEST_NAME;\n+ loadTestConfiguration(getTestConfiguration(TEST_NAME));\n+ String HOME = SCRIPT_DIR + TEST_DIR;\n+ try {\n+ loadTestConfiguration(getTestConfiguration(TEST_NAME));\n+ fullDMLScriptName = HOME + dml_test_name + \".dml\";\n+ programArgs = new String[]{\"-explain\", \"-args\", input(\"AA\"), input(\"B\")};\n+ double[][] A = TestUtils.ceil(getRandomMatrix(rows, cols, 0, 10, 1, 7));\n+ double[][] B = TestUtils.ceil(getRandomMatrix(10, 1, 0, 10, 1.0, 3));\n+ double[][] As = new double[rows][cols];\n+ double [] Ys = new double[rows];\n+ double Y[] = new double[rows];\n+\n+ //Y = X %*% B\n+ for (int i = 0; i < rows; i++)\n+ for (int k = 0; k < cols; k++)\n+ Y[i] += A[i][k] * B[k][0];\n+\n+ double AA[][] = new double[rows][cols+1];\n+\n+ switch (test) {\n+ case 1:\n+ AA = modifyValue(A, Y,7,5);\n+ break;\n+ case 2:\n+ AA = modifyValue(A, Y, 6, 3);\n+ for(int i = 0;i<rows;i++){\n+ for(int j = 0; j < cols+1;j++){\n+ if(j == cols )\n+ Ys[i] = (int) AA[i][j];\n+ else\n+ As[i][j] = AA[i][j];\n+ }\n+ }\n+ AA = modifyValue(As,Ys,3,3);\n+ break;\n+ case 3:\n+ AA = modifyValue(A, Y, 6, 3);\n+ for(int i = 0;i<rows;i++){\n+ for(int j = 0; j < cols+1;j++){\n+ if(j == cols ){\n+ Ys[i] = (int) AA[i][j];\n+ }else{\n+ As[i][j] = AA[i][j];\n+ }\n+ }\n+ }\n+ AA = modifyValue(As,Ys,3,7);\n+ break;\n+ }\n+\n+ writeInputMatrixWithMTD(\"AA\", AA, true);\n+ writeInputMatrixWithMTD(\"B\", B, true);\n+\n+ runTest(true, false, null, -1);\n+\n+ }\n+ finally {\n+ rtplatform = platformOld;\n+ }\n+ }\n+\n+ private double[][] randomizeArray(double[][]y){\n+ Random rgen=new Random();\n+ for(int i=0; i<y.length; i++){\n+ int randomPosition=rgen.nextInt(y.length);\n+ double temp=y[i][0];\n+ y[i][0]=y[randomPosition][0];\n+ y[randomPosition][0]=temp;\n+ }\n+ return y;\n+ }\n+\n+ private double[][] modifyValue(double[][] A, double[] Y, int value, int coll){\n+ int counter = 0;\n+ double nvec[][] = new double[rows][1];\n+ for (int i = 0; i < rows; i++) {\n+ if (A[i][coll] == value) {\n+ nvec[counter][0] = Y[i];\n+ counter++;\n+ }\n+ }\n+ double[][] y = new double[counter][1];\n+ for (int i = 0; i < counter; i++)\n+ y[i][0] = nvec[i][0];\n+\n+ double[][] yy = randomizeArray(y);\n+ double AA [][] = new double[rows][cols + 1];\n+ counter = 0;\n+\n+ for(int i = 0; i<rows; i++) {\n+ for(int j = 0; j < cols + 1;j++)\n+ AA[i][j] = (j == cols ) ? Y[i] : A[i][j];\n+ if(A[i][coll] == value) { // this condition changes the values you choose\n+ AA[i][10] = yy[counter][0];\n+ counter++;\n+ }\n+ }\n+ return AA;\n+ }\n+}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/builtin/slicefinder.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1);\n+w = read($2);\n+ress = slicefinder(X0 = X,W = w, k = 5);\n+\n+#TODO write and check automatically\n+print(toString(ress));\n\\ No newline at end of file\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -80,10 +80,12 @@ SYSTEMDS-110 New Builtin Functions\n* 112 Image data augmentation builtin functions OK\n* 113 Builtin functions for linear regression algorithms OK\n* 114 Builtin function for stepwise regression OK\n+ * 115 Builtin function for model debugging (slice finder) OK\nSYSTEMDS-120 Performance Features\n* 121 Avoid spark context creation on parfor result merge OK\n* 122 Reduce thread contention on parfor left indexing OK\n+ * 123 Avoid unnecessary spark context creation on explain OK\nSYSTEMDS-130 IPA and Size Propagation\n* 131 New IPA pass for function call forwarding OK\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/Hop.java", "new_path": "src/main/java/org/tugraz/sysds/hops/Hop.java", "diff": "@@ -28,6 +28,7 @@ import org.tugraz.sysds.common.Types.DataType;\nimport org.tugraz.sysds.common.Types.ExecMode;\nimport org.tugraz.sysds.common.Types.ValueType;\nimport org.tugraz.sysds.conf.ConfigurationManager;\n+import org.tugraz.sysds.hops.recompile.Recompiler;\nimport org.tugraz.sysds.hops.recompile.Recompiler.ResetType;\nimport org.tugraz.sysds.lops.Binary;\nimport org.tugraz.sysds.lops.BinaryScalar;\n@@ -51,6 +52,7 @@ import org.tugraz.sysds.runtime.controlprogram.parfor.util.IDSequence;\nimport org.tugraz.sysds.runtime.instructions.gpu.context.GPUContextPool;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport org.tugraz.sysds.runtime.meta.DataCharacteristics;\n+import org.tugraz.sysds.runtime.meta.MatrixCharacteristics;\nimport org.tugraz.sysds.runtime.util.UtilFunctions;\nimport java.util.ArrayList;\n@@ -313,22 +315,10 @@ public abstract class Hop implements ParseInfo\nprivate void constructAndSetCheckpointLopIfRequired() {\n//determine execution type\nExecType et = ExecType.CP;\n- if( OptimizerUtils.isSparkExecutionMode()\n- && getDataType()!=DataType.SCALAR )\n- {\n- //conditional checkpoint based on memory estimate in order to\n- //(1) avoid unnecessary persist and unpersist calls, and\n- //(2) avoid unnecessary creation of spark context (incl executors)\n- if( (OptimizerUtils.isHybridExecutionMode() && hasValidCPDimsAndSize()\n- && !OptimizerUtils.exceedsCachingThreshold(getDim2(), _outputMemEstimate))\n- || _etypeForced == ExecType.CP )\n- {\n- et = ExecType.CP;\n- }\n- else //default case\n- {\n- et = ExecType.SPARK;\n- }\n+ if( OptimizerUtils.isSparkExecutionMode() && getDataType()!=DataType.SCALAR ) {\n+ //conditional checkpoint based on memory estimate\n+ et = ( Recompiler.checkCPCheckpoint(getDataCharacteristics() )\n+ || _etypeForced == ExecType.CP ) ? ExecType.CP : ExecType.SPARK;\n}\n//add checkpoint lop to output if required\n@@ -932,6 +922,11 @@ public abstract class Hop implements ParseInfo\nreturn OptimizerUtils.getSparsity(_dim1, _dim2, _nnz);\n}\n+ public DataCharacteristics getDataCharacteristics() {\n+ return new MatrixCharacteristics(\n+ _dim1, _dim2, _rows_in_block, _cols_in_block, _nnz);\n+ }\n+\nprotected void setOutputDimensions(Lop lop) {\nlop.getOutputParameters().setDimensions(\ngetDim1(), getDim2(), getRowsInBlock(), getColsInBlock(), getNnz(), getUpdateType());\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/recompile/Recompiler.java", "new_path": "src/main/java/org/tugraz/sysds/hops/recompile/Recompiler.java", "diff": "@@ -1528,6 +1528,12 @@ public class Recompiler\nreturn (estFilesize < cpThreshold);\n}\n+ public static boolean checkCPCheckpoint(DataCharacteristics dc) {\n+ return OptimizerUtils.isHybridExecutionMode()\n+ && OptimizerUtils.isValidCPDimensions(dc.getRows(), dc.getCols())\n+ && !OptimizerUtils.exceedsCachingThreshold(dc.getCols(), OptimizerUtils.estimateSize(dc));\n+ }\n+\npublic static void executeInMemoryMatrixReblock(ExecutionContext ec, String varin, String varout) {\nMatrixObject in = ec.getMatrixObject(varin);\nMatrixObject out = ec.getMatrixObject(varout);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/CSVReblockSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/CSVReblockSPInstruction.java", "diff": "@@ -43,6 +43,7 @@ import org.tugraz.sysds.runtime.matrix.data.MatrixIndexes;\nimport org.tugraz.sysds.runtime.matrix.operators.Operator;\nimport org.tugraz.sysds.runtime.meta.DataCharacteristics;\nimport org.tugraz.sysds.runtime.meta.MetaDataFormat;\n+import org.tugraz.sysds.utils.Statistics;\npublic class CSVReblockSPInstruction extends UnarySPInstruction {\nprivate int _brlen;\n@@ -109,6 +110,7 @@ public class CSVReblockSPInstruction extends UnarySPInstruction {\nRecompiler.executeInMemoryMatrixReblock(sec, input1.getName(), output.getName());\nelse if( input1.getDataType() == DataType.FRAME )\nRecompiler.executeInMemoryFrameReblock(sec, input1.getName(), output.getName());\n+ Statistics.decrementNoOfExecutedSPInst();\nreturn;\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/CheckpointSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/CheckpointSPInstruction.java", "diff": "@@ -25,6 +25,7 @@ import org.apache.spark.api.java.JavaPairRDD;\nimport org.apache.spark.storage.StorageLevel;\nimport org.tugraz.sysds.common.Types.DataType;\nimport org.tugraz.sysds.hops.OptimizerUtils;\n+import org.tugraz.sysds.hops.recompile.Recompiler;\nimport org.tugraz.sysds.lops.Checkpoint;\nimport org.tugraz.sysds.runtime.controlprogram.caching.CacheableData;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\n@@ -44,6 +45,7 @@ import org.tugraz.sysds.runtime.matrix.data.MatrixIndexes;\nimport org.tugraz.sysds.runtime.matrix.operators.Operator;\nimport org.tugraz.sysds.runtime.meta.DataCharacteristics;\nimport org.tugraz.sysds.runtime.util.UtilFunctions;\n+import org.tugraz.sysds.utils.Statistics;\npublic class CheckpointSPInstruction extends UnarySPInstruction {\n// default storage level\n@@ -86,14 +88,15 @@ public class CheckpointSPInstruction extends UnarySPInstruction {\n//(for csv input files with unknown dimensions, we might have generated a checkpoint after\n//csvreblock although not necessary because the csvreblock was subject to in-memory reblock)\nCacheableData<?> obj = sec.getCacheableData(input1.getName());\n- if( obj.isCached(true) ) { //available in memory\n+ DataCharacteristics mcIn = sec.getDataCharacteristics( input1.getName() );\n+ if( obj.isCached(true) || Recompiler.checkCPCheckpoint(mcIn) ) { //available in memory\nsec.setVariable(output.getName(), obj);\n+ Statistics.decrementNoOfExecutedSPInst();\nreturn;\n}\n//get input rdd handle (for matrix or frame)\nJavaPairRDD<?,?> in = sec.getRDDHandleForVariable(input1.getName(), InputInfo.BinaryBlockInputInfo, -1, true);\n- DataCharacteristics mcIn = sec.getDataCharacteristics( input1.getName() );\n// Step 2: Checkpoint given rdd (only if currently in different storage level to prevent redundancy)\n// -------\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/utils/Explain.java", "new_path": "src/main/java/org/tugraz/sysds/utils/Explain.java", "diff": "@@ -60,6 +60,7 @@ import org.tugraz.sysds.runtime.instructions.Instruction;\nimport org.tugraz.sysds.runtime.instructions.cp.CPInstruction;\nimport org.tugraz.sysds.runtime.instructions.gpu.GPUInstruction;\nimport org.tugraz.sysds.runtime.instructions.spark.CSVReblockSPInstruction;\n+import org.tugraz.sysds.runtime.instructions.spark.CheckpointSPInstruction;\nimport org.tugraz.sysds.runtime.instructions.spark.ReblockSPInstruction;\nimport org.tugraz.sysds.runtime.instructions.spark.SPInstruction;\nimport org.tugraz.sysds.runtime.lineage.LineageItem;\n@@ -93,6 +94,7 @@ public class Explain\npublic int numCPInst = 0;\npublic int numJobs = 0;\npublic int numReblocks = 0;\n+ public int numChkpts = 0;\n}\n//////////////\n@@ -113,18 +115,16 @@ public class Explain\nreturn explainMemoryBudget(new ExplainCounts());\n}\n- public static String explainMemoryBudget(ExplainCounts counts)\n- {\n+ public static String explainMemoryBudget(ExplainCounts counts) {\nStringBuilder sb = new StringBuilder();\n-\nsb.append( \"# Memory Budget local/remote = \" );\nsb.append( OptimizerUtils.toMB(OptimizerUtils.getLocalMemBudget()) );\nsb.append( \"MB/\" );\n- if( OptimizerUtils.isSparkExecutionMode() )\n- {\n- if( counts.numJobs-counts.numReblocks == 0 ) {\n+ if( OptimizerUtils.isSparkExecutionMode() ) {\n//avoid unnecessary lazy spark context creation on access to memory configurations\n+ if( counts.numJobs-counts.numReblocks-counts.numChkpts <= 0\n+ || !SparkExecutionContext.isSparkContextCreated() ) {\nsb.append( \"?MB/?MB/?MB\" );\n}\nelse { //default\n@@ -136,8 +136,7 @@ public class Explain\nsb.append( \"MB\" );\n}\n}\n- else\n- {\n+ else {\nsb.append( OptimizerUtils.toMB(OptimizerUtils.getRemoteMemBudgetMap()) );\nsb.append( \"MB/\" );\nsb.append( OptimizerUtils.toMB(OptimizerUtils.getRemoteMemBudgetReduce()) );\n@@ -147,23 +146,20 @@ public class Explain\nreturn sb.toString();\n}\n- public static String explainDegreeOfParallelism()\n- {\n+ public static String explainDegreeOfParallelism() {\nreturn explainDegreeOfParallelism(new ExplainCounts());\n}\n- public static String explainDegreeOfParallelism(ExplainCounts counts)\n- {\n+ public static String explainDegreeOfParallelism(ExplainCounts counts) {\nint lk = InfrastructureAnalyzer.getLocalParallelism();\n-\nStringBuilder sb = new StringBuilder();\nsb.append( \"# Degree of Parallelism (vcores) local/remote = \" );\nsb.append( lk );\nsb.append( \"/\" );\n- if( OptimizerUtils.isSparkExecutionMode() ) //SP\n- {\n- if( counts.numJobs-counts.numReblocks == 0 ) {\n+ if( OptimizerUtils.isSparkExecutionMode() ) {\n+ if( counts.numJobs-counts.numReblocks-counts.numChkpts <= 0\n+ || !SparkExecutionContext.isSparkContextCreated() ) {\n//avoid unnecessary lazy spark context creation on access to memory configurations\nsb.append( \"?\" );\n}\n@@ -179,8 +175,7 @@ public class Explain\nreturn explain(prog, rtprog, type, null);\n}\n- public static String explain(DMLProgram prog, Program rtprog, ExplainType type, ExplainCounts counts)\n- {\n+ public static String explain(DMLProgram prog, Program rtprog, ExplainType type, ExplainCounts counts) {\n//dispatch to individual explain utils\nswitch( type ) {\n//explain hops with stats\n@@ -246,7 +241,7 @@ public class Explain\nboolean sparkExec = OptimizerUtils.isSparkExecutionMode();\nif( counts == null ) {\ncounts = new ExplainCounts();\n- countCompiledInstructions(rtprog, counts, !sparkExec, true, sparkExec);\n+ countCompiledInstructions(rtprog, counts, true, sparkExec);\n}\nStringBuilder sb = new StringBuilder();\n@@ -428,11 +423,7 @@ public class Explain\n*/\npublic static ExplainCounts countDistributedOperations( Program rtprog ) {\nExplainCounts counts = new ExplainCounts();\n- if( OptimizerUtils.isSparkExecutionMode() )\n- Explain.countCompiledInstructions(rtprog, counts, false, true, true);\n- else\n- Explain.countCompiledInstructions(rtprog, counts, true, true, false);\n-\n+ Explain.countCompiledInstructions(rtprog, counts, true, true);\nreturn counts;\n}\n@@ -801,15 +792,15 @@ public class Explain\nreturn sb.toString();\n}\n- private static void countCompiledInstructions( Program rtprog, ExplainCounts counts, boolean MR, boolean CP, boolean SP )\n+ private static void countCompiledInstructions( Program rtprog, ExplainCounts counts, boolean CP, boolean SP )\n{\n//analyze DML-bodied functions\nfor( FunctionProgramBlock fpb : rtprog.getFunctionProgramBlocks().values() )\n- countCompiledInstructions( fpb, counts, MR, CP, SP );\n+ countCompiledInstructions( fpb, counts, CP, SP );\n//analyze main program\nfor( ProgramBlock pb : rtprog.getProgramBlocks() )\n- countCompiledInstructions( pb, counts, MR, CP, SP );\n+ countCompiledInstructions( pb, counts, CP, SP );\n}\n/**\n@@ -818,43 +809,42 @@ public class Explain\n*\n* @param pb program block\n* @param counts explain countst\n- * @param MR if true, count Hadoop instructions\n* @param CP if true, count CP instructions\n* @param SP if true, count Spark instructions\n*/\n- private static void countCompiledInstructions(ProgramBlock pb, ExplainCounts counts, boolean MR, boolean CP, boolean SP)\n+ private static void countCompiledInstructions(ProgramBlock pb, ExplainCounts counts, boolean CP, boolean SP)\n{\nif (pb instanceof WhileProgramBlock) {\nWhileProgramBlock tmp = (WhileProgramBlock)pb;\n- countCompiledInstructions(tmp.getPredicate(), counts, MR, CP, SP);\n+ countCompiledInstructions(tmp.getPredicate(), counts, CP, SP);\nfor (ProgramBlock pb2 : tmp.getChildBlocks())\n- countCompiledInstructions(pb2, counts, MR, CP, SP);\n+ countCompiledInstructions(pb2, counts, CP, SP);\n}\nelse if (pb instanceof IfProgramBlock) {\nIfProgramBlock tmp = (IfProgramBlock)pb;\n- countCompiledInstructions(tmp.getPredicate(), counts, MR, CP, SP);\n+ countCompiledInstructions(tmp.getPredicate(), counts, CP, SP);\nfor( ProgramBlock pb2 : tmp.getChildBlocksIfBody() )\n- countCompiledInstructions(pb2, counts, MR, CP, SP);\n+ countCompiledInstructions(pb2, counts, CP, SP);\nfor( ProgramBlock pb2 : tmp.getChildBlocksElseBody() )\n- countCompiledInstructions(pb2, counts, MR, CP, SP);\n+ countCompiledInstructions(pb2, counts, CP, SP);\n}\nelse if (pb instanceof ForProgramBlock) { //includes ParFORProgramBlock\nForProgramBlock tmp = (ForProgramBlock)pb;\n- countCompiledInstructions(tmp.getFromInstructions(), counts, MR, CP, SP);\n- countCompiledInstructions(tmp.getToInstructions(), counts, MR, CP, SP);\n- countCompiledInstructions(tmp.getIncrementInstructions(), counts, MR, CP, SP);\n+ countCompiledInstructions(tmp.getFromInstructions(), counts, CP, SP);\n+ countCompiledInstructions(tmp.getToInstructions(), counts, CP, SP);\n+ countCompiledInstructions(tmp.getIncrementInstructions(), counts, CP, SP);\nfor( ProgramBlock pb2 : tmp.getChildBlocks() )\n- countCompiledInstructions(pb2, counts, MR, CP, SP);\n+ countCompiledInstructions(pb2, counts, CP, SP);\n//additional parfor jobs counted during runtime\n}\nelse if ( pb instanceof FunctionProgramBlock ) {\nFunctionProgramBlock fpb = (FunctionProgramBlock)pb;\nfor( ProgramBlock pb2 : fpb.getChildBlocks() )\n- countCompiledInstructions(pb2, counts, MR, CP, SP);\n+ countCompiledInstructions(pb2, counts, CP, SP);\n}\nelse if( pb instanceof BasicProgramBlock ) {\nBasicProgramBlock bpb = (BasicProgramBlock) pb;\n- countCompiledInstructions(bpb.getInstructions(), counts, MR, CP, SP);\n+ countCompiledInstructions(bpb.getInstructions(), counts, CP, SP);\n}\n}\n@@ -867,15 +857,13 @@ public class Explain\n* list of instructions\n* @param counts\n* explain counts\n- * @param MR\n- * if true, count Hadoop instructions\n* @param CP\n* if true, count CP instructions\n* @param SP\n* if true, count Spark instructions and Spark reblock\n* instructions\n*/\n- private static void countCompiledInstructions( ArrayList<Instruction> instSet, ExplainCounts counts, boolean MR, boolean CP, boolean SP )\n+ private static void countCompiledInstructions( ArrayList<Instruction> instSet, ExplainCounts counts, boolean CP, boolean SP )\n{\nfor( Instruction inst : instSet )\n{\n@@ -887,6 +875,8 @@ public class Explain\n//keep track of reblocks (in order to prevent unnecessary spark context creation)\nif( SP && (inst instanceof CSVReblockSPInstruction || inst instanceof ReblockSPInstruction) )\ncounts.numReblocks++;\n+ if( SP && inst instanceof CheckpointSPInstruction )\n+ counts.numChkpts++;\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/utils/Statistics.java", "new_path": "src/main/java/org/tugraz/sysds/utils/Statistics.java", "diff": "@@ -178,6 +178,10 @@ public class Statistics\nnumCompiledSPInst.increment();\n}\n+ public static boolean createdSparkContext() {\n+ return sparkCtxCreateTime > 0;\n+ }\n+\npublic static long getTotalUIPVar() {\nreturn lTotalUIPVar.longValue();\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/AutomatedTestBase.java", "new_path": "src/test/java/org/tugraz/sysds/test/AutomatedTestBase.java", "diff": "@@ -245,6 +245,10 @@ public abstract class AutomatedTestBase\navailableTestConfigurations.put(testName, config);\n}\n+ protected void addTestConfiguration(TestConfiguration config) {\n+ availableTestConfigurations.put(config.getTestScript(), config);\n+ }\n+\n/**\n* <p>\n* Adds a test configuration to the list of available test configurations based\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/TestConfiguration.java", "new_path": "src/test/java/org/tugraz/sysds/test/TestConfiguration.java", "diff": "@@ -34,9 +34,6 @@ import java.util.HashMap;\n*/\npublic class TestConfiguration\n{\n-\n-\n-\n/** directory where the test can be found */\nprivate String testDirectory = null;\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/io/csv/CSVParametersTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/io/csv/CSVParametersTest.java", "diff": "@@ -195,5 +195,4 @@ public class CSVParametersTest extends AutomatedTestBase\ndouble dmlScalar = TestUtils.readDMLScalar(scalarFile);\nTestUtils.compareScalars(dmlScalar, 0.0, eps);\n}\n-\n}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/CSVReadInFunctionTest.java", "diff": "+/*\n+ * Modifications Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed to the Apache Software Foundation (ASF) under one\n+ * or more contributor license agreements. See the NOTICE file\n+ * distributed with this work for additional information\n+ * regarding copyright ownership. The ASF licenses this file\n+ * to you under the Apache License, Version 2.0 (the\n+ * \"License\"); you may not use this file except in compliance\n+ * with the License. You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing,\n+ * software distributed under the License is distributed on an\n+ * \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+ * KIND, either express or implied. See the License for the\n+ * specific language governing permissions and limitations\n+ * under the License.\n+ */\n+\n+package org.tugraz.sysds.test.functions.recompile;\n+\n+import java.util.HashMap;\n+\n+import org.junit.Assert;\n+import org.junit.Test;\n+import org.tugraz.sysds.common.Types.ValueType;\n+import org.tugraz.sysds.runtime.io.MatrixWriter;\n+import org.tugraz.sysds.runtime.io.MatrixWriterFactory;\n+import org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\n+import org.tugraz.sysds.runtime.matrix.data.OutputInfo;\n+import org.tugraz.sysds.runtime.matrix.data.MatrixValue.CellIndex;\n+import org.tugraz.sysds.runtime.util.DataConverter;\n+import org.tugraz.sysds.runtime.util.HDFSTool;\n+import org.tugraz.sysds.test.AutomatedTestBase;\n+import org.tugraz.sysds.test.TestConfiguration;\n+import org.tugraz.sysds.test.TestUtils;\n+import org.tugraz.sysds.utils.Statistics;\n+\n+public class CSVReadInFunctionTest extends AutomatedTestBase {\n+\n+ private final static String TEST_NAME1 = \"csv_read_function1\";\n+ private final static String TEST_NAME2 = \"csv_read_function2\";\n+ private final static String TEST_DIR = \"functions/recompile/\";\n+ private final static String TEST_CLASS_DIR = TEST_DIR + CSVReadInFunctionTest.class.getSimpleName() + \"/\";\n+\n+ private final static int rows = 123;\n+ private final static int cols = 45;\n+\n+ @Override\n+ public void setUp() {\n+ TestUtils.clearAssertionInformation();\n+ addTestConfiguration(new TestConfiguration(TEST_CLASS_DIR, TEST_NAME1, new String[]{\"X\"}));\n+ addTestConfiguration(new TestConfiguration(TEST_CLASS_DIR, TEST_NAME2, new String[]{\"X\"}));\n+ }\n+\n+ @Test\n+ public void testCSVReadNoFunctionNoMTD() {\n+ runCSVReadInFunctionTest(TEST_NAME1, false);\n+ }\n+\n+ @Test\n+ public void testCSVReadFunctionNoMTD() {\n+ runCSVReadInFunctionTest(TEST_NAME2, false);\n+ }\n+\n+ @Test\n+ public void testCSVReadNoFunctionMTD() {\n+ runCSVReadInFunctionTest(TEST_NAME1, true);\n+ }\n+\n+ @Test\n+ public void testCSVReadFunctionMTD() {\n+ runCSVReadInFunctionTest(TEST_NAME2, true);\n+ }\n+\n+ private void runCSVReadInFunctionTest(String testname, boolean withMtD) {\n+ try {\n+ getAndLoadTestConfiguration(testname);\n+\n+ String HOME = SCRIPT_DIR + TEST_DIR;\n+ fullDMLScriptName = HOME + testname + \".dml\";\n+ programArgs = new String[]{\"-stats\", \"-explain\",\n+ \"-args\", input(\"A\"), input(\"B\"), output(\"R\") };\n+\n+ //write csv matrix without size information (no mtd file)\n+ double[][] A = getRandomMatrix(rows, cols, -1, 1, 1.0d, 7);\n+ MatrixBlock mbA = DataConverter.convertToMatrixBlock(A);\n+ MatrixWriter writer = MatrixWriterFactory.createMatrixWriter(OutputInfo.CSVOutputInfo);\n+ writer.writeMatrixToHDFS(mbA, input(\"A\"), rows, cols, -1, -1, mbA.getNonZeros());\n+\n+ double[][] B = getRandomMatrix(rows, 1, -1, 1, 1.0d, 7);\n+ MatrixBlock mbB = DataConverter.convertToMatrixBlock(B);\n+ MatrixWriter writer2 = MatrixWriterFactory.createMatrixWriter(OutputInfo.CSVOutputInfo);\n+ writer2.writeMatrixToHDFS(mbB, input(\"B\"), rows, 1, -1, -1, mbB.getNonZeros());\n+\n+ if( withMtD ) {\n+ HDFSTool.writeMetaDataFile(input(\"A\")+\".mtd\", ValueType.FP64,\n+ mbA.getDataCharacteristics(), OutputInfo.CSVOutputInfo);\n+ HDFSTool.writeMetaDataFile(input(\"B\")+\".mtd\", ValueType.FP64,\n+ mbB.getDataCharacteristics(), OutputInfo.CSVOutputInfo);\n+ }\n+\n+ runTest(true, false, null, -1);\n+\n+ //compare matrices\n+ HashMap<CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(\"R\");\n+ Assert.assertEquals(dmlfile.get(new CellIndex(1,1)), new Double(mbA.sum()+mbB.sum()));\n+\n+ //check no executed spark instructions\n+ Assert.assertEquals(Statistics.getNoOfExecutedSPInst(), 0);\n+ Assert.assertTrue(!Statistics.createdSparkContext());\n+ }\n+ catch(Exception ex) {\n+ throw new RuntimeException(ex);\n+ }\n+ finally {\n+ try {\n+ HDFSTool.deleteFileIfExistOnHDFS(input(\"A\"));\n+ HDFSTool.deleteFileIfExistOnHDFS(input(\"A\")+\".mtd\");\n+ HDFSTool.deleteFileIfExistOnHDFS(input(\"B\"));\n+ HDFSTool.deleteFileIfExistOnHDFS(input(\"B\")+\".mtd\");\n+ } catch(Exception ex) {} //ignore\n+ }\n+ }\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/CSVReadUnknownSizeTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/CSVReadUnknownSizeTest.java", "diff": "@@ -139,19 +139,16 @@ public class CSVReadUnknownSizeTest extends AutomatedTestBase {\n//check expected number of compiled and executed MR jobs\n//note: with algebraic rewrites - unary op in reducer prevents job-level recompile\n- //TODO investigate current number of spark instructions\n- int expectedNumCompiled = (rewrites && !splitDags) ? 5 : 5; //reblock, GMR\n- int expectedNumExecuted = splitDags ? 2 : rewrites ? 5 : 5;\n+ int expectedNumCompiled = (rewrites && !splitDags) ? 5 : 5;\n+ int expectedNumExecuted = splitDags ? 0 : rewrites ? 3 : 3;\ncheckNumCompiledSparkInst(expectedNumCompiled);\ncheckNumExecutedSparkInst(expectedNumExecuted);\n}\n- catch(Exception ex)\n- {\n+ catch(Exception ex) {\nthrow new RuntimeException(ex);\n}\n- finally\n- {\n+ finally {\nOptimizerUtils.ALLOW_SPLIT_HOP_DAGS = oldFlagSplit;\nOptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = oldFlagRewrites;\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/PredicateRecompileTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/PredicateRecompileTest.java", "diff": "@@ -44,16 +44,11 @@ public class PredicateRecompileTest extends AutomatedTestBase\nprivate final static int val = 7;\n@Override\n- public void setUp()\n- {\n- addTestConfiguration(TEST_NAME1,\n- new TestConfiguration(TEST_CLASS_DIR, TEST_NAME1, new String[] { \"Rout\" }) );\n- addTestConfiguration(TEST_NAME2,\n- new TestConfiguration(TEST_CLASS_DIR, TEST_NAME2, new String[] { \"Rout\" }) );\n- addTestConfiguration(TEST_NAME3,\n- new TestConfiguration(TEST_CLASS_DIR, TEST_NAME3, new String[] { \"Rout\" }) );\n- addTestConfiguration(TEST_NAME4,\n- new TestConfiguration(TEST_CLASS_DIR, TEST_NAME4, new String[] { \"Rout\" }) );\n+ public void setUp() {\n+ addTestConfiguration(new TestConfiguration(TEST_CLASS_DIR, TEST_NAME1, new String[] { \"Rout\" }) );\n+ addTestConfiguration(new TestConfiguration(TEST_CLASS_DIR, TEST_NAME2, new String[] { \"Rout\" }) );\n+ addTestConfiguration(new TestConfiguration(TEST_CLASS_DIR, TEST_NAME3, new String[] { \"Rout\" }) );\n+ addTestConfiguration(new TestConfiguration(TEST_CLASS_DIR, TEST_NAME4, new String[] { \"Rout\" }) );\n}\n@Test\n@@ -285,8 +280,8 @@ public class PredicateRecompileTest extends AutomatedTestBase\nif( IPA ) {\nint expected = (testname.equals(TEST_NAME1) ?\n4 - ((evalExpr||constFold)?4:0) :\n- 3 - ((evalExpr||constFold)?3:0))\n- + ((!testname.equals(TEST_NAME2)&&!(evalExpr||constFold))?1:0); //loop checkpoint\n+ 3 - ((evalExpr||constFold)?3:0));\n+ //+ ((!testname.equals(TEST_NAME2)&&!(evalExpr||constFold))?1:0); //loop checkpoint\nAssert.assertEquals(\"Unexpected number of executed Spark instructions.\",\nexpected, Statistics.getNoOfExecutedSPInst());\n}\n@@ -294,8 +289,8 @@ public class PredicateRecompileTest extends AutomatedTestBase\n//old expected numbers before IPA\nint expected = (testname.equals(TEST_NAME1) ?\n4 - ((evalExpr||constFold)?1:0) :\n- 3 - ((evalExpr||constFold)?1:0))\n- + (!testname.equals(TEST_NAME2)?1:0); //loop checkpoint\n+ 3 - ((evalExpr||constFold)?1:0));\n+ //+ (!testname.equals(TEST_NAME2)?1:0); //loop checkpoint\nAssert.assertEquals(\"Unexpected number of executed Spark instructions.\",\nexpected, Statistics.getNoOfExecutedSPInst());\n}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/recompile/csv_read_function1.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Licensed to the Apache Software Foundation (ASF) under one\n+# or more contributor license agreements. See the NOTICE file\n+# distributed with this work for additional information\n+# regarding copyright ownership. The ASF licenses this file\n+# to you under the Apache License, Version 2.0 (the\n+# \"License\"); you may not use this file except in compliance\n+# with the License. You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing,\n+# software distributed under the License is distributed on an\n+# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+# KIND, either express or implied. See the License for the\n+# specific language governing permissions and limitations\n+# under the License.\n+#\n+#-------------------------------------------------------------\n+\n+\n+A = read($1, data_type=\"matrix\", format=\"csv\");\n+B = read($2, data_type=\"matrix\", format=\"csv\");\n+\n+R = as.matrix(sum(A) + sum(B));\n+\n+write(R, $3);\n+\n\\ No newline at end of file\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/recompile/csv_read_function2.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Licensed to the Apache Software Foundation (ASF) under one\n+# or more contributor license agreements. See the NOTICE file\n+# distributed with this work for additional information\n+# regarding copyright ownership. The ASF licenses this file\n+# to you under the Apache License, Version 2.0 (the\n+# \"License\"); you may not use this file except in compliance\n+# with the License. You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing,\n+# software distributed under the License is distributed on an\n+# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+# KIND, either express or implied. See the License for the\n+# specific language governing permissions and limitations\n+# under the License.\n+#\n+#-------------------------------------------------------------\n+\n+readCSV = function(Integer N, Integer M, Double sp) return(Matrix[Double] A) {\n+ if( N == 1 )\n+ A = read($1, data_type=\"matrix\", format=\"csv\");\n+ else\n+ A = read($2, data_type=\"matrix\", format=\"csv\");\n+}\n+\n+A = readCSV(1, -1, 1.0);\n+B = readCSV(2, -1, 1.0);\n+\n+R = as.matrix(sum(A) + sum(B));\n+\n+write(R, $3);\n+\n\\ No newline at end of file\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/utils/Statistics.java", "new_path": "src/main/java/org/tugraz/sysds/utils/Statistics.java", "diff": "@@ -441,6 +441,8 @@ public class Statistics\nparforInitTime = 0;\nparforMergeTime = 0;\n+ sparkCtxCreateTime = 0;\n+\nlTotalLix.reset();\nlTotalLixUIP.reset();\nlTotalUIPVar.reset();\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/CSVReadInFunctionTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/recompile/CSVReadInFunctionTest.java", "diff": "@@ -78,6 +78,7 @@ public class CSVReadInFunctionTest extends AutomatedTestBase {\nprivate void runCSVReadInFunctionTest(String testname, boolean withMtD) {\ntry {\ngetAndLoadTestConfiguration(testname);\n+ Statistics.reset();\nString HOME = SCRIPT_DIR + TEST_DIR;\nfullDMLScriptName = HOME + testname + \".dml\";\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/transform/TransformCSVFrameEncodeDecodeTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/transform/TransformCSVFrameEncodeDecodeTest.java", "diff": "@@ -64,12 +64,7 @@ public class TransformCSVFrameEncodeDecodeTest extends AutomatedTestBase\npublic void testHomesRecodeIDsHybridCSV() {\nrunTransformTest(ExecMode.HYBRID, \"csv\");\n}\n- /**\n- *\n- * @param rt\n- * @param ofmt\n- * @param dataset\n- */\n+\nprivate void runTransformTest( ExecMode rt, String ofmt )\n{\n//set runtime platform\n@@ -106,7 +101,7 @@ public class TransformCSVFrameEncodeDecodeTest extends AutomatedTestBase\nif( rt == ExecMode.HYBRID ) {\nAssert.assertEquals(\"Wrong number of executed Spark instructions: \" +\n- Statistics.getNoOfExecutedSPInst(), new Long(2), new Long(Statistics.getNoOfExecutedSPInst()));\n+ Statistics.getNoOfExecutedSPInst(), new Long(0), new Long(Statistics.getNoOfExecutedSPInst()));\n}\n}\ncatch(Exception ex) {\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/transform/TransformFrameEncodeApplyTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/transform/TransformFrameEncodeApplyTest.java", "diff": "@@ -335,7 +335,7 @@ public class TransformFrameEncodeApplyTest extends AutomatedTestBase\nif( rt == ExecMode.HYBRID ) {\nAssert.assertEquals(\"Wrong number of executed Spark instructions: \" +\n- Statistics.getNoOfExecutedSPInst(), new Long(2), new Long(Statistics.getNoOfExecutedSPInst()));\n+ Statistics.getNoOfExecutedSPInst(), new Long(0), new Long(Statistics.getNoOfExecutedSPInst()));\n}\n//additional checks for binning as encode-decode impossible\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/transform/TransformFrameEncodeDecodeTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/transform/TransformFrameEncodeDecodeTest.java", "diff": "@@ -182,7 +182,7 @@ public class TransformFrameEncodeDecodeTest extends AutomatedTestBase\nif( rt == ExecMode.HYBRID ) {\nAssert.assertEquals(\"Wrong number of executed Spark instructions: \" +\n- Statistics.getNoOfExecutedSPInst(), new Long(2), new Long(Statistics.getNoOfExecutedSPInst()));\n+ Statistics.getNoOfExecutedSPInst(), new Long(0), new Long(Statistics.getNoOfExecutedSPInst()));\n}\n}\ncatch(Exception ex) {\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/transform/TransformFrameEncodeDecodeTokenTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/transform/TransformFrameEncodeDecodeTokenTest.java", "diff": "@@ -70,12 +70,6 @@ public class TransformFrameEncodeDecodeTokenTest extends AutomatedTestBase\nrunTransformTest(ExecMode.HYBRID, \"csv\");\n}\n- /**\n- *\n- * @param rt\n- * @param ofmt\n- * @param dataset\n- */\nprivate void runTransformTest( ExecMode rt, String ofmt )\n{\n//set runtime platform\n@@ -116,7 +110,7 @@ public class TransformFrameEncodeDecodeTokenTest extends AutomatedTestBase\nif( rt == ExecMode.HYBRID ) {\nAssert.assertEquals(\"Wrong number of executed Spark instructions: \" +\n- Statistics.getNoOfExecutedSPInst(), new Long(2), new Long(Statistics.getNoOfExecutedSPInst()));\n+ Statistics.getNoOfExecutedSPInst(), new Long(0), new Long(Statistics.getNoOfExecutedSPInst()));\n}\n}\ncatch(Exception ex) {\n" } ]

[ { "change_type": "MODIFY", "old_path": ".gitignore", "new_path": ".gitignore", "diff": "# Mac\n.DS_Store\n+# KDE\n+.directory\n+\n# Eclipse\n.classpath\n.project\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/utils/NativeHelper.java", "new_path": "src/main/java/org/tugraz/sysds/utils/NativeHelper.java", "diff": "@@ -96,8 +96,8 @@ public class NativeHelper {\n/**\n* Initialize the native library before executing the DML program\n*\n- * @param customLibPath specified by sysml.native.blas.directory\n- * @param userSpecifiedBLAS specified by sysml.native.blas\n+ * @param customLibPath specified by sysds.native.blas.directory\n+ * @param userSpecifiedBLAS specified by sysds.native.blas\n*/\npublic static void initialize(String customLibPath, String userSpecifiedBLAS) {\nif(isBLASLoaded() && isSupportedBLAS(userSpecifiedBLAS) && !blasType.equalsIgnoreCase(userSpecifiedBLAS)) {\n@@ -117,7 +117,7 @@ public class NativeHelper {\n/**\n* Return true if the given BLAS type is supported.\n*\n- * @param userSpecifiedBLAS BLAS type specified via sysml.native.blas property\n+ * @param userSpecifiedBLAS BLAS type specified via sysds.native.blas property\n* @return true if the userSpecifiedBLAS is auto | mkl | openblas, else false\n*/\nprivate static boolean isSupportedBLAS(String userSpecifiedBLAS) {\n@@ -167,7 +167,7 @@ public class NativeHelper {\nif(!SystemUtils.IS_OS_LINUX)\nreturn;\n- // attemptedLoading variable ensures that we don't try to load SystemML and other dependencies\n+ // attemptedLoading variable ensures that we don't try to load SystemDS and other dependencies\n// again and again especially in the parfor (hence the double-checking with synchronized).\nif(shouldReload(customLibPath) && isSupportedBLAS(userSpecifiedBLAS) && isSupportedArchitecture()) {\nlong start = System.nanoTime();\n@@ -181,7 +181,7 @@ public class NativeHelper {\n}\n- if(checkAndLoadBLAS(customLibPath, blas) && loadLibraryHelper(\"libsystemml_\" + blasType + \"-Linux-x86_64.so\")) {\n+ if(checkAndLoadBLAS(customLibPath, blas) && loadLibraryHelper(\"libsystemds_\" + blasType + \"-Linux-x86_64.so\")) {\nLOG.info(\"Using native blas: \" + blasType + getNativeBLASPath());\nCURRENT_NATIVE_BLAS_STATE = NativeBlasState.SUCCESSFULLY_LOADED_NATIVE_BLAS_AND_IN_USE;\n}\n@@ -192,7 +192,7 @@ public class NativeHelper {\nLOG.warn(\"Time to load native blas: \" + timeToLoadInMilliseconds + \" milliseconds.\");\n}\nelse if(LOG.isDebugEnabled() && !isSupportedBLAS(userSpecifiedBLAS)) {\n- LOG.debug(\"Using internal Java BLAS as native BLAS support the configuration 'sysml.native.blas'=\" + userSpecifiedBLAS + \".\");\n+ LOG.debug(\"Using internal Java BLAS as native BLAS support the configuration 'sysds.native.blas'=\" + userSpecifiedBLAS + \".\");\n}\n}\n@@ -207,7 +207,7 @@ public class NativeHelper {\nisLoaded = loadBLAS(customLibPath, \"mkl_rt\", null);\n}\nelse if(blas.equalsIgnoreCase(\"openblas\")) {\n- boolean isGompLoaded = loadBLAS(customLibPath, \"gomp\", \"gomp required for loading OpenBLAS-enabled SystemML library\");\n+ boolean isGompLoaded = loadBLAS(customLibPath, \"gomp\", \"gomp required for loading OpenBLAS-enabled SystemDS library\");\nif(isGompLoaded) {\nisLoaded = loadBLAS(customLibPath, \"openblas\", null);\n}\n" } ]

[ { "change_type": "ADD", "old_path": null, "new_path": "bin/sparkDML2.sh", "diff": "+#!/bin/bash\n+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+#set -x\n+\n+# This script is a simplified version of sparkDML.sh in order to\n+# allow a simple drop-in replacement for 'hadoop jar' without\n+# the need to change any command line arguments.\n+\n+#export HADOOP_CONF_DIR=/etc/hadoop/conf\n+#SPARK_HOME=../spark-2.3.1-bin-hadoop2.7\n+#export HADOOP_HOME=${HADOOP_HOME:-/usr/hdp/2.5.0.0-1245/hadoop}\n+#HADOOP_CONF_DIR=${HADOOP_CONF_DIR:-/usr/hdp/2.5.0.0-1245/hadoop/conf}\n+\n+export SPARK_MAJOR_VERSION=2\n+\n+#$SPARK_HOME/bin/spark-submit \\\n+spark-submit \\\n+ --master yarn \\\n+ --driver-memory 80g \\\n+ --num-executors 1 \\\n+ --executor-memory 60g \\\n+ --executor-cores 19 \\\n+ --conf \"spark.yarn.am.extraJavaOptions -Dhdp.version=2.5.0.0-1245\" \\\n+ \"$@\"\n+\n+# # run spark submit locally\n+# spark-submit \\\n+# \"$@\"\n" }, { "change_type": "ADD", "old_path": null, "new_path": "bin/systemds-standalone.py", "diff": "+#!/usr/bin/env python\n+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+import os\n+import sys\n+from os.path import join\n+import argparse\n+import platform\n+from utils import get_env_systemds_root, find_dml_file, log4j_path, config_path\n+\n+\n+def default_classpath(systemds_root):\n+ \"\"\"\n+ Classpath information required for excution\n+\n+ return: String\n+ Classpath location of build, library and hadoop directories\n+ \"\"\"\n+ build_lib = join(systemds_root, 'target', '*')\n+ lib_lib = join(systemds_root, 'target', 'lib', '*')\n+ hadoop_lib = join(systemds_root, 'target', 'lib', 'hadoop', '*')\n+ sysds_jar = join(systemds_root, 'target', 'SystemDS.jar')\n+ return build_lib, lib_lib, hadoop_lib, sysds_jar\n+\n+\n+def standalone_execution_entry(nvargs, args, config, explain, debug, stats, gpu, heapmem, f):\n+ \"\"\"\n+ This function is responsible for the execution of arguments via\n+ subprocess call in singlenode mode\n+ \"\"\"\n+\n+ systemds_root = get_env_systemds_root()\n+ script_file = find_dml_file(systemds_root, f)\n+\n+ if platform.system() == 'Windows':\n+ default_cp = ';'.join(default_classpath(systemds_root))\n+ else:\n+ default_cp = ':'.join(default_classpath(systemds_root))\n+\n+ java_memory = '-Xmx' + heapmem + ' -Xms4g -Xmn1g'\n+\n+ # Log4j\n+ log4j = log4j_path(systemds_root)\n+ log4j_properties_path = '-Dlog4j.configuration=file:{}'.format(log4j)\n+\n+ # Config\n+ if config is None:\n+ default_config = config_path(systemds_root)\n+ else:\n+ default_config = config\n+\n+ ds_options = []\n+ if nvargs is not None:\n+ ds_options.append('-nvargs')\n+ ds_options.append(' '.join(nvargs))\n+ if args is not None:\n+ ds_options.append('-args')\n+ ds_options.append(' '.join(args))\n+ if explain is not None:\n+ ds_options.append('-explain')\n+ ds_options.append(explain)\n+ if debug is not False:\n+ ds_options.append('-debug')\n+ if stats is not None:\n+ ds_options.append('-stats')\n+ ds_options.append(stats)\n+ if gpu is not None:\n+ ds_options.append('-gpu')\n+ ds_options.append(gpu)\n+\n+ os.environ['HADOOP_HOME'] = '/tmp/systemds'\n+\n+ cmd = ['java', java_memory, log4j_properties_path,\n+ '-cp', default_cp, 'org.tugraz.sysds.api.DMLScript',\n+ '-f', script_file, '-exec', 'singlenode', '-config', default_config,\n+ ' '.join(ds_options)]\n+\n+ cmd = ' '.join(cmd)\n+ print(cmd)\n+\n+ return_code = os.system(cmd)\n+ return return_code\n+\n+\n+if __name__ == '__main__':\n+\n+ fn = sys.argv[0]\n+ if os.path.exists(fn):\n+ #print(os.path.basename(fn))\n+ print(fn[:fn.rfind('/')])\n+\n+ cparser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter,\n+ description='System-DS Standalone Script')\n+\n+ # SYSTEM-DS Options\n+ cparser.add_argument('-nvargs', help='List of attributeName-attributeValue pairs', nargs='+', metavar='')\n+ cparser.add_argument('-args', help='List of positional argument values', metavar='', nargs='+')\n+ cparser.add_argument('-config', help='System-DS configuration file (e.g SystemDS-config.xml)', metavar='')\n+ cparser.add_argument('-explain', help='explains plan levels can be hops, runtime, '\n+ 'recompile_hops, recompile_runtime', nargs='?', const='runtime', metavar='')\n+ cparser.add_argument('-debug', help='runs in debug mode', action='store_true')\n+ cparser.add_argument('-stats', help='Monitor and report caching/recompilation statistics, '\n+ 'heavy hitter <count> is 10 unless overridden', nargs='?', const='10',\n+ metavar='')\n+ cparser.add_argument('-gpu', help='uses CUDA instructions when reasonable, '\n+ 'set <force> option to skip conservative memory estimates '\n+ 'and use GPU wherever possible', nargs='?')\n+ cparser.add_argument('-heapmem', help='maximum JVM heap memory', metavar='', default='8g')\n+ cparser.add_argument('-f', required=True, help='specifies dml file to execute; '\n+ 'path can be local/hdfs/gpfs', metavar='')\n+\n+ args = cparser.parse_args()\n+ arg_dict = vars(args)\n+ return_code = standalone_execution_entry(**arg_dict)\n+\n+ if return_code != 0:\n+ print('Failed to run SystemDS. Exit code :' + str(return_code))\n" }, { "change_type": "ADD", "old_path": null, "new_path": "bin/systemds-standalone.sh", "diff": "+#!/usr/bin/env bash\n+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+\n+# error help print\n+printSimpleUsage()\n+{\n+cat << EOF\n+Usage: $0 <dml-filename> [arguments] [-help]\n+ -help - Print detailed help message\n+EOF\n+ exit 1\n+}\n+\n+# Script internally invokes 'java -Xmx4g -Xms4g -Xmn400m [Custom-Java-Options] -jar StandaloneSystemDS.jar -f <dml-filename> -exec singlenode -config=SystemDS-config.xml [Optional-Arguments]'\n+\n+if [ -z \"$1\" ] ; then\n+ echo \"Wrong Usage.\";\n+ printSimpleUsage\n+fi\n+\n+if [ ! -z $SYSTEMDS_ROOT ]; then\n+ PROJECT_ROOT_DIR=\"$SYSTEMDS_ROOT\"\n+ echo \"SYTEMDS_ROOT is set to:\" $SYSTEMDS_ROOT\n+else\n+ # find the systemDS root path which contains the bin folder, the script folder and the target folder\n+ # tolerate path with spaces\n+ SCRIPT_DIR=$( dirname \"$0\" )\n+ PROJECT_ROOT_DIR=$( cd \"${SCRIPT_DIR}/..\" ; pwd -P )\n+fi\n+\n+USER_DIR=$PWD\n+\n+BUILD_DIR=${PROJECT_ROOT_DIR}/target\n+HADOOP_LIB_DIR=${BUILD_DIR}/lib\n+DML_SCRIPT_CLASS=${BUILD_DIR}/classes/org/tugraz/sysds/api/DMLScript.class\n+\n+BUILD_ERR_MSG=\"You must build the project before running this script.\"\n+BUILD_DIR_ERR_MSG=\"Could not find target directory \\\"${BUILD_DIR}\\\". ${BUILD_ERR_MSG}\"\n+HADOOP_LIB_ERR_MSG=\"Could not find required libraries \\\"${HADOOP_LIB_DIR}/*\\\". ${BUILD_ERR_MSG}\"\n+DML_SCRIPT_ERR_MSG=\"Could not find \\\"${DML_SCRIPT_CLASS}\\\". ${BUILD_ERR_MSG}\"\n+\n+# check if the project had been built and the jar files exist\n+if [ ! -d \"${BUILD_DIR}\" ]; then echo \"${BUILD_DIR_ERR_MSG}\"; exit 1; fi\n+if [ ! -d \"${HADOOP_LIB_DIR}\" ]; then echo \"${HADOOP_LIB_ERR_MSG}\"; exit 1; fi\n+if [ ! -f \"${DML_SCRIPT_CLASS}\" ]; then echo \"${DML_SCRIPT_ERR_MSG}\"; exit 1; fi\n+\n+\n+echo \"================================================================================\"\n+\n+# if the present working directory is the project root or bin folder, then use the temp folder as user.dir\n+if [ \"$USER_DIR\" = \"$PROJECT_ROOT_DIR\" ] || [ \"$USER_DIR\" = \"$PROJECT_ROOT_DIR/bin\" ]\n+then\n+ USER_DIR=${PROJECT_ROOT_DIR}/temp\n+ echo \"Output dir: $USER_DIR\"\n+fi\n+\n+\n+# if the SystemDS-config.xml does not exist, create it from the template\n+if [ ! -f \"${PROJECT_ROOT_DIR}/conf/SystemDS-config.xml\" ]\n+then\n+ cp \"${PROJECT_ROOT_DIR}/conf/SystemDS-config.xml.template\" \\\n+ \"${PROJECT_ROOT_DIR}/conf/SystemDS-config.xml\"\n+ echo \"... created ${PROJECT_ROOT_DIR}/conf/SystemDS-config.xml\"\n+fi\n+\n+# if the log4j.properties do not exis, create them from the template\n+if [ ! -f \"${PROJECT_ROOT_DIR}/conf/log4j.properties\" ]\n+then\n+ cp \"${PROJECT_ROOT_DIR}/conf/log4j.properties.template\" \\\n+ \"${PROJECT_ROOT_DIR}/conf/log4j.properties\"\n+ echo \"... created ${PROJECT_ROOT_DIR}/conf/log4j.properties\"\n+fi\n+\n+\n+\n+\n+# add hadoop libraries which were generated by the build to the classpath\n+CLASSPATH=\\\"${BUILD_DIR}/lib/*\\\"\n+\n+#SYSTEM_DS_JAR=$( find $PROJECT_ROOT_DIR/target/system-ds-*-SNAPSHOT.jar )\n+SYSTEM_DS_JAR=\\\"${BUILD_DIR}/classes\\\"\n+\n+CLASSPATH=${CLASSPATH}:${SYSTEM_DS_JAR}\n+\n+echo \"================================================================================\"\n+\n+# Set default Java options\n+SYSTEMDS_DEFAULT_JAVA_OPTS=\"\\\n+-Xmx8g -Xms4g -Xmn1g \\\n+-cp $CLASSPATH \\\n+-Dlog4j.configuration=file:'$PROJECT_ROOT_DIR/conf/log4j.properties' \\\n+-Duser.dir='$USER_DIR'\"\n+\n+# Add any custom Java options set by the user at command line, overriding defaults as necessary.\n+if [ ! -z \"${SYSTEMDS_JAVA_OPTS}\" ]; then\n+ SYSTEMDS_DEFAULT_JAVA_OPTS+=\" ${SYSTEMDS_JAVA_OPTS}\"\n+ unset SYSTEMDS_JAVA_OPTS\n+fi\n+\n+# Add any custom Java options set by the user in the environment variables file, overriding defaults as necessary.\n+if [ -f \"${PROJECT_ROOT_DIR}/conf/systemds-env.sh\" ]; then\n+ . \"${PROJECT_ROOT_DIR}/conf/systemds-env.sh\"\n+ if [ ! -z \"${SYSTEMDS_JAVA_OPTS}\" ]; then\n+ SYSTEMDS_DEFAULT_JAVA_OPTS+=\" ${SYSTEMDS_JAVA_OPTS}\"\n+ fi\n+fi\n+\n+\n+printUsageExit()\n+{\n+CMD=\"\\\n+java ${SYSTEMDS_DEFAULT_JAVA_OPTS} \\\n+org.tugraz.sysds.api.DMLScript \\\n+-help\"\n+# echo ${CMD}\n+eval ${CMD}\n+exit 0\n+}\n+\n+while getopts \"h:f:\" options; do\n+ case $options in\n+ h ) echo Warning: Help requested. Will exit after usage message\n+ printUsageExit\n+ ;;\n+ \\? ) echo Warning: Help requested. Will exit after usage message\n+ printUsageExit\n+ ;;\n+ f ) #echo \"Shifting args due to -f\"\n+ shift\n+ ;;\n+ * ) echo Error: Unexpected error while processing options\n+ esac\n+done\n+\n+# Peel off first argument so that $@ contains arguments to DML script\n+SCRIPT_FILE=$1\n+shift\n+\n+# if the script file path was omitted, try to complete the script path\n+if [ ! -f \"$SCRIPT_FILE\" ]\n+then\n+ SCRIPT_FILE_NAME=$(basename $SCRIPT_FILE)\n+ SCRIPT_FILE_FOUND=$(find \"$PROJECT_ROOT_DIR/scripts\" -name \"$SCRIPT_FILE_NAME\")\n+ if [ ! \"$SCRIPT_FILE_FOUND\" ]\n+ then\n+ echo \"Could not find DML script: $SCRIPT_FILE\"\n+ printSimpleUsage\n+ else\n+ SCRIPT_FILE=$SCRIPT_FILE_FOUND\n+ echo \"DML script: $SCRIPT_FILE\"\n+ fi\n+fi\n+\n+\n+# Invoke the jar with options and arguments\n+CMD=\"\\\n+java ${SYSTEMDS_DEFAULT_JAVA_OPTS} \\\n+org.tugraz.sysds.api.DMLScript \\\n+-f '$SCRIPT_FILE' \\\n+-exec singlenode \\\n+-config '$PROJECT_ROOT_DIR/conf/SystemDS-config.xml' \\\n+$@\"\n+\n+export HADOOP_HOME=/tmp/systemds\n+eval ${CMD}\n+\n+RETURN_CODE=$?\n+\n+# if there was an error, display the full java command (in case some of the variable substitutions broke it)\n+if [ $RETURN_CODE -ne 0 ]\n+then\n+ echo \"Failed to run SystemDS. Exit code: $RETURN_CODE\"\n+ LF=$'\\n'\n+\n+\n+ # keep empty lines above for the line breaks\n+ echo \" ${CMD// /$LF }\"\n+fi\n+\n" }, { "change_type": "ADD", "old_path": null, "new_path": "bin/utils.py", "diff": "+#!/usr/bin/env python\n+# -------------------------------------------------------------\n+#\n+# Licensed to the Apache Software Foundation (ASF) under one\n+# or more contributor license agreements. See the NOTICE file\n+# distributed with this work for additional information\n+# regarding copyright ownership. The ASF licenses this file\n+# to you under the Apache License, Version 2.0 (the\n+# \"License\"); you may not use this file except in compliance\n+# with the License. You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing,\n+# software distributed under the License is distributed on an\n+# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+# KIND, either express or implied. See the License for the\n+# specific language governing permissions and limitations\n+# under the License.\n+#\n+# -------------------------------------------------------------\n+\n+import sys\n+import os\n+from os.path import join, exists\n+from os import environ\n+import shutil\n+\n+\n+def get_env_systemds_root():\n+ \"\"\"\n+ Env variable error check and path location\n+\n+ return: String\n+ Location of SYSTEMDS_ROOT\n+ \"\"\"\n+ systemds_root = os.environ.get('SYSTEMDS_ROOT')\n+ if systemds_root is None:\n+ #print('SYSTEMDS_ROOT not found')\n+ #sys.exit()\n+ fn = sys.argv[0]\n+ systemds_root = fn[:fn.rfind('/')] + '/..'\n+\n+ return systemds_root\n+\n+\n+def get_env_spark_root():\n+ \"\"\"\n+ Env variable error check and path location\n+\n+ return: String\n+ Location of SPARK_ROOT\n+ \"\"\"\n+ spark_root = environ.get('SPARK_ROOT')\n+ if spark_root is None:\n+ print('SPARK_ROOT not found')\n+ sys.exit()\n+\n+ return spark_root\n+\n+\n+def find_file(name, path):\n+ \"\"\"\n+ Responsible for finding a specific file recursively given a location\n+ \"\"\"\n+ for root, dirs, files in os.walk(path):\n+ if name in files:\n+ return join(root, name)\n+\n+\n+def find_dml_file(systemds_root, script_file):\n+ \"\"\"\n+ Find the location of DML script being executed\n+\n+ return: String\n+ Location of the dml script\n+ \"\"\"\n+ scripts_dir = join(systemds_root, 'scripts')\n+ if not exists(script_file):\n+ script_file_path = find_file(script_file, scripts_dir)\n+ if script_file_path is not None:\n+ return script_file_path\n+ else:\n+ print('Could not find DML script: ' + script_file)\n+ sys.exit()\n+\n+ return script_file\n+\n+def log4j_path(systemds_root):\n+ \"\"\"\n+ Create log4j.properties from the template if not exist\n+\n+ return: String\n+ Location of log4j.properties path\n+ \"\"\"\n+ log4j_properties_path = join(systemds_root, 'conf', 'log4j.properties')\n+ log4j_template_properties_path = join(systemds_root, 'conf', 'log4j.properties.template')\n+ if not (exists(log4j_properties_path)):\n+ shutil.copyfile(log4j_template_properties_path, log4j_properties_path)\n+ print('... created ' + log4j_properties_path)\n+ return log4j_properties_path\n+\n+\n+def config_path(systemds_root):\n+ \"\"\"\n+ Create SystemDS-config from the template if not exist\n+\n+ return: String\n+ Location of SystemDS-config.xml\n+ \"\"\"\n+ systemds_config_path = join(systemds_root, 'conf', 'SystemDS-config.xml')\n+ systemds_template_config_path = join(systemds_root, 'conf', 'SystemDS-config.xml.template')\n+ if not (exists(systemds_config_path)):\n+ shutil.copyfile(systemds_template_config_path, systemds_config_path)\n+ print('... created ' + systemds_config_path)\n+ return systemds_config_path\n" }, { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -26,8 +26,8 @@ SYSTEMDS-20 New Data Model\n*\nSYSTEMDS-30 Builtin and Packaging\n- * 31 Shell script for local runs\n- * 32 Shell script for spark runs\n+ * 31 Shell script for local runs OK\n+ * 32 Shell script for spark runs OK\n* 33 Cleanup hadoop dependency for local runs\n* 34 Wrapper blocks for sequence files\n* 35 Replace unnecessary dependencies w/ custom\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/ProgramBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/ProgramBlock.java", "diff": "@@ -208,9 +208,9 @@ public abstract class ProgramBlock implements ParseInfo\nif( !LineageCache.reuse(tmp, ec) ) {\n// process actual instruction\ntmp.processInstruction(ec);\n+\n// cache result\nLineageCache.put(tmp, ec);\n- }\n// post-process instruction (debug)\ntmp.postprocessInstruction( ec );\n@@ -220,6 +220,7 @@ public abstract class ProgramBlock implements ParseInfo\nStatistics.maintainCPHeavyHitters(\ntmp.getExtendedOpcode(), System.nanoTime()-t0);\n}\n+ }\n// optional trace information (instruction and runtime)\nif( LOG.isTraceEnabled() ) {\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/Instruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/Instruction.java", "diff": "@@ -217,5 +217,6 @@ public abstract class Instruction\n* @param ec execution context\n*/\npublic void postprocessInstruction(ExecutionContext ec) {\n+ //do nothing\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "diff": "@@ -51,9 +51,7 @@ public class LineageCache {\npublic static void put(Instruction inst, ExecutionContext ec) {\nif (!DMLScript.LINEAGE_REUSE)\nreturn;\n- //FIXME: move applicability check in isolated mapping class\n- if (inst instanceof ComputationCPInstruction\n- && inst.getOpcode().equalsIgnoreCase(\"tsmm\")) {\n+ if (inst instanceof ComputationCPInstruction && isReusable(inst) ) {\nLineageItem[] items = ((LineageTraceable) inst).getLineageItems(ec);\nfor (LineageItem item : items) {\nMatrixObject mo = ec.getMatrixObject(((ComputationCPInstruction) inst).output);\n@@ -108,7 +106,7 @@ public class LineageCache {\nif (!DMLScript.LINEAGE_REUSE)\nreturn false;\n- if (inst instanceof ComputationCPInstruction && LineageCache.ifReusable(inst)) {\n+ if (inst instanceof ComputationCPInstruction && LineageCache.isReusable(inst)) {\nboolean reused = true;\nLineageItem[] items = ((ComputationCPInstruction) inst).getLineageItems(ec);\nfor (LineageItem item : items) {\n@@ -138,9 +136,10 @@ public class LineageCache {\nreturn readFromLocalFS(inst, key);\n}\n- public static boolean ifReusable (Instruction inst) {\n+ public static boolean isReusable (Instruction inst) {\n// TODO: Move this to the new class LineageCacheConfig and extend\n- return (inst.getOpcode().equalsIgnoreCase(\"tsmm\"));\n+ return (inst.getOpcode().equalsIgnoreCase(\"tsmm\")\n+ || inst.getOpcode().equalsIgnoreCase(\"ba+*\"));\n}\n//---------------- CACHE SPACE MANAGEMENT METHODS -----------------\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/matrix/data/LibCommonsMath.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/matrix/data/LibCommonsMath.java", "diff": "package org.tugraz.sysds.runtime.matrix.data;\nimport org.apache.commons.math3.linear.Array2DRowRealMatrix;\n+import org.apache.commons.math3.linear.BlockRealMatrix;\nimport org.apache.commons.math3.linear.CholeskyDecomposition;\nimport org.apache.commons.math3.linear.DecompositionSolver;\nimport org.apache.commons.math3.linear.EigenDecomposition;\n@@ -93,8 +94,10 @@ public class LibCommonsMath\n* @return matrix block\n*/\nprivate static MatrixBlock computeSolve(MatrixBlock in1, MatrixBlock in2) {\n- Array2DRowRealMatrix matrixInput = DataConverter.convertToArray2DRowRealMatrix(in1);\n- Array2DRowRealMatrix vectorInput = DataConverter.convertToArray2DRowRealMatrix(in2);\n+ //convert to commons math BlockRealMatrix instead of Array2DRowRealMatrix\n+ //to avoid unnecessary conversion as QR internally creates a BlockRealMatrix\n+ BlockRealMatrix matrixInput = DataConverter.convertToBlockRealMatrix(in1);\n+ BlockRealMatrix vectorInput = DataConverter.convertToBlockRealMatrix(in2);\n/*LUDecompositionImpl ludecompose = new LUDecompositionImpl(matrixInput);\nDecompositionSolver lusolver = ludecompose.getSolver();\n@@ -106,7 +109,7 @@ public class LibCommonsMath\n// Invoke solve\nRealMatrix solutionMatrix = solver.solve(vectorInput);\n- return DataConverter.convertToMatrixBlock(solutionMatrix.getData());\n+ return DataConverter.convertToMatrixBlock(solutionMatrix);\n}\n/**\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/util/DataConverter.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/util/DataConverter.java", "diff": "@@ -23,6 +23,8 @@ package org.tugraz.sysds.runtime.util;\nimport org.apache.commons.lang.StringUtils;\nimport org.apache.commons.math3.linear.Array2DRowRealMatrix;\n+import org.apache.commons.math3.linear.BlockRealMatrix;\n+import org.apache.commons.math3.linear.RealMatrix;\nimport org.tugraz.sysds.common.Types;\nimport org.tugraz.sysds.common.Types.ValueType;\nimport org.tugraz.sysds.runtime.DMLRuntimeException;\n@@ -232,14 +234,12 @@ public class DataConverter\n* @param mb matrix block\n* @return 2d double array\n*/\n- public static double[][] convertToDoubleMatrix( MatrixBlock mb )\n- {\n+ public static double[][] convertToDoubleMatrix( MatrixBlock mb ) {\nint rows = mb.getNumRows();\nint cols = mb.getNumColumns();\ndouble[][] ret = new double[rows][cols]; //0-initialized\n- if( mb.getNonZeros() > 0 )\n- {\n+ if( mb.getNonZeros() > 0 ) {\nif( mb.isInSparseFormat() ) {\nIterator<IJV> iter = mb.getSparseBlockIterator();\nwhile( iter.hasNext() ) {\n@@ -254,7 +254,6 @@ public class DataConverter\nret[i][j] = a[ix];\n}\n}\n-\nreturn ret;\n}\n@@ -830,6 +829,36 @@ public class DataConverter\nreturn new Array2DRowRealMatrix(data, false);\n}\n+ public static BlockRealMatrix convertToBlockRealMatrix(MatrixBlock mb) {\n+ BlockRealMatrix ret = new BlockRealMatrix(mb.getNumRows(), mb.getNumColumns());\n+ if( mb.getNonZeros() > 0 ) {\n+ if( mb.isInSparseFormat() ) {\n+ Iterator<IJV> iter = mb.getSparseBlockIterator();\n+ while( iter.hasNext() ) {\n+ IJV cell = iter.next();\n+ ret.setEntry(cell.getI(), cell.getJ(), cell.getV());\n+ }\n+ }\n+ else {\n+ double[] a = mb.getDenseBlockValues();\n+ for( int i=0, ix=0; i<mb.getNumRows(); i++ )\n+ for( int j=0; j<mb.getNumColumns(); j++, ix++ )\n+ ret.setEntry(i, j, a[ix]);\n+ }\n+ }\n+ return ret;\n+ }\n+\n+ public static MatrixBlock convertToMatrixBlock(RealMatrix rm) {\n+ MatrixBlock ret = new MatrixBlock(rm.getRowDimension(),\n+ rm.getColumnDimension(), false).allocateDenseBlock();\n+ for(int i=0; i<ret.getNumRows(); i++)\n+ for(int j=0; j<ret.getNumColumns(); j++)\n+ ret.quickSetValue(i, j, rm.getEntry(i, j));\n+ ret.examSparsity();\n+ return ret;\n+ }\n+\npublic static void copyToDoubleVector( MatrixBlock mb, double[] dest, int destPos )\n{\nif( mb.isEmptyBlock(false) )\n" } ]

[ { "change_type": "ADD", "old_path": null, "new_path": "scripts/builtin/kmeans.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+# Builtin function that implements the k-Means clustering algorithm\n+#\n+# INPUT PARAMETERS:\n+# ----------------------------------------------------------------------------\n+# NAME TYPE DEFAULT MEANING\n+# ----------------------------------------------------------------------------\n+# X String --- Location to read matrix X with the input data records\n+# k Int --- Number of centroids\n+# runs Int 10 Number of runs (with different initial centroids)\n+# max_iter Int 1000 Maximum number of iterations per run\n+# eps Double 0.000001 Tolerance (epsilon) for WCSS change ratio\n+# is_verbose Boolean FALSE do not print per-iteration stats\n+# avg_sample_size_per_centroid Int 50 Average number of records per centroid in data samples\n+#\n+#\n+# RETURN VALUES\n+# ----------------------------------------------------------------------------\n+# Y String \"Y.mtx\" Location to store the mapping of records to centroids\n+# C String \"C.mtx\" Location to store the output matrix with the centroids\n+# ----------------------------------------------------------------------------\n+\n+\n+m_kmeans = function(Matrix[Double] X, Integer k = 0, Integer runs = 10, Integer max_iter = 1000,\n+ Double eps = 0.000001, Boolean is_verbose = FALSE, Integer avg_sample_size_per_centroid = 50)\n+ return (Matrix[Double] C, Matrix[Double] Y)\n+{\n+ print (\"BEGIN K-MEANS SCRIPT\");\n+ num_records = nrow (X);\n+ num_features = ncol (X);\n+ num_centroids = k;\n+ num_runs = runs;\n+\n+ if(is_verbose == TRUE)\n+ print(\"dim X=\" + nrow(X) + \"x\" + ncol(X))\n+\n+ sumXsq = sum (X ^ 2);\n+\n+ # STEP 1: INITIALIZE CENTROIDS FOR ALL RUNS FROM DATA SAMPLES:\n+\n+ print (\"Taking data samples for initialization...\");\n+\n+ [sample_maps, samples_vs_runs_map, sample_block_size] =\n+ get_sample_maps (num_records, num_runs, num_centroids * avg_sample_size_per_centroid);\n+\n+ is_row_in_samples = rowSums (sample_maps);\n+ X_samples = sample_maps %*% X;\n+ X_samples_sq_norms = rowSums (X_samples ^ 2);\n+\n+ print (\"Initializing the centroids for all runs...\");\n+ All_Centroids = matrix (0, num_runs * num_centroids, num_features);\n+\n+ # We select centroids according to the k-Means++ heuristic applied to a sample of X\n+ # Loop invariant: min_distances ~ sq.distances from X_sample rows to nearest centroids,\n+ # with the out-of-range X_sample positions in min_distances set to 0.0\n+\n+ min_distances = is_row_in_samples; # Pick the 1-st centroids uniformly at random\n+\n+ for (i in 1 : num_centroids)\n+ {\n+ # \"Matricize\" and prefix-sum to compute the cumulative distribution function:\n+ min_distances_matrix_form = matrix (min_distances, rows = sample_block_size, cols = num_runs, byrow = FALSE);\n+ cdf_min_distances = cumsum (min_distances_matrix_form);\n+\n+ # Select the i-th centroid in each sample as a random sample row id with\n+ # probability ~ min_distances:\n+ random_row = rand(rows = 1, cols = num_runs);\n+ threshold_matrix = random_row * cdf_min_distances [sample_block_size, ];\n+ centroid_ids = t(colSums (cdf_min_distances < threshold_matrix)) + 1;\n+\n+ # Place the selected centroids together, one per run, into a matrix:\n+ centroid_placer = table (seq (1, num_runs),\n+ sample_block_size * seq (0, num_runs - 1) + centroid_ids, num_runs, sample_block_size * num_runs);\n+ centroids = centroid_placer %*% X_samples;\n+\n+ # Place the selected centroids into their appropriate slots in All_Centroids:\n+ centroid_placer = table (seq (i, num_centroids * (num_runs - 1) + i, num_centroids),\n+ seq (1, num_runs, 1), nrow (All_Centroids), num_runs);\n+ All_Centroids = All_Centroids + centroid_placer %*% centroids;\n+\n+ # Update min_distances to preserve the loop invariant:\n+ distances = X_samples_sq_norms + samples_vs_runs_map %*% rowSums (centroids ^ 2)\n+ - 2 * rowSums (X_samples * (samples_vs_runs_map %*% centroids));\n+ min_distances = ifelse(i==1, is_row_in_samples*distances, min(min_distances,distances));\n+ }\n+\n+ # STEP 2: PERFORM K-MEANS ITERATIONS FOR ALL RUNS:\n+\n+ termination_code = matrix (0, rows = num_runs, cols = 1);\n+ final_wcss = matrix (0, rows = num_runs, cols = 1);\n+ num_iterations = matrix (0, rows = num_runs, cols = 1);\n+\n+ print (\"Performing k-means iterations for all runs...\");\n+\n+ parfor (run_index in 1 : num_runs, check = 0)\n+ {\n+ C = All_Centroids [(num_centroids * (run_index - 1) + 1) : (num_centroids * run_index), ];\n+ C_old = C;\n+ iter_count = 0;\n+ term_code = 0;\n+ wcss = Inf\n+\n+ while (term_code == 0)\n+ {\n+ # Compute Euclidean squared distances from records (X rows) to centroids (C rows)\n+ # without the C-independent term, then take the minimum for each record\n+ D = -2 * (X %*% t(C)) + t(rowSums (C ^ 2));\n+ minD = rowMins (D);\n+ # Compute the current centroid-based within-cluster sum of squares (WCSS)\n+ wcss_old = wcss;\n+ wcss = sumXsq + sum (minD);\n+ if( is_verbose ) {\n+ if (iter_count == 0)\n+ print (\"Run \" + run_index + \", At Start-Up: Centroid WCSS = \" + wcss);\n+ else\n+ print (\"Run \" + run_index + \", Iteration \" + iter_count + \": Centroid WCSS = \" + wcss\n+ + \"; Centroid change (avg.sq.dist.) = \" + (sum ((C - C_old) ^ 2) / num_centroids));\n+ }\n+\n+ # Find the closest centroid for each record\n+ P = D <= minD;\n+ # If some records belong to multiple centroids, share them equally\n+ P = P / rowSums (P);\n+ # Compute the column normalization factor for P\n+ P_denom = colSums (P);\n+ # Compute new centroids as weighted averages over the records\n+ C_new = (t(P) %*% X) / t(P_denom);\n+\n+ # Check if convergence or maximum iteration has been reached\n+ iter_count = iter_count + 1\n+ if(wcss_old - wcss < eps * wcss)\n+ term_code = 1; # Convergence reached\n+ else if(iter_count >= max_iter)\n+ term_code = 2; # Max iteration reached\n+ else if(sum (P_denom <= 0) > 0)\n+ term_code = 3; # \"Runaway\" centroid (0.0 denom)\n+ else\n+ C_old = C; C = C_new;\n+ }\n+\n+ if(is_verbose == TRUE)\n+ print (\"Run \" + run_index + \", Iteration \" + iter_count + \": Terminated with code = \"\n+ + term_code + \", Centroid WCSS = \" + wcss);\n+\n+ All_Centroids [(num_centroids * (run_index - 1) + 1) : (num_centroids * run_index), ] = C;\n+ final_wcss [run_index, 1] = wcss;\n+ termination_code [run_index, 1] = term_code;\n+ num_iterations [run_index, 1] = iter_count;\n+ }\n+\n+ # STEP 3: SELECT THE RUN WITH BEST CENTROID-WCSS AND OUTPUT ITS CENTROIDS:\n+\n+ termination_bitmap = matrix (0, num_runs, 3);\n+ termination_bitmap_raw = table (seq (1, num_runs, 1), termination_code);\n+ termination_bitmap [, 1 : ncol(termination_bitmap_raw)] = termination_bitmap_raw;\n+ termination_stats = colSums (termination_bitmap);\n+\n+ print (\"Number of successful runs = \" + as.integer (as.scalar (termination_stats [1, 1])));\n+ print (\"Number of incomplete runs = \" + as.integer (as.scalar (termination_stats [1, 2])));\n+ print (\"Number of failed runs (with lost centroids) = \" + as.integer (as.scalar (termination_stats [1, 3])));\n+\n+ num_successful_runs = as.scalar (termination_stats [1, 1]);\n+\n+ if (num_successful_runs > 0)\n+ {\n+ final_wcss_successful = final_wcss * termination_bitmap [, 1];\n+ worst_wcss = max (final_wcss_successful);\n+ best_wcss = min (final_wcss_successful + (10 * worst_wcss + 10) * (1 - termination_bitmap [, 1]));\n+ avg_wcss = sum (final_wcss_successful) / num_successful_runs;\n+ best_index_vector = (final_wcss_successful == best_wcss);\n+ aggr_best_index_vector = cumsum (best_index_vector);\n+ best_index = as.integer (sum (aggr_best_index_vector == 0) + 1);\n+\n+ print (\"Successful runs: Best run is \" + best_index + \" with Centroid WCSS = \" + best_wcss\n+ + \"; Avg WCSS = \" + avg_wcss + \"; Worst WCSS = \" + worst_wcss);\n+\n+ C = All_Centroids [(num_centroids * (best_index - 1) + 1) : (num_centroids * best_index), ];\n+ D = -2 * (X %*% t(C)) + t(rowSums (C ^ 2));\n+ P = (D <= rowMins (D));\n+ aggr_P = t(cumsum (t(P)));\n+ Y = rowSums (aggr_P == 0) + 1\n+\n+ if(is_verbose == TRUE)\n+ print(\"dim C=\" + nrow(C) + \"x\" + ncol(C) + \", dim Y=\" + nrow(Y) + \"x\" + ncol(Y))\n+ print (\"DONE.\");\n+ }\n+ else\n+ stop (\"No output is produced. Try increasing the number of iterations and/or runs.\");\n+\n+}\n+\n+\n+get_sample_maps = function (int num_records, int num_samples, int approx_sample_size)\n+ return (Matrix[double] sample_maps, Matrix[double] sample_col_map, int sample_block_size)\n+{\n+ if (approx_sample_size < num_records)\n+ {\n+ # Input value \"approx_sample_size\" is the average sample size; increase it by ~10 std.dev's\n+ # to get the sample block size (to allocate space):\n+ sample_block_size = as.integer (approx_sample_size + round (10 * sqrt (approx_sample_size)));\n+ num_rows = sample_block_size * num_samples;\n+\n+ # Generate all samples in parallel by converting uniform random values into random\n+ # integer skip-ahead intervals and prefix-summing them:\n+ sample_rec_ids = Rand (rows = sample_block_size, cols = num_samples, min = 0.0, max = 1.0);\n+ sample_rec_ids = round (log (sample_rec_ids) / log (1.0 - approx_sample_size / num_records) + 0.5);\n+ # Prob [k-1 < log(uniform)/log(1-p) < k] = p*(1-p)^(k-1) = Prob [k-1 zeros before a one]\n+ sample_rec_ids = cumsum (sample_rec_ids); # (skip to next one) --> (skip to i-th one)\n+\n+ # Replace all sample record ids over \"num_records\" (i.e. out of range) by \"num_records + 1\":\n+ is_sample_rec_id_within_range = (sample_rec_ids <= num_records);\n+ sample_rec_ids = sample_rec_ids * is_sample_rec_id_within_range\n+ + (num_records + 1) * (1 - is_sample_rec_id_within_range);\n+\n+ # Rearrange all samples (and their out-of-range indicators) into one column-vector:\n+ sample_rec_ids = matrix (sample_rec_ids, rows = num_rows, cols = 1, byrow = FALSE);\n+ is_row_in_samples = matrix (is_sample_rec_id_within_range, rows = num_rows, cols = 1, byrow = FALSE);\n+\n+ # Use contingency table to create the \"sample_maps\" matrix that is a vertical concatenation\n+ # of 0-1-matrices, one per sample, each with 1s at (i, sample_record[i]) and 0s elsewhere:\n+ sample_maps = table (seq (1, num_rows), sample_rec_ids, num_rows, num_records);\n+\n+ # Create a 0-1-matrix that maps each sample column ID into all row positions of the\n+ # corresponding sample; map out-of-sample-range positions to row id = num_rows + 1:\n+ sample_positions = (num_rows + 1) - is_row_in_samples * seq (num_rows, 1, -1);\n+ # Column ID positions = 1, 1, ..., 1, 2, 2, ..., 2, . . . , n_c, n_c, ..., n_c:\n+ col_positions = round (0.5 + seq (0, num_rows - 1, 1) / sample_block_size);\n+ sample_col_map = table (sample_positions, col_positions);\n+ # Remove the out-of-sample-range positions by cutting off the last row:\n+ sample_col_map = sample_col_map [1 : (num_rows), ];\n+ }\n+ else {\n+ one_per_record = matrix (1, num_records, 1);\n+ sample_block_size = num_records;\n+ sample_maps = matrix (0, (num_records * num_samples), num_records);\n+ sample_col_map = matrix (0, (num_records * num_samples), num_samples);\n+ for (i in 1:num_samples) {\n+ sample_maps [(num_records * (i - 1) + 1) : (num_records * i), ] = diag (one_per_record);\n+ sample_col_map [(num_records * (i - 1) + 1) : (num_records * i), i] = one_per_record;\n+ }\n+ }\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "new_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "diff": "@@ -89,6 +89,7 @@ public enum Builtins {\nINTERQUANTILE(\"interQuantile\", false),\nINVERSE(\"inv\", \"inverse\", false),\nIQM(\"interQuartileMean\", false),\n+ KMEANS(\"kmeans\", true),\nLENGTH(\"length\", false),\nLINEAGE(\"lineage\", false),\nLIST(\"list\", false), //note: builtin and parbuiltin\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinKmeansTest.java", "diff": "+package org.tugraz.sysds.test.functions.builtin;\n+\n+import org.junit.Test;\n+import org.tugraz.sysds.api.DMLScript;\n+import org.tugraz.sysds.common.Types;\n+import org.tugraz.sysds.hops.OptimizerUtils;\n+import org.tugraz.sysds.lops.LopProperties;\n+import org.tugraz.sysds.test.AutomatedTestBase;\n+import org.tugraz.sysds.test.TestConfiguration;\n+import org.tugraz.sysds.test.TestUtils;\n+\n+public class BuiltinKmeansTest extends AutomatedTestBase\n+{\n+ private final static String TEST_NAME = \"kmeans\";\n+ private final static String TEST_DIR = \"functions/builtin/\";\n+ private static final String TEST_CLASS_DIR = TEST_DIR + BuiltinKmeansTest.class.getSimpleName() + \"/\";\n+ private final static double eps = 1e-10;\n+ private final static int rows = 3972;\n+ private final static int cols = 972;\n+ private final static double spSparse = 0.3;\n+ private final static double spDense = 0.7;\n+ private final static double max_iter = 10;\n+\n+ @Override\n+ public void setUp() {\n+ TestUtils.clearAssertionInformation();\n+ addTestConfiguration(TEST_NAME,new TestConfiguration(TEST_CLASS_DIR, TEST_NAME,new String[]{\"C\"}));\n+ }\n+\n+ @Test\n+ public void testKMeansDenseBinSingleRewritesCP() {\n+ runKMeansTest(false, 2, 1, true, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansSparseBinSingleRewritesCP() {\n+ runKMeansTest(true,2, 1, true, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansDenseBinSingleCP() {\n+ runKMeansTest(false,2, 1, false, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansSparseBinSingleCP() {\n+ runKMeansTest(true, 2, 1, false, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansDenseBinMultiRewritesCP() {\n+ runKMeansTest(false, 2, 10, true, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansSparseBinMultiRewritesCP() {\n+ runKMeansTest(true, 2, 10, true, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansDenseBinMultiCP() {\n+ runKMeansTest(false, 2, 10, false, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansSparseBinMultiCP() {\n+ runKMeansTest(true, 2, 10, false, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansDenseMulSingleRewritesCP() {\n+ runKMeansTest(false, 20, 1, true, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansSparseMulSingleRewritesCP() {\n+ runKMeansTest(true, 20, 1, true, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansDenseMulSingleCP() {\n+ runKMeansTest(false, 20, 1, false, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansSparseMulSingleCP() {\n+ runKMeansTest(true, 20, 1, false, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansDenseMulMultiRewritesCP() {\n+ runKMeansTest( false, 20, 10, true, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansSparseMulMultiRewritesCP() {\n+ runKMeansTest(true, 20, 10, true, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansDenseMulMultiCP() {\n+ runKMeansTest(false, 20, 10, false, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testKMeansSparseMulMultiCP() {\n+ runKMeansTest(true, 20, 10, false, LopProperties.ExecType.CP);\n+ }\n+\n+ private void runKMeansTest(boolean sparse, int centroids, int runs,\n+ boolean rewrites, LopProperties.ExecType instType)\n+ {\n+ Types.ExecMode platformOld = setExecMode(instType);\n+\n+ boolean oldFlag = OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION;\n+ boolean sparkConfigOld = DMLScript.USE_LOCAL_SPARK_CONFIG;\n+\n+ try\n+ {\n+ loadTestConfiguration(getTestConfiguration(TEST_NAME));\n+\n+ double sparsity = sparse ? spSparse : spDense;\n+\n+ String HOME = SCRIPT_DIR + TEST_DIR;\n+\n+ fullDMLScriptName = HOME + TEST_NAME + \".dml\";\n+ programArgs = new String[]{ \"-explain\", \"-stats\",\n+ \"-nvargs\", \"X=\" + input(\"X\"), \"Y=\" + output(\"Y\"), \"C=\" + output(\"C\"),\n+ \"k=\" + centroids, \"runs=\" + runs,\n+ \"eps=\" + eps, \"max_iter=\" + max_iter};\n+\n+ OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = rewrites;\n+\n+ //generate actual datasets\n+ double[][] X = getRandomMatrix(rows, cols, 0, 1, sparsity, 714);\n+ writeInputMatrixWithMTD(\"X\", X, true);\n+\n+ runTest(true, false, null, -1);\n+ }\n+ finally {\n+ rtplatform = platformOld;\n+ DMLScript.USE_LOCAL_SPARK_CONFIG = sparkConfigOld;\n+ OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = oldFlag;\n+ OptimizerUtils.ALLOW_AUTO_VECTORIZATION = true;\n+ OptimizerUtils.ALLOW_OPERATOR_FUSION = true;\n+ }\n+ }\n+}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/builtin/kmeans.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($X)\n+[C, Y] = kmeans(X, $k, $runs, $max_iter, $eps, TRUE, 50)\n+write(C, $C)\n+write(Y, $Y)\n" } ]

[ { "change_type": "MODIFY", "old_path": "bin/systemds-standalone.sh", "new_path": "bin/systemds-standalone.sh", "diff": "@@ -30,6 +30,10 @@ EOF\n# Script internally invokes 'java -Xmx4g -Xms4g -Xmn400m [Custom-Java-Options] -jar StandaloneSystemDS.jar -f <dml-filename> -exec singlenode -config=SystemDS-config.xml [Optional-Arguments]'\n+# This path can stay in *nix separator style even when using cygwin/msys.\n+export HADOOP_HOME=/tmp/systemds\n+mkdir -p $HADOOP_HOME\n+\nif [ -z \"$1\" ] ; then\necho \"Wrong Usage.\";\nprintSimpleUsage\n@@ -39,17 +43,31 @@ if [ ! -z $SYSTEMDS_ROOT ]; then\nPROJECT_ROOT_DIR=\"$SYSTEMDS_ROOT\"\necho \"SYTEMDS_ROOT is set to:\" $SYSTEMDS_ROOT\nelse\n+ # Paths that go to the java executable need to be Windows style in cygwin/msys.\n+ DIR_SEP=/\n# find the systemDS root path which contains the bin folder, the script folder and the target folder\n# tolerate path with spaces\nSCRIPT_DIR=$( dirname \"$0\" )\n+\n+ if [ \"$OSTYPE\" == \"win32\" ] || [ \"$OSTYPE\" == \"msys\" ] ; then\n+ echo \"Constructing Windows-style classpath\"\n+ DIR_SEP=\\\\\n+ PROJECT_ROOT_DIR=`cygpath -w -p $( cd \"${SCRIPT_DIR}/..\" ; pwd -P )`\n+\n+ if [ ! -f ${HADOOP_HOME}${DIR_SEP}bin ]; then\n+ cp -a ${PROJECT_ROOT_DIR}${DIR_SEP}target${DIR_SEP}lib${DIR_SEP}hadoop${DIR_SEP}bin ${HADOOP_HOME}\n+ fi\n+ USER_DIR=`cygpath -w -p ${PWD}`\n+ else\nPROJECT_ROOT_DIR=$( cd \"${SCRIPT_DIR}/..\" ; pwd -P )\n+ USER_DIR=$PWD\n+ fi\nfi\n-USER_DIR=$PWD\n-BUILD_DIR=${PROJECT_ROOT_DIR}/target\n-HADOOP_LIB_DIR=${BUILD_DIR}/lib\n-DML_SCRIPT_CLASS=${BUILD_DIR}/classes/org/tugraz/sysds/api/DMLScript.class\n+BUILD_DIR=${PROJECT_ROOT_DIR}${DIR_SEP}target\n+HADOOP_LIB_DIR=${BUILD_DIR}${DIR_SEP}lib\n+DML_SCRIPT_CLASS=${BUILD_DIR}${DIR_SEP}classes${DIR_SEP}org${DIR_SEP}tugraz${DIR_SEP}sysds${DIR_SEP}api${DIR_SEP}DMLScript.class\nBUILD_ERR_MSG=\"You must build the project before running this script.\"\nBUILD_DIR_ERR_MSG=\"Could not find target directory \\\"${BUILD_DIR}\\\". ${BUILD_ERR_MSG}\"\n@@ -67,37 +85,33 @@ echo \"==========================================================================\n# if the present working directory is the project root or bin folder, then use the temp folder as user.dir\nif [ \"$USER_DIR\" = \"$PROJECT_ROOT_DIR\" ] || [ \"$USER_DIR\" = \"$PROJECT_ROOT_DIR/bin\" ]\nthen\n- USER_DIR=${PROJECT_ROOT_DIR}/temp\n+ USER_DIR=${PROJECT_ROOT_DIR}${DIR_SEP}temp\necho \"Output dir: $USER_DIR\"\nfi\n# if the SystemDS-config.xml does not exist, create it from the template\n-if [ ! -f \"${PROJECT_ROOT_DIR}/conf/SystemDS-config.xml\" ]\n+if [ ! -f \"${PROJECT_ROOT_DIR}${DIR_SEP}conf${DIR_SEP}SystemDS-config.xml\" ]\nthen\n- cp \"${PROJECT_ROOT_DIR}/conf/SystemDS-config.xml.template\" \\\n- \"${PROJECT_ROOT_DIR}/conf/SystemDS-config.xml\"\n- echo \"... created ${PROJECT_ROOT_DIR}/conf/SystemDS-config.xml\"\n+ cp \"${PROJECT_ROOT_DIR}${DIR_SEP}conf${DIR_SEP}SystemDS-config.xml.template\" \\\n+ \"${PROJECT_ROOT_DIR}${DIR_SEP}conf${DIR_SEP}SystemDS-config.xml\"\n+ echo \"... created ${PROJECT_ROOT_DIR}${DIR_SEP}conf${DIR_SEP}SystemDS-config.xml\"\nfi\n# if the log4j.properties do not exis, create them from the template\n-if [ ! -f \"${PROJECT_ROOT_DIR}/conf/log4j.properties\" ]\n+if [ ! -f \"${PROJECT_ROOT_DIR}${DIR_SEP}conf${DIR_SEP}log4j.properties\" ]\nthen\n- cp \"${PROJECT_ROOT_DIR}/conf/log4j.properties.template\" \\\n- \"${PROJECT_ROOT_DIR}/conf/log4j.properties\"\n- echo \"... created ${PROJECT_ROOT_DIR}/conf/log4j.properties\"\n+ cp \"${PROJECT_ROOT_DIR}${DIR_SEP}conf${DIR_SEP}log4j.properties.template\" \\\n+ \"${PROJECT_ROOT_DIR}${DIR_SEP}conf${DIR_SEP}log4j.properties\"\n+ echo \"... created ${PROJECT_ROOT_DIR}${DIR_SEP}conf${DIR_SEP}log4j.properties\"\nfi\n-\n-\n-\n-# add hadoop libraries which were generated by the build to the classpath\n-CLASSPATH=\\\"${BUILD_DIR}/lib/*\\\"\n-\n#SYSTEM_DS_JAR=$( find $PROJECT_ROOT_DIR/target/system-ds-*-SNAPSHOT.jar )\n-SYSTEM_DS_JAR=\\\"${BUILD_DIR}/classes\\\"\n-\n-CLASSPATH=${CLASSPATH}:${SYSTEM_DS_JAR}\n+SYSTEM_DS_JAR=${BUILD_DIR}${DIR_SEP}classes\n+# add hadoop libraries which were generated by the build to the classpath\n+CLASSPATH=${BUILD_DIR}${DIR_SEP}lib${DIR_SEP}*\n+CLASSPATH=\\\"${CLASSPATH}\\;${SYSTEM_DS_JAR}\\\"\n+#echo ${CLASSPATH}\necho \"================================================================================\"\n@@ -105,7 +119,7 @@ echo \"==========================================================================\nSYSTEMDS_DEFAULT_JAVA_OPTS=\"\\\n-Xmx8g -Xms4g -Xmn1g \\\n-cp $CLASSPATH \\\n--Dlog4j.configuration=file:'$PROJECT_ROOT_DIR/conf/log4j.properties' \\\n+-Dlog4j.configuration=file:'$PROJECT_ROOT_DIR${DIR_SEP}conf${DIR_SEP}log4j.properties' \\\n-Duser.dir='$USER_DIR'\"\n# Add any custom Java options set by the user at command line, overriding defaults as necessary.\n@@ -115,8 +129,8 @@ if [ ! -z \"${SYSTEMDS_JAVA_OPTS}\" ]; then\nfi\n# Add any custom Java options set by the user in the environment variables file, overriding defaults as necessary.\n-if [ -f \"${PROJECT_ROOT_DIR}/conf/systemds-env.sh\" ]; then\n- . \"${PROJECT_ROOT_DIR}/conf/systemds-env.sh\"\n+if [ -f \"${PROJECT_ROOT_DIR}${DIR_SEP}conf${DIR_SEP}systemds-env.sh\" ]; then\n+ . \"${PROJECT_ROOT_DIR}${DIR_SEP}conf${DIR_SEP}systemds-env.sh\"\nif [ ! -z \"${SYSTEMDS_JAVA_OPTS}\" ]; then\nSYSTEMDS_DEFAULT_JAVA_OPTS+=\" ${SYSTEMDS_JAVA_OPTS}\"\nfi\n@@ -129,7 +143,6 @@ CMD=\"\\\njava ${SYSTEMDS_DEFAULT_JAVA_OPTS} \\\norg.tugraz.sysds.api.DMLScript \\\n-help\"\n-# echo ${CMD}\neval ${CMD}\nexit 0\n}\n@@ -157,7 +170,7 @@ shift\nif [ ! -f \"$SCRIPT_FILE\" ]\nthen\nSCRIPT_FILE_NAME=$(basename $SCRIPT_FILE)\n- SCRIPT_FILE_FOUND=$(find \"$PROJECT_ROOT_DIR/scripts\" -name \"$SCRIPT_FILE_NAME\")\n+ SCRIPT_FILE_FOUND=$(find \"$PROJECT_ROOT_DIR${DIR_SEP}scripts\" -name \"$SCRIPT_FILE_NAME\")\nif [ ! \"$SCRIPT_FILE_FOUND\" ]\nthen\necho \"Could not find DML script: $SCRIPT_FILE\"\n@@ -175,10 +188,10 @@ java ${SYSTEMDS_DEFAULT_JAVA_OPTS} \\\norg.tugraz.sysds.api.DMLScript \\\n-f '$SCRIPT_FILE' \\\n-exec singlenode \\\n--config '$PROJECT_ROOT_DIR/conf/SystemDS-config.xml' \\\n+-config '$PROJECT_ROOT_DIR${DIR_SEP}conf${DIR_SEP}SystemDS-config.xml' \\\n$@\"\n-export HADOOP_HOME=/tmp/systemds\n+echo \"Executing \"${CMD}\neval ${CMD}\nRETURN_CODE=$?\n" }, { "change_type": "ADD", "old_path": null, "new_path": "bin/systemds.bat", "diff": "+@ECHO OFF\n+::-------------------------------------------------------------\n+::\n+:: Copyright 2019 Graz University of Technology\n+::\n+:: Licensed under the Apache License, Version 2.0 (the \"License\");\n+:: you may not use this file except in compliance with the License.\n+:: You may obtain a copy of the License at\n+::\n+:: http://www.apache.org/licenses/LICENSE-2.0\n+::\n+:: Unless required by applicable law or agreed to in writing, software\n+:: distributed under the License is distributed on an \"AS IS\" BASIS,\n+:: WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+:: See the License for the specific language governing permissions and\n+:: limitations under the License.\n+::\n+::-------------------------------------------------------------\n+\n+echo ================================================================================\n+\n+IF \"%~1\" == \"\" GOTO Err\n+IF \"%~1\" == \"-help\" GOTO Msg\n+IF \"%~1\" == \"-h\" GOTO Msg\n+\n+setLocal EnableDelayedExpansion\n+\n+SET USER_DIR=%CD%\n+\n+pushd %~dp0..\n+SET PROJECT_ROOT_DIR=%CD%\n+popd\n+\n+SET BUILD_DIR=%PROJECT_ROOT_DIR%\\target\n+SET HADOOP_LIB_DIR=%BUILD_DIR%\\lib\n+SET DML_SCRIPT_CLASS=%BUILD_DIR%\\classes\\org\\tugraz\\sysds\\api\\DMLScript.class\n+\n+SET BUILD_ERR_MSG=You must build the project before running this script.\n+SET BUILD_DIR_ERR_MSG=Could not find target directory \"%BUILD_DIR%\". %BUILD_ERR_MSG%\n+SET HADOOP_LIB_ERR_MSG=Could not find required libraries \"%HADOOP_LIB_DIR%\\*\". %BUILD_ERR_MSG%\n+SET DML_SCRIPT_ERR_MSG=Could not find \"%DML_SCRIPT_CLASS%\". %BUILD_ERR_MSG%\n+\n+:: check if the project had been built and the jar files exist\n+IF NOT EXIST \"%BUILD_DIR%\" ( echo %BUILD_DIR_ERR_MSG% & GOTO ExitErr )\n+IF NOT EXIST \"%HADOOP_LIB_DIR%\" ( echo %HADOOP_LIB_ERR_MSG% & GOTO ExitErr )\n+IF NOT EXIST \"%DML_SCRIPT_CLASS%\" ( echo %DML_SCRIPT_ERR_MSG% & GOTO ExitErr )\n+\n+\n+:: if the present working directory is the project root or the bin folder, then use the temp folder as user.dir\n+IF \"%USER_DIR%\" == \"%PROJECT_ROOT_DIR%\" (\n+ SET USER_DIR=%PROJECT_ROOT_DIR%\\temp\n+ ECHO Output dir: \"!USER_DIR!\"\n+)\n+IF \"%USER_DIR%\" == \"%PROJECT_ROOT_DIR%\\bin\" (\n+ SET USER_DIR=%PROJECT_ROOT_DIR%\\temp\n+ ECHO Output dir: \"!USER_DIR!\"\n+)\n+\n+\n+:: if the SystemDS-config.xml does not exist, create it from the template\n+IF NOT EXIST \"%PROJECT_ROOT_DIR%\\conf\\SystemDS-config.xml\" (\n+ copy \"%PROJECT_ROOT_DIR%\\conf\\SystemDS-config.xml.template\" ^\n+ \"%PROJECT_ROOT_DIR%\\conf\\SystemDS-config.xml\" > nul\n+ echo ...created \"%PROJECT_ROOT_DIR%\\conf\\SystemDS-config.xml\"\n+)\n+\n+:: if the log4j.properties do not exis, create them from the template\n+IF NOT EXIST \"%PROJECT_ROOT_DIR%\\conf\\log4j.properties\" (\n+ copy \"%PROJECT_ROOT_DIR%\\conf\\log4j.properties.template\" ^\n+ \"%PROJECT_ROOT_DIR%\\conf\\log4j.properties\" > nul\n+ echo ...created \"%PROJECT_ROOT_DIR%\\conf\\log4j.properties\"\n+)\n+\n+SET SCRIPT_FILE=%1\n+\n+:: if the script file path was omitted, try to complete the script path\n+IF NOT EXIST %SCRIPT_FILE% (\n+ FOR /R \"%PROJECT_ROOT_DIR%\" %%f IN (%SCRIPT_FILE%) DO IF EXIST %%f ( SET \"SCRIPT_FILE_FOUND=%%f\" )\n+)\n+\n+IF NOT EXIST %SCRIPT_FILE% IF NOT DEFINED SCRIPT_FILE_FOUND (\n+ echo Could not find DML script: \"%SCRIPT_FILE%\"\n+ GOTO Err\n+) ELSE (\n+ SET SCRIPT_FILE=\"%SCRIPT_FILE_FOUND%\"\n+ echo DML script: \"%SCRIPT_FILE_FOUND%\"\n+)\n+\n+\n+:: the hadoop winutils\n+SET HADOOP_HOME=%PROJECT_ROOT_DIR%\\target\\lib\\hadoop\n+\n+:: add dependent libraries to classpath (since Java 1.6 we can use wildcards)\n+set CLASSPATH=%PROJECT_ROOT_DIR%\\target\\lib\\*\n+\n+:: add compiled SystemDS classes to classpath\n+set CLASSPATH=%CLASSPATH%;%PROJECT_ROOT_DIR%\\target\\classes\n+\n+ECHO Classpath: \"%CLASSPATH%\"\n+\n+:: remove the DML script file from the list of arguments\n+:: allow for dml script path with spaces, enclosed in quotes\n+rem for /f \"tokens=1,* delims= \" %%a in (\"%*\") do set DML_OPT_ARGS=%%b\n+rem for /f tokens^=1^,*^ delims^=^\" %%a in (\"%*\") do set DML_OPT_ARGS=%%b\n+set ARGS=%*\n+set DML_OPT_ARGS=!ARGS:%1 =!\n+\n+echo ================================================================================\n+\n+:: construct the java command with options and arguments\n+set CMD=java -Xmx4g -Xms2g -Xmn400m ^\n+ -cp \"%CLASSPATH%\" ^\n+ -Dlog4j.configuration=file:\"%PROJECT_ROOT_DIR%\\conf\\log4j.properties\" ^\n+ -Duser.dir=\"%USER_DIR%\" ^\n+ org.tugraz.sysds.api.DMLScript ^\n+ -f %SCRIPT_FILE% ^\n+ -exec singlenode ^\n+ -config \"%PROJECT_ROOT_DIR%\\conf\\SystemDS-config.xml\" ^\n+ %DML_OPT_ARGS%\n+\n+:: execute the java command\n+%CMD%\n+\n+:: if there was an error, display the full java command (in case some of the variable substitutions broke it)\n+IF ERRORLEVEL 1 (\n+ ECHO Failed to run SystemDS. Exit code: %ERRORLEVEL%\n+ SET LF=^\n+\n+\n+ :: keep empty lines above for the line breaks\n+ ECHO %CMD: =!LF! %\n+ EXIT /B %ERRORLEVEL%\n+)\n+GOTO End\n+\n+:Err\n+ECHO Wrong Usage. Please provide DML filename to be executed.\n+GOTO Msg\n+\n+:Msg\n+ECHO Usage: systemds.bat ^<dml-filename^> [arguments] [-help]\n+ECHO Script internally invokes 'java -Xmx4g -Xms4g -Xmn400m -jar SystemDS.jar -f ^<dml-filename^> -exec singlenode -config SystemDS-config.xml [Optional-Arguments]'\n+GOTO ExitErr\n+\n+:ExitErr\n+EXIT /B 1\n+\n+:End\n+EXIT /B 0\n\\ No newline at end of file\n" }, { "change_type": "MODIFY", "old_path": "conf/SystemDS-config.xml", "new_path": "conf/SystemDS-config.xml", "diff": "<!--\n- * Licensed to the Apache Software Foundation (ASF) under one\n- * or more contributor license agreements. See the NOTICE file\n- * distributed with this work for additional information\n- * regarding copyright ownership. The ASF licenses this file\n- * to you under the Apache License, Version 2.0 (the\n- * \"License\"); you may not use this file except in compliance\n- * with the License. You may obtain a copy of the License at\n+ ********************************************************************\n+ *\n+ * Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed under the Apache License, Version 2.0 (the \"License\");\n+ * you may not use this file except in compliance with the License.\n+ * You may obtain a copy of the License at\n*\n* http://www.apache.org/licenses/LICENSE-2.0\n*\n- * Unless required by applicable law or agreed to in writing,\n- * software distributed under the License is distributed on an\n- * \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n- * KIND, either express or implied. See the License for the\n- * specific language governing permissions and limitations\n- * under the License.\n+ * Unless required by applicable law or agreed to in writing, software\n+ * distributed under the License is distributed on an \"AS IS\" BASIS,\n+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+ * See the License for the specific language governing permissions and\n+ * limitations under the License.\n+ *\n+ ********************************************************************\n-->\n+\n<root>\n<!-- local fs tmp working directory-->\n<sysml.localtmpdir>/tmp/systemds</sysml.localtmpdir>\n" }, { "change_type": "MODIFY", "old_path": "conf/SystemDS-config.xml.template", "new_path": "conf/SystemDS-config.xml.template", "diff": "-<!--\n- * Licensed to the Apache Software Foundation (ASF) under one\n- * or more contributor license agreements. See the NOTICE file\n- * distributed with this work for additional information\n- * regarding copyright ownership. The ASF licenses this file\n- * to you under the Apache License, Version 2.0 (the\n- * \"License\"); you may not use this file except in compliance\n- * with the License. You may obtain a copy of the License at\n+<!---------------------------------------------------------------\n+ *\n+ * Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed under the Apache License, Version 2.0 (the \"License\");\n+ * you may not use this file except in compliance with the License.\n+ * You may obtain a copy of the License at\n*\n* http://www.apache.org/licenses/LICENSE-2.0\n*\n- * Unless required by applicable law or agreed to in writing,\n- * software distributed under the License is distributed on an\n- * \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n- * KIND, either express or implied. See the License for the\n- * specific language governing permissions and limitations\n- * under the License.\n--->\n+ * Unless required by applicable law or agreed to in writing, software\n+ * distributed under the License is distributed on an \"AS IS\" BASIS,\n+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+ * See the License for the specific language governing permissions and\n+ * limitations under the License.\n+ *\n+ * --------------------------------------------------------------->\n+\n<root>\n<!-- local fs tmp working directory-->\n" }, { "change_type": "MODIFY", "old_path": "conf/log4j-silent.properties", "new_path": "conf/log4j-silent.properties", "diff": "#-------------------------------------------------------------\n#\n-# Licensed to the Apache Software Foundation (ASF) under one\n-# or more contributor license agreements. See the NOTICE file\n-# distributed with this work for additional information\n-# regarding copyright ownership. The ASF licenses this file\n-# to you under the Apache License, Version 2.0 (the\n-# \"License\"); you may not use this file except in compliance\n-# with the License. You may obtain a copy of the License at\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n#\n# http://www.apache.org/licenses/LICENSE-2.0\n#\n-# Unless required by applicable law or agreed to in writing,\n-# software distributed under the License is distributed on an\n-# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n-# KIND, either express or implied. See the License for the\n-# specific language governing permissions and limitations\n-# under the License.\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n#\n#-------------------------------------------------------------\n" }, { "change_type": "MODIFY", "old_path": "conf/log4j.properties.template", "new_path": "conf/log4j.properties.template", "diff": "#-------------------------------------------------------------\n#\n-# Licensed to the Apache Software Foundation (ASF) under one\n-# or more contributor license agreements. See the NOTICE file\n-# distributed with this work for additional information\n-# regarding copyright ownership. The ASF licenses this file\n-# to you under the Apache License, Version 2.0 (the\n-# \"License\"); you may not use this file except in compliance\n-# with the License. You may obtain a copy of the License at\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n#\n# http://www.apache.org/licenses/LICENSE-2.0\n#\n-# Unless required by applicable law or agreed to in writing,\n-# software distributed under the License is distributed on an\n-# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n-# KIND, either express or implied. See the License for the\n-# specific language governing permissions and limitations\n-# under the License.\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n#\n#-------------------------------------------------------------\n" }, { "change_type": "ADD", "old_path": null, "new_path": "conf/systemds-env.sh.template", "diff": "+#!/usr/bin/env bash\n+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+# This file is sourced when running the bin/systemds-standalone script.\n+# Copy it as systemds-env.sh and edit it to configure SystemDS.\n+\n+# Example of adding additional Java execution options, which will\n+# override defaults as necessary.\n+#SYSTEMDS_JAVA_OPTS=\"-Xmx12g -Xms8g\"\n+\n" }, { "change_type": "DELETE", "old_path": "conf/systemml-env.sh.template", "new_path": null, "diff": "-#!/usr/bin/env bash\n-#-------------------------------------------------------------\n-#\n-# Licensed to the Apache Software Foundation (ASF) under one\n-# or more contributor license agreements. See the NOTICE file\n-# distributed with this work for additional information\n-# regarding copyright ownership. The ASF licenses this file\n-# to you under the Apache License, Version 2.0 (the\n-# \"License\"); you may not use this file except in compliance\n-# with the License. You may obtain a copy of the License at\n-#\n-# http://www.apache.org/licenses/LICENSE-2.0\n-#\n-# Unless required by applicable law or agreed to in writing,\n-# software distributed under the License is distributed on an\n-# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n-# KIND, either express or implied. See the License for the\n-# specific language governing permissions and limitations\n-# under the License.\n-#\n-#-------------------------------------------------------------\n-\n-# This file is sourced when running the bin/systemml script.\n-# Copy it as systemml-env.sh and edit it to configure SystemML.\n-\n-# Example of adding additional Java execution options, which will\n-# override defaults as necessary.\n-#SYSTEMML_JAVA_OPTS=\"-Xmx12g -Xms8g\"\n-\n" }, { "change_type": "ADD", "old_path": "src/main/cpp/lib/systemds_mkl-Windows-AMD64.dll", "new_path": "src/main/cpp/lib/systemds_mkl-Windows-AMD64.dll", "diff": "Binary files /dev/null and b/src/main/cpp/lib/systemds_mkl-Windows-AMD64.dll differ\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/matrix/data/LibMatrixNative.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/matrix/data/LibMatrixNative.java", "diff": "@@ -130,6 +130,7 @@ public class LibMatrixNative\nif( NativeHelper.tsmm(m1.getDenseBlockValues(),\nret.getDenseBlockValues(), m1.rlen, m1.clen, leftTrans, k) )\n{\n+ LOG.info(\"Using native TSMM()\");\nlong nnz = (ret.clen==1) ? ret.recomputeNonZeros() :\nLibMatrixMult.copyUpperToLowerTriangle(ret);\nret.setNonZeros(nnz);\n@@ -137,6 +138,7 @@ public class LibMatrixNative\nreturn;\n}\n//fallback to default java implementation\n+ LOG.info(\"Falling back to java TSMM()\");\nStatistics.incrementNativeFailuresCounter();\n}\nif( k > 1 )\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/utils/NativeHelper.java", "new_path": "src/main/java/org/tugraz/sysds/utils/NativeHelper.java", "diff": "@@ -164,8 +164,8 @@ public class NativeHelper {\n// Performing loading in a method instead of a static block will throw a detailed stack trace in case of fatal errors\nprivate static void performLoading(String customLibPath, String userSpecifiedBLAS) {\n// Only Linux supported for BLAS\n- if(!SystemUtils.IS_OS_LINUX)\n- return;\n+// if(!SystemUtils.IS_OS_LINUX)\n+// return;\n// attemptedLoading variable ensures that we don't try to load SystemDS and other dependencies\n// again and again especially in the parfor (hence the double-checking with synchronized).\n@@ -181,12 +181,25 @@ public class NativeHelper {\n}\n+ if(SystemUtils.IS_OS_WINDOWS) {\n+ if (checkAndLoadBLAS(customLibPath + \"\\\\lib\", blas) &&\n+// loadLibraryHelper(customLibPath + \"\\\\bin\\\\systemds_\" + blasType + \"-Windows-AMD64.dll\")) {\n+// loadLibraryHelper(\"systemds_\" + blasType + \"-Windows-AMD64.lib\")) {\n+// loadLibraryHelper(customLibPath + \"\\\\systemds_\" + blasType + \"-Windows-AMD64.dll\")) {\n+ loadBLAS(customLibPath, \"systemds_\" + blasType + \"-Windows-AMD64\", \"\"))\n+ {\n+ LOG.info(\"Using native blas: \" + blasType + getNativeBLASPath());\n+ CURRENT_NATIVE_BLAS_STATE = NativeBlasState.SUCCESSFULLY_LOADED_NATIVE_BLAS_AND_IN_USE;\n+ }\n+ }\n+ else {\nif (checkAndLoadBLAS(customLibPath, blas) && loadLibraryHelper(\"libsystemds_\" + blasType + \"-Linux-x86_64.so\")) {\nLOG.info(\"Using native blas: \" + blasType + getNativeBLASPath());\nCURRENT_NATIVE_BLAS_STATE = NativeBlasState.SUCCESSFULLY_LOADED_NATIVE_BLAS_AND_IN_USE;\n}\n}\n}\n+ }\ndouble timeToLoadInMilliseconds = (System.nanoTime()-start)*1e-6;\nif(timeToLoadInMilliseconds > 1000)\nLOG.warn(\"Time to load native blas: \" + timeToLoadInMilliseconds + \" milliseconds.\");\n@@ -204,6 +217,10 @@ public class NativeHelper {\nfor(int i = 0; i < listBLAS.length; i++) {\nString blas = listBLAS[i];\nif(blas.equalsIgnoreCase(\"mkl\")) {\n+ if(SystemUtils.IS_OS_WINDOWS)\n+ isLoaded = true;\n+// isLoaded = loadBLAS(customLibPath, \"mkl_rt.dll\", null);\n+ else\nisLoaded = loadBLAS(customLibPath, \"mkl_rt\", null);\n}\nelse if(blas.equalsIgnoreCase(\"openblas\")) {\n@@ -295,9 +312,11 @@ public class NativeHelper {\nprivate static boolean loadLibraryHelper(String path) {\nOutputStream out = null;\n- try( InputStream in = NativeHelper.class.getResourceAsStream(\"/lib/\"+path) ) {\n+ try(InputStream in = NativeHelper.class.getResourceAsStream(\"/lib/\"+path))\n+ {\n// This logic is added because Java does not allow to load library from a resource file.\n- if(in != null) {\n+ if(in != null)\n+ {\nFile temp = File.createTempFile(path, \"\");\ntemp.deleteOnExit();\nout = FileUtils.openOutputStream(temp);\n" }, { "change_type": "MODIFY", "old_path": "src/test/config/SystemDS-config.xml", "new_path": "src/test/config/SystemDS-config.xml", "diff": "<!--\n- * Licensed to the Apache Software Foundation (ASF) under one\n- * or more contributor license agreements. See the NOTICE file\n- * distributed with this work for additional information\n- * regarding copyright ownership. The ASF licenses this file\n- * to you under the Apache License, Version 2.0 (the\n- * \"License\"); you may not use this file except in compliance\n- * with the License. You may obtain a copy of the License at\n+ ********************************************************************\n+ *\n+ * Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed under the Apache License, Version 2.0 (the \"License\");\n+ * you may not use this file except in compliance with the License.\n+ * You may obtain a copy of the License at\n*\n* http://www.apache.org/licenses/LICENSE-2.0\n*\n- * Unless required by applicable law or agreed to in writing,\n- * software distributed under the License is distributed on an\n- * \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n- * KIND, either express or implied. See the License for the\n- * specific language governing permissions and limitations\n- * under the License.\n+ * Unless required by applicable law or agreed to in writing, software\n+ * distributed under the License is distributed on an \"AS IS\" BASIS,\n+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+ * See the License for the specific language governing permissions and\n+ * limitations under the License.\n+ *\n+ ********************************************************************\n-->\n<root>\n<!-- enables multi-threaded read/write in singlenode control program -->\n<sysml.cp.parallel.io>true</sysml.cp.parallel.io>\n+\n+ <!-- enables native blas for matrix multiplication and convolution, experimental feature (options: auto, mkl, openblas, none) -->\n+ <sysml.native.blas>none</sysml.native.blas>\n+\n+ <!-- custom directory where BLAS libraries are available, experimental feature (options: absolute directory path or none). If set to none, we use standard LD_LIBRARY_PATH. -->\n+ <sysml.native.blas.directory>none</sysml.native.blas.directory>\n+\n</root>\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/binary/matrix_full_other/FullMatrixMultiplicationTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/binary/matrix_full_other/FullMatrixMultiplicationTest.java", "diff": "@@ -184,7 +184,7 @@ public class FullMatrixMultiplicationTest extends AutomatedTestBase\n/* This is for running the junit test the new way, i.e., construct the arguments directly */\nString HOME = SCRIPT_DIR + TEST_DIR;\nfullDMLScriptName = HOME + TEST_NAME + \".dml\";\n- programArgs = new String[]{\"-args\",\n+ programArgs = new String[]{\"-stats\",\"-args\",\ninput(\"A\"), Integer.toString(rowsA), Integer.toString(colsA),\ninput(\"B\"), Integer.toString(rowsB), Integer.toString(colsB), output(\"C\") };\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/binary/matrix_full_other/FullMatrixMultiplicationTransposeSelfTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/binary/matrix_full_other/FullMatrixMultiplicationTransposeSelfTest.java", "diff": "@@ -167,7 +167,7 @@ public class FullMatrixMultiplicationTransposeSelfTest extends AutomatedTestBase\n/* This is for running the junit test the new way, i.e., construct the arguments directly */\nString HOME = SCRIPT_DIR + TEST_DIR;\nfullDMLScriptName = HOME + TEST_NAME + \".dml\";\n- programArgs = new String[]{\"-args\", input(\"A\"),\n+ programArgs = new String[]{\"-stats\",\"-args\", input(\"A\"),\nInteger.toString(rows), Integer.toString(cols), output(\"B\") };\nfullRScriptName = HOME + TEST_NAME + \".R\";\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinLmTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinLmTest.java", "diff": "@@ -127,7 +127,7 @@ public class BuiltinLmTest extends AutomatedTestBase\nfullDMLScriptName = HOME + dml_test_name + \".dml\";\n- programArgs = new String[]{\"-explain\", \"-args\", input(\"A\"), input(\"B\"), output(\"C\") };\n+ programArgs = new String[]{\"-explain\", \"-stats\", \"-args\", input(\"A\"), input(\"B\"), output(\"C\") };\nfullRScriptName = HOME + TEST_NAME + \".R\";\nrCmd = \"Rscript\" + \" \" + fullRScriptName + \" \" + inputDir() + \" \" + expectedDir();\n" }, { "change_type": "MODIFY", "old_path": "src/test/resources/log4j.properties", "new_path": "src/test/resources/log4j.properties", "diff": "#-------------------------------------------------------------\n# Define some default values that can be overridden by system properties\n-hadoop.root.logger=INFO,console\n+hadoop.root.logger=DEBUG,console\nhadoop.log.dir=.\nhadoop.log.file=hadoop.log\nhadoop.security.logger=OFF\n" } ]

[ { "change_type": "ADD", "old_path": null, "new_path": "bin/Readme.md", "diff": "+## Scripts to run SystemDS\n+This directory contains various scripts to run SystemDS locally via the java executable or distributed via spark-submit\n+\n+#### Local run scripts\n+* Windows Batch Script\n+ * `$ systemds-standalone.bat <script.dml> -args <parameters>`\n+\n+* Bash Shell Script\n+ * `$ systemds-standalone.sh -f <script.dml> -args <parameters>`\n+\n+* Python 3.x Script\n+ * `$ python systemds-standalone.py -f <script.dml> -args <parameters>`\n+\n+#### Spark run script\n+Run `systemds-spark.sh -h` for help. For convenience it is recommended to have the SystemDS\n+source tree checked out and SYSTEMDS_ROOT set.\n+\n+* Bash Shell Script\n+ * `$ systemds-spark.sh -f <script.dml> --args <parameters>`\n+\n+\n+## Setting SYSTEMDS_ROOT environment variable\n+In order to run SystemDS from your development directory and leave the\n+SystemDS source tree untouched, the following setup could be used (example for bash):\n+ ```shell script\n+$ export SYSTEMDS_ROOT=/home/<username>/systemds\n+$ export PATH=$SYSTEMDS_ROOT/bin:$PATH\n+```\n+Alternatively, if the run scripts are invoked from the root of the\n+SystemDS source tree, `SYSTEMDS_ROOT` does not need to be set (example for python):\n+\n+`$ python bin/systemds-standalone.py -f <script.dml> <parameters>`\n+\n+The DML scripts residing in the directory `$SYSTEMDS_ROOT/scripts` will be found automatically by the run scripts.\n+\n+## Running a first example:\n+To see SystemDS in action a simple example using the `Univar-stats.dml`\n+script can be executed. This example is taken from the\n+[SystemML documentation](http://apache.github.io/systemml/standalone-guide).\n+The relevant commands to run this example with SystemDS will be listed here.\n+See their documentation for further details.\n+\n+#### Example preparations\n+```shell script\n+# download test data\n+$ wget -P data/ http://archive.ics.uci.edu/ml/machine-learning-databases/haberman/haberman.data\n+\n+# generate a metadata file for the dataset\n+$ echo '{\"rows\": 306, \"cols\": 4, \"format\": \"csv\"}' > data/haberman.data.mtd\n+\n+# generate type description for the data\n+$ echo '1,1,1,2' > data/types.csv\n+$ echo '{\"rows\": 1, \"cols\": 4, \"format\": \"csv\"}' > data/types.csv.mtd\n+```\n+#### Executing the DML script\n+```shell script\n+$ systemds-standalone.sh Univar-Stats.dml -nvargs X=data/haberman.data TYPES=data/types.csv STATS=data/univarOut.mtx CONSOLE_OUTPUT=TRUE\n+```\n+\n+#### Using Intel MKL native instructions\n+To use the MKL acceleration download and install the latest MKL libtary from [1]\n\\ No newline at end of file\n" }, { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -98,3 +98,4 @@ SYSTEMDS-140 Distributed Tensor Operations\nSYSTEMDS-150 Release 0.1\n* 151 Fixing native BLAS instructions (MKL,OpenBLAS) OK\n+ * 152 Run script documentation OK\n\\ No newline at end of file\n" } ]

[ { "change_type": "MODIFY", "old_path": "scripts/builtin/slicefinder.dml", "new_path": "scripts/builtin/slicefinder.dml", "diff": "#\n#-------------------------------------------------------------\n+#-------------------------------------------------------------\n+# X Input matrix\n+# W beta in lm\n+# Y matrix column for training\n+# k top-K subsets / slices\n+# paq amount of values wanted for each col, if = 1 then its off\n+# S amount of subsets to combine (for now supported only 2/1)\n+# ------------------------------------------------------------\n-m_slicefinder = function(Matrix[Double] X0, Matrix[Double] W, Integer k = 1) return(Matrix[Double] result) {\n+m_slicefinder = function(Matrix[Double] X, Matrix[Double] W, Matrix[Double] Y, Integer k = 1, Integer paq = 1, Integer S = 2) return(Matrix[Double] result) {\n- f = 2;\n+ X0 = cbind(X, Y);\nbeta = W;\n- # number of features combined\ncol = ncol(X0);\nrow = nrow(X0);\n+\nval_matrix = matrix(0, rows = 2, cols = col - 1);\nvcol = ncol(val_matrix);\nempty_row = matrix(0, rows = 1, cols = col - 1);\n- print(\"col: \" + col + \" row = \" + row);\n- #first scan, spot data, make first slices\n+\n+ #first scan, making val_matrix with different values from the each col.\n+ #first row or this matrix indicates how many different values are in each col.\nfor (j in 1:col - 1) {\nvector = order(target = X0[, j], by = 1, decreasing = FALSE);\nval_matrix[2, j] = vector[1, 1];\nval_counter = 1;\n- print(\"Col \" + j);\nfor (i in 1:row) {\nif (as.scalar(val_matrix[val_counter + 1, j]) != as.scalar(vector[i, 1])) {\nif (nrow(val_matrix) == val_counter + 1)\n@@ -45,13 +53,17 @@ m_slicefinder = function(Matrix[Double] X0, Matrix[Double] W, Integer k = 1) ret\n}\nval_matrix[1, j] = val_counter;\n- #here I can add some condition to split the values from each column if val_counter is too big;\n+\n+ #here I add some condition to split the values from each column if val_counter is too big;\n################################################\n#this code relates to large datasets\n- /* if (val_counter > k) {\n- position = floor(val_counter / k);\n- for (a in 1:k) {\n- if (a == k) {\n+ #packing values according to paq value\n+ ## TODO -- this if needs to be checked, is not working properly with all the paq values\n+ if (paq != 1) {\n+\n+ position = floor(val_counter / paq);\n+ for (a in 1:paq) {\n+ if (a == paq) {\npos = as.scalar(val_matrix[1, j]) + 1;\ntresh = val_matrix[pos, j];\nval_matrix[a + 1, j] = tresh;\n@@ -61,56 +73,41 @@ m_slicefinder = function(Matrix[Double] X0, Matrix[Double] W, Integer k = 1) ret\nval_matrix[a + 1, j] = tresh;\n}\n}\n+\n+ val_matrix = val_matrix[1:paq + 1,];\n+\n}\n- */\n##################################################\n}\n-\n- # now val_matrix[1:4,]) is a treshhold matrix that define clear slices\n- print(toString(val_matrix));\n-\n- #start selecting slices\nvrow = nrow(val_matrix);\nvcol = ncol(val_matrix);\ntotalrows = (vrow - 1) * vcol;\n- print(\"vrow: \" + vrow);\n- print(\"vcol: \" + vcol);\n- print(\"totalrows: \" + totalrows);\n#######################################\nY0 = X0[1:nrow(X0), ncol(X0)];\nY = lmpredict(X = X0[1:nrow(X0), 1:col - 1], w = beta, icpt = 0);\n[error0, diff0] = standart_error(Y, Y0);\n- print(\"Error0: \" + error0);\n- print(\"diff0: \" + diff0);\n-\n#####################################################\n+ # set_matrix will be the matrix with all slices and combination of them\n+ #acctually supporting only combination of 2 slices\n+ set_matrix = matrix(0, rows = 1, cols = 2 + (9 * S));\n+ set_row = matrix(0, rows = 1, cols = 2 + (9 * S));\n- set_matrix = matrix(0, rows = 1, cols = 2 + (8 * f));\n- set_row = matrix(0, rows = 1, cols = 2 + (8 * f));\n- cont = 1;\n+ # first_slices is returning in slice_matrix single subsets\n- b0 = 1;\n- b1 = col - 1;\n- slice_number = 0;\n- pointer_col = 1;\n- pointer_row = 2;\n-\n- set_matrix = first_slices(val_matrix, set_matrix, X0,set_row, beta);\n-\n- ress = order(target = set_matrix, by = 1, decreasing = TRUE);\n- set_matrix = double_features(val_matrix, set_matrix, X0,Y, set_row, beta);\n+ set_matrix = first_slices(val_matrix, set_matrix, X0, set_row, beta, paq, S);\n+ #double_features returns subsets that cover 2 values from the same or different feature\n+ if (S == 2)\n+ set_matrix = double_features(val_matrix, set_matrix, X0, Y, set_row, beta, paq);\nress = order(target = set_matrix, by = 1, decreasing = TRUE);\nset_rows = nrow(set_matrix);\nset_cols = ncol(set_matrix);\n- print(\"Second ress\");\n- print(toString(ress));\n- print(set_rows);\n- result = ress;\n+ #checking values by ordering set_matrix col 1 or 2\n+ result = ress[1:k,];\n}\nstandart_error = function(matrix[double] Y, matrix[double] Y0) return(double error, double diff) {\n@@ -118,6 +115,7 @@ standart_error = function(matrix[double] Y, matrix[double] Y0) return(double err\nerror = sqrt(sum((Y0 - Y) ^ 2) / (nrow(Y) - 2));\n}\n+#index = binary search\nindex = function(matrix[double] X, Integer column, double value, Integer mode) return(Integer pos) {\nbegin = 1;\ne = nrow(X) + 1;\n@@ -138,7 +136,7 @@ index = function(matrix[double] X, Integer column, double value, Integer mode) r\n}\n}\n-first_slices = function(Matrix[Double] val_matrix, Matrix[Double] set_matrix, Matrix[Double] X0, Matrix[Double] set_row, Matrix[Double] beta) return(Matrix[Double] set_matrix) {\n+first_slices = function(Matrix[Double] val_matrix, Matrix[Double] set_matrix, Matrix[Double] X0, Matrix[Double] set_row, Matrix[Double] beta, Integer paq, Integer S) return(Matrix[Double] set_matrix) {\ncol = ncol(X0);\nrow = nrow(X0);\nvrow = nrow(val_matrix);\n@@ -149,23 +147,48 @@ first_slices = function(Matrix[Double] val_matrix, Matrix[Double] set_matrix, Ma\nfor (j in 1:vcol) {\nnum_value = as.scalar(val_matrix[1, j]);\n+\n+ if (paq != 1)\n+ num_value = paq;\nx = order(target = X0, by = j, decreasing = FALSE);\n- print(\"my col: \" + j)\n+\nfor (i in 2:num_value + 1) {\n+ value1 = as.scalar(val_matrix[i, j]);\n+\n+ if (paq != 1) {\n+ if (i == 2) {\n+ a0 = 1;\nswich = 1;\n+ value0 = value1;\n+ }\n+ else if (as.scalar(val_matrix[i - 1, j]) <= as.scalar(val_matrix[i, j])) {\n+ value0 = as.scalar(val_matrix[i - 1, j]);\n+ a0 = index(x, j, value0, 1);\n+ swich = 1;\n+ }\n+ }\n+ else {\n+ swich = 1;\n+ value0 = value1;\n+ a0 = index(x, j, value0, 0);\n+ }\n+\nif (nrow(set_matrix) < cont)\nset_matrix = rbind(set_matrix, set_row);\nif (swich == 1) {\n- value = as.scalar(val_matrix[i, j]);\n- a0 = index(x, j, value, 0);\n- a1 = index(x, j, value, 1);\n+ a1 = index(x, j, value1, 1);\nslice_matrix = x[a0:a1, b0:b1];\nY0 = x[a0:a1, col];\nY = lmpredict(X = slice_matrix, w = beta, icpt = 0);\n[error, diff] = standart_error(Y, Y0);\n- set_matrix[cont,1:10] = t(as.matrix(list(diff, error, value,\n- j, nrow(slice_matrix), ncol(slice_matrix), a0, a1, b0, b1)))\n+ ## TODO - mylist needs to be modified in order to show the total rows of the slice\n+ if (S == 1)\n+ mylist = as.matrix(list(diff, error, value0, value1, j, nrow(slice_matrix), ncol(slice_matrix), a0, a1, b0, b1))\n+ else\n+ mylist = as.matrix(list(diff, error, value0, value1, j, nrow(slice_matrix), ncol(slice_matrix), a0, a1, b0, b1, 0, 0, 0, 0, 0, 0, 0, 0, 0))\n+\n+ set_matrix[cont, 1:ncol(set_matrix)] = t(mylist)\ncont = cont + 1;\nswich = 0;\n}\n@@ -174,7 +197,7 @@ first_slices = function(Matrix[Double] val_matrix, Matrix[Double] set_matrix, Ma\n}\n-double_features = function(Matrix[Double] val_matrix, Matrix[Double] set_matrix, Matrix[Double] X0,Matrix[Double] Y, Matrix[Double] set_row, Matrix[Double] beta) return(Matrix[Double] set_matrix) {\n+double_features = function(Matrix[Double] val_matrix, Matrix[Double] set_matrix, Matrix[Double] X0, Matrix[Double] Y, Matrix[Double] set_row, Matrix[Double] beta, Integer paq) return(Matrix[Double] set_matrix) {\nvrow = nrow(val_matrix);\nvcol = ncol(val_matrix);\n@@ -186,44 +209,59 @@ double_features = function(Matrix[Double] val_matrix, Matrix[Double] set_matrix,\nb1 = col - 1;\nslice_number = 2;\n+ #combining subsets from set_matrix with the ones from val_matrix\n+ #avoiding repeating subsets, taking in account the amount of values in val_matrix or the paq value if activated.\n+ #new subsets checked are stored in set_matrix\n+\nfor (j in 1:vcol) {\nnum_value = as.scalar(val_matrix[1, j]);\nx = order(target = X0, by = j, decreasing = FALSE);\n+\n+ if (paq != 1)\n+ num_value = paq;\nif (j == num_value + 1)\nvrow = vrow - 1;\n- for (i in 2:vrow) {\n+ for (i in 2:num_value + 1) {\nif (i > 2 | j > 1)\nslice_number = slice_number + 1;\nfor (a in slice_number:totalrows) {\n- num_col = as.scalar(set_matrix[a, 4]);\n+ num_col = as.scalar(set_matrix[a, 5]);\nx_x = order(target = X0, by = num_col, decreasing = FALSE);\n- value_A = as.scalar(set_matrix[a, 3]);\n- a00 = as.scalar(set_matrix[a,7]);\n- a11 = as.scalar(set_matrix[a,8]);\n- #print(\"a0 y a1: \" + a00 + \" \" + a11)\n+ value_A0 = as.scalar(set_matrix[a, 3]);\n+ value_A1 = as.scalar(set_matrix[a, 4]);\n+ a00 = as.scalar(set_matrix[a, 8]);\n+ a11 = as.scalar(set_matrix[a, 9]);\nA = x_x[a00:a11, b0:b1];\nYa = x_x[a00:a11, col];\nif (nrow(set_matrix) <= cont)\nset_matrix = rbind(set_matrix, set_row);\n- value_B = as.scalar(val_matrix[i,j]);\n- a0 = index(x, j, value_B, 0);\n- a1 = index(x, j, value_B, 1);\n+ value_B1 = as.scalar(val_matrix[i, j]);\n+\n+ if (i == 2) {\n+ a0 = 1;\n+ value_B0 = value_B1;\n+ }\n+ else if (as.scalar(val_matrix[i - 1, j]) <= as.scalar(val_matrix[i, j])) {\n+ value_B0 = as.scalar(val_matrix[i - 1, j]);\n+ a0 = index(x, j, value_B0, 1);\n+ }\n+\n+ a1 = index(x, j, value_B1, 1);\nB = x[a0:a1, b0:b1];\nslice_matrix = rbind(A, B);\nYb = x[a0:a1, col];\nY0 = rbind(Ya, Yb);\nY = lmpredict(X = slice_matrix, w = beta, icpt = 0);\n-\n[error, diff] = standart_error(Y, Y0);\n-\n- set_matrix[cont, ] = t(as.matrix(list(diff, error, value_A, num_col, nrow(A),\n- ncol(A), a00, a11, b0, b1, value_B, j, nrow(B), ncol(B), a0, a1, b0, b1)));\n+ ## TODO - next code needs to be modified in order to show the total rows of the slice (as in previous function)\n+ set_matrix[cont, 1:ncol(set_matrix)] = t(as.matrix(list(diff, error, value_A0, value_A1, num_col, nrow(A),\n+ ncol(A), a00, a11, b0, b1, value_B0, value_B1, j, nrow(B), ncol(B), a0, a1, b0, b1)));\ncont = cont + 1;\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinSliceFinderTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinSliceFinderTest.java", "diff": "package org.tugraz.sysds.test.functions.builtin;\n+import java.util.HashMap;\n+import java.util.Map.Entry;\n+import java.util.Random;\n+\nimport org.junit.Test;\nimport org.tugraz.sysds.common.Types.ExecMode;\nimport org.tugraz.sysds.lops.LopProperties.ExecType;\n+import org.tugraz.sysds.runtime.matrix.data.MatrixValue;\nimport org.tugraz.sysds.test.AutomatedTestBase;\nimport org.tugraz.sysds.test.TestConfiguration;\nimport org.tugraz.sysds.test.TestUtils;\n-//package io;\n-import java.util.*;\n-\npublic class BuiltinSliceFinderTest extends AutomatedTestBase {\nprivate final static String TEST_NAME = \"slicefinder\";\nprivate final static String TEST_DIR = \"functions/builtin/\";\nprivate static final String TEST_CLASS_DIR = TEST_DIR + BuiltinSliceFinderTest.class.getSimpleName() + \"/\";\n- private final static int rows = 32000;\n+ private final static int rows = 2000;\nprivate final static int cols = 10;\n@Override\n@@ -61,10 +63,11 @@ public class BuiltinSliceFinderTest extends AutomatedTestBase {\nString dml_test_name = TEST_NAME;\nloadTestConfiguration(getTestConfiguration(TEST_NAME));\nString HOME = SCRIPT_DIR + TEST_DIR;\n+\ntry {\nloadTestConfiguration(getTestConfiguration(TEST_NAME));\nfullDMLScriptName = HOME + dml_test_name + \".dml\";\n- programArgs = new String[]{\"-explain\", \"-args\", input(\"AA\"), input(\"B\")};\n+ programArgs = new String[]{\"-explain\", \"-args\", input(\"A\"), input(\"B\"), input(\"Y0\"), output(\"C\")};\ndouble[][] A = TestUtils.ceil(getRandomMatrix(rows, cols, 0, 10, 1, 7));\ndouble[][] B = TestUtils.ceil(getRandomMatrix(10, 1, 0, 10, 1.0, 3));\ndouble[][] As = new double[rows][cols];\n@@ -78,12 +81,17 @@ public class BuiltinSliceFinderTest extends AutomatedTestBase {\ndouble AA[][] = new double[rows][cols+1];\n+ int value0 = 7;\n+ int value1 = 2;\n+ int coll0 = 5;\n+ int coll1 = 3;\n+\nswitch (test) {\ncase 1:\n- AA = modifyValue(A, Y,7,5);\n+ AA = modifyValue(A, Y,value0,coll0);\nbreak;\ncase 2:\n- AA = modifyValue(A, Y, 6, 3);\n+ AA = modifyValue(A, Y, value0, coll0);\nfor(int i = 0;i<rows;i++){\nfor(int j = 0; j < cols+1;j++){\nif(j == cols )\n@@ -92,10 +100,10 @@ public class BuiltinSliceFinderTest extends AutomatedTestBase {\nAs[i][j] = AA[i][j];\n}\n}\n- AA = modifyValue(As,Ys,3,3);\n+ AA = modifyValue(As,Ys,value1,coll0);\nbreak;\ncase 3:\n- AA = modifyValue(A, Y, 6, 3);\n+ AA = modifyValue(A, Y, value0, coll0);\nfor(int i = 0;i<rows;i++){\nfor(int j = 0; j < cols+1;j++){\nif(j == cols ){\n@@ -104,16 +112,38 @@ public class BuiltinSliceFinderTest extends AutomatedTestBase {\nAs[i][j] = AA[i][j];\n}\n}\n+\n}\n- AA = modifyValue(As,Ys,3,7);\n+ AA = modifyValue(As,Ys,value1,coll1);\nbreak;\n}\n-\n- writeInputMatrixWithMTD(\"AA\", AA, true);\n+ double[][] Y0 = new double[rows][1];\n+ for(int i = 0; i< rows;i++){\n+ Y0[i][0]= AA[i][10];\n+ }\n+ writeInputMatrixWithMTD(\"A\", A, true);\nwriteInputMatrixWithMTD(\"B\", B, true);\n-\n+ writeInputMatrixWithMTD(\"Y0\", Y0, true);\nrunTest(true, false, null, -1);\n+ HashMap<MatrixValue.CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(\"C\");\n+ double[][] ress = new double [5][20];\n+ for (Entry<MatrixValue.CellIndex, Double> a : dmlfile.entrySet()) {\n+ MatrixValue.CellIndex ci = a.getKey();\n+ ress[ci.row-1][ci.column-1] = a.getValue();\n+ }\n+ for(int i = 0; i < 5; i++){\n+ if(test == 1 ){\n+ if(ress[i][3] == value0 && ress[i][4] == coll0+1 && ress[i][12] == 0 && ress[i][13] == 0){\n+ System.out.print(\"Test passed!\");\n+ }\n+ }else{\n+ if(((ress[i][3] == value0 | ress[i][3] == value1) && (ress[i][4] == coll0+1 | ress[i][4] == coll1 +1)) && ((ress[i][12] == value0 | ress[i][12] == value1)\n+ && (ress[i][13] == coll0+1 | ress[i][13] == coll1+1))){\n+ System.out.print(\"Test passed!\");\n+ }\n+ }\n+ }\n}\nfinally {\nrtplatform = platformOld;\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/builtin/slicefinder.dml", "new_path": "src/test/scripts/functions/builtin/slicefinder.dml", "diff": "X = read($1);\nw = read($2);\n-ress = slicefinder(X0 = X,W = w, k = 5);\n-\n-#TODO write and check automatically\n-print(toString(ress));\n\\ No newline at end of file\n+y = read($3)\n+ress = slicefinder(X = X,W = w, Y = y, k = 5, paq = 1, S = 2);\n+write(ress, $4);\n\\ No newline at end of file\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -81,6 +81,7 @@ SYSTEMDS-110 New Builtin Functions\n* 113 Builtin functions for linear regression algorithms OK\n* 114 Builtin function for stepwise regression OK\n* 115 Builtin function for model debugging (slice finder) OK\n+ * 116 Builtin function for kmeans\nSYSTEMDS-120 Performance Features\n* 121 Avoid spark context creation on parfor result merge OK\n" }, { "change_type": "MODIFY", "old_path": "src/test/resources/log4j.properties", "new_path": "src/test/resources/log4j.properties", "diff": "#-------------------------------------------------------------\n# Define some default values that can be overridden by system properties\n-hadoop.root.logger=DEBUG,console\n+hadoop.root.logger=INFO,console\nhadoop.log.dir=.\nhadoop.log.file=hadoop.log\nhadoop.security.logger=OFF\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/builtins-reference.md", "new_path": "docs/builtins-reference.md", "diff": "@@ -29,6 +29,8 @@ limitations under the License.\n* [`lmDS`-Function](#lmds-function)\n* [`lmCG`-Function](#lmcg-function)\n* [`lmpredict`-Function](#lmpredict-function)\n+ * [`steplm`-Function](#steplm-function)\n+ * [`slicefinder`-Function](#slicefinder-function)\n# Introduction\n@@ -242,3 +244,81 @@ w = lm(X = X, y = y)\nyp = lmpredict(X, w)\n```\n+## `steplm`-Function\n+\n+The `steplm`-function (stepwise linear regression) implements a classical forward feature selection method.\n+This method iteratively runs what-if scenarios and greedily selects the next best feature until the Akaike\n+information criterion (AIC) does not improve anymore. Each configuration trains a regression model via `lm`,\n+which in turn calls either the closed form `lmDS` or iterative `lmGC`.\n+\n+### Usage\n+```r\n+steplm(X, y, icpt);\n+```\n+\n+### Arguments\n+| Name | Type | Default | Description |\n+| :------ | :------------- | -------- | :---------- |\n+| X | Matrix[Double] | required | Matrix of feature vectors. |\n+| y | Matrix[Double] | required | 1-column matrix of response values. |\n+| icpt | Integer | `0` | Intercept presence, shifting and rescaling the columns of X ([Details](#icpt-argument))|\n+| reg | Double | `1e-7` | Regularization constant (lambda) for L2-regularization. set to nonzero for highly dependent/sparse/numerous features|\n+| tol | Double | `1e-7` | Tolerance (epsilon); conjugate gradient procedure terminates early if L2 norm of the beta-residual is less than tolerance * its initial norm|\n+| maxi | Integer | `0` | Maximum number of conjugate gradient iterations. 0 = no maximum |\n+| verbose | Boolean | `TRUE` | If `TRUE` print messages are activated |\n+\n+### Returns\n+| Type | Description |\n+| :------------- | :---------- |\n+| Matrix[Double] | Matrix of regression parameters (the betas) and its size depend on `icpt` input value. (C in the example)|\n+| Matrix[Double] | Matrix of `selected` features ordered as computed by the algorithm. (S in the example)|\n+\n+##### `icpt`-Argument\n+\n+The *icpt-arg* can be set to 2 modes:\n+\n+ * 0 = no intercept, no shifting, no rescaling\n+ * 1 = add intercept, but neither shift nor rescale X\n+\n+##### `selected`-Output\n+\n+If the best AIC is achieved without any features the matrix of *selected* features contains 0. Moreover, in this case no further statistics will be produced\n+\n+### Example\n+```r\n+X = rand (rows = 50, cols = 10)\n+y = X %*% rand(rows=ncol(X), 1)\n+[C, S] = steplm(X = X, y = y, icpt = 1);\n+```\n+\n+## `slicefinder`-Function\n+\n+The `slicefinder`-function returns top-k worst performing subsets according to a model calculation.\n+\n+### Usage\n+```r\n+slicefinder(X,W, y, k, paq, S);\n+```\n+\n+### Arguments\n+| Name | Type | Default | Description |\n+| :------ | :------------- | -------- | :---------- |\n+| X | Matrix[Double] | required | Recoded dataset into Matrix |\n+| W | Matrix[Double] | required | Trained model |\n+| y | Matrix[Double] | required | 1-column matrix of response values. |\n+| k | Integer | 1 | Number of subsets required |\n+| paq | Integer | 1 | amount of values wanted for each col, if paq = 1 then its off |\n+| S | Integer | 2 | amount of subsets to combine (for now supported only 1 and 2) |\n+\n+### Returns\n+| Type | Description |\n+| :------------- | :---------- |\n+| Matrix[Double] | Matrix containing the information of top_K slices (relative error, standart error, value0, value1, col_number(sort), rows, cols,range_row,range_cols, value00, value01,col_number2(sort), rows2, cols2,range_row2,range_cols2) |\n+\n+### Usage\n+```r\n+X = rand (rows = 50, cols = 10)\n+y = X %*% rand(rows=ncol(X), 1)\n+w = lm(X = X, y = y)\n+ress = slicefinder(X = X,W = w, Y = y, k = 5, paq = 1, S = 2);\n+```\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/DataGenCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/DataGenCPInstruction.java", "diff": "@@ -370,8 +370,8 @@ public class DataGenCPInstruction extends UnaryCPInstruction {\nif (getSeed() == DataGenOp.UNSPECIFIED_SEED) {\n//generate pseudo-random seed (because not specified)\nif (runtimeSeed == null)\n- runtimeSeed = DataGenOp.generateRandomSeed();\n-\n+ runtimeSeed = (minValue == maxValue && sparsity == 1) ?\n+ DataGenOp.UNSPECIFIED_SEED : DataGenOp.generateRandomSeed();\nint position = (method == DataGenMethod.RAND) ? SEED_POSITION_RAND + 1 :\n(method == DataGenMethod.SAMPLE) ? SEED_POSITION_SAMPLE + 1 : 0;\ntmpInstStr = InstructionUtils.replaceOperand(\n" } ]

[ { "change_type": "MODIFY", "old_path": "bin/Readme.md", "new_path": "bin/Readme.md", "diff": "@@ -23,13 +23,13 @@ source tree checked out and SYSTEMDS_ROOT set.\nIn order to run SystemDS from your development directory and leave the\nSystemDS source tree untouched, the following setup could be used (example for bash):\n```shell script\n-$ export SYSTEMDS_ROOT=/home/<username>/systemds\n+$ export SYSTEMDS_ROOT=/home/$USER/systemds\n$ export PATH=$SYSTEMDS_ROOT/bin:$PATH\n```\nAlternatively, if the run scripts are invoked from the root of the\nSystemDS source tree, `SYSTEMDS_ROOT` does not need to be set (example for python):\n-`$ python bin/systemds-standalone.py -f <script.dml> <parameters>`\n+`$ python bin/systemds-standalone.py -f <script.dml> -args <parameters>`\nThe DML scripts residing in the directory `$SYSTEMDS_ROOT/scripts` will be found automatically by the run scripts.\n" }, { "change_type": "MODIFY", "old_path": "bin/systemds-standalone.py", "new_path": "bin/systemds-standalone.py", "diff": "-#!/usr/bin/env python\n+#!/usr/bin/env python3\n#-------------------------------------------------------------\n#\n# Copyright 2019 Graz University of Technology\n@@ -35,7 +35,7 @@ def default_classpath(systemds_root):\nbuild_lib = join(systemds_root, 'target', '*')\nlib_lib = join(systemds_root, 'target', 'lib', '*')\nhadoop_lib = join(systemds_root, 'target', 'lib', 'hadoop', '*')\n- sysds_jar = join(systemds_root, 'target', 'SystemDS.jar')\n+ sysds_jar = join(systemds_root, 'target', 'classes')\nreturn build_lib, lib_lib, hadoop_lib, sysds_jar\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinLmTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinLmTest.java", "diff": "@@ -28,7 +28,7 @@ import java.util.HashMap;\npublic class BuiltinLmTest extends AutomatedTestBase\n{\n- private final static String TEST_NAME = \"Lm\";\n+ private final static String TEST_NAME = \"lm\";\nprivate final static String TEST_DIR = \"functions/builtin/\";\nprivate static final String TEST_CLASS_DIR = TEST_DIR + BuiltinLmTest.class.getSimpleName() + \"/\";\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinNormalizeTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinNormalizeTest.java", "diff": "@@ -28,7 +28,7 @@ import org.tugraz.sysds.test.TestUtils;\npublic class BuiltinNormalizeTest extends AutomatedTestBase\n{\n- private final static String TEST_NAME = \"Normalize\";\n+ private final static String TEST_NAME = \"normalize\";\nprivate final static String TEST_DIR = \"functions/builtin/\";\nprivate static final String TEST_CLASS_DIR = TEST_DIR + BuiltinNormalizeTest.class.getSimpleName() + \"/\";\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinOutlierTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinOutlierTest.java", "diff": "@@ -28,7 +28,7 @@ import org.tugraz.sysds.test.TestUtils;\npublic class BuiltinOutlierTest extends AutomatedTestBase\n{\n- private final static String TEST_NAME = \"Outlier\";\n+ private final static String TEST_NAME = \"outlier\";\nprivate final static String TEST_DIR = \"functions/builtin/\";\nprivate static final String TEST_CLASS_DIR = TEST_DIR + BuiltinOutlierTest.class.getSimpleName() + \"/\";\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinScaleTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinScaleTest.java", "diff": "@@ -28,7 +28,7 @@ import org.tugraz.sysds.test.TestUtils;\npublic class BuiltinScaleTest extends AutomatedTestBase\n{\n- private final static String TEST_NAME = \"Scale\";\n+ private final static String TEST_NAME = \"scale\";\nprivate final static String TEST_DIR = \"functions/builtin/\";\nprivate final static String TEST_CLASS_DIR = TEST_DIR + BuiltinScaleTest.class.getSimpleName() + \"/\";\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinSigmoidTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinSigmoidTest.java", "diff": "@@ -28,7 +28,7 @@ import org.tugraz.sysds.test.TestUtils;\npublic class BuiltinSigmoidTest extends AutomatedTestBase\n{\n- private final static String TEST_NAME = \"Sigmoid\";\n+ private final static String TEST_NAME = \"sigmoid\";\nprivate final static String TEST_DIR = \"functions/builtin/\";\nprivate static final String TEST_CLASS_DIR = TEST_DIR + BuiltinSigmoidTest.class.getSimpleName() + \"/\";\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinWinsorizeTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinWinsorizeTest.java", "diff": "@@ -28,7 +28,7 @@ import org.tugraz.sysds.test.TestUtils;\npublic class BuiltinWinsorizeTest extends AutomatedTestBase\n{\n- private final static String TEST_NAME = \"Winsorize\";\n+ private final static String TEST_NAME = \"winsorize\";\nprivate final static String TEST_DIR = \"functions/builtin/\";\nprivate static final String TEST_CLASS_DIR = TEST_DIR + BuiltinWinsorizeTest.class.getSimpleName() + \"/\";\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/MultipleBuiltinsTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/MultipleBuiltinsTest.java", "diff": "@@ -28,7 +28,7 @@ import org.tugraz.sysds.test.TestUtils;\npublic class MultipleBuiltinsTest extends AutomatedTestBase\n{\n- private final static String TEST_NAME = \"MultipleBuiltins\";\n+ private final static String TEST_NAME = \"multipleBuiltins\";\nprivate final static String TEST_DIR = \"functions/builtin/\";\nprivate static final String TEST_CLASS_DIR = TEST_DIR + MultipleBuiltinsTest.class.getSimpleName() + \"/\";\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/data/TimeTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/data/TimeTest.java", "diff": "@@ -25,7 +25,7 @@ import org.tugraz.sysds.test.TestConfiguration;\npublic class TimeTest extends AutomatedTestBase\n{\n- private final static String TEST_NAME = \"Time\";\n+ private final static String TEST_NAME = \"time\";\nprivate final static String TEST_DIR = \"functions/data/\";\nprivate final static String TEST_CLASS_DIR = TEST_DIR + TimeTest.class.getSimpleName() + \"/\";\n" }, { "change_type": "RENAME", "old_path": "src/test/scripts/functions/builtin/MultipleBuiltins.R", "new_path": "src/test/scripts/functions/builtin/multipleBuiltins.R", "diff": "" }, { "change_type": "RENAME", "old_path": "src/test/scripts/functions/builtin/MultipleBuiltins.dml", "new_path": "src/test/scripts/functions/builtin/multipleBuiltins.dml", "diff": "" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/installDependencies.R", "new_path": "src/test/scripts/installDependencies.R", "diff": "@@ -33,4 +33,6 @@ custom_install(\"batch\");\ncustom_install(\"matrixStats\");\ncustom_install(\"outliers\");\ncustom_install(\"caret\");\n+custom_install(\"Sigmoid\");\n+custom_install(\"DescTools\");\n" } ]

[ { "change_type": "MODIFY", "old_path": "bin/Readme.md", "new_path": "bin/Readme.md", "diff": "@@ -58,4 +58,4 @@ $ systemds-standalone.sh Univar-Stats.dml -nvargs X=data/haberman.data TYPES=dat\n```\n#### Using Intel MKL native instructions\n-To use the MKL acceleration download and install the latest MKL libtary from [1]\n\\ No newline at end of file\n+To use the MKL acceleration download and install the latest MKL library from [1]\n\\ No newline at end of file\n" }, { "change_type": "MODIFY", "old_path": "pom.xml", "new_path": "pom.xml", "diff": "<version>18</version>\n</parent>\n<groupId>org.tugraz.systemds</groupId>\n- <version>0.1.0</version>\n+ <version>0.1.0-SNAPSHOT</version>\n<artifactId>systemds</artifactId>\n<packaging>jar</packaging>\n<name>SystemDS</name>\n<groupId>org.apache.maven.plugins</groupId>\n<artifactId>maven-failsafe-plugin</artifactId>\n- <version>2.17</version>\n-\n<executions>\n<execution>\n<goals>\n" } ]

[ { "change_type": "ADD", "old_path": null, "new_path": "src/assembly/extra.xml", "diff": "+<!--\n+ * Modifications Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed to the Apache Software Foundation (ASF) under one\n+ * or more contributor license agreements. See the NOTICE file\n+ * distributed with this work for additional information\n+ * regarding copyright ownership. The ASF licenses this file\n+ * to you under the Apache License, Version 2.0 (the\n+ * \"License\"); you may not use this file except in compliance\n+ * with the License. You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing,\n+ * software distributed under the License is distributed on an\n+ * \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+ * KIND, either express or implied. See the License for the\n+ * specific language governing permissions and limitations\n+ * under the License.\n+-->\n+<assembly\n+ xmlns=\"http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.3\"\n+ xmlns:xsi=\"http://www.w3.org/2001/XMLSchema-instance\"\n+ xsi:schemaLocation=\"http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.3 http://maven.apache.org/xsd/assembly-1.1.3.xsd\">\n+ <!-- Assembly file for the SystemML extra jar artifact. -->\n+ <id>extra</id>\n+\n+ <formats>\n+ <format>jar</format>\n+ </formats>\n+\n+ <includeBaseDirectory>false</includeBaseDirectory>\n+\n+ <fileSets>\n+ <fileSet>\n+ <directory>${basedir}/src/assembly/extra</directory>\n+ <includes>\n+ <include>LICENSE</include>\n+ <include>NOTICE</include>\n+ </includes>\n+ <outputDirectory>./META-INF</outputDirectory>\n+ </fileSet>\n+\n+ </fileSets>\n+\n+ <!-- Include JCuda Jars -->\n+ <dependencySets>\n+ <dependencySet>\n+ <includes>\n+ <include>org.jcuda:jcuda:jar:${jcuda.version}</include>\n+ <include>org.jcuda:jcublas:jar:${jcuda.version}</include>\n+ <include>org.jcuda:jcusparse:jar:${jcuda.version}</include>\n+ <include>org.jcuda:jcusolver:jar:${jcuda.version}</include>\n+ <include>org.jcuda:jcudnn:jar:${jcuda.version}</include>\n+\n+ <!-- windows specific jcuda jars -->\n+ <include>org.jcuda:jcuda-natives:jar:windows-x86_64:${jcuda.version}</include>\n+ <include>org.jcuda:jcublas-natives:jar:windows-x86_64:${jcuda.version}</include>\n+ <include>org.jcuda:jcusparse-natives:jar:windows-x86_64:${jcuda.version}</include>\n+ <include>org.jcuda:jcusolver-natives:jar:windows-x86_64:${jcuda.version}</include>\n+ <include>org.jcuda:jcudnn-natives:jar:windows-x86_64:${jcuda.version}</include>\n+\n+ <!-- linux x86_64 specific jcuda jars -->\n+ <include>org.jcuda:jcuda-natives:jar:linux-x86_64:${jcuda.version}</include>\n+ <include>org.jcuda:jcublas-natives:jar:linux-x86_64:${jcuda.version}</include>\n+ <include>org.jcuda:jcusparse-natives:jar:linux-x86_64:${jcuda.version}</include>\n+ <include>org.jcuda:jcusolver-natives:jar:linux-x86_64:${jcuda.version}</include>\n+ <include>org.jcuda:jcudnn-natives:jar:linux-x86_64:${jcuda.version}</include>\n+\n+ <!-- linux ppc_64le specific jcuda jars -->\n+ <include>org.jcuda:jcuda-natives:jar:linux-ppc_64:${jcuda.version}</include>\n+ <include>org.jcuda:jcublas-natives:jar:linux-ppc_64:${jcuda.version}</include>\n+ <include>org.jcuda:jcusparse-natives:jar:linux-ppc_64:${jcuda.version}</include>\n+ <include>org.jcuda:jcusolver-natives:jar:linux-ppc_64:${jcuda.version}</include>\n+ <include>org.jcuda:jcudnn-natives:jar:linux-ppc_64:${jcuda.version}</include>\n+ </includes>\n+ <unpack>true</unpack>\n+ <scope>compile</scope>\n+ </dependencySet>\n+ </dependencySets>\n+\n+</assembly>\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/assembly/extra/LICENSE", "diff": "+ Apache License\n+ Version 2.0, January 2004\n+ http://www.apache.org/licenses/\n+\n+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION\n+\n+ 1. Definitions.\n+\n+ \"License\" shall mean the terms and conditions for use, reproduction,\n+ and distribution as defined by Sections 1 through 9 of this document.\n+\n+ \"Licensor\" shall mean the copyright owner or entity authorized by\n+ the copyright owner that is granting the License.\n+\n+ \"Legal Entity\" shall mean the union of the acting entity and all\n+ other entities that control, are controlled by, or are under common\n+ control with that entity. For the purposes of this definition,\n+ \"control\" means (i) the power, direct or indirect, to cause the\n+ direction or management of such entity, whether by contract or\n+ otherwise, or (ii) ownership of fifty percent (50%) or more of the\n+ outstanding shares, or (iii) beneficial ownership of such entity.\n+\n+ \"You\" (or \"Your\") shall mean an individual or Legal Entity\n+ exercising permissions granted by this License.\n+\n+ \"Source\" form shall mean the preferred form for making modifications,\n+ including but not limited to software source code, documentation\n+ source, and configuration files.\n+\n+ \"Object\" form shall mean any form resulting from mechanical\n+ transformation or translation of a Source form, including but\n+ not limited to compiled object code, generated documentation,\n+ and conversions to other media types.\n+\n+ \"Work\" shall mean the work of authorship, whether in Source or\n+ Object form, made available under the License, as indicated by a\n+ copyright notice that is included in or attached to the work\n+ (an example is provided in the Appendix below).\n+\n+ \"Derivative Works\" shall mean any work, whether in Source or Object\n+ form, that is based on (or derived from) the Work and for which the\n+ editorial revisions, annotations, elaborations, or other modifications\n+ represent, as a whole, an original work of authorship. For the purposes\n+ of this License, Derivative Works shall not include works that remain\n+ separable from, or merely link (or bind by name) to the interfaces of,\n+ the Work and Derivative Works thereof.\n+\n+ \"Contribution\" shall mean any work of authorship, including\n+ the original version of the Work and any modifications or additions\n+ to that Work or Derivative Works thereof, that is intentionally\n+ submitted to Licensor for inclusion in the Work by the copyright owner\n+ or by an individual or Legal Entity authorized to submit on behalf of\n+ the copyright owner. For the purposes of this definition, \"submitted\"\n+ means any form of electronic, verbal, or written communication sent\n+ to the Licensor or its representatives, including but not limited to\n+ communication on electronic mailing lists, source code control systems,\n+ and issue tracking systems that are managed by, or on behalf of, the\n+ Licensor for the purpose of discussing and improving the Work, but\n+ excluding communication that is conspicuously marked or otherwise\n+ designated in writing by the copyright owner as \"Not a Contribution.\"\n+\n+ \"Contributor\" shall mean Licensor and any individual or Legal Entity\n+ on behalf of whom a Contribution has been received by Licensor and\n+ subsequently incorporated within the Work.\n+\n+ 2. Grant of Copyright License. Subject to the terms and conditions of\n+ this License, each Contributor hereby grants to You a perpetual,\n+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable\n+ copyright license to reproduce, prepare Derivative Works of,\n+ publicly display, publicly perform, sublicense, and distribute the\n+ Work and such Derivative Works in Source or Object form.\n+\n+ 3. Grant of Patent License. Subject to the terms and conditions of\n+ this License, each Contributor hereby grants to You a perpetual,\n+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable\n+ (except as stated in this section) patent license to make, have made,\n+ use, offer to sell, sell, import, and otherwise transfer the Work,\n+ where such license applies only to those patent claims licensable\n+ by such Contributor that are necessarily infringed by their\n+ Contribution(s) alone or by combination of their Contribution(s)\n+ with the Work to which such Contribution(s) was submitted. If You\n+ institute patent litigation against any entity (including a\n+ cross-claim or counterclaim in a lawsuit) alleging that the Work\n+ or a Contribution incorporated within the Work constitutes direct\n+ or contributory patent infringement, then any patent licenses\n+ granted to You under this License for that Work shall terminate\n+ as of the date such litigation is filed.\n+\n+ 4. Redistribution. You may reproduce and distribute copies of the\n+ Work or Derivative Works thereof in any medium, with or without\n+ modifications, and in Source or Object form, provided that You\n+ meet the following conditions:\n+\n+ (a) You must give any other recipients of the Work or\n+ Derivative Works a copy of this License; and\n+\n+ (b) You must cause any modified files to carry prominent notices\n+ stating that You changed the files; and\n+\n+ (c) You must retain, in the Source form of any Derivative Works\n+ that You distribute, all copyright, patent, trademark, and\n+ attribution notices from the Source form of the Work,\n+ excluding those notices that do not pertain to any part of\n+ the Derivative Works; and\n+\n+ (d) If the Work includes a \"NOTICE\" text file as part of its\n+ distribution, then any Derivative Works that You distribute must\n+ include a readable copy of the attribution notices contained\n+ within such NOTICE file, excluding those notices that do not\n+ pertain to any part of the Derivative Works, in at least one\n+ of the following places: within a NOTICE text file distributed\n+ as part of the Derivative Works; within the Source form or\n+ documentation, if provided along with the Derivative Works; or,\n+ within a display generated by the Derivative Works, if and\n+ wherever such third-party notices normally appear. The contents\n+ of the NOTICE file are for informational purposes only and\n+ do not modify the License. You may add Your own attribution\n+ notices within Derivative Works that You distribute, alongside\n+ or as an addendum to the NOTICE text from the Work, provided\n+ that such additional attribution notices cannot be construed\n+ as modifying the License.\n+\n+ You may add Your own copyright statement to Your modifications and\n+ may provide additional or different license terms and conditions\n+ for use, reproduction, or distribution of Your modifications, or\n+ for any such Derivative Works as a whole, provided Your use,\n+ reproduction, and distribution of the Work otherwise complies with\n+ the conditions stated in this License.\n+\n+ 5. Submission of Contributions. Unless You explicitly state otherwise,\n+ any Contribution intentionally submitted for inclusion in the Work\n+ by You to the Licensor shall be under the terms and conditions of\n+ this License, without any additional terms or conditions.\n+ Notwithstanding the above, nothing herein shall supersede or modify\n+ the terms of any separate license agreement you may have executed\n+ with Licensor regarding such Contributions.\n+\n+ 6. Trademarks. This License does not grant permission to use the trade\n+ names, trademarks, service marks, or product names of the Licensor,\n+ except as required for reasonable and customary use in describing the\n+ origin of the Work and reproducing the content of the NOTICE file.\n+\n+ 7. Disclaimer of Warranty. Unless required by applicable law or\n+ agreed to in writing, Licensor provides the Work (and each\n+ Contributor provides its Contributions) on an \"AS IS\" BASIS,\n+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or\n+ implied, including, without limitation, any warranties or conditions\n+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A\n+ PARTICULAR PURPOSE. You are solely responsible for determining the\n+ appropriateness of using or redistributing the Work and assume any\n+ risks associated with Your exercise of permissions under this License.\n+\n+ 8. Limitation of Liability. In no event and under no legal theory,\n+ whether in tort (including negligence), contract, or otherwise,\n+ unless required by applicable law (such as deliberate and grossly\n+ negligent acts) or agreed to in writing, shall any Contributor be\n+ liable to You for damages, including any direct, indirect, special,\n+ incidental, or consequential damages of any character arising as a\n+ result of this License or out of the use or inability to use the\n+ Work (including but not limited to damages for loss of goodwill,\n+ work stoppage, computer failure or malfunction, or any and all\n+ other commercial damages or losses), even if such Contributor\n+ has been advised of the possibility of such damages.\n+\n+ 9. Accepting Warranty or Additional Liability. While redistributing\n+ the Work or Derivative Works thereof, You may choose to offer,\n+ and charge a fee for, acceptance of support, warranty, indemnity,\n+ or other liability obligations and/or rights consistent with this\n+ License. However, in accepting such obligations, You may act only\n+ on Your own behalf and on Your sole responsibility, not on behalf\n+ of any other Contributor, and only if You agree to indemnify,\n+ defend, and hold each Contributor harmless for any liability\n+ incurred by, or claims asserted against, such Contributor by reason\n+ of your accepting any such warranty or additional liability.\n+\n+ END OF TERMS AND CONDITIONS\n+\n+ APPENDIX: How to apply the Apache License to your work.\n+\n+ To apply the Apache License to your work, attach the following\n+ boilerplate notice, with the fields enclosed by brackets \"{}\"\n+ replaced with your own identifying information. (Don't include\n+ the brackets!) The text should be enclosed in the appropriate\n+ comment syntax for the file format. We also recommend that a\n+ file or class name and description of purpose be included on the\n+ same \"printed page\" as the copyright notice for easier\n+ identification within third-party archives.\n+\n+ Copyright {yyyy} {name of copyright owner}\n+\n+ Licensed under the Apache License, Version 2.0 (the \"License\");\n+ you may not use this file except in compliance with the License.\n+ You may obtain a copy of the License at\n+\n+ http://www.apache.org/licenses/LICENSE-2.0\n+\n+ Unless required by applicable law or agreed to in writing, software\n+ distributed under the License is distributed on an \"AS IS\" BASIS,\n+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+ See the License for the specific language governing permissions and\n+ limitations under the License.\n+\n+===============================================================================\n+\n+The proto file (src/main/proto/caffe/caffe.proto) is part of Caffe project,\n+which is used to generate caffe java package.\n+Caffe are distributed under the below license.\n+\n+COPYRIGHT\n+\n+All contributions by the University of California:\n+Copyright (c) 2014-2017 The Regents of the University of California (Regents)\n+All rights reserved.\n+\n+All other contributions:\n+Copyright (c) 2014-2017, the respective contributors\n+All rights reserved.\n+\n+Caffe uses a shared copyright model: each contributor holds copyright over\n+their contributions to Caffe. The project versioning records all such\n+contribution and copyright details. If a contributor wants to further mark\n+their specific copyright on a particular contribution, they should indicate\n+their copyright solely in the commit message of the change when it is\n+committed.\n+\n+LICENSE\n+\n+Redistribution and use in source and binary forms, with or without\n+modification, are permitted provided that the following conditions are met:\n+\n+1. Redistributions of source code must retain the above copyright notice, this\n+ list of conditions and the following disclaimer.\n+2. Redistributions in binary form must reproduce the above copyright notice,\n+ this list of conditions and the following disclaimer in the documentation\n+ and/or other materials provided with the distribution.\n+\n+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\" AND\n+ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED\n+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE\n+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR\n+ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES\n+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;\n+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND\n+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT\n+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS\n+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\n+\n+CONTRIBUTION AGREEMENT\n+\n+By contributing to the BVLC/caffe repository through pull-request, comment,\n+or otherwise, the contributor releases their content to the\n+license and copyright terms herein.\n+\n+===============================================================================\n+\n+The following compile-scope dependencies come under the MIT License\n+\n+JCuda (jcuda.org)\n+\n+org.jcuda:jcuda:0.9.0\n+org.jcuda:jcublas:0.9.0\n+org.jcuda:jcufft:0.9.0\n+org.jcuda:jcusparse:0.9.0\n+org.jcuda:jcusolver:0.9.0\n+org.jcuda:jcurand:0.9.0\n+org.jcuda:jnvgraph:0.9.0\n+org.jcuda:jcudnn:0.9.0\n+org.jcuda:jcuda-natives:0.9.0\n+org.jcuda:jcublas-natives:0.9.0\n+org.jcuda:jcufft-natives:0.9.0\n+org.jcuda:jcusparse-natives:0.9.0\n+org.jcuda:jcusolver-natives:0.9.0\n+org.jcuda:jcurand-natives:0.9.0\n+org.jcuda:jnvgraph-natives:0.9.0\n+org.jcuda:jcudnn-natives:0.9.0\n+\n+\n+The MIT License (MIT)\n+\n+Copyright (c) 2008-2016 Marco Hutter - http://www.jcuda.org\n+\n+Permission is hereby granted, free of charge, to any person obtaining a copy\n+of this software and associated documentation files (the \"Software\"), to deal\n+in the Software without restriction, including without limitation the rights\n+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell\n+copies of the Software, and to permit persons to whom the Software is\n+furnished to do so, subject to the following conditions:\n+\n+The above copyright notice and this permission notice shall be included in all\n+copies or substantial portions of the Software.\n+\n+THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\n+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\n+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\n+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\n+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\n+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\n+SOFTWARE.\n+\n+===============================================================================\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/assembly/extra/NOTICE", "diff": "+Apache SystemDS\n+Copyright [2015-2018] The Apache Software Foundation\n+\n+This product includes software developed at\n+The Apache Software Foundation (http://www.apache.org/).\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/api/jmlc/Connection.java", "new_path": "src/main/java/org/tugraz/sysds/api/jmlc/Connection.java", "diff": "@@ -69,10 +69,9 @@ import org.tugraz.sysds.runtime.util.UtilFunctions;\n/**\n* Interaction with SystemDS using the JMLC (Java Machine Learning Connector) API is initiated with\n- * a {@link Connection} object. The JMLC API is patterned\n- * after JDBC. A DML script is precompiled by calling\n- * the {@link #prepareScript(String, String[], String[], boolean)}\n- * method or the {@link #prepareScript(String, Map, String[], String[], boolean)}\n+ * a {@link Connection} object. The JMLC API is designed after JDBC. A DML script is precompiled by calling\n+ * the {@link #prepareScript(String, String[], String[])}\n+ * method or the {@link #prepareScript(String, Map, String[], String[])}\n* method on the {@link Connection} object, which returns a\n* {@link PreparedScript} object. Note that this is similar to calling\n* a {@code prepareStatement} method on a JDBC {@code Connection} object.\n@@ -86,14 +85,6 @@ import org.tugraz.sysds.runtime.util.UtilFunctions;\n* {@code ResultSet}. Data can be read from a {@link ResultVariables} object by calling\n* its {@link ResultVariables#getFrame(String) getFrame} and\n* {@link ResultVariables#getMatrix(String) getMatrix} methods.\n- *\n- * <p>\n- * For examples, please see the following:\n- * <ul>\n- * <li>JMLC JUnit test cases (org.tugraz.sysds.test.integration.functions.jmlc)</li>\n- * <li><a target=\"_blank\" href=\"http://apache.github.io/systemml/jmlc.html\">JMLC section\n- * of SystemDS online documentation</a></li>\n- * </ul>\n*/\npublic class Connection implements Closeable\n{\n" } ]

[ { "change_type": "MODIFY", "old_path": "dev/release/release-build.sh", "new_path": "dev/release/release-build.sh", "diff": "@@ -311,7 +311,7 @@ if [[ \"$RELEASE_PREPARE\" == \"true\" ]]; then\nexit 0\nfi\n-ToDo: fix release deployment\n+#ToDo: fix release deployment\nif [[ \"$RELEASE_PUBLISH\" == \"true\" ]]; then\necho \"Preparing release $RELEASE_VERSION\"\n# Checkout code\n" }, { "change_type": "MODIFY", "old_path": "src/assembly/bin.xml", "new_path": "src/assembly/bin.xml", "diff": "<outputDirectory>scripts</outputDirectory>\n</fileSet>\n+ <fileSet>\n+ <directory>${basedir}/conf</directory>\n+ <includes>\n+ <include>log4j.properties</include>\n+ <include>SystemDS-config.xml</include>\n+ </includes>\n+ <outputDirectory>.</outputDirectory>\n+ </fileSet>\n+\n<fileSet>\n<directory>${basedir}/src/main/standalone</directory>\n<includes>\n<include>log4j.properties</include>\n- <include>README.txt</include>\n<include>SystemDS-config.xml</include>\n</includes>\n<outputDirectory>.</outputDirectory>\n</fileSet>\n+\n<fileSet>\n<directory>${basedir}/src/test/config/hadoop_bin_windows/bin</directory>\n<includes>\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/main/standalone/runStandaloneSystemDS.bat", "diff": "+::-------------------------------------------------------------\n+::\n+:: Modifications Copyright 2019 Graz University of Technology\n+::\n+:: Licensed to the Apache Software Foundation (ASF) under one\n+:: or more contributor license agreements. See the NOTICE file\n+:: distributed with this work for additional information\n+:: regarding copyright ownership. The ASF licenses this file\n+:: to you under the Apache License, Version 2.0 (the\n+:: \"License\"); you may not use this file except in compliance\n+:: with the License. You may obtain a copy of the License at\n+::\n+:: http://www.apache.org/licenses/LICENSE-2.0\n+::\n+:: Unless required by applicable law or agreed to in writing,\n+:: software distributed under the License is distributed on an\n+:: \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+:: KIND, either express or implied. See the License for the\n+:: specific language governing permissions and limitations\n+:: under the License.\n+::\n+::-------------------------------------------------------------\n+\n+@ECHO OFF\n+\n+IF \"%~1\" == \"\" GOTO Err\n+IF \"%~1\" == \"-help\" GOTO Msg\n+IF \"%~1\" == \"-h\" GOTO Msg\n+\n+setLocal EnableDelayedExpansion\n+\n+SET HADOOP_HOME=%CD%/lib/hadoop\n+\n+set CLASSPATH=./lib/*\n+\n+set LOG4JPROP=log4j.properties\n+\n+for /f \"tokens=1,* delims= \" %%a in (\"%*\") do set ALLBUTFIRST=%%b\n+\n+IF \"%SYSTEMDS_STANDALONE_OPTS%\" == \"\" (\n+ SET SYSTEMDS_STANDALONE_OPTS=-Xmx4g -Xms4g -Xmn400m\n+)\n+\n+:: construct the java command with options and arguments\n+set CMD=java %SYSTEMDS_STANDALONE_OPTS% ^\n+ -cp %CLASSPATH% ^\n+ -Dlog4j.configuration=file:%LOG4JPROP% ^\n+ org.apache.sysml.api.DMLScript ^\n+ -f %1 ^\n+ -exec singlenode ^\n+ -config SystemDS-config.xml ^\n+ %ALLBUTFIRST%\n+\n+:: execute the java command\n+%CMD%\n+\n+:: if there was an error, display the full java command\n+::IF ERRORLEVEL 1 (\n+:: ECHO Failed to run SystemDS. Exit code: %ERRORLEVEL%\n+:: SET LF=^\n+::\n+::\n+:: :: keep empty lines above for the line breaks\n+:: ECHO %CMD: =!LF! %\n+:: EXIT /B %ERRORLEVEL%\n+::)\n+\n+GOTO End\n+\n+:Err\n+ECHO Wrong Usage. Please provide DML filename to be executed.\n+GOTO Msg\n+\n+:Msg\n+ECHO Usage: runStandaloneSystemDS.bat ^<dml-filename^> [arguments] [-help]\n+ECHO Default Java options (-Xmx4g -Xms4g -Xmn400m) can be overridden by setting SYSTEMDS_STANDALONE_OPTS.\n+ECHO Script internally invokes 'java [SYSTEMDS_STANDALONE_OPTS] -cp ./lib/* -Dlog4j.configuration=file:log4j.properties org.apache.sysml.api.DMLScript -f ^<dml-filename^> -exec singlenode -config SystemDS-config.xml [arguments]'\n+\n+:End\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/main/standalone/runStandaloneSystemDS.sh", "diff": "+#!/bin/bash\n+#-------------------------------------------------------------\n+#\n+# Modifications Copyright 2019 Graz University of Technology\n+#\n+# Licensed to the Apache Software Foundation (ASF) under one\n+# or more contributor license agreements. See the NOTICE file\n+# distributed with this work for additional information\n+# regarding copyright ownership. The ASF licenses this file\n+# to you under the Apache License, Version 2.0 (the\n+# \"License\"); you may not use this file except in compliance\n+# with the License. You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing,\n+# software distributed under the License is distributed on an\n+# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+# KIND, either express or implied. See the License for the\n+# specific language governing permissions and limitations\n+# under the License.\n+#\n+#-------------------------------------------------------------\n+\n+# error help print\n+printUsageExit()\n+{\n+cat << EOF\n+Usage: $0 <dml-filename> [arguments] [-help]\n+ -help - Print this usage message and exit\n+Default Java options (-Xmx4g -Xms4g -Xmn400m) can be overridden by setting SYSTEMDS_STANDALONE_OPTS.\n+EOF\n+ exit 1\n+}\n+# Script internally invokes 'java [SYSTEMDS_STANDALONE_OPTS] -jar StandaloneSystemDS.jar -f <dml-filename> -exec singlenode -config=SystemDS-config.xml [arguments]'\n+\n+while getopts \"h:\" options; do\n+ case $options in\n+ h ) echo Warning: Help requested. Will exit after usage message;\n+ printUsageExit\n+ ;;\n+ \\? ) echo Warning: Help requested. Will exit after usage message;\n+ printUsageExit\n+ ;;\n+ * ) echo Error: Unexpected error while processing options;\n+ esac\n+done\n+\n+if [ -z $1 ] ; then\n+ echo \"Wrong Usage.\";\n+ printUsageExit;\n+fi\n+\n+# Peel off first argument so that $@ contains arguments to DML script\n+SCRIPT_FILE=$1\n+shift\n+\n+# Build up a classpath with all included libraries\n+CURRENT_PATH=$( cd $(dirname $0) ; pwd -P )\n+\n+CLASSPATH=\"\"\n+for f in ${CURRENT_PATH}/lib/*.jar; do\n+ CLASSPATH=${CLASSPATH}:$f;\n+done\n+\n+LOG4JPROP=log4j.properties\n+\n+# set default java opts if none supplied\n+if [ -z \"$SYSTEMDS_STANDALONE_OPTS\" ] ; then\n+ SYSTEMDS_STANDALONE_OPTS=\"-Xmx4g -Xms4g -Xmn400m\"\n+fi;\n+\n+# invoke the jar with options and arguments\n+CMD=\"\\\n+java ${SYSTEMDS_STANDALONE_OPTS} \\\n+-cp ${CLASSPATH} \\\n+-Dlog4j.configuration=file:${LOG4JPROP} \\\n+org.apache.sysml.api.DMLScript \\\n+-f ${SCRIPT_FILE} \\\n+-exec singlenode \\\n+-config $CURRENT_PATH\"/SystemDS-config.xml\" \\\n+$@\"\n+\n+$CMD\n" } ]

[ { "change_type": "MODIFY", "old_path": "conf/SystemDS-config.xml", "new_path": "conf/SystemDS-config.xml", "diff": "<!-- compiler optimization level, valid values: 0 | 1 | 2 | 3 | 4, default: 2 -->\n<sysds.optlevel>2</sysds.optlevel>\n- <!-- default number of reduce tasks per MR job, default: 2 x number of nodes -->\n- <sysds.numreducers>10</sysds.numreducers>\n-\n- <!-- override jvm reuse flag for specific MR jobs, valid values: true | false -->\n- <sysds.jvmreuse>false</sysds.jvmreuse>\n-\n<!-- default block dim for binary block files -->\n<sysds.defaultblocksize>1000</sysds.defaultblocksize>\n- <!-- run systemds control program as yarn appmaster, in case of MR1 always falls back to client, please disable for debug mode -->\n- <sysds.yarn.appmaster>false</sysds.yarn.appmaster>\n-\n- <!-- maximum jvm heap size of the dml yarn appmaster in MB, the requested memory is 1.5x this parameter -->\n- <sysds.yarn.appmaster.mem>2048</sysds.yarn.appmaster.mem>\n-\n- <!-- maximum jvm heap size of the map/reduce tasks in MB, the requested memory is 1.5x this parameter, negative values ignored -->\n- <sysds.yarn.mapreduce.mem>2048</sysds.yarn.mapreduce.mem>\n-\n- <!-- yarn application submission queue, relevant for default capacity scheduler -->\n- <sysds.yarn.app.queue>default</sysds.yarn.app.queue>\n-\n<!-- enables multi-threaded operations in singlenode control program -->\n<sysds.cp.parallel.ops>true</sysds.cp.parallel.ops>\n<!-- custom directory where BLAS libraries are available, experimental feature (options: absolute directory path or none). If set to none, we use standard LD_LIBRARY_PATH. -->\n<sysds.native.blas.directory>none</sysds.native.blas.directory>\n- <!-- prints finegrained statistics information (includes extra GPU information and extra statistics information for Deep Neural Networks done in CP mode) -->\n- <sysds.stats.finegrained>false</sysds.stats.finegrained>\n-\n<!-- sets the GPUs to use per process, -1 for all GPUs, a specific GPU number (5), a range (eg: 0-2) or a comma separated list (eg: 0,2,4)-->\n<sysds.gpu.availableGPUs>-1</sysds.gpu.availableGPUs>\n<!-- maximum wrap length for instruction and miscellaneous timer column of statistics -->\n<sysds.stats.maxWrapLength>30</sysds.stats.maxWrapLength>\n- <!-- Advanced optimization: fraction of driver memory to use for caching (default: 0.15) -->\n- <sysds.caching.bufferSize>0.15</sysds.caching.bufferSize>\n-\n<!-- Advanced optimization: fraction of driver memory to use for GPU shadow buffer. This optimization is ignored for double precision.\nBy default, it is disabled (hence set to 0.0). If you intend to train network larger than GPU memory size, consider using single precision and setting this to 0.1 -->\n<sysds.gpu.eviction.shadow.bufferSize>0.0</sysds.gpu.eviction.shadow.bufferSize>\n" }, { "change_type": "MODIFY", "old_path": "conf/SystemDS-config.xml.template", "new_path": "conf/SystemDS-config.xml.template", "diff": "<!-- compiler optimization level, valid values: 0 | 1 | 2 | 3 | 4, default: 2 -->\n<sysds.optlevel>2</sysds.optlevel>\n- <!-- default number of reduce tasks per MR job, default: 2 x number of nodes -->\n- <sysds.numreducers>10</sysds.numreducers>\n-\n- <!-- override jvm reuse flag for specific MR jobs, valid values: true | false -->\n- <sysds.jvmreuse>false</sysds.jvmreuse>\n-\n<!-- default block dim for binary block files -->\n<sysds.defaultblocksize>1000</sysds.defaultblocksize>\n- <!-- run systemds control program as yarn appmaster, in case of MR1 always falls back to client, please disable for debug mode -->\n- <sysds.yarn.appmaster>false</sysds.yarn.appmaster>\n-\n- <!-- maximum jvm heap size of the dml yarn appmaster in MB, the requested memory is 1.5x this parameter -->\n- <sysds.yarn.appmaster.mem>2048</sysds.yarn.appmaster.mem>\n-\n- <!-- maximum jvm heap size of the map/reduce tasks in MB, the requested memory is 1.5x this parameter, negative values ignored -->\n- <sysds.yarn.mapreduce.mem>2048</sysds.yarn.mapreduce.mem>\n-\n- <!-- yarn application submission queue, relevant for default capacity scheduler -->\n- <sysds.yarn.app.queue>default</sysds.yarn.app.queue>\n-\n<!-- enables multi-threaded operations in singlenode control program -->\n<sysds.cp.parallel.ops>true</sysds.cp.parallel.ops>\n<!-- custom directory where BLAS libraries are available, experimental feature (options: absolute directory path or none). If set to none, we use standard LD_LIBRARY_PATH. -->\n<sysds.native.blas.directory>none</sysds.native.blas.directory>\n- <!-- prints finegrained statistics information (includes extra GPU information and extra statistics information for Deep Neural Networks done in CP mode) -->\n- <sysds.stats.finegrained>false</sysds.stats.finegrained>\n-\n<!-- sets the GPUs to use per process, -1 for all GPUs, a specific GPU number (5), a range (eg: 0-2) or a comma separated list (eg: 0,2,4)-->\n<sysds.gpu.availableGPUs>-1</sysds.gpu.availableGPUs>\n<!-- maximum wrap length for instruction and miscellaneous timer column of statistics -->\n<sysds.stats.maxWrapLength>30</sysds.stats.maxWrapLength>\n- <!-- Advanced optimization: fraction of driver memory to use for caching (default: 0.15) -->\n- <sysds.caching.bufferSize>0.15</sysds.caching.bufferSize>\n-\n<!-- Advanced optimization: fraction of driver memory to use for GPU shadow buffer. This optimization is ignored for double precision.\nBy default, it is disabled (hence set to 0.0). If you intend to train network larger than GPU memory size, consider using single precision and setting this to 0.1 -->\n<sysds.gpu.eviction.shadow.bufferSize>0.0</sysds.gpu.eviction.shadow.bufferSize>\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/api/DMLScript.java", "new_path": "src/main/java/org/tugraz/sysds/api/DMLScript.java", "diff": "@@ -417,13 +417,6 @@ public class DMLScript\n// Sets the GPUs to use for this process (a range, all GPUs, comma separated list or a specific GPU)\nGPUContextPool.AVAILABLE_GPUS = dmlconf.getTextValue(DMLConfig.AVAILABLE_GPUS);\n-\n- // Whether extra statistics useful for developers and others interested\n- // in digging into performance problems are recorded and displayed\n- CacheableData.CACHING_BUFFER_SIZE = dmlconf.getDoubleValue(DMLConfig.CACHING_BUFFER_SIZE);\n- if(CacheableData.CACHING_BUFFER_SIZE < 0 || CacheableData.CACHING_BUFFER_SIZE > 1)\n- throw new RuntimeException(\"Incorrect value (\" + CacheableData.CACHING_BUFFER_SIZE + \") for the configuration \" + DMLConfig.CACHING_BUFFER_SIZE);\n-\nDMLScript.STATISTICS_MAX_WRAP_LEN = dmlconf.getIntValue(DMLConfig.STATS_MAX_WRAP_LEN);\nNativeHelper.initialize(dmlconf.getTextValue(DMLConfig.NATIVE_BLAS_DIR), dmlconf.getTextValue(DMLConfig.NATIVE_BLAS).trim());\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/conf/ConfigurationManager.java", "new_path": "src/main/java/org/tugraz/sysds/conf/ConfigurationManager.java", "diff": "@@ -163,10 +163,6 @@ public class ConfigurationManager\nreturn getCompilerConfig().getInt(ConfigType.BLOCK_SIZE);\n}\n- public static int getNumReducers() {\n- return getDMLConfig().getIntValue(DMLConfig.NUM_REDUCERS);\n- }\n-\npublic static boolean isDynamicRecompilation() {\nreturn getCompilerConfigFlag(ConfigType.ALLOW_DYN_RECOMPILATION);\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/conf/DMLConfig.java", "new_path": "src/main/java/org/tugraz/sysds/conf/DMLConfig.java", "diff": "@@ -26,7 +26,6 @@ import java.io.FileNotFoundException;\nimport java.io.IOException;\nimport java.io.StringWriter;\nimport java.util.HashMap;\n-import java.util.Map;\nimport javax.xml.parsers.DocumentBuilder;\nimport javax.xml.parsers.DocumentBuilderFactory;\n@@ -65,13 +64,7 @@ public class DMLConfig\npublic static final String LOCAL_TMP_DIR = \"sysds.localtmpdir\";\npublic static final String SCRATCH_SPACE = \"sysds.scratch\";\npublic static final String OPTIMIZATION_LEVEL = \"sysds.optlevel\";\n- public static final String NUM_REDUCERS = \"sysds.numreducers\";\n- public static final String JVM_REUSE = \"sysds.jvmreuse\";\npublic static final String DEFAULT_BLOCK_SIZE = \"sysds.defaultblocksize\";\n- public static final String YARN_APPMASTER = \"sysds.yarn.appmaster\";\n- public static final String YARN_APPMASTERMEM = \"sysds.yarn.appmaster.mem\";\n- public static final String YARN_MAPREDUCEMEM = \"sysds.yarn.mapreduce.mem\";\n- public static final String YARN_APPQUEUE = \"sysds.yarn.app.queue\";\npublic static final String CP_PARALLEL_OPS = \"sysds.cp.parallel.ops\";\npublic static final String CP_PARALLEL_IO = \"sysds.cp.parallel.io\";\npublic static final String NATIVE_BLAS = \"sysds.native.blas\";\n@@ -81,8 +74,6 @@ public class DMLConfig\npublic static final String CODEGEN_OPTIMIZER = \"sysds.codegen.optimizer\"; //see SpoofCompiler.PlanSelector\npublic static final String CODEGEN_PLANCACHE = \"sysds.codegen.plancache\"; //boolean\npublic static final String CODEGEN_LITERALS = \"sysds.codegen.literals\"; //1..heuristic, 2..always\n- public static final String CACHING_BUFFER_SIZE = \"sysds.caching.bufferSize\"; //double: default:0.15\n- public static final String EXTRA_FINEGRAINED_STATS = \"sysds.stats.finegrained\"; //boolean\npublic static final String STATS_MAX_WRAP_LEN = \"sysds.stats.maxWrapLength\"; //int\npublic static final String AVAILABLE_GPUS = \"sysds.gpu.availableGPUs\"; // String to specify which GPUs to use (a range, all GPUs, comma separated list or a specific GPU)\npublic static final String SYNCHRONIZE_GPU = \"sysds.gpu.sync.postProcess\"; // boolean: whether to synchronize GPUs after every instruction\n@@ -97,10 +88,6 @@ public class DMLConfig\npublic static final String PRINT_GPU_MEMORY_INFO = \"sysds.gpu.print.memoryInfo\";\npublic static final String EVICTION_SHADOW_BUFFERSIZE = \"sysds.gpu.eviction.shadow.bufferSize\";\n- // supported prefixes for custom map/reduce configurations\n- public static final String PREFIX_MAPRED = \"mapred\";\n- public static final String PREFIX_MAPREDUCE = \"mapreduce\";\n-\n//internal config\npublic static final String DEFAULT_SHARED_DIR_PERMISSION = \"777\"; //for local fs and DFS\npublic static String LOCAL_MR_MODE_STAGING_DIR = null;\n@@ -119,13 +106,7 @@ public class DMLConfig\n_defaultVals.put(LOCAL_TMP_DIR, \"/tmp/systemds\" );\n_defaultVals.put(SCRATCH_SPACE, \"scratch_space\" );\n_defaultVals.put(OPTIMIZATION_LEVEL, String.valueOf(OptimizerUtils.DEFAULT_OPTLEVEL.ordinal()) );\n- _defaultVals.put(NUM_REDUCERS, \"10\" );\n- _defaultVals.put(JVM_REUSE, \"false\" );\n_defaultVals.put(DEFAULT_BLOCK_SIZE, String.valueOf(OptimizerUtils.DEFAULT_BLOCKSIZE) );\n- _defaultVals.put(YARN_APPMASTER, \"false\" );\n- _defaultVals.put(YARN_APPMASTERMEM, \"2048\" );\n- _defaultVals.put(YARN_MAPREDUCEMEM, \"-1\" );\n- _defaultVals.put(YARN_APPQUEUE, \"default\" );\n_defaultVals.put(CP_PARALLEL_OPS, \"true\" );\n_defaultVals.put(CP_PARALLEL_IO, \"true\" );\n_defaultVals.put(CODEGEN, \"false\" );\n@@ -135,7 +116,6 @@ public class DMLConfig\n_defaultVals.put(CODEGEN_LITERALS, \"1\" );\n_defaultVals.put(NATIVE_BLAS, \"none\" );\n_defaultVals.put(NATIVE_BLAS_DIR, \"none\" );\n- _defaultVals.put(EXTRA_FINEGRAINED_STATS,\"false\" );\n_defaultVals.put(PRINT_GPU_MEMORY_INFO, \"false\" );\n_defaultVals.put(EVICTION_SHADOW_BUFFERSIZE, \"0.0\" );\n_defaultVals.put(STATS_MAX_WRAP_LEN, \"30\" );\n@@ -144,7 +124,6 @@ public class DMLConfig\n_defaultVals.put(AVAILABLE_GPUS, \"-1\");\n_defaultVals.put(GPU_EVICTION_POLICY, \"min_evict\");\n_defaultVals.put(SYNCHRONIZE_GPU, \"false\" );\n- _defaultVals.put(CACHING_BUFFER_SIZE, \"0.15\" );\n_defaultVals.put(EAGER_CUDA_FREE, \"false\" );\n_defaultVals.put(FLOATING_POINT_PRECISION, \"double\" );\n}\n@@ -311,36 +290,6 @@ public class DMLConfig\n}\n}\n- /**\n- * Get a map of key/value pairs of all configurations w/ the prefix 'mapred'\n- * or 'mapreduce'.\n- *\n- * @return map of mapred and mapreduce key/value pairs\n- */\n- public Map<String, String> getCustomMRConfig()\n- {\n- HashMap<String, String> ret = new HashMap<>();\n-\n- //check for non-existing config xml tree\n- if( _xmlRoot == null )\n- return ret;\n-\n- //get all mapred.* and mapreduce.* tag / value pairs\n- NodeList list = _xmlRoot.getElementsByTagName(\"*\");\n- for( int i=0; list!=null && i<list.getLength(); i++ ) {\n- if( list.item(i) instanceof Element &&\n- ( ((Element)list.item(i)).getNodeName().startsWith(PREFIX_MAPRED)\n- || ((Element)list.item(i)).getNodeName().startsWith(PREFIX_MAPREDUCE)) )\n- {\n- Element elem = (Element) list.item(i);\n- ret.put(elem.getNodeName(),\n- elem.getFirstChild().getNodeValue());\n- }\n- }\n-\n- return ret;\n- }\n-\npublic synchronized String serializeDMLConfig()\n{\nString ret = null;\n@@ -418,22 +367,18 @@ public class DMLConfig\nreturn config;\n}\n- public String getConfigInfo()\n- {\n+ public String getConfigInfo() {\nString[] tmpConfig = new String[] {\n- LOCAL_TMP_DIR,SCRATCH_SPACE,OPTIMIZATION_LEVEL,\n- NUM_REDUCERS, DEFAULT_BLOCK_SIZE,\n- YARN_APPMASTER, YARN_APPMASTERMEM, YARN_MAPREDUCEMEM,\n+ LOCAL_TMP_DIR,SCRATCH_SPACE,OPTIMIZATION_LEVEL, DEFAULT_BLOCK_SIZE,\nCP_PARALLEL_OPS, CP_PARALLEL_IO, NATIVE_BLAS, NATIVE_BLAS_DIR,\nCODEGEN, CODEGEN_COMPILER, CODEGEN_OPTIMIZER, CODEGEN_PLANCACHE, CODEGEN_LITERALS,\n- EXTRA_FINEGRAINED_STATS, STATS_MAX_WRAP_LEN, PRINT_GPU_MEMORY_INFO, CACHING_BUFFER_SIZE,\n- AVAILABLE_GPUS, SYNCHRONIZE_GPU, EAGER_CUDA_FREE, FLOATING_POINT_PRECISION, GPU_EVICTION_POLICY, EVICTION_SHADOW_BUFFERSIZE,\n- GPU_MEMORY_ALLOCATOR, GPU_MEMORY_UTILIZATION_FACTOR\n+ STATS_MAX_WRAP_LEN, PRINT_GPU_MEMORY_INFO,\n+ AVAILABLE_GPUS, SYNCHRONIZE_GPU, EAGER_CUDA_FREE, FLOATING_POINT_PRECISION, GPU_EVICTION_POLICY,\n+ EVICTION_SHADOW_BUFFERSIZE, GPU_MEMORY_ALLOCATOR, GPU_MEMORY_UTILIZATION_FACTOR\n};\nStringBuilder sb = new StringBuilder();\n- for( String tmp : tmpConfig )\n- {\n+ for( String tmp : tmpConfig ) {\nsb.append(\"INFO: \");\nsb.append(tmp);\nsb.append(\": \");\n@@ -444,23 +389,6 @@ public class DMLConfig\nreturn sb.toString();\n}\n- public void updateYarnMemorySettings(String amMem, String mrMem)\n- {\n- //app master memory\n- NodeList list1 = _xmlRoot.getElementsByTagName(YARN_APPMASTERMEM);\n- if (list1 != null && list1.getLength() > 0) {\n- Element elem = (Element) list1.item(0);\n- elem.getFirstChild().setNodeValue(String.valueOf(amMem));\n- }\n-\n- //mapreduce memory\n- NodeList list2 = _xmlRoot.getElementsByTagName(YARN_MAPREDUCEMEM);\n- if (list2 != null && list2.getLength() > 0) {\n- Element elem = (Element) list2.item(0);\n- elem.getFirstChild().setNodeValue(String.valueOf(mrMem));\n- }\n- }\n-\npublic static String getDefaultTextValue( String key ) {\nreturn _defaultVals.get( key );\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/AggBinaryOp.java", "new_path": "src/main/java/org/tugraz/sysds/hops/AggBinaryOp.java", "diff": "@@ -1040,122 +1040,6 @@ public class AggBinaryOp extends MultiThreadedHop\nreturn footprint;\n}\n- /**\n- * Optimization that chooses between two methods to perform matrix multiplication on map-reduce.\n- *\n- * More details on the cost-model used: refer ICDE 2011 paper.\n- *\n- * @param m1_rows m1 rows\n- * @param m1_cols m1 cols\n- * @param m1_rpb m1 rows per block\n- * @param m1_cpb m1 cols per block\n- * @param m1_nnz m1 num non-zeros\n- * @param m2_rows m2 rows\n- * @param m2_cols m2 cols\n- * @param m2_rpb m2 rows per block\n- * @param m2_cpb m2 cols per block\n- * @param m2_nnz m2 num non-zeros\n- * @param mmtsj the MMTSJType\n- * @param chainType the chain type\n- * @param leftPMInput the left pm input\n- * @return the MMultMethod\n- */\n- @SuppressWarnings(\"unused\")\n- private static MMultMethod optFindMMultMethodMR ( long m1_rows, long m1_cols, long m1_rpb, long m1_cpb, long m1_nnz,\n- long m2_rows, long m2_cols, long m2_rpb, long m2_cpb, long m2_nnz,\n- MMTSJType mmtsj, ChainType chainType, boolean leftPMInput )\n- {\n- double memBudget = MAPMULT_MEM_MULTIPLIER * OptimizerUtils.getRemoteMemBudgetMap(true);\n-\n- // Step 0: check for forced mmultmethod\n- if( FORCED_MMULT_METHOD !=null )\n- return FORCED_MMULT_METHOD;\n-\n- // Step 1: check TSMM\n- // If transpose self pattern and result is single block:\n- // use specialized TSMM method (always better than generic jobs)\n- if( ( mmtsj == MMTSJType.LEFT && m2_cols>=0 && m2_cols <= m2_cpb )\n- || ( mmtsj == MMTSJType.RIGHT && m1_rows>=0 && m1_rows <= m1_rpb ) )\n- {\n- return MMultMethod.TSMM;\n- }\n-\n- // Step 2: check MapMultChain\n- // If mapmultchain pattern and result is a single block:\n- // use specialized mapmult method\n- if( OptimizerUtils.ALLOW_SUM_PRODUCT_REWRITES )\n- {\n- //matmultchain if dim2(X)<=blocksize and all vectors fit in mappers\n- //(X: m1_cols x m1_rows, v: m1_rows x m2_cols, w: m1_cols x m2_cols)\n- //NOTE: generalization possibe: m2_cols>=0 && m2_cols<=m2_cpb\n- if( chainType!=ChainType.NONE && m1_rows>=0 && m1_rows<= m1_rpb && m2_cols==1 )\n- {\n- if( chainType==ChainType.XtXv && m1_rows>=0 && m2_cols>=0\n- && OptimizerUtils.estimateSize(m1_rows, m2_cols ) < memBudget )\n- {\n- return MMultMethod.MAPMM_CHAIN;\n- }\n- else if( (chainType==ChainType.XtwXv || chainType==ChainType.XtXvy )\n- && m1_rows>=0 && m2_cols>=0 && m1_cols>=0\n- && OptimizerUtils.estimateSize(m1_rows, m2_cols )\n- + OptimizerUtils.estimateSize(m1_cols, m2_cols) < memBudget )\n- {\n- return MMultMethod.MAPMM_CHAIN;\n- }\n- }\n- }\n-\n- // Step 3: check for PMM (permutation matrix needs to fit into mapper memory)\n- // (needs to be checked before mapmult for consistency with removeEmpty compilation\n- double footprintPM1 = getMapmmMemEstimate(m1_rows, 1, m1_rpb, m1_cpb, m1_nnz, m2_rows, m2_cols, m2_rpb, m2_cpb, m2_nnz, 1, true);\n- double footprintPM2 = getMapmmMemEstimate(m2_rows, 1, m1_rpb, m1_cpb, m1_nnz, m2_rows, m2_cols, m2_rpb, m2_cpb, m2_nnz, 1, true);\n- if( (footprintPM1 < memBudget && m1_rows>=0 || footprintPM2 < memBudget && m2_rows>=0 )\n- && leftPMInput )\n- {\n- return MMultMethod.PMM;\n- }\n-\n- // Step 4: check MapMult\n- // If the size of one input is small, choose a method that uses distributed cache\n- // (with awareness of output size because one input block might generate many output blocks)\n- //memory estimates for local partitioning (mb -> partitioned mb)\n- double m1SizeP = OptimizerUtils.estimatePartitionedSizeExactSparsity(m1_rows, m1_cols, m1_rpb, m1_cpb, m1_nnz); //m1 partitioned\n- double m2SizeP = OptimizerUtils.estimatePartitionedSizeExactSparsity(m2_rows, m2_cols, m2_rpb, m2_cpb, m2_nnz); //m2 partitioned\n-\n- //memory estimates for remote execution (broadcast and outputs)\n- double footprint1 = getMapmmMemEstimate(m1_rows, m1_cols, m1_rpb, m1_cpb, m1_nnz, m2_rows, m2_cols, m2_rpb, m2_cpb, m2_nnz, 1, false);\n- double footprint2 = getMapmmMemEstimate(m1_rows, m1_cols, m1_rpb, m1_cpb, m1_nnz, m2_rows, m2_cols, m2_rpb, m2_cpb, m2_nnz, 2, false);\n-\n- if ( (footprint1 < memBudget && m1_rows>=0 && m1_cols>=0)\n- || (footprint2 < memBudget && m2_rows>=0 && m2_cols>=0) )\n- {\n- //apply map mult if one side fits in remote task memory\n- //(if so pick smaller input for distributed cache)\n- if( m1SizeP < m2SizeP && m1_rows>=0 && m1_cols>=0)\n- return MMultMethod.MAPMM_L;\n- else\n- return MMultMethod.MAPMM_R;\n- }\n-\n- // Step 5: check for unknowns\n- // If the dimensions are unknown at compilation time, simply assume\n- // the worst-case scenario and produce the most robust plan -- which is CPMM\n- if ( m1_rows == -1 || m1_cols == -1 || m2_rows == -1 || m2_cols == -1 )\n- return MMultMethod.CPMM;\n-\n- // Step 6: Decide CPMM vs RMM based on io costs\n-\n- //estimate shuffle costs weighted by parallelism\n- double rmm_costs = getRMMCostEstimate(m1_rows, m1_cols, m1_rpb, m1_cpb, m2_rows, m2_cols, m2_rpb, m2_cpb);\n- double cpmm_costs = getCPMMCostEstimate(m1_rows, m1_cols, m1_rpb, m1_cpb, m2_rows, m2_cols, m2_rpb, m2_cpb);\n-\n- //final mmult method decision\n- if ( cpmm_costs < rmm_costs )\n- return MMultMethod.CPMM;\n- else\n- return MMultMethod.RMM;\n- }\n-\nprivate static MMultMethod optFindMMultMethodCP( long m1_rows, long m1_cols, long m2_rows, long m2_cols, MMTSJType mmtsj, ChainType chainType, boolean leftPM )\n{\n//step 1: check for TSMM pattern\n@@ -1363,7 +1247,7 @@ public class AggBinaryOp extends MultiThreadedHop\n// CPMM phase 1\ndouble cpmm_shuffle1 = m1_size + m2_size;\ndouble cpmm_nred1 = Math.min( m1_ncb, //max used reducers\n- numReducersCPMM); //available reducers\n+ numReducersCPMM); //available reducer\ndouble cpmm_io1 = m1_size + m2_size + cpmm_nred1 * result_size;\n// CPMM phase 2\ndouble cpmm_shuffle2 = cpmm_nred1 * result_size;\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/OptimizerUtils.java", "new_path": "src/main/java/org/tugraz/sysds/hops/OptimizerUtils.java", "diff": "@@ -522,7 +522,7 @@ public class OptimizerUtils\nif( isSparkExecutionMode() )\nreturn SparkExecutionContext.getDefaultParallelism(false);\n- int ret = ConfigurationManager.getNumReducers();\n+ int ret = 2 * InfrastructureAnalyzer.getLocalParallelism();\nif( !configOnly ) {\nret = Math.min(ret,InfrastructureAnalyzer.getRemoteParallelReduceTasks());\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/ParForProgramBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/ParForProgramBlock.java", "diff": "@@ -27,7 +27,6 @@ import org.tugraz.sysds.common.Types.DataType;\nimport org.tugraz.sysds.common.Types.ValueType;\nimport org.tugraz.sysds.conf.CompilerConfig;\nimport org.tugraz.sysds.conf.ConfigurationManager;\n-import org.tugraz.sysds.conf.DMLConfig;\nimport org.tugraz.sysds.hops.OptimizerUtils;\nimport org.tugraz.sysds.hops.recompile.Recompiler;\nimport org.tugraz.sysds.lops.Lop;\n@@ -288,8 +287,6 @@ public class ParForProgramBlock extends ForProgramBlock\npublic static final boolean USE_RANGE_TASKS_IF_USEFUL = true; // use range tasks whenever size>3, false, otherwise wrong split order in remote\npublic static final boolean USE_STREAMING_TASK_CREATION = true; // start working while still creating tasks, prevents blocking due to too small task queue\npublic static final boolean ALLOW_NESTED_PARALLELISM = true; // if not, transparently change parfor to for on program conversions (local,remote)\n- public static boolean ALLOW_REUSE_MR_JVMS = true; // potential benefits: less setup costs per task, NOTE> cannot be used MR4490 in Hadoop 1.0.3, still not fixed in 1.1.1\n- public static boolean ALLOW_REUSE_MR_PAR_WORKER = ALLOW_REUSE_MR_JVMS; //potential benefits: less initialization, reuse in-memory objects and result consolidation!\npublic static final boolean USE_PARALLEL_RESULT_MERGE = false; // if result merge is run in parallel or serial\npublic static final boolean USE_PARALLEL_RESULT_MERGE_REMOTE = true; // if remote result merge should be run in parallel for multiple result vars\npublic static final boolean ALLOW_DATA_COLOCATION = true;\n@@ -380,9 +377,6 @@ public class ParForProgramBlock extends ForProgramBlock\n{\nsuper(prog, iterPredVar);\n- //init internal flags according to DML config\n- initInternalConfigurations(ConfigurationManager.getDMLConfig());\n-\n//ID generation and setting\nsetParForProgramBlockIDs( ID );\n_resultVars = resultVars;\n@@ -555,11 +549,6 @@ public class ParForProgramBlock extends ForProgramBlock\nreturn _hasFunctions;\n}\n- public static void initInternalConfigurations( DMLConfig conf ) {\n- ALLOW_REUSE_MR_JVMS = conf.getBooleanValue(DMLConfig.JVM_REUSE);\n- ALLOW_REUSE_MR_PAR_WORKER = ALLOW_REUSE_MR_JVMS;\n- }\n-\n@Override\npublic void execute(ExecutionContext ec)\n{\n@@ -1268,8 +1257,7 @@ public class ParForProgramBlock extends ForProgramBlock\n//determine max degree of parallelism\nint numReducers = OptimizerUtils.isSparkExecutionMode() ?\n- SparkExecutionContext.getDefaultParallelism(false) :\n- ConfigurationManager.getNumReducers();\n+ SparkExecutionContext.getDefaultParallelism(false) : 1;\nint maxNumRed = InfrastructureAnalyzer.getRemoteParallelReduceTasks();\n//correction max number of reducers on yarn clusters\nif( InfrastructureAnalyzer.isYarnEnabled() )\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/caching/CacheableData.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/caching/CacheableData.java", "diff": "@@ -80,7 +80,7 @@ public abstract class CacheableData<T extends CacheBlock> extends Data\n// global constant configuration parameters\npublic static final long CACHING_THRESHOLD = (long)Math.max(4*1024, //obj not s.t. caching\n1e-5 * InfrastructureAnalyzer.getLocalMaxMemory()); //if below threshold [in bytes]\n- public static double CACHING_BUFFER_SIZE = 0.15;\n+ public static final double CACHING_BUFFER_SIZE = 0.15;\npublic static final RPolicy CACHING_BUFFER_POLICY = RPolicy.FIFO;\npublic static final boolean CACHING_BUFFER_PAGECACHE = false;\npublic static final boolean CACHING_WRITE_CACHE_ON_READ = false;\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/parfor/RemoteParForUtils.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/parfor/RemoteParForUtils.java", "diff": "@@ -124,8 +124,7 @@ public class RemoteParForUtils\nMatrixObject mo = (MatrixObject) dat;\nif( mo.isDirty() )\n{\n- if( ParForProgramBlock.ALLOW_REUSE_MR_PAR_WORKER && rvarFnames!=null )\n- {\n+ if( rvarFnames!=null ) {\nString fname = rvarFnames.get( rvar._name );\nif( fname!=null )\nmo.setFileName( fname );\n@@ -134,8 +133,7 @@ public class RemoteParForUtils\nmo.exportData(); //note: this is equivalent to doing it in close (currently not required because 1 Task=1Map tasks, hence only one map invocation)\nrvarFnames.put(rvar._name, mo.getFileName());\n}\n- else\n- {\n+ else {\n//export result var (iff actually modified in parfor)\nmo.exportData(); //note: this is equivalent to doing it in close (currently not required because 1 Task=1Map tasks, hence only one map invocation)\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/matrix/mapred/MRJobConfiguration.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/matrix/mapred/MRJobConfiguration.java", "diff": "package org.tugraz.sysds.runtime.matrix.mapred;\nimport java.util.ArrayList;\n-import java.util.Map;\n-import java.util.Map.Entry;\nimport org.apache.hadoop.conf.Configuration;\nimport org.apache.hadoop.mapreduce.filecache.DistributedCache;\nimport org.tugraz.sysds.api.DMLScript;\nimport org.tugraz.sysds.conf.ConfigurationManager;\n-import org.tugraz.sysds.conf.DMLConfig;\nimport org.tugraz.sysds.lops.Lop;\nimport org.tugraz.sysds.runtime.controlprogram.parfor.stat.InfrastructureAnalyzer;\nimport org.tugraz.sysds.runtime.controlprogram.parfor.util.IDSequence;\n@@ -44,7 +41,6 @@ import org.apache.hadoop.mapred.JobConf;\n@SuppressWarnings({\"deprecation\" })\npublic class MRJobConfiguration\n{\n-\n//internal param: custom deserializer/serializer (usually 30% faster than WritableSerialization)\npublic static final boolean USE_BINARYBLOCK_SERIALIZATION = true;\n@@ -242,18 +238,4 @@ public class MRJobConfiguration\nString frameworkClassBB = BinaryBlockSerialization.class.getCanonicalName();\njob.set(MRConfigurationNames.IO_SERIALIZATIONS, frameworkClassBB+\",\"+frameworkList);\n}\n-\n- /**\n- * Set all configurations with prefix mapred or mapreduce that exist in the given\n- * DMLConfig into the given JobConf.\n- *\n- * @param job job configuration\n- * @param config dml configuration\n- */\n- public static void setupCustomMRConfigurations( JobConf job, DMLConfig config ) {\n- Map<String,String> map = config.getCustomMRConfig();\n- for( Entry<String,String> e : map.entrySet() ) {\n- job.set(e.getKey(), e.getValue());\n- }\n- }\n}\n\\ No newline at end of file\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/util/ProgramConverter.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/util/ProgramConverter.java", "diff": "@@ -1222,9 +1222,6 @@ public class ProgramConverter\nJobConf job = ConfigurationManager.getCachedJobConf();\nif( !InfrastructureAnalyzer.isLocalMode(job) ) {\nhandleDMLConfig(confStr);\n- //init internal configuration w/ parsed or default config\n- ParForProgramBlock.initInternalConfigurations(\n- ConfigurationManager.getDMLConfig());\n}\n//handle additional configs\n" }, { "change_type": "MODIFY", "old_path": "src/test/config/SystemDS-config.xml", "new_path": "src/test/config/SystemDS-config.xml", "diff": "<!-- compiler optimization level, valid values: 0 | 1 | 2 | 3 | 4, default: 2 -->\n<sysds.optlevel>2</sysds.optlevel>\n- <!-- default number of reduce tasks per MR job, default: 2 x number of nodes -->\n- <sysds.numreducers>10</sysds.numreducers>\n-\n- <!-- override jvm reuse flag for specific MR jobs, valid values: true | false -->\n- <sysds.jvmreuse>false</sysds.jvmreuse>\n-\n<!-- default block dim for binary block files -->\n<sysds.defaultblocksize>1000</sysds.defaultblocksize>\n- <!-- run systemds control program as yarn appmaster, in case of MR1 always falls back to client, please disable for debug mode -->\n- <sysds.yarn.appmaster>false</sysds.yarn.appmaster>\n-\n- <!-- maximum jvm heap size of the dml yarn appmaster in MB, the requested memory is 1.5x this parameter -->\n- <sysds.yarn.appmaster.mem>2048</sysds.yarn.appmaster.mem>\n-\n- <!-- maximum jvm heap size of the map/reduce tasks in MB, the requested memory is 1.5x this parameter, negative values ignored -->\n- <sysds.yarn.mapreduce.mem>2048</sysds.yarn.mapreduce.mem>\n-\n- <!-- yarn application submission queue, relevant for default capacity scheduler -->\n- <sysds.yarn.app.queue>default</sysds.yarn.app.queue>\n-\n<!-- enables multi-threaded matrix operations in singlenode control program -->\n<sysds.cp.parallel.ops>true</sysds.cp.parallel.ops>\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/dmlscript/SystemML-config.xml", "new_path": "src/test/scripts/functions/dmlscript/SystemML-config.xml", "diff": "* under the License.\n-->\n<root>\n-<sysds.numreducers>10</sysds.numreducers>\n<sysds.scratch>scratch_space</sysds.scratch>\n<sysds.defaultblocksize>1000</sysds.defaultblocksize>\n<sysds.cp.parallel.ops>true</sysds.cp.parallel.ops>\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/context/SparkExecutionContext.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/context/SparkExecutionContext.java", "diff": "@@ -478,7 +478,7 @@ public class SparkExecutionContext extends ExecutionContext\nfromFile = true;*/\n} else { //default case\nTensorBlock tb = to.acquireRead(); //pin matrix in memory\n- int blen = dc.getBlockSize();\n+ int blen = dc.getBlocksize();\nrdd = toTensorJavaPairRDD(sc, tb, blen, numParts, inclEmpty);\nto.release(); //unpin matrix\n_parRDDs.registerRDD(rdd.id(), OptimizerUtils.estimatePartitionedSizeExactSparsity(dc), true);\n@@ -817,7 +817,7 @@ public class SparkExecutionContext extends ExecutionContext\n}\nbret = new PartitionedBroadcast<>(ret, new MatrixCharacteristics(\n- fo.getDataCharacteristics()).setBlockSize(blen));\n+ fo.getDataCharacteristics()).setBlocksize(blen));\nif (fo.getBroadcastHandle() == null)\nfo.setBroadcastHandle(new BroadcastObject<FrameBlock>());\n@@ -960,7 +960,7 @@ public class SparkExecutionContext extends ExecutionContext\ntry {\n//compute block indexes\nlong[] blockIx = new long[tc.getNumDims()];\n- int blocksize = tc.getBlockSize();\n+ int blocksize = tc.getBlocksize();\nint[] outDims = new int[tc.getNumDims()];\nint[] offset = new int[tc.getNumDims()];\nfor (int i = tc.getNumDims() - 1; i >= 0; i--) {\n@@ -1212,7 +1212,7 @@ public class SparkExecutionContext extends ExecutionContext\nint[] lower = new int[ix.getNumDims()];\nint[] upper = new int[ix.getNumDims()];\nfor (int i = 0; i < lower.length; i++) {\n- lower[i] = (int) ((ix.getIndex(i) - 1) * dc.getBlockSize());\n+ lower[i] = (int) ((ix.getIndex(i) - 1) * dc.getBlocksize());\nupper[i] = lower[i] + block.getDim(i) - 1;\n}\nupper[upper.length - 1]++;\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/VariableCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/VariableCPInstruction.java", "diff": "@@ -366,7 +366,7 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\nelse if (parts.length >= 11) {\n// matrix characteristics\nmc.setDimension(Long.parseLong(parts[6]), Long.parseLong(parts[7]));\n- mc.setBlockSize(Integer.parseInt(parts[8]));\n+ mc.setBlocksize(Integer.parseInt(parts[8]));\nmc.setNonZeros(Long.parseLong(parts[10]));\n}\nelse {\n@@ -383,7 +383,7 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\n// TODO correct sizes\ntc.setDim(0, Long.parseLong(parts[6]));\ntc.setDim(1, Long.parseLong(parts[7]));\n- tc.setBlockSize(Integer.parseInt(parts[8]));\n+ tc.setBlocksize(Integer.parseInt(parts[8]));\n}\nelse {\nthrow new DMLRuntimeException(\"Invalid number of operands in createvar instruction: \" + str);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/CastSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/CastSPInstruction.java", "diff": "@@ -69,7 +69,7 @@ public class CastSPInstruction extends UnarySPInstruction {\n//convert frame-matrix / matrix-frame and set output\nif( opcode.equals(UnaryCP.CAST_AS_MATRIX_OPCODE) ) {\nDataCharacteristics mcOut = new MatrixCharacteristics(mcIn);\n- mcOut.setBlockSize(ConfigurationManager.getBlocksize());\n+ mcOut.setBlocksize(ConfigurationManager.getBlocksize());\nout = FrameRDDConverterUtils.binaryBlockToMatrixBlock(\n(JavaPairRDD<Long, FrameBlock>)in, mcIn, mcOut);\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/ParameterizedBuiltinSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/ParameterizedBuiltinSPInstruction.java", "diff": "@@ -927,7 +927,7 @@ public class ParameterizedBuiltinSPInstruction extends ComputationSPInstruction\nelse {\nout = SparkUtils.cacheBinaryCellRDD(out);\nmcOut.set(SparkUtils.computeDataCharacteristics(out));\n- mcOut.setBlockSize(-1); //grouped aggregate with cell output\n+ mcOut.setBlocksize(-1); //grouped aggregate with cell output\n}\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/data/PartitionedBroadcast.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/data/PartitionedBroadcast.java", "diff": "@@ -109,8 +109,8 @@ public class PartitionedBroadcast<T extends CacheBlock> implements Serializable\nint pix = 0;\nif( _pbc.length > 1 ) { //compute partition index\nlong[] dims = _dc.getDims();\n- int blen = _dc.getBlockSize();\n- int numPerPart = computeBlocksPerPartition(dims, _dc.getBlockSize());\n+ int blen = _dc.getBlocksize();\n+ int numPerPart = computeBlocksPerPartition(dims, _dc.getBlocksize());\npix = (int) (UtilFunctions.computeBlockNumber(ix, dims, blen) / numPerPart);\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/utils/FrameRDDConverterUtils.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/utils/FrameRDDConverterUtils.java", "diff": "@@ -195,7 +195,7 @@ public class FrameRDDConverterUtils\nif(dcIn.getCols() > dcIn.getBlocksize()) {\n//split matrix blocks into extended matrix blocks\nin = in.flatMapToPair(new MatrixFrameReblockFunction(dcIn));\n- mc.setBlockSize(MatrixFrameReblockFunction.computeBlockSize(mc));\n+ mc.setBlocksize(MatrixFrameReblockFunction.computeBlockSize(mc));\n//shuffle matrix blocks (instead of frame blocks) in order to exploit\n//sparse formats (for sparse or wide matrices) during shuffle\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/utils/RDDConverterUtils.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/utils/RDDConverterUtils.java", "diff": "@@ -223,7 +223,7 @@ public class RDDConverterUtils\n//ensure valid blocksizes\nif( mc.getBlocksize()<=1 || mc.getBlocksize()<=1 ) {\n- mc.setBlockSize(ConfigurationManager.getBlocksize());\n+ mc.setBlocksize(ConfigurationManager.getBlocksize());\n}\n//construct or reuse row ids\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/meta/DataCharacteristics.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/meta/DataCharacteristics.java", "diff": "@@ -73,8 +73,9 @@ public abstract class DataCharacteristics implements Serializable {\nreturn _blocksize;\n}\n- public void setBlocksize(int blen){\n+ public DataCharacteristics setBlocksize(int blen){\n_blocksize = blen;\n+ return this;\n}\npublic long getNumBlocks() {\n@@ -113,14 +114,6 @@ public abstract class DataCharacteristics implements Serializable {\nthrow new DMLRuntimeException(\"DataCharacteristics.setDims(long[]): should never get called in the base class\");\n}\n- public int getBlockSize() {\n- throw new DMLRuntimeException(\"DataCharacteristics.getBlockSize(): should never get called in the base class\");\n- }\n-\n- public DataCharacteristics setBlockSize(int blen) {\n- throw new DMLRuntimeException(\"DataCharacteristics.setBlockSize(int): should never get called in the base class\");\n- }\n-\npublic long getNumBlocks(int i) {\nthrow new DMLRuntimeException(\"DataCharacteristics.getNumBlocks(i): should never get called in the base class\");\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/meta/TensorCharacteristics.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/meta/TensorCharacteristics.java", "diff": "@@ -64,7 +64,7 @@ public class TensorCharacteristics extends DataCharacteristics\n@Override\npublic DataCharacteristics set(DataCharacteristics that) {\nlong[] dims = that.getDims().clone();\n- set(dims, that.getBlockSize(), that.getNonZeros());\n+ set(dims, that.getBlocksize(), that.getNonZeros());\nreturn this;\n}\n@@ -134,7 +134,7 @@ public class TensorCharacteristics extends DataCharacteristics\n@Override\npublic long getNumBlocks(int i) {\n- return Math.max((long) Math.ceil((double)getDim(i) / getBlockSize()), 1);\n+ return Math.max((long) Math.ceil((double)getDim(i) / getBlocksize()), 1);\n}\n@Override\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/frame/FrameConverterTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/frame/FrameConverterTest.java", "diff": "@@ -337,7 +337,7 @@ public class FrameConverterTest extends AutomatedTestBase\n//write matrix data to hdfs\nMatrixWriter matWriter = MatrixWriterFactory.createMatrixWriter(oinfo);\nmatWriter.writeMatrixToHDFS(matrixBlock1, input(\"A\"), rows, schema.length,\n- mcMatrix.getBlockSize(), mcMatrix.getNonZeros());\n+ mcMatrix.getBlocksize(), mcMatrix.getNonZeros());\n}\nelse {\n//initialize the frame data.\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/lops/CSVReBlock.java", "new_path": "src/main/java/org/tugraz/sysds/lops/CSVReBlock.java", "diff": "@@ -108,7 +108,6 @@ public class CSVReBlock extends Lop\ngetInputs().get(0).prepInputOperand(input1),\nprepOutputOperand(output),\nString.valueOf(_blocksize),\n- String.valueOf(_blocksize),\nprepCSVProperties());\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/lops/Data.java", "new_path": "src/main/java/org/tugraz/sysds/lops/Data.java", "diff": "@@ -461,8 +461,6 @@ public class Data extends Lop\nsb.append( OPERAND_DELIMITOR );\nsb.append( oparams.getBlocksize() );\nsb.append( OPERAND_DELIMITOR );\n- sb.append( oparams.getBlocksize() );\n- sb.append( OPERAND_DELIMITOR );\nsb.append( oparams.getNnz() );\nsb.append( OPERAND_DELIMITOR );\nsb.append( oparams.getUpdateType().toString().toLowerCase() );\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/lops/DataGen.java", "new_path": "src/main/java/org/tugraz/sysds/lops/DataGen.java", "diff": "@@ -84,15 +84,14 @@ public class DataGen extends Lop\nreturn method;\n}\n- public void init(DataIdentifier id, String baseDir, ExecType et)\n- {\n- this.getOutputParameters().setFormat(Format.BINARY);\n- this.getOutputParameters().setBlocked(true);\n+ public void init(DataIdentifier id, String baseDir, ExecType et) {\n+ getOutputParameters().setFormat(Format.BINARY);\n+ getOutputParameters().setBlocked(true);\n// TODO size for tensor\n- this.getOutputParameters().setNumRows(id.getDim1());\n- this.getOutputParameters().setNumCols(id.getDim2());\n- this.getOutputParameters().setNnz(-1);\n- this.getOutputParameters().setBlocksize(id.getBlocksize());\n+ getOutputParameters().setNumRows(id.getDim1());\n+ getOutputParameters().setNumCols(id.getDim2());\n+ getOutputParameters().setNnz(-1);\n+ getOutputParameters().setBlocksize(id.getBlocksize());\nlps.setProperties( inputs, et);\n}\n@@ -161,9 +160,6 @@ public class DataGen extends Lop\nsb.append(getOutputParameters().getBlocksize());\nsb.append(OPERAND_DELIMITOR);\n- sb.append(getOutputParameters().getBlocksize());\n- sb.append(OPERAND_DELIMITOR);\n-\niLop = _inputParams.get(DataExpression.RAND_MIN);\nsb.append(iLop.prepScalarLabel());\nsb.append(OPERAND_DELIMITOR);\n@@ -251,8 +247,6 @@ public class DataGen extends Lop\nsb.append(OPERAND_DELIMITOR);\nsb.append(blen);\nsb.append(OPERAND_DELIMITOR);\n- sb.append(blen);\n- sb.append(OPERAND_DELIMITOR);\nsb.append(minString);\nsb.append(OPERAND_DELIMITOR);\nsb.append(prepOutputOperand(output));\n@@ -278,7 +272,6 @@ public class DataGen extends Lop\nlreplace.prepScalarLabel(),\nlseed.prepScalarLabel(),\nString.valueOf(getOutputParameters().getBlocksize()),\n- String.valueOf(getOutputParameters().getBlocksize()),\nprepOutputOperand(output));\n}\n@@ -334,8 +327,6 @@ public class DataGen extends Lop\nsb.append( OPERAND_DELIMITOR );\nsb.append( blen );\nsb.append( OPERAND_DELIMITOR );\n- sb.append( blen );\n- sb.append( OPERAND_DELIMITOR );\nsb.append( fromString );\nsb.append( OPERAND_DELIMITOR );\nsb.append( toString );\n@@ -348,17 +339,16 @@ public class DataGen extends Lop\n}\n@Override\n- public String toString()\n- {\n+ public String toString() {\nStringBuilder sb = new StringBuilder();\nsb.append(method.toString());\n- sb.append(\" ; num_rows=\" + this.getOutputParameters().getNumRows());\n- sb.append(\" ; num_cols=\" + this.getOutputParameters().getNumCols());\n- sb.append(\" ; nnz=\" + this.getOutputParameters().getNnz());\n- sb.append(\" ; blocksize=\" + this.getOutputParameters().getBlocksize());\n- sb.append(\" ; format=\" + this.getOutputParameters().getFormat());\n- sb.append(\" ; blocked=\" + this.getOutputParameters().isBlocked());\n- sb.append(\" ; dir=\" + this.baseDir);\n+ sb.append(\" ; num_rows=\" + getOutputParameters().getNumRows());\n+ sb.append(\" ; num_cols=\" + getOutputParameters().getNumCols());\n+ sb.append(\" ; nnz=\" + getOutputParameters().getNnz());\n+ sb.append(\" ; blocksize=\" + getOutputParameters().getBlocksize());\n+ sb.append(\" ; format=\" + getOutputParameters().getFormat());\n+ sb.append(\" ; blocked=\" + getOutputParameters().isBlocked());\n+ sb.append(\" ; dir=\" + baseDir);\nreturn sb.toString();\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/lops/ReBlock.java", "new_path": "src/main/java/org/tugraz/sysds/lops/ReBlock.java", "diff": "@@ -66,7 +66,6 @@ public class ReBlock extends Lop\ngetInputs().get(0).prepInputOperand(input1),\nprepOutputOperand(output),\nString.valueOf(_blocksize),\n- String.valueOf(_blocksize),\nString.valueOf(_outputEmptyBlocks));\n}\n@@ -96,7 +95,7 @@ public class ReBlock extends Lop\n* => the input lop (if it is other than Data) is always executed in a different job\n*/\n// return getChildFormat(node.getInputs().get(0));\n- return node.getInputs().get(0).getOutputParameters().getFormat(); }\n-\n+ return node.getInputs().get(0).getOutputParameters().getFormat();\n+ }\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/DataGenCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/DataGenCPInstruction.java", "diff": "@@ -61,7 +61,7 @@ public class DataGenCPInstruction extends UnaryCPInstruction {\nprivate final int numThreads;\n// seed positions\n- private static final int SEED_POSITION_RAND = 9;\n+ private static final int SEED_POSITION_RAND = 8;\nprivate static final int SEED_POSITION_SAMPLE = 4;\nprivate DataGenCPInstruction(Operator op, DataGenMethod mthd, CPOperand in, CPOperand out,\n@@ -176,17 +176,17 @@ public class DataGenCPInstruction extends UnaryCPInstruction {\nif ( opcode.equalsIgnoreCase(DataGen.RAND_OPCODE) ) {\nmethod = DataGenMethod.RAND;\n- InstructionUtils.checkNumFields ( s, 13 );\n+ InstructionUtils.checkNumFields ( s, 12 );\n}\nelse if ( opcode.equalsIgnoreCase(DataGen.SEQ_OPCODE) ) {\nmethod = DataGenMethod.SEQ;\n- // 8 operands: rows, cols, rpb, cpb, from, to, incr, outvar\n- InstructionUtils.checkNumFields ( s, 8 );\n+ // 8 operands: rows, cols, blen, from, to, incr, outvar\n+ InstructionUtils.checkNumFields ( s, 7 );\n}\nelse if ( opcode.equalsIgnoreCase(DataGen.SAMPLE_OPCODE) ) {\nmethod = DataGenMethod.SAMPLE;\n- // 7 operands: range, size, replace, seed, rpb, cpb, outvar\n- InstructionUtils.checkNumFields ( s, 7 );\n+ // 7 operands: range, size, replace, seed, blen, outvar\n+ InstructionUtils.checkNumFields ( s, 6 );\n}\nelse if ( opcode.equalsIgnoreCase(DataGen.TIME_OPCODE) ) {\nmethod = DataGenMethod.TIME;\n@@ -203,24 +203,24 @@ public class DataGenCPInstruction extends UnaryCPInstruction {\nCPOperand cols = new CPOperand(s[2]);\nCPOperand dims = new CPOperand(s[3]);\nint blen = Integer.parseInt(s[4]);\n- double sparsity = !s[8].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n- Double.valueOf(s[8]) : -1;\n+ double sparsity = !s[7].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ Double.valueOf(s[7]) : -1;\nlong seed = !s[SEED_POSITION_RAND].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\nLong.valueOf(s[SEED_POSITION_RAND]) : -1;\n- String pdf = s[10];\n- String pdfParams = !s[11].contains( Lop.VARIABLE_NAME_PLACEHOLDER) ?\n- s[11] : null;\n- int k = Integer.parseInt(s[12]);\n+ String pdf = s[9];\n+ String pdfParams = !s[10].contains( Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ s[10] : null;\n+ int k = Integer.parseInt(s[11]);\n- return new DataGenCPInstruction(op, method, null, out, rows, cols, dims, blen, s[6], s[7],\n- sparsity, seed, pdf, pdfParams, k, opcode, str);\n+ return new DataGenCPInstruction(op, method, null, out, rows, cols, dims, blen,\n+ s[5], s[6], sparsity, seed, pdf, pdfParams, k, opcode, str);\n}\nelse if ( method == DataGenMethod.SEQ)\n{\nint blen = Integer.parseInt(s[3]);\n- CPOperand from = new CPOperand(s[5]);\n- CPOperand to = new CPOperand(s[6]);\n- CPOperand incr = new CPOperand(s[7]);\n+ CPOperand from = new CPOperand(s[4]);\n+ CPOperand to = new CPOperand(s[5]);\n+ CPOperand incr = new CPOperand(s[6]);\nreturn new DataGenCPInstruction(op, method, null, out, null, null, null, blen, from, to, incr, opcode, str);\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/StringInitCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/StringInitCPInstruction.java", "diff": "@@ -39,7 +39,7 @@ public class StringInitCPInstruction extends UnaryCPInstruction {\nprivate final String _data;\nprivate StringInitCPInstruction(Operator op, CPOperand in, CPOperand out, long rows, long cols,\n- int rpb, int cpb, String data, String opcode, String inst) {\n+ int blen, String data, String opcode, String inst) {\nsuper(CPType.StringInit, op, in, out, opcode, inst);\n_rlen = rows;\n_clen = cols;\n@@ -60,15 +60,14 @@ public class StringInitCPInstruction extends UnaryCPInstruction {\nthrow new DMLRuntimeException(\"Unsupported opcode: \"+opcode);\n//parse instruction\nString[] s = InstructionUtils.getInstructionPartsWithValueType ( str );\n- InstructionUtils.checkNumFields( s, 7 );\n+ InstructionUtils.checkNumFields( s, 6 );\nCPOperand out = new CPOperand(s[s.length-1]); // output is specified by the last operand\nlong rows = (s[1].contains( Lop.VARIABLE_NAME_PLACEHOLDER)?-1:Double.valueOf(s[1]).longValue());\nlong cols = (s[2].contains( Lop.VARIABLE_NAME_PLACEHOLDER)?-1:Double.valueOf(s[2]).longValue());\n// Ignore dims\n- int rpb = Integer.parseInt(s[4]);\n- int cpb = Integer.parseInt(s[5]);\n- String data = s[6];\n- return new StringInitCPInstruction(null, null, out, rows, cols, rpb, cpb, data, opcode, str);\n+ int blen = Integer.parseInt(s[4]);\n+ String data = s[5];\n+ return new StringInitCPInstruction(null, null, out, rows, cols, blen, data, opcode, str);\n}\n@Override\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/VariableCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/VariableCPInstruction.java", "diff": "@@ -334,7 +334,7 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\n// variable name\nDataType dt = DataType.valueOf(parts[4]);\nValueType vt = dt==DataType.MATRIX ? ValueType.FP64 : ValueType.STRING;\n- int extSchema = (dt==DataType.FRAME && parts.length>=13) ? 1 : 0;\n+ int extSchema = (dt==DataType.FRAME && parts.length>=12) ? 1 : 0;\nin1 = new CPOperand(parts[1], vt, dt);\n// file name\nin2 = new CPOperand(parts[2], ValueType.STRING, DataType.SCALAR);\n@@ -347,11 +347,11 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\n// Cretevar instructions for CSV format either has 13 or 14 inputs.\n// 13 inputs: createvar corresponding to WRITE -- includes properties hasHeader, delim, and sparse\n// 14 inputs: createvar corresponding to READ -- includes properties hasHeader, delim, fill, and fillValue\n- if ( parts.length < 15+extSchema || parts.length > 17+extSchema )\n+ if ( parts.length < 14+extSchema || parts.length > 16+extSchema )\nthrow new DMLRuntimeException(\"Invalid number of operands in createvar instruction: \" + str);\n}\nelse {\n- if ( parts.length != 6 && parts.length != 12+extSchema )\n+ if ( parts.length != 6 && parts.length != 11+extSchema )\nthrow new DMLRuntimeException(\"Invalid number of operands in createvar instruction: \" + str);\n}\nOutputInfo oi = OutputInfo.stringToOutputInfo(fmt);\n@@ -363,11 +363,11 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\nif (parts.length == 6) {\n// do nothing\n}\n- else if (parts.length >= 11) {\n+ else if (parts.length >= 10) {\n// matrix characteristics\nmc.setDimension(Long.parseLong(parts[6]), Long.parseLong(parts[7]));\nmc.setBlocksize(Integer.parseInt(parts[8]));\n- mc.setNonZeros(Long.parseLong(parts[10]));\n+ mc.setNonZeros(Long.parseLong(parts[9]));\n}\nelse {\nthrow new DMLRuntimeException(\"Invalid number of operands in createvar instruction: \" + str);\n@@ -379,7 +379,7 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\nif (parts.length == 6) {\n// do nothing\n}\n- else if (parts.length >= 11) {\n+ else if (parts.length >= 10) {\n// TODO correct sizes\ntc.setDim(0, Long.parseLong(parts[6]));\ntc.setDim(1, Long.parseLong(parts[7]));\n@@ -391,31 +391,32 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\niimd = new MetaDataFormat(tc, oi, ii);\n}\nUpdateType updateType = UpdateType.COPY;\n- if ( parts.length >= 12 )\n- updateType = UpdateType.valueOf(parts[11].toUpperCase());\n+ if ( parts.length >= 11 )\n+ updateType = UpdateType.valueOf(parts[10].toUpperCase());\n//handle frame schema\n- String schema = (dt==DataType.FRAME && parts.length>=13) ? parts[parts.length-1] : null;\n+ String schema = (dt==DataType.FRAME && parts.length>=12) ? parts[parts.length-1] : null;\nif ( fmt.equalsIgnoreCase(\"csv\") ) {\n// Cretevar instructions for CSV format either has 13 or 14 inputs.\n// 13 inputs: createvar corresponding to WRITE -- includes properties hasHeader, delim, and sparse\n// 14 inputs: createvar corresponding to READ -- includes properties hasHeader, delim, fill, and fillValue\nFileFormatProperties fmtProperties = null;\n- if ( parts.length == 15+extSchema ) {\n- boolean hasHeader = Boolean.parseBoolean(parts[12]);\n- String delim = parts[13];\n- boolean sparse = Boolean.parseBoolean(parts[14]);\n+ int curPos = 11;\n+ if ( parts.length == 14+extSchema ) {\n+ boolean hasHeader = Boolean.parseBoolean(parts[curPos]);\n+ String delim = parts[curPos+1];\n+ boolean sparse = Boolean.parseBoolean(parts[curPos+2]);\nfmtProperties = new FileFormatPropertiesCSV(hasHeader, delim, sparse) ;\n}\nelse {\n- boolean hasHeader = Boolean.parseBoolean(parts[12]);\n- String delim = parts[13];\n- boolean fill = Boolean.parseBoolean(parts[14]);\n- double fillValue = UtilFunctions.parseToDouble(parts[15]);\n+ boolean hasHeader = Boolean.parseBoolean(parts[curPos]);\n+ String delim = parts[curPos+1];\n+ boolean fill = Boolean.parseBoolean(parts[curPos+2]);\n+ double fillValue = UtilFunctions.parseToDouble(parts[curPos+3]);\nString naStrings = null;\n- if ( parts.length == 17+extSchema )\n- naStrings = parts[16];\n+ if ( parts.length == 16+extSchema )\n+ naStrings = parts[curPos+4];\nfmtProperties = new FileFormatPropertiesCSV(hasHeader, delim, fill, fillValue, naStrings) ;\n}\nreturn new VariableCPInstruction(VariableOperationCode.CreateVariable, in1, in2, in3, iimd, updateType, fmtProperties, schema, opcode, str);\n@@ -1126,8 +1127,6 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\nsb.append(Lop.OPERAND_DELIMITOR);\nsb.append(mc.getBlocksize());\nsb.append(Lop.OPERAND_DELIMITOR);\n- sb.append(mc.getBlocksize());\n- sb.append(Lop.OPERAND_DELIMITOR);\nsb.append(mc.getNonZeros());\nsb.append(Lop.OPERAND_DELIMITOR);\nsb.append(update.toString().toLowerCase());\n@@ -1148,8 +1147,6 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\nsb.append(Lop.OPERAND_DELIMITOR);\nsb.append(mc.getBlocksize());\nsb.append(Lop.OPERAND_DELIMITOR);\n- sb.append(mc.getBlocksize());\n- sb.append(Lop.OPERAND_DELIMITOR);\nsb.append(mc.getNonZeros());\nsb.append(Lop.OPERAND_DELIMITOR);\nsb.append(update.toString().toLowerCase());\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/AppendGSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/AppendGSPInstruction.java", "diff": "@@ -159,12 +159,11 @@ public class AppendGSPInstruction extends BinarySPInstruction {\nprivate int _blen;\nprivate long _outlen;\n- public ShiftMatrix(DataCharacteristics mc1, DataCharacteristics mc2, boolean cbind)\n- {\n+ public ShiftMatrix(DataCharacteristics mc1, DataCharacteristics mc2, boolean cbind) {\n_cbind = cbind;\n_startIx = cbind ? UtilFunctions.computeBlockIndex(mc1.getCols(), mc1.getBlocksize()) :\nUtilFunctions.computeBlockIndex(mc1.getRows(), mc1.getBlocksize());\n- _blen = (int) (cbind ? mc1.getBlocksize() : mc1.getBlocksize());\n+ _blen = mc1.getBlocksize();\n_shiftBy = (int) (cbind ? mc1.getCols()%_blen : mc1.getRows()%_blen);\n_outlen = cbind ? mc1.getCols()+mc2.getCols() : mc1.getRows()+mc2.getRows();\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/BuiltinNarySPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/BuiltinNarySPInstruction.java", "diff": "@@ -123,8 +123,7 @@ public class BuiltinNarySPInstruction extends SPInstruction\nprivate static DataCharacteristics computeAppendOutputDataCharacteristics(SparkExecutionContext sec, CPOperand[] inputs, boolean cbind) {\nDataCharacteristics mcIn1 = sec.getDataCharacteristics(inputs[0].getName());\n- DataCharacteristics mcOut = new MatrixCharacteristics(\n- 0, 0, mcIn1.getBlocksize(), 0);\n+ DataCharacteristics mcOut = new MatrixCharacteristics(0, 0, mcIn1.getBlocksize(), 0);\nfor( CPOperand input : inputs ) {\nDataCharacteristics mcIn = sec.getDataCharacteristics(input.getName());\nupdateAppendDataCharacteristics(mcIn, mcOut, cbind);\n@@ -163,18 +162,18 @@ public class BuiltinNarySPInstruction extends SPInstruction\npublic Tuple2<MatrixIndexes, MatrixBlock> call(Tuple2<MatrixIndexes, MatrixBlock> arg0) throws Exception {\nMatrixIndexes ix = arg0._1();\nMatrixBlock mb = arg0._2();\n- int blen = UtilFunctions.computeBlockSize(_mcOut.getRows(), ix.getRowIndex(), _mcOut.getBlocksize());\n- //int lblen = UtilFunctions.computeBlockSize(_mcOut.getCols(), ix.getColumnIndex(), _mcOut.getBlocksize());\n+ int brlen = UtilFunctions.computeBlockSize(_mcOut.getRows(), ix.getRowIndex(), _mcOut.getBlocksize());\n+ int bclen = UtilFunctions.computeBlockSize(_mcOut.getCols(), ix.getColumnIndex(), _mcOut.getBlocksize());\n//check for pass-through\n- if( blen == mb.getNumRows() && blen == mb.getNumColumns() )\n+ if( brlen == mb.getNumRows() && bclen == mb.getNumColumns() )\nreturn arg0;\n//cbind or rbind to pad to right blocksize\n- if( blen > mb.getNumRows() ) //rbind\n- mb = mb.append(new MatrixBlock(blen-mb.getNumRows(),blen,true), new MatrixBlock(), false);\n- else if( blen > mb.getNumColumns() ) //cbind\n- mb = mb.append(new MatrixBlock(blen,blen-mb.getNumColumns(),true), new MatrixBlock(), true);\n+ if( brlen > mb.getNumRows() ) //rbind\n+ mb = mb.append(new MatrixBlock(brlen-mb.getNumRows(),bclen,true), new MatrixBlock(), false);\n+ else if( bclen > mb.getNumColumns() ) //cbind\n+ mb = mb.append(new MatrixBlock(brlen,bclen-mb.getNumColumns(),true), new MatrixBlock(), true);\nreturn new Tuple2<>(ix, mb);\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/CSVReblockSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/CSVReblockSPInstruction.java", "diff": "@@ -76,10 +76,10 @@ public class CSVReblockSPInstruction extends UnarySPInstruction {\nCPOperand in = new CPOperand(parts[1]);\nCPOperand out = new CPOperand(parts[2]);\nint blen = Integer.parseInt(parts[3]);\n- boolean hasHeader = Boolean.parseBoolean(parts[5]);\n- String delim = parts[6];\n- boolean fill = Boolean.parseBoolean(parts[7]);\n- double fillValue = Double.parseDouble(parts[8]);\n+ boolean hasHeader = Boolean.parseBoolean(parts[4]);\n+ String delim = parts[5];\n+ boolean fill = Boolean.parseBoolean(parts[6]);\n+ double fillValue = Double.parseDouble(parts[7]);\nreturn new CSVReblockSPInstruction(null, in, out, blen, blen,\nhasHeader, delim, fill, fillValue, opcode, str);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/RandSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/RandSPInstruction.java", "diff": "@@ -187,17 +187,17 @@ public class RandSPInstruction extends UnarySPInstruction {\nDataGenMethod method = DataGenMethod.INVALID;\nif ( opcode.equalsIgnoreCase(DataGen.RAND_OPCODE) ) {\nmethod = DataGenMethod.RAND;\n- InstructionUtils.checkNumFields ( str, 13 );\n+ InstructionUtils.checkNumFields ( str, 12 );\n}\nelse if ( opcode.equalsIgnoreCase(DataGen.SEQ_OPCODE) ) {\nmethod = DataGenMethod.SEQ;\n- // 8 operands: rows, cols, rpb, cpb, from, to, incr, outvar\n- InstructionUtils.checkNumFields ( str, 8 );\n+ // 8 operands: rows, cols, blen, from, to, incr, outvar\n+ InstructionUtils.checkNumFields ( str, 7 );\n}\nelse if ( opcode.equalsIgnoreCase(DataGen.SAMPLE_OPCODE) ) {\nmethod = DataGenMethod.SAMPLE;\n- // 7 operands: range, size, replace, seed, rpb, cpb, outvar\n- InstructionUtils.checkNumFields ( str, 7 );\n+ // 7 operands: range, size, replace, seed, blen, outvar\n+ InstructionUtils.checkNumFields ( str, 6 );\n}\nOperator op = null;\n@@ -209,23 +209,23 @@ public class RandSPInstruction extends UnarySPInstruction {\nCPOperand cols = new CPOperand(s[2]);\nCPOperand dims = new CPOperand(s[3]);\nint blen = Integer.parseInt(s[4]);\n- double sparsity = !s[8].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n- Double.valueOf(s[8]).doubleValue() : -1;\n- long seed = !s[9].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n- Long.valueOf(s[9]).longValue() : -1;\n- String dir = s[10];\n- String pdf = s[11];\n- String pdfParams = !s[12].contains( Lop.VARIABLE_NAME_PLACEHOLDER) ?\n- s[12] : null;\n+ double sparsity = !s[7].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ Double.valueOf(s[7]).doubleValue() : -1;\n+ long seed = !s[8].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ Long.valueOf(s[8]).longValue() : -1;\n+ String dir = s[9];\n+ String pdf = s[10];\n+ String pdfParams = !s[11].contains( Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ s[11] : null;\nreturn new RandSPInstruction(op, method, null, out, rows, cols, dims,\n- blen, s[6], s[7], sparsity, seed, dir, pdf, pdfParams, opcode, str);\n+ blen, s[5], s[6], sparsity, seed, dir, pdf, pdfParams, opcode, str);\n}\nelse if ( method == DataGenMethod.SEQ) {\nint blen = Integer.parseInt(s[3]);\n- CPOperand from = new CPOperand(s[5]);\n- CPOperand to = new CPOperand(s[6]);\n- CPOperand incr = new CPOperand(s[7]);\n+ CPOperand from = new CPOperand(s[4]);\n+ CPOperand to = new CPOperand(s[5]);\n+ CPOperand incr = new CPOperand(s[6]);\nCPOperand in = null;\nreturn new RandSPInstruction(op, method, in, out, null,\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/ReblockSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/ReblockSPInstruction.java", "diff": "@@ -74,7 +74,7 @@ public class ReblockSPInstruction extends UnarySPInstruction {\nCPOperand in = new CPOperand(parts[1]);\nCPOperand out = new CPOperand(parts[2]);\nint blen=Integer.parseInt(parts[3]);\n- boolean outputEmptyBlocks = Boolean.parseBoolean(parts[5]);\n+ boolean outputEmptyBlocks = Boolean.parseBoolean(parts[4]);\nOperator op = null; // no operator for ReblockSPInstruction\nreturn new ReblockSPInstruction(op, in, out, blen, blen, outputEmptyBlocks, opcode, str);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/util/ProgramConverter.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/util/ProgramConverter.java", "diff": "@@ -848,18 +848,17 @@ public class ProgramConverter\nvalue = mo.getFileName();\nPartitionFormat partFormat = (mo.getPartitionFormat()!=null) ? new PartitionFormat(\nmo.getPartitionFormat(),mo.getPartitionSize()) : PartitionFormat.NONE;\n- metaData = new String[11];\n+ metaData = new String[10];\nmetaData[0] = String.valueOf( dc.getRows() );\nmetaData[1] = String.valueOf( dc.getCols() );\nmetaData[2] = String.valueOf( dc.getBlocksize() );\n- metaData[3] = String.valueOf( dc.getBlocksize() );\n- metaData[4] = String.valueOf( dc.getNonZeros() );\n- metaData[5] = InputInfo.inputInfoToString( md.getInputInfo() );\n- metaData[6] = OutputInfo.outputInfoToString( md.getOutputInfo() );\n- metaData[7] = String.valueOf( partFormat );\n- metaData[8] = String.valueOf( mo.getUpdateType() );\n- metaData[9] = String.valueOf(mo.isHDFSFileExists());\n- metaData[10] = String.valueOf(mo.isCleanupEnabled());\n+ metaData[3] = String.valueOf( dc.getNonZeros() );\n+ metaData[4] = InputInfo.inputInfoToString( md.getInputInfo() );\n+ metaData[5] = OutputInfo.outputInfoToString( md.getOutputInfo() );\n+ metaData[6] = String.valueOf( partFormat );\n+ metaData[7] = String.valueOf( mo.getUpdateType() );\n+ metaData[8] = String.valueOf(mo.isHDFSFileExists());\n+ metaData[9] = String.valueOf(mo.isCleanupEnabled());\nbreak;\ncase LIST:\n// SCHEMA: <name>|<datatype>|<valuetype>|value|<metadata>|<tab>element1<tab>element2<tab>element3 (this is the list)\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/mlcontext/MLContextParforDatasetTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/mlcontext/MLContextParforDatasetTest.java", "diff": "@@ -46,7 +46,6 @@ import org.tugraz.sysds.test.TestUtils;\npublic class MLContextParforDatasetTest extends MLContextTestBase\n{\n-\nprivate final static int rows = 100;\nprivate final static int cols = 1600;\nprivate final static double sparsity = 0.7;\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/mlcontext/MLContextTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/mlcontext/MLContextTest.java", "diff": "@@ -1171,7 +1171,7 @@ public class MLContextTest extends MLContextTestBase {\n@Test\npublic void testCSVMatrixFromURLSumDML() throws MalformedURLException {\nSystem.out.println(\"MLContextTest - CSV matrix from URL sum DML\");\n- String csv = \"https://raw.githubusercontent.com/tugraz-isds/systemds/master/src/test/scripts/org/tugraz/sysds/api/mlcontext/1234.csv\";\n+ String csv = \"https://raw.githubusercontent.com/tugraz-isds/systemds/master/src/test/scripts/functions/mlcontext/1234.csv\";\nURL url = new URL(csv);\nScript script = dml(\"print('sum: ' + sum(M));\").in(\"M\", url);\nsetExpectedStdOut(\"sum: 10.0\");\n@@ -1181,7 +1181,7 @@ public class MLContextTest extends MLContextTestBase {\n@Test\npublic void testIJVMatrixFromURLSumDML() throws MalformedURLException {\nSystem.out.println(\"MLContextTest - IJV matrix from URL sum DML\");\n- String ijv = \"https://raw.githubusercontent.com/tugraz-isds/systemds/master/src/test/scripts/org/tugraz/sysds/api/mlcontext/1234.ijv\";\n+ String ijv = \"https://raw.githubusercontent.com/tugraz-isds/systemds/master/src/test/scripts/functions/mlcontext/1234.ijv\";\nURL url = new URL(ijv);\nMatrixMetadata mm = new MatrixMetadata(MatrixFormat.IJV, 2, 2);\nScript script = dml(\"print('sum: ' + sum(M));\").in(\"M\", url, mm);\n" } ]

[ { "change_type": "MODIFY", "old_path": ".gitignore", "new_path": ".gitignore", "diff": "@@ -44,6 +44,11 @@ docs/api\n# Excluded sources\n# (we're including the generated files to simply IDE setup w/o mvn build)\n+# Since the gitignore file does not ignore tracked files, the auto-generated\n+# parser files need to be ignored by the following command:\n+# git update-index --assume-unchanged [<file> ...]\n+# This can be undone by git update-index --no-assume-unchanged [<file> ...]\n+# Use the command above on these files:\n# src/main/java/Dml.tokens\n# src/main/java/DmlLexer.tokens\n# src/main/java/org/tugraz/sysds/parser/dml/DmlBaseListener.java\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -82,6 +82,7 @@ SYSTEMDS-110 New Builtin Functions\n* 114 Builtin function for stepwise regression OK\n* 115 Builtin function for model debugging (slice finder) OK\n* 116 Builtin function for kmeans OK\n+ * 117 Builtin function for lm cross validation OK\nSYSTEMDS-120 Performance Features\n* 121 Avoid spark context creation on parfor result merge OK\n" }, { "change_type": "ADD", "old_path": null, "new_path": "scripts/builtin/cvlm.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+m_cvlm = function(Matrix[Double] X, Matrix[Double] y, Integer k, Integer icpt = 0, Double reg = 1e-7) return (Matrix[Double] y_predict, Matrix[Double] allbeta)\n+{\n+ M = nrow(X);\n+ lim = as.integer(M/k);\n+ y_predict = y;\n+ allbeta = matrix(0, rows=k, cols=ncol(X));\n+\n+ for (i in 1:k)\n+ {\n+ testS = ifelse(i==1, 1, ((i-1) * lim)+1)\n+ testE = i * lim;\n+ testSet = X[testS:testE,];\n+\n+ if (i == 1) {\n+ trainSet = X[testE+1:M,];\n+ trainRes = y[testE+1:M,];\n+ }\n+ else {\n+ trainSet = rbind(X[1:testS-1,], X[testE+1:M,]);\n+ trainRes = rbind(y[1:testS-1,], y[testE+1:M,]);\n+ }\n+\n+ beta = lm(X=trainSet, y=trainRes, icpt=icpt, reg=reg);\n+ pred = lmpredict(X=testSet, w=beta, icpt=icpt);\n+ y_predict[testS:testE,] = pred;\n+ allbeta[i,] = t(beta);\n+ }\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "new_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "diff": "@@ -174,6 +174,7 @@ public enum Builtins {\nRMEMPTY(\"removeEmpty\", false, true),\nSCALE(\"scale\", true, false), //TODO parameterize center & scale\nTIME(\"time\", false),\n+ CVLM(\"cvlm\", true, false),\nTOSTRING(\"toString\", false, true),\nTRANSFORMAPPLY(\"transformapply\", false, true),\nTRANSFORMCOLMAP(\"transformcolmap\", false, true),\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinCVLmTest.java", "diff": "+/*\n+ * Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed under the Apache License, Version 2.0 (the \"License\");\n+ * you may not use this file except in compliance with the License.\n+ * You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing, software\n+ * distributed under the License is distributed on an \"AS IS\" BASIS,\n+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+ * See the License for the specific language governing permissions and\n+ * limitations under the License.\n+ */\n+\n+package org.tugraz.sysds.test.functions.builtin;\n+\n+import java.util.ArrayList;\n+import java.util.List;\n+\n+import org.junit.Test;\n+import org.tugraz.sysds.test.AutomatedTestBase;\n+import org.tugraz.sysds.test.TestConfiguration;\n+\n+\n+public class BuiltinCVLmTest extends AutomatedTestBase\n+{\n+ private final static String TEST_NAME = \"cvlm\";\n+ private final static String TEST_DIR = \"functions/builtin/\";\n+ private final static String TEST_CLASS_DIR = TEST_DIR + BuiltinCVLmTest.class.getSimpleName() + \"/\";\n+\n+ private final static int rows = 100;\n+ private final static int cols = 10;\n+\n+ @Override\n+ public void setUp() {\n+ addTestConfiguration(TEST_NAME,new TestConfiguration(TEST_CLASS_DIR, TEST_NAME,new String[]{\"B\"}));\n+ }\n+\n+ @Test\n+ public void testlmCV() {\n+ runtestlmCV();\n+ }\n+\n+ private void runtestlmCV()\n+ {\n+ loadTestConfiguration(getTestConfiguration(TEST_NAME));\n+ String HOME = SCRIPT_DIR + TEST_DIR;\n+ fullDMLScriptName = HOME + TEST_NAME + \".dml\";\n+ List<String> proArgs = new ArrayList<String>();\n+\n+ int k = 3;\n+ proArgs.add(\"-explain\");\n+ proArgs.add(\"-stats\");\n+ proArgs.add(\"-args\");\n+ proArgs.add(input(\"X\"));\n+ proArgs.add(input(\"y\"));\n+ proArgs.add(String.valueOf(k));\n+ proArgs.add(output(\"y_predict\"));\n+ proArgs.add(output(\"beta\"));\n+ programArgs = proArgs.toArray(new String[proArgs.size()]);\n+ double[][] X = getRandomMatrix(rows, cols, 0, 1, 0.8, -1);\n+ double[][] y = getRandomMatrix(rows, 1, 0, 1, 0.8, -1);\n+ writeInputMatrixWithMTD(\"X\", X, true);\n+ writeInputMatrixWithMTD(\"y\", y, true);\n+\n+ runTest(true, EXCEPTION_NOT_EXPECTED, null, -1);\n+ }\n+}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/builtin/cvlm.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1);\n+y = read($2);\n+[Y, beta] = cvlm(X=X, y=y, k=$3);\n+write(Y, $4)\n+write(beta, $5)\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -61,6 +61,7 @@ SYSTEMDS-70 Lineage Tracing and Reuse OK\n* 76 Generate runtime plan from lineage trace OK\n* 77 New builtin function for obtaining lineage OK\n* 78 Extended lineage tracing (parfor, funs, codegen) OK\n+ * 79 Reuse cache configuration options\nSYSTEMDS-80 Improved distributed operations\n* 81 Avoid unnecessary replication on rmm\n@@ -108,3 +109,11 @@ SYSTEMDS-150 Releases\nSYSTEMDS-160 Tensor Compiler/Runtime\n* 161 Local readers and writers for tensors (all formats)\n* 162 Spark binary tensor instructions\n+\n+ SYSTEMDS-170 Lineage full and partial reuse\n+ * 171 Initial version of partial rewrites\n+ * 172 Parfor integration (blocked waiting for results)\n+ * 173 Improved cost estimates and ba+* support\n+ * 174 Reuse rewrite for rbind-tsmm\n+ * 175 Reuse rewrite for cbind/rbind-elementwise */+\n+ * 176 Reuse rewrite for aggregate\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/recompile/Recompiler.java", "new_path": "src/main/java/org/tugraz/sysds/hops/recompile/Recompiler.java", "diff": "@@ -57,6 +57,7 @@ import org.tugraz.sysds.parser.DMLProgram;\nimport org.tugraz.sysds.parser.DataExpression;\nimport org.tugraz.sysds.parser.ForStatementBlock;\nimport org.tugraz.sysds.parser.IfStatementBlock;\n+import org.tugraz.sysds.parser.ParseInfo;\nimport org.tugraz.sysds.parser.Statement;\nimport org.tugraz.sysds.parser.StatementBlock;\nimport org.tugraz.sysds.parser.WhileStatementBlock;\n@@ -395,12 +396,13 @@ public class Recompiler\n}\nprivate static void logExplainDAG(StatementBlock sb, ArrayList<Hop> hops, ArrayList<Instruction> inst) {\n+ ParseInfo pi = (sb != null) ? sb : hops.get(0);\nif( DMLScript.EXPLAIN == ExplainType.RECOMPILE_HOPS ) {\n- System.out.println(\"EXPLAIN RECOMPILE \\nGENERIC (lines \"+sb.getBeginLine()+\"-\"+sb.getEndLine()+\"):\\n\" +\n+ System.out.println(\"EXPLAIN RECOMPILE \\nGENERIC (lines \"+pi.getBeginLine()+\"-\"+pi.getEndLine()+\"):\\n\" +\nExplain.explainHops(hops, 1));\n}\nif( DMLScript.EXPLAIN == ExplainType.RECOMPILE_RUNTIME ) {\n- System.out.println(\"EXPLAIN RECOMPILE \\nGENERIC (lines \"+sb.getBeginLine()+\"-\"+sb.getEndLine()+\"):\\n\" +\n+ System.out.println(\"EXPLAIN RECOMPILE \\nGENERIC (lines \"+pi.getBeginLine()+\"-\"+pi.getEndLine()+\"):\\n\" +\nExplain.explain(inst, 1));\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/rewrite/HopRewriteUtils.java", "new_path": "src/main/java/org/tugraz/sysds/hops/rewrite/HopRewriteUtils.java", "diff": "@@ -65,6 +65,8 @@ import org.tugraz.sysds.parser.IfStatementBlock;\nimport org.tugraz.sysds.parser.Statement;\nimport org.tugraz.sysds.parser.StatementBlock;\nimport org.tugraz.sysds.parser.WhileStatementBlock;\n+import org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject;\n+import org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject.UpdateType;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\nimport org.tugraz.sysds.runtime.instructions.cp.ScalarObject;\nimport org.tugraz.sysds.runtime.instructions.cp.ScalarObjectFactory;\n@@ -535,6 +537,26 @@ public class HopRewriteUtils\nreturn tread;\n}\n+ public static DataOp createTransientRead(String name, MatrixBlock mb) {\n+ DataOp tread = new DataOp(name, DataType.MATRIX, ValueType.FP64, DataOpTypes.TRANSIENTREAD,\n+ null, mb.getNumRows(), mb.getNumColumns(), mb.getNonZeros(), UpdateType.COPY,\n+ ConfigurationManager.getBlocksize());\n+ tread.setVisited();\n+ copyLineNumbers(mb, tread);\n+ tread.setFileName(name);\n+ return tread;\n+ }\n+\n+ public static DataOp createTransientRead(String name, MatrixObject mo) {\n+ DataOp tread = new DataOp(name, DataType.MATRIX, ValueType.FP64, DataOpTypes.TRANSIENTREAD,\n+ null, mo.getNumRows(), mo.getNumColumns(), mo.getNnz(), UpdateType.COPY,\n+ (int)mo.getBlocksize());\n+ tread.setVisited();\n+ copyLineNumbers(mo, tread);\n+ tread.setFileName(name);\n+ return tread;\n+ }\n+\npublic static DataOp createTransientWrite(String name, Hop in) {\nreturn createDataOp(name, in, DataOpTypes.TRANSIENTWRITE);\n}\n@@ -776,6 +798,20 @@ public class HopRewriteUtils\ndest.setParseInfo(src);\n}\n+ public static void copyLineNumbers(MatrixBlock mb, Hop tread) {\n+ tread.setBeginLine(1);\n+ tread.setEndLine(mb.getNumRows());\n+ tread.setBeginColumn(1);\n+ tread.setEndColumn(mb.getNumColumns());\n+ }\n+\n+ public static void copyLineNumbers(MatrixObject mo, Hop tread) {\n+ tread.setBeginLine(1);\n+ tread.setEndLine((int)mo.getNumRows());\n+ tread.setBeginColumn(1);\n+ tread.setEndColumn((int)mo.getNumColumns());\n+ }\n+\npublic static void updateHopCharacteristics( Hop hop, int blen, Hop src ) {\nupdateHopCharacteristics(hop, blen, new MemoTable(), src);\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/Lineage.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/Lineage.java", "diff": "@@ -20,6 +20,7 @@ import org.tugraz.sysds.runtime.controlprogram.ForProgramBlock;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.instructions.Instruction;\nimport org.tugraz.sysds.runtime.instructions.cp.CPOperand;\n+import org.tugraz.sysds.runtime.lineage.LineageCacheConfig.CacheType;\nimport java.util.HashMap;\nimport java.util.Map;\n@@ -98,4 +99,20 @@ public class Lineage {\nLineageItem.resetIDSequence();\nLineageCache.resetCache();\n}\n+\n+ public static void setLinReusePartial() {\n+ LineageCacheConfig.setConfigTsmmCbind(CacheType.PARTIAL);\n+ }\n+\n+ public static void setLinReuseFull() {\n+ LineageCacheConfig.setConfigTsmmCbind(CacheType.FULL);\n+ }\n+\n+ public static void setLinReuseFullAndPartial() {\n+ LineageCacheConfig.setConfigTsmmCbind(CacheType.HYBRID_FULL_PARTIAL);\n+ }\n+\n+ public static void setLinReuseNone() {\n+ LineageCacheConfig.setConfigTsmmCbind(CacheType.NONE);\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "diff": "@@ -105,7 +105,14 @@ public class LineageCache {\npublic static boolean reuse(Instruction inst, ExecutionContext ec) {\nif (!DMLScript.LINEAGE_REUSE)\nreturn false;\n+ if (LineageCacheConfig.getCacheType().isFullReuse())\n+ return fullReuse(inst, ec);\n+ if (LineageCacheConfig.getCacheType().isPartialReuse())\n+ return RewriteCPlans.executeRewrites(inst, ec);\n+ return false;\n+ }\n+ private static boolean fullReuse (Instruction inst, ExecutionContext ec) {\nif (inst instanceof ComputationCPInstruction && LineageCache.isReusable(inst)) {\nboolean reused = true;\nLineageItem[] items = ((ComputationCPInstruction) inst).getLineageItems(ec);\n@@ -138,8 +145,9 @@ public class LineageCache {\npublic static boolean isReusable (Instruction inst) {\n// TODO: Move this to the new class LineageCacheConfig and extend\n- return (inst.getOpcode().equalsIgnoreCase(\"tsmm\")\n- || inst.getOpcode().equalsIgnoreCase(\"ba+*\"));\n+ return (inst.getOpcode().equalsIgnoreCase(\"tsmm\"));\n+ //|| inst.getOpcode().equalsIgnoreCase(\"ba+*\"));\n+ // TODO: Fix getRecomputeEstimate to support ba+* before enabling above code.\n}\n//---------------- CACHE SPACE MANAGEMENT METHODS -----------------\n@@ -205,6 +213,7 @@ public class LineageCache {\nswitch (cptype)\n{\ncase MMTSJ:\n+ //case AggregateBinary:\nMMTSJType type = ((MMTSJCPInstruction)inst).getMMTSJType();\nif (type.isLeft())\nnflops = !sparse ? (r * c * s * c /2):(r * c * s * c * s /2);\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCacheConfig.java", "diff": "+/*\n+ * Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed under the Apache License, Version 2.0 (the \"License\");\n+ * you may not use this file except in compliance with the License.\n+ * You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing, software\n+ * distributed under the License is distributed on an \"AS IS\" BASIS,\n+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+ * See the License for the specific language governing permissions and\n+ * limitations under the License.\n+ */\n+\n+package org.tugraz.sysds.runtime.lineage;\n+\n+import org.tugraz.sysds.api.DMLScript;\n+import java.util.ArrayList;\n+\n+public class LineageCacheConfig {\n+ public enum CacheType {\n+ FULL, // no rewrites\n+ PARTIAL,\n+ HYBRID_FULL_PARTIAL,\n+ NONE;\n+ public boolean isFullReuse() {\n+ return this == FULL || this == HYBRID_FULL_PARTIAL;\n+ }\n+ public boolean isPartialReuse() {\n+ return this == PARTIAL || this == HYBRID_FULL_PARTIAL;\n+ }\n+ }\n+\n+ public ArrayList<String> _MMult = new ArrayList<String>();\n+\n+ public enum CachedItemHead {\n+ TSMM,\n+ ALL\n+ }\n+\n+ public enum CachedItemTail {\n+ CBIND,\n+ RBIND,\n+ INDEX,\n+ ALL\n+ }\n+\n+ private static CacheType _cacheType = null;\n+ private static CachedItemHead _itemH = null;\n+ private static CachedItemTail _itemT = null;\n+\n+ public static void setConfigTsmmCbind(CacheType ct) {\n+ _cacheType = ct;\n+ _itemH = CachedItemHead.TSMM;\n+ _itemT = CachedItemTail.CBIND;\n+ }\n+\n+ public static void setConfig(CacheType ct, CachedItemHead ith, CachedItemTail itt) {\n+ _cacheType = ct;\n+ _itemH = ith;\n+ _itemT = itt;\n+ }\n+\n+ public static void shutdownReuse() {\n+ DMLScript.LINEAGE = false;\n+ DMLScript.LINEAGE_REUSE = false;\n+ }\n+\n+ public static void restartReuse() {\n+ DMLScript.LINEAGE = true;\n+ DMLScript.LINEAGE_REUSE = true;\n+ }\n+\n+ public static CacheType getCacheType() {\n+ return _cacheType;\n+ }\n+\n+ public static CachedItemHead getCachedItemHead() {\n+ return _itemH;\n+ }\n+\n+ public static CachedItemTail getCachedItemTail() {\n+ return _itemT;\n+ }\n+}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/RewriteCPlans.java", "diff": "+/*\n+ * Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed under the Apache License, Version 2.0 (the \"License\");\n+ * you may not use this file except in compliance with the License.\n+ * You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing, software\n+ * distributed under the License is distributed on an \"AS IS\" BASIS,\n+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+ * See the License for the specific language governing permissions and\n+ * limitations under the License.\n+ */\n+\n+package org.tugraz.sysds.runtime.lineage;\n+\n+import java.util.ArrayList;\n+\n+import org.apache.commons.logging.Log;\n+import org.apache.commons.logging.LogFactory;\n+import org.tugraz.sysds.api.DMLScript;\n+import org.tugraz.sysds.common.Types.ValueType;\n+import org.tugraz.sysds.hops.AggBinaryOp;\n+import org.tugraz.sysds.hops.BinaryOp;\n+import org.tugraz.sysds.hops.DataOp;\n+import org.tugraz.sysds.hops.Hop;\n+import org.tugraz.sysds.hops.Hop.OpOp2;\n+import org.tugraz.sysds.hops.Hop.OpOpN;\n+import org.tugraz.sysds.hops.IndexingOp;\n+import org.tugraz.sysds.hops.LiteralOp;\n+import org.tugraz.sysds.hops.NaryOp;\n+import org.tugraz.sysds.hops.ReorgOp;\n+import org.tugraz.sysds.hops.recompile.Recompiler;\n+import org.tugraz.sysds.hops.rewrite.HopRewriteUtils;\n+import org.tugraz.sysds.runtime.DMLRuntimeException;\n+import org.tugraz.sysds.runtime.controlprogram.BasicProgramBlock;\n+import org.tugraz.sysds.runtime.controlprogram.Program;\n+import org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject;\n+import org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\n+import org.tugraz.sysds.runtime.controlprogram.context.ExecutionContextFactory;\n+import org.tugraz.sysds.runtime.instructions.Instruction;\n+import org.tugraz.sysds.runtime.instructions.cp.ComputationCPInstruction;\n+import org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\n+import org.tugraz.sysds.runtime.meta.MetaData;\n+import org.tugraz.sysds.utils.Explain;\n+import org.tugraz.sysds.utils.Explain.ExplainType;\n+\n+public class RewriteCPlans\n+{\n+ private static final String LR_VAR = \"__lrwrt\";\n+ private static BasicProgramBlock _lrPB = null;\n+ private static ExecutionContext _lrEC = null;\n+ private static final Log LOG = LogFactory.getLog(RewriteCPlans.class.getName());\n+\n+ public static boolean executeRewrites (Instruction curr, ExecutionContext ec)\n+ {\n+ boolean oneappend = false;\n+ boolean twoappend = false;\n+ MatrixBlock lastResult = null;\n+ if (curr.getOpcode().equalsIgnoreCase(\"tsmm\"))\n+ {\n+ // If the input to tsmm came from cbind, look for both the inputs in cache.\n+ LineageItem[] items = ((ComputationCPInstruction) curr).getLineageItems(ec);\n+ LineageItem item = items[0];\n+\n+ // TODO restructuring of rewrites to make them all\n+ // independent of each other and this opening condition here\n+ for (LineageItem source : item.getInputs())\n+ if (source.getOpcode().equalsIgnoreCase(\"append\")) {\n+ for (LineageItem input : source.getInputs()) {\n+ // create tsmm lineage on top of the input of last append\n+ LineageItem tmp = new LineageItem(\"toProbe\", curr.getOpcode(), new LineageItem[] {input});\n+ if (LineageCache.probe(tmp)) {\n+ oneappend = true; // at least one entry to reuse\n+ if (lastResult == null)\n+ lastResult = LineageCache.get(curr, tmp);\n+ }\n+ }\n+ if (oneappend)\n+ break; // no need to look for the next append\n+\n+ // if not found in cache, look for two consecutive cbinds\n+ LineageItem input = source.getInputs()[0];\n+ if (input.getOpcode().equalsIgnoreCase(\"append\")) {\n+ for (LineageItem L2appin : input.getInputs()) {\n+ LineageItem tmp = new LineageItem(\"comb\", \"append\", new LineageItem[] {L2appin, source.getInputs()[1]});\n+ LineageItem toProbe = new LineageItem(\"toProbe\", curr.getOpcode(), new LineageItem[] {tmp});\n+ if (LineageCache.probe(toProbe)) {\n+ twoappend = true;\n+ if (lastResult == null)\n+ lastResult = LineageCache.get(curr, toProbe);\n+ }\n+ }\n+ }\n+ }\n+ }\n+ else\n+ return false;\n+\n+ if (!oneappend && !twoappend)\n+ return false;\n+\n+ ExecutionContext lrwec = getExecutionContext();\n+ ExplainType et = DMLScript.EXPLAIN;\n+ // Disable explain not to print unnecessary logs\n+ // TODO extend recompiler to allow use without explain output\n+ DMLScript.EXPLAIN = ExplainType.NONE;\n+\n+ try {\n+ ArrayList<Instruction> newInst = oneappend ? rewriteCbindTsmm(curr, ec, lrwec, lastResult) :\n+ twoappend ? rewrite2CbindTsmm(curr, ec, lrwec, lastResult) : null;\n+ //execute instructions\n+ BasicProgramBlock pb = getProgramBlock();\n+ pb.setInstructions(newInst);\n+ LineageCacheConfig.shutdownReuse();\n+ pb.execute(lrwec);\n+ LineageCacheConfig.restartReuse();\n+ ec.setVariable(((ComputationCPInstruction)curr).output.getName(), lrwec.getVariable(LR_VAR));\n+ // add this to cache\n+ LineageCache.put(curr, ec);\n+ }\n+ catch (Exception e) {\n+ throw new DMLRuntimeException(\"Error evaluating instruction: \" + curr.toString() , e);\n+ }\n+ DMLScript.EXPLAIN = et;\n+ return true;\n+ }\n+\n+ private static ArrayList<Instruction> rewriteCbindTsmm(Instruction curr, ExecutionContext ec, ExecutionContext lrwec, MatrixBlock lastResult)\n+ {\n+ // Create a transient read op over the last tsmm result\n+ MetaData md = new MetaData(lastResult.getDataCharacteristics());\n+ MatrixObject newmo = new MatrixObject(ValueType.FP64, \"lastResult\", md);\n+ newmo.acquireModify(lastResult);\n+ newmo.release();\n+ lrwec.setVariable(\"lastResult\", newmo);\n+ DataOp lastRes = HopRewriteUtils.createTransientRead(\"lastResult\", lastResult);\n+ // Create rightIndex op to find the last matrix and the appended column\n+ // TODO: For now assumption is that a single column is being appended in a loop\n+ // Need to go down the lineage to find number of columns are being appended.\n+ MatrixObject mo = ec.getMatrixObject(((ComputationCPInstruction)curr).input1);\n+ lrwec.setVariable(\"oldMatrix\", mo);\n+ DataOp newMatrix = HopRewriteUtils.createTransientRead(\"oldMatrix\", mo);\n+ IndexingOp oldMatrix = HopRewriteUtils.createIndexingOp(newMatrix, new LiteralOp(1),\n+ new LiteralOp(mo.getNumRows()), new LiteralOp(1), new LiteralOp(mo.getNumColumns()-1));\n+ IndexingOp lastCol = HopRewriteUtils.createIndexingOp(newMatrix, new LiteralOp(1),\n+ new LiteralOp(mo.getNumRows()), new LiteralOp(mo.getNumColumns()),\n+ new LiteralOp(mo.getNumColumns()));\n+ // cell topRight = t(oldMatrix) %*% lastCol\n+ ReorgOp tOldM = HopRewriteUtils.createTranspose(oldMatrix);\n+ AggBinaryOp topRight = HopRewriteUtils.createMatrixMultiply(tOldM, lastCol);\n+ // cell bottomLeft = t(lastCol) %*% oldMatrix\n+ ReorgOp tLastCol = HopRewriteUtils.createTranspose(lastCol);\n+ AggBinaryOp bottomLeft = HopRewriteUtils.createMatrixMultiply(tLastCol, oldMatrix);\n+ // bottomRight = t(lastCol) %*% lastCol\n+ AggBinaryOp bottomRight = HopRewriteUtils.createMatrixMultiply(tLastCol, lastCol);\n+ // rowOne = cbind(lastRes, topRight)\n+ BinaryOp rowOne = HopRewriteUtils.createBinary(lastRes, topRight, OpOp2.CBIND);\n+ // rowTwo = cbind(bottomLeft, bottomRight)\n+ BinaryOp rowTwo = HopRewriteUtils.createBinary(bottomLeft, bottomRight, OpOp2.CBIND);\n+ // rbind(rowOne, rowTwo)\n+ BinaryOp lrwHop= HopRewriteUtils.createBinary(rowOne, rowTwo, OpOp2.RBIND);\n+ DataOp lrwWrite = HopRewriteUtils.createTransientWrite(LR_VAR, lrwHop);\n+\n+ // generate runtime instructions\n+ LOG.debug(\"LINEAGE REWRITE rewriteCbindTsmm APPLIED\");\n+ return genInst(lrwWrite, lrwec);\n+ }\n+\n+ private static ArrayList<Instruction> rewrite2CbindTsmm(Instruction curr, ExecutionContext ec, ExecutionContext lrwec, MatrixBlock lastResult)\n+ {\n+ // Create a transient read op over the last tsmm result\n+ MetaData md = new MetaData(lastResult.getDataCharacteristics());\n+ MatrixObject newmo = new MatrixObject(ValueType.FP64, \"lastResult\", md);\n+ newmo.acquireModify(lastResult);\n+ newmo.release();\n+ lrwec.setVariable(\"lastResult\", newmo);\n+ DataOp lastRes = HopRewriteUtils.createTransientRead(\"lastResult\", lastResult);\n+ MatrixObject mo = ec.getMatrixObject(((ComputationCPInstruction)curr).input1);\n+ lrwec.setVariable(\"oldMatrix\", mo);\n+ DataOp newMatrix = HopRewriteUtils.createTransientRead(\"oldMatrix\", mo);\n+\n+ // pull out the newly added column(2nd last) from the input matrix\n+ IndexingOp lastCol = HopRewriteUtils.createIndexingOp(newMatrix, new LiteralOp(1), new LiteralOp(mo.getNumRows()),\n+ new LiteralOp(mo.getNumColumns()-1), new LiteralOp(mo.getNumColumns()-1));\n+ // apply t(lastCol) on i/p matrix to get the result vectors.\n+ ReorgOp tlastCol = HopRewriteUtils.createTranspose(lastCol);\n+ AggBinaryOp newCol = HopRewriteUtils.createMatrixMultiply(tlastCol, newMatrix);\n+ ReorgOp tnewCol = HopRewriteUtils.createTranspose(newCol);\n+\n+ // push the result row & column inside the cashed block as 2nd last row and col respectively.\n+ IndexingOp topLeft = HopRewriteUtils.createIndexingOp(lastRes, new LiteralOp(1), new LiteralOp(newmo.getNumRows()-1),\n+ new LiteralOp(1), new LiteralOp(newmo.getNumColumns()-1));\n+ IndexingOp topRight = HopRewriteUtils.createIndexingOp(lastRes, new LiteralOp(1), new LiteralOp(newmo.getNumRows()-1),\n+ new LiteralOp(newmo.getNumColumns()), new LiteralOp(newmo.getNumColumns()));\n+ IndexingOp bottomLeft = HopRewriteUtils.createIndexingOp(lastRes, new LiteralOp(newmo.getNumRows()),\n+ new LiteralOp(newmo.getNumRows()), new LiteralOp(1), new LiteralOp(newmo.getNumColumns()-1));\n+ IndexingOp bottomRight = HopRewriteUtils.createIndexingOp(lastRes, new LiteralOp(newmo.getNumRows()),\n+ new LiteralOp(newmo.getNumRows()), new LiteralOp(newmo.getNumColumns()), new LiteralOp(newmo.getNumColumns()));\n+ IndexingOp topCol = HopRewriteUtils.createIndexingOp(tnewCol, new LiteralOp(1), new LiteralOp(mo.getNumColumns()-2),\n+ new LiteralOp(1), new LiteralOp(1));\n+ IndexingOp bottomCol = HopRewriteUtils.createIndexingOp(tnewCol, new LiteralOp(mo.getNumColumns()),\n+ new LiteralOp(mo.getNumColumns()), new LiteralOp(1), new LiteralOp(1));\n+ NaryOp rowOne = HopRewriteUtils.createNary(OpOpN.CBIND, topLeft, topCol, topRight);\n+ NaryOp rowTwo = HopRewriteUtils.createNary(OpOpN.CBIND, bottomLeft, bottomCol, bottomRight);\n+ NaryOp lrwHop = HopRewriteUtils.createNary(OpOpN.RBIND, rowOne, newCol, rowTwo);\n+ DataOp lrwWrite = HopRewriteUtils.createTransientWrite(LR_VAR, lrwHop);\n+\n+ // generate runtime instructions\n+ LOG.debug(\"LINEAGE REWRITE rewrite2CbindTsmm APPLIED\");\n+ return genInst(lrwWrite, lrwec);\n+ }\n+\n+ private static ArrayList<Instruction> genInst(Hop hops, ExecutionContext ec) {\n+ ArrayList<Instruction> newInst = Recompiler.recompileHopsDag(hops, ec.getVariables(), null, true, true, 0);\n+ LOG.debug(\"EXPLAIN LINEAGE REWRITE \\nGENERIC (line \"+hops.getBeginLine()+\"):\\n\" + Explain.explain(hops,1));\n+ LOG.debug(\"EXPLAIN LINEAGE REWRITE \\nGENERIC (line \"+hops.getBeginLine()+\"):\\n\" + Explain.explain(newInst,1));\n+ return newInst;\n+ }\n+\n+ private static ExecutionContext getExecutionContext() {\n+ if( _lrEC == null )\n+ _lrEC = ExecutionContextFactory.createContext();\n+ return _lrEC;\n+ }\n+\n+ private static BasicProgramBlock getProgramBlock() {\n+ if( _lrPB == null )\n+ _lrPB = new BasicProgramBlock( new Program() );\n+ return _lrPB;\n+ }\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/FullReusePerfTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/FullReusePerfTest.java", "diff": "@@ -90,7 +90,9 @@ public class FullReusePerfTest extends AutomatedTestBase\nwriteInputMatrixWithMTD(\"X\", X, true);\nLineage.resetInternalState();\n+ Lineage.setLinReuseFull();\nrunTest(true, EXCEPTION_NOT_EXPECTED, null, -1);\n+ Lineage.setLinReuseNone();\nString X_lineage = readDMLLineageFromHDFS(\"X\");\nLineageItem X_li = LineageParser.parseLineageTrace(X_lineage);\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/FullReuseTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/FullReuseTest.java", "diff": "@@ -95,8 +95,10 @@ public class FullReuseTest extends AutomatedTestBase {\nprogramArgs = proArgs.toArray(new String[proArgs.size()]);\nLineage.resetInternalState();\n+ Lineage.setLinReuseFull();\nrunTest(true, EXCEPTION_NOT_EXPECTED, null, -1);\nHashMap<MatrixValue.CellIndex, Double> X_reused = readDMLMatrixFromHDFS(\"X\");\n+ Lineage.setLinReuseNone();\nTestUtils.compareMatrices(X_orig, X_reused, 1e-6, \"Origin\", \"Reused\");\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageRewriteTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageRewriteTest.java", "diff": "package org.tugraz.sysds.test.functions.lineage;\nimport java.util.ArrayList;\n+import java.util.HashMap;\nimport java.util.List;\nimport org.junit.Test;\n+import org.tugraz.sysds.hops.recompile.Recompiler;\n+import org.tugraz.sysds.runtime.lineage.Lineage;\n+import org.tugraz.sysds.runtime.matrix.data.MatrixValue;\nimport org.tugraz.sysds.test.AutomatedTestBase;\nimport org.tugraz.sysds.test.TestConfiguration;\nimport org.tugraz.sysds.test.TestUtils;\npublic class LineageRewriteTest extends AutomatedTestBase {\nprotected static final String TEST_DIR = \"functions/lineage/\";\n- protected static final String TEST_NAME1 = \"RewriteTest1\";\n+ protected static final String TEST_NAME1 = \"RewriteTest3\";\n+ protected static final String TEST_NAME2 = \"RewriteTest2\";\nprotected String TEST_CLASS_DIR = TEST_DIR + LineageRewriteTest.class.getSimpleName() + \"/\";\n- protected static final int numRecords = 10;\n- protected static final int numFeatures = 6;\n+ protected static final int numRecords = 1000;\n+ protected static final int numFeatures = 100;\n@Override\npublic void setUp() {\nTestUtils.clearAssertionInformation();\naddTestConfiguration(TEST_NAME1, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME1));\n+ addTestConfiguration(TEST_NAME2, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME2));\n}\n@Test\n@@ -44,22 +50,50 @@ public class LineageRewriteTest extends AutomatedTestBase {\ntestRewrite(TEST_NAME1);\n}\n+ @Test\n+ public void testRewrite2() {\n+ testRewrite(TEST_NAME2);\n+ }\n+\nprivate void testRewrite(String testname) {\n+ try {\ngetAndLoadTestConfiguration(testname);\n- List<String> proArgs = new ArrayList<>();\n+ List<String> proArgs = new ArrayList<String>();\nproArgs.add(\"-explain\");\n+ proArgs.add(\"-stats\");\nproArgs.add(\"-lineage\");\nproArgs.add(\"-args\");\nproArgs.add(input(\"X\"));\n- proArgs.add(output(\"Res1\"));\n- proArgs.add(output(\"Res2\"));\n+ proArgs.add(output(\"Res\"));\nprogramArgs = proArgs.toArray(new String[proArgs.size()]);\nfullDMLScriptName = getScript();\ndouble[][] X = getRandomMatrix(numRecords, numFeatures, 0, 1, 0.8, -1);\nwriteInputMatrixWithMTD(\"X\", X, true);\n+ runTest(true, EXCEPTION_NOT_EXPECTED, null, -1);\n+ HashMap<MatrixValue.CellIndex, Double> R_orig = readDMLMatrixFromHDFS(\"Res\");\n- //run the test\n+ proArgs.clear();\n+ proArgs.add(\"-explain\");\n+ proArgs.add(\"recompile_hops\");\n+ proArgs.add(\"-stats\");\n+ proArgs.add(\"-lineage\");\n+ proArgs.add(\"reuse\");\n+ proArgs.add(\"-args\");\n+ proArgs.add(input(\"X\"));\n+ proArgs.add(output(\"Res\"));\n+ programArgs = proArgs.toArray(new String[proArgs.size()]);\n+ fullDMLScriptName = getScript();\n+ writeInputMatrixWithMTD(\"X\", X, true);\n+ Lineage.resetInternalState();\n+ Lineage.setLinReusePartial();\nrunTest(true, EXCEPTION_NOT_EXPECTED, null, -1);\n+ Lineage.setLinReuseNone();\n+ HashMap<MatrixValue.CellIndex, Double> R_reused = readDMLMatrixFromHDFS(\"Res\");\n+ TestUtils.compareMatrices(R_orig, R_reused, 1e-6, \"Origin\", \"Reused\");\n+ }\n+ finally {\n+ Recompiler.reinitRecompiler();\n+ }\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/lineage/RewriteTest1.dml", "new_path": "src/test/scripts/functions/lineage/RewriteTest1.dml", "diff": "@@ -20,6 +20,7 @@ X = read($1);\n# calculate and same the trans-mult of X\nRes1 = t(X) %*% X;\n+while(FALSE) {}\n# find vector V as the last column and matrix X1 as the rest\nnc = ncol(X);\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/lineage/RewriteTest2.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1);\n+\n+sum = 0;\n+tmp = X[,1];\n+R = matrix(0, 1, ncol(X));\n+\n+for (i in 2:ncol(X)) {\n+ Res1 = t(tmp) %*% tmp;\n+ tmp = cbind(tmp, X[,i]);\n+ while(FALSE) {}\n+ R[1,i] = sum(Res1);\n+ sum = sum + sum(Res1);\n+}\n+\n+write(R, $2, format=\"text\");\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/lineage/RewriteTest3.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1);\n+\n+sum = 0;\n+tmp = X[,1];\n+tmp1 = matrix(0, rows=nrow(X), cols=1);\n+R = matrix(0, 1, ncol(X));\n+\n+for (i in 2:ncol(X)) {\n+ Res1 = t(tmp1) %*% tmp1;\n+ tmp = cbind(tmp, X[,i]);\n+ while(FALSE) {};\n+ ones_n = matrix(1, rows=nrow(X), cols=1);\n+ tmp1 = cbind(tmp, ones_n);\n+ R[1,i] = sum(Res1);\n+ sum = sum + sum(Res1);\n+}\n+\n+write(R, $2, format=\"text\");\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/lineage/RewriteTest4.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1);\n+\n+tmp = X[,1];\n+for (i in 2:ncol(X)) {\n+ Res1 = t(tmp) %*% tmp;\n+ tmp = cbind(tmp, X[,i]);\n+ #while(FALSE) {}\n+ #print(sum(Res1));\n+ print(toString(Res1));\n+}\n+\n+tmp = X[,1];\n+for (i in 2:ncol(X)) {\n+ Res1 = t(tmp) %*% tmp;\n+\n+ if (i > 2) {\n+ while (FALSE) {}\n+ #tmp = tmp[,1:(i-2)];\n+ r1 = cbind((t(tmp[,1:(i-2)]) %*% tmp[,1:(i-2)]), (t(tmp[,1:(i-2)]) %*% X[,i-1]));\n+ r2 = cbind((t(X[,i-1]) %*% tmp[,1:(i-2)]), (t(X[,i-1]) %*% X[,i-1]));\n+ Res1 = rbind(r1, r2);\n+ while(FALSE) {}\n+ }\n+\n+ tmp = cbind(tmp, X[,i]);\n+ while(FALSE) {}\n+ #print(sum(Res1));\n+ print(toString(Res1));\n+}\n+\n+write(Res1, $2, format=\"text\");\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/lineage/RewriteTest6.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1);\n+\n+# calculate and same the trans-mult of X\n+Res1 = t(X) %*% X;\n+while(FALSE) {}\n+\n+# find vector V as the last column and matrix X1 as the rest\n+nc = ncol(X);\n+V = X[,nc];\n+X1 = X[, 1:nc-1];\n+while(FALSE) {}\n+\n+# use X1 and V to derive trans-mult of X\n+r1c1 = t(X1) %*% X1;\n+while(FALSE) {}\n+r1c2 = t(X1) %*% V;\n+while(FALSE) {}\n+r2c1 = t(V) %*% X1;\n+while(FALSE) {}\n+r2c2 = t(V) %*% V;\n+while(FALSE) {}\n+r1 = cbind(r1c1, r1c2);\n+while(FALSE) {}\n+r2 = cbind(r2c1, r2c2);\n+while(FALSE) {}\n+#r1 = cbind((t(X1) %*% X1), (t(X1) %*% V));\n+#r2 = cbind((t(V) %*% X1), (t(V) %*% V));\n+Res2 = rbind(r1, r2);\n+\n+# verify if the sums are same\n+while(FALSE) {}\n+sum1 = sum(Res1);\n+sum2 = sum(Res2);\n+#print(\"original matrix sum = \" + sum1);\n+#print(\"rewritten matrix sum = \" + sum2);\n+\n+write(Res1, $2, format=\"text\");\n+write(Res2, $3, format=\"text\");\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -61,7 +61,7 @@ SYSTEMDS-70 Lineage Tracing and Reuse OK\n* 76 Generate runtime plan from lineage trace OK\n* 77 New builtin function for obtaining lineage OK\n* 78 Extended lineage tracing (parfor, funs, codegen) OK\n- * 79 Reuse cache configuration options\n+ * 79 Reuse cache configuration options (cmd line options)\nSYSTEMDS-80 Improved distributed operations\n* 81 Avoid unnecessary replication on rmm\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/api/DMLScript.java", "new_path": "src/main/java/org/tugraz/sysds/api/DMLScript.java", "diff": "@@ -70,6 +70,8 @@ import org.tugraz.sysds.runtime.controlprogram.parfor.stat.InfrastructureAnalyze\nimport org.tugraz.sysds.runtime.controlprogram.parfor.util.IDHandler;\nimport org.tugraz.sysds.runtime.instructions.gpu.context.GPUContextPool;\nimport org.tugraz.sysds.runtime.io.IOUtilFunctions;\n+import org.tugraz.sysds.runtime.lineage.LineageCacheConfig;\n+import org.tugraz.sysds.runtime.lineage.LineageCacheConfig.CacheType;\nimport org.tugraz.sysds.runtime.matrix.mapred.MRConfigurationNames;\nimport org.tugraz.sysds.runtime.matrix.mapred.MRJobConfiguration;\nimport org.tugraz.sysds.runtime.util.LocalFileUtils;\n@@ -219,6 +221,11 @@ public class DMLScript\nreturn true;\n}\n+ if( LINEAGE_REUSE ) {\n+ //TODO proper cmd line configuration (SYSTEMDS-79)\n+ LineageCacheConfig.setConfig(CacheType.FULL);\n+ }\n+\nString dmlScriptStr = readDMLScript(isFile, fileOrScript);\nMap<String, String> argVals = dmlOptions.argVals;\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "diff": "@@ -145,8 +145,9 @@ public class LineageCache {\npublic static boolean isReusable (Instruction inst) {\n// TODO: Move this to the new class LineageCacheConfig and extend\n- return (inst.getOpcode().equalsIgnoreCase(\"tsmm\"));\n- //|| inst.getOpcode().equalsIgnoreCase(\"ba+*\"));\n+ return inst.getOpcode().equalsIgnoreCase(\"tsmm\")\n+ || (LineageCacheConfig.getCacheType().isFullReuse()\n+ && inst.getOpcode().equalsIgnoreCase(\"ba+*\"));\n// TODO: Fix getRecomputeEstimate to support ba+* before enabling above code.\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCacheConfig.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCacheConfig.java", "diff": "@@ -33,7 +33,7 @@ public class LineageCacheConfig {\n}\n}\n- public ArrayList<String> _MMult = new ArrayList<String>();\n+ public ArrayList<String> _MMult = new ArrayList<>();\npublic enum CachedItemHead {\nTSMM,\n@@ -57,6 +57,10 @@ public class LineageCacheConfig {\n_itemT = CachedItemTail.CBIND;\n}\n+ public static void setConfig(CacheType ct) {\n+ _cacheType = ct;\n+ }\n+\npublic static void setConfig(CacheType ct, CachedItemHead ith, CachedItemTail itt) {\n_cacheType = ct;\n_itemH = ith;\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/DataOp.java", "new_path": "src/main/java/org/tugraz/sysds/hops/DataOp.java", "diff": "@@ -404,15 +404,15 @@ public class DataOp extends Hop\nprotected DataCharacteristics inferOutputCharacteristics( MemoTable memo ) {\nDataCharacteristics ret = null;\nif( _dataop == DataOpTypes.PERSISTENTWRITE || _dataop == DataOpTypes.TRANSIENTWRITE ) {\n- DataCharacteristics dc = memo.getAllInputStats(getInput().get(0));\n- if( dc.dimsKnown() )\n- ret = _dc;\n+ DataCharacteristics tmp = memo.getAllInputStats(getInput().get(0));\n+ if( tmp.dimsKnown() )\n+ ret = tmp;\n}\nelse if( _dataop == DataOpTypes.TRANSIENTREAD ) {\n//prepare statistics, passed from cross-dag transient writes\n- DataCharacteristics dc = memo.getAllInputStats(this);\n- if( dc.dimsKnown() )\n- ret = _dc;\n+ DataCharacteristics tmp = memo.getAllInputStats(this);\n+ if( tmp.dimsKnown() )\n+ ret = tmp;\n}\nreturn ret;\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/matrix/data/MatrixBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/matrix/data/MatrixBlock.java", "diff": "@@ -2272,9 +2272,8 @@ public class MatrixBlock extends MatrixValue implements CacheBlock, Externalizab\n{\n//fast deserialize of dense/sparse blocks\nObjectInputStream ois = (ObjectInputStream)is;\n- try(FastBufferedDataInputStream fis = new FastBufferedDataInputStream(ois)) {\n- readFields(fis);\n- }\n+ FastBufferedDataInputStream fis = new FastBufferedDataInputStream(ois);\n+ readFields(fis); //note: cannot close fos as this would close oos\n}\nelse {\n//default deserialize (general case)\n@@ -2301,11 +2300,10 @@ public class MatrixBlock extends MatrixValue implements CacheBlock, Externalizab\n&& !(os instanceof MatrixBlockDataOutput) ) {\n//fast serialize of dense/sparse blocks\nObjectOutputStream oos = (ObjectOutputStream)os;\n- try(FastBufferedDataOutputStream fos = new FastBufferedDataOutputStream(oos)) {\n- write(fos);\n+ FastBufferedDataOutputStream fos = new FastBufferedDataOutputStream(oos);\n+ write(fos); //note: cannot close fos as this would close oos\nfos.flush();\n}\n- }\nelse {\n//default serialize (general case)\nwrite(os);\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "diff": "@@ -38,7 +38,7 @@ import java.util.Map;\npublic class LineageCache {\nprivate static final Map<LineageItem, Entry> _cache = new HashMap<>();\n- private static final Map<LineageItem, String> _spillList = new HashMap<>();\n+ private static final Map<LineageItem, SpilledItem> _spillList = new HashMap<>();\nprivate static final HashSet<LineageItem> _removelist = new HashSet<>();\nprivate static final long CACHELIMIT= (long)512*1024*1024; // 500MB\nprivate static String outdir = null;\n@@ -66,7 +66,7 @@ public class LineageCache {\n//create a placeholder if no reuse to avoid redundancy\n//(e.g., concurrent threads that try to start the computation)\nif( ! reuse )\n- putIntern(inst, item, null);\n+ putIntern(inst, item, null, 0);\n}\n}\n@@ -80,7 +80,7 @@ public class LineageCache {\nLineageItem item = ((LineageTraceable) inst).getLineageItems(ec)[0];\nMatrixObject mo = ec.getMatrixObject(((ComputationCPInstruction) inst).output);\nsynchronized( _cache ) {\n- putIntern(inst, item, mo.acquireReadAndRelease());\n+ putIntern(inst, item, mo.acquireReadAndRelease(), getRecomputeEstimate(inst, ec));\n}\n}\n}\n@@ -92,7 +92,7 @@ public class LineageCache {\nLineageItem item = ((LineageTraceable) inst).getLineageItems(ec)[0];\nMatrixObject mo = ec.getMatrixObject(((ComputationCPInstruction) inst).output);\nMatrixBlock value = mo.acquireReadAndRelease();\n- _cache.get(item).setValue(value); //outside sync to prevent deadlocks\n+ _cache.get(item).setValue(value, getRecomputeEstimate(inst, ec)); //outside sync to prevent deadlocks\nsynchronized( _cache ) {\nif( !isBelowThreshold(value) )\n@@ -102,12 +102,12 @@ public class LineageCache {\n}\n}\n- private static void putIntern(Instruction inst, LineageItem key, MatrixBlock value) {\n+ private static void putIntern(Instruction inst, LineageItem key, MatrixBlock value, double compcost) {\nif (_cache.containsKey(key))\nthrow new DMLRuntimeException(\"Redundant lineage caching detected: \"+inst);\n// Create a new entry.\n- Entry newItem = new Entry(key, value);\n+ Entry newItem = new Entry(key, value, compcost);\n// Make space by removing or spilling LRU entries.\nif( value != null ) {\n@@ -164,12 +164,11 @@ public class LineageCache {\nreturn readFromLocalFS(inst, key);\n}\n- private static boolean isReusable (Instruction inst) {\n+ public static boolean isReusable (Instruction inst) {\n// TODO: Move this to the new class LineageCacheConfig and extend\nreturn inst.getOpcode().equalsIgnoreCase(\"tsmm\")\n|| (LineageCacheConfig.getCacheType().isFullReuse()\n&& inst.getOpcode().equalsIgnoreCase(\"ba+*\"));\n- // TODO: Fix getRecomputeEstimate to support ba+* before enabling above code.\n}\n//---------------- CACHE SPACE MANAGEMENT METHODS -----------------\n@@ -187,14 +186,7 @@ public class LineageCache {\nwhile ((valSize+_cachesize) > CACHELIMIT)\n{\ndouble reduction = _cache.get(_end._key).getValue().getInMemorySize();\n- long t0 = DMLScript.STATISTICS ? System.nanoTime() : 0;\n- double spill = getDiskSpillEstimate();\n- double comp = getRecomputeEstimate(inst);\n- if (DMLScript.STATISTICS) {\n- long t1 = System.nanoTime();\n- LineageCacheStatistics.incrementCostingTime(t1-t0);\n- }\n- if (comp > spill)\n+ if (_cache.get(_end._key)._compEst > getDiskSpillEstimate())\nspillToLocalFS(); // If re-computation is more expensive, spill data to disk.\nremoveEntry(reduction);\n@@ -212,6 +204,7 @@ public class LineageCache {\nprivate static double getDiskSpillEstimate() {\n// This includes sum of writing to and reading from disk\n+ long t0 = DMLScript.STATISTICS ? System.nanoTime() : 0;\nMatrixBlock mb = _cache.get(_end._key).getValue();\nlong r = mb.getNumRows();\nlong c = mb.getNumColumns();\n@@ -219,24 +212,25 @@ public class LineageCache {\ndouble s = OptimizerUtils.getSparsity(r, c, nnz);\ndouble loadtime = CostEstimatorStaticRuntime.getFSReadTime(r, c, s);\ndouble writetime = CostEstimatorStaticRuntime.getFSWriteTime(r, c, s);\n+ if (DMLScript.STATISTICS)\n+ LineageCacheStatistics.incrementCostingTime(System.nanoTime() - t0);\nreturn loadtime+writetime;\n}\n- private static double getRecomputeEstimate(Instruction inst) {\n- MatrixBlock mb = _cache.get(_end._key).getValue();\n- long r = mb.getNumRows();\n- long c = mb.getNumColumns();\n- long nnz = mb.getNonZeros();\n- double s = OptimizerUtils.getSparsity(r, c, nnz);\n- boolean sparse = MatrixBlock.evalSparseFormatInMemory(r, c, nnz);\n-\n+ private static double getRecomputeEstimate(Instruction inst, ExecutionContext ec) {\n+ long t0 = DMLScript.STATISTICS ? System.nanoTime() : 0;\ndouble nflops = 0;\nCPType cptype = CPInstructionParser.String2CPInstructionType.get(inst.getOpcode());\n//TODO: All other relevant instruction types.\nswitch (cptype)\n{\n- case MMTSJ:\n- //case AggregateBinary:\n+ case MMTSJ: //tsmm\n+ MatrixObject mo = ec.getMatrixObject(((ComputationCPInstruction)inst).input1);\n+ long r = mo.getNumRows();\n+ long c = mo.getNumColumns();\n+ long nnz = mo.getNnz();\n+ double s = OptimizerUtils.getSparsity(r, c, nnz);\n+ boolean sparse = MatrixBlock.evalSparseFormatInMemory(r, c, nnz);\nMMTSJType type = ((MMTSJCPInstruction)inst).getMMTSJType();\nif (type.isLeft())\nnflops = !sparse ? (r * c * s * c /2):(r * c * s * c * s /2);\n@@ -244,9 +238,37 @@ public class LineageCache {\nnflops = !sparse ? ((double)r * c * r/2):(r*c*s + r*c*s*c*s /2);\nbreak;\n+ case AggregateBinary: //ba+*\n+ MatrixObject mo1 = ec.getMatrixObject(((ComputationCPInstruction)inst).input1);\n+ MatrixObject mo2 = ec.getMatrixObject(((ComputationCPInstruction)inst).input2);\n+ long r1 = mo1.getNumRows();\n+ long c1 = mo1.getNumColumns();\n+ long nnz1 = mo1.getNnz();\n+ double s1 = OptimizerUtils.getSparsity(r1, c1, nnz1);\n+ boolean lsparse = MatrixBlock.evalSparseFormatInMemory(r1, c1, nnz1);\n+ long r2 = mo2.getNumRows();\n+ long c2 = mo2.getNumColumns();\n+ long nnz2 = mo2.getNnz();\n+ double s2 = OptimizerUtils.getSparsity(r2, c2, nnz2);\n+ boolean rsparse = MatrixBlock.evalSparseFormatInMemory(r2, c2, nnz2);\n+ if( !lsparse && !rsparse )\n+ nflops = 2 * (r1 * c1 * ((c2>1)?s1:1.0) * c2) /2;\n+ else if( !lsparse && rsparse )\n+ nflops = 2 * (r1 * c1 * s1 * c2 * s2) /2;\n+ else if( lsparse && !rsparse )\n+ nflops = 2 * (r1 * c1 * s1 * c2) /2;\n+ else //lsparse && rsparse\n+ nflops = 2 * (r1 * c1 * s1 * c2 * s2) /2;\n+ break;\n+\ndefault:\nthrow new DMLRuntimeException(\"Lineage Cache: unsupported instruction: \"+inst.getOpcode());\n}\n+\n+ if (DMLScript.STATISTICS) {\n+ long t1 = System.nanoTime();\n+ LineageCacheStatistics.incrementCostingTime(t1-t0);\n+ }\nreturn nflops / (2L * 1024 * 1024 * 1024);\n}\n@@ -270,7 +292,7 @@ public class LineageCache {\nLineageCacheStatistics.incrementFSWrites();\n}\n- _spillList.put(_end._key, outfile);\n+ _spillList.put(_end._key, new SpilledItem(outfile, _end._compEst));\n}\nprivate static MatrixBlock readFromLocalFS(Instruction inst, LineageItem key) {\n@@ -278,14 +300,14 @@ public class LineageCache {\nMatrixBlock mb = null;\n// Read from local FS\ntry {\n- mb = LocalFileUtils.readMatrixBlockFromLocal(_spillList.get(key));\n+ mb = LocalFileUtils.readMatrixBlockFromLocal(_spillList.get(key)._outfile);\n} catch (IOException e) {\n- throw new DMLRuntimeException (\"Read from \" + _spillList.get(key) + \" failed.\", e);\n+ throw new DMLRuntimeException (\"Read from \" + _spillList.get(key)._outfile + \" failed.\", e);\n}\n// Restore to cache\n- LocalFileUtils.deleteFileIfExists(_spillList.get(key), true);\n+ LocalFileUtils.deleteFileIfExists(_spillList.get(key)._outfile, true);\n+ putIntern(inst, key, mb, _spillList.get(key)._compEst);\n_spillList.remove(key);\n- putIntern(inst, key, mb);\nif (DMLScript.STATISTICS) {\nlong t1 = System.nanoTime();\nLineageCacheStatistics.incrementFSReadTime(t1-t0);\n@@ -328,12 +350,14 @@ public class LineageCache {\nprivate static class Entry {\nprivate final LineageItem _key;\nprivate MatrixBlock _val;\n+ double _compEst;\nprivate Entry _prev;\nprivate Entry _next;\n- public Entry(LineageItem key, MatrixBlock value) {\n+ public Entry(LineageItem key, MatrixBlock value, double computecost) {\n_key = key;\n_val = value;\n+ _compEst = computecost;\n}\npublic synchronized MatrixBlock getValue() {\n@@ -350,9 +374,20 @@ public class LineageCache {\n}\n}\n- public synchronized void setValue(MatrixBlock val) {\n+ public synchronized void setValue(MatrixBlock val, double compEst) {\n_val = val;\n+ _compEst = compEst;\nnotifyAll();\n}\n}\n+\n+ private static class SpilledItem {\n+ String _outfile;\n+ double _compEst;\n+\n+ public SpilledItem(String outfile, double computecost) {\n+ this._outfile = outfile;\n+ this._compEst = computecost;\n+ }\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCacheStatistics.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCacheStatistics.java", "diff": "@@ -28,16 +28,23 @@ public class LineageCacheStatistics {\nprivate static final LongAdder _numHitsDel = new LongAdder();\nprivate static final LongAdder _numWritesMem = new LongAdder();\nprivate static final LongAdder _numWritesFS = new LongAdder();\n+ private static final LongAdder _numRewrites = new LongAdder();\nprivate static final LongAdder _ctimeFSRead = new LongAdder(); //in nano sec\nprivate static final LongAdder _ctimeFSWrite = new LongAdder(); //in nano sec\nprivate static final LongAdder _ctimeCosting = new LongAdder(); //in nano sec\n+ private static final LongAdder _ctimeRewrite = new LongAdder(); //in nano sec\npublic static void reset() {\n_numHitsMem.reset();\n_numHitsFS.reset();\n+ _numHitsDel.reset();\n+ _numWritesMem.reset();\n_numWritesFS.reset();\n+ _numRewrites.reset();\n_ctimeFSRead.reset();\n_ctimeFSWrite.reset();\n+ _ctimeCosting.reset();\n+ _ctimeRewrite.reset();\n}\npublic static void incrementMemHits() {\n@@ -60,6 +67,11 @@ public class LineageCacheStatistics {\n_numWritesMem.increment();\n}\n+ public static void incrementPRewrites() {\n+ // Number of times written in local FS.\n+ _numRewrites.increment();\n+ }\n+\npublic static void incrementFSWrites() {\n// Number of times written in local FS.\n_numWritesFS.increment();\n@@ -80,6 +92,11 @@ public class LineageCacheStatistics {\n_ctimeCosting.add(delta);\n}\n+ public static void incrementPRewriteTime(long delta) {\n+ // Total time spent estimating computation and disk spill costs.\n+ _ctimeRewrite.add(delta);\n+ }\n+\npublic static String displayHits() {\nStringBuilder sb = new StringBuilder();\nsb.append(_numHitsMem.longValue());\n@@ -98,6 +115,12 @@ public class LineageCacheStatistics {\nreturn sb.toString();\n}\n+ public static String displayRewrites() {\n+ StringBuilder sb = new StringBuilder();\n+ sb.append(_numRewrites.longValue());\n+ return sb.toString();\n+ }\n+\npublic static String displayTime() {\nStringBuilder sb = new StringBuilder();\nsb.append(String.format(\"%.3f\", ((double)_ctimeFSRead.longValue())/1000000000)); //in sec\n@@ -111,4 +134,10 @@ public class LineageCacheStatistics {\nsb.append(String.format(\"%.3f\", ((double)_ctimeCosting.longValue())/1000000000)); //in sec\nreturn sb.toString();\n}\n+\n+ public static String displayRewriteTime() {\n+ StringBuilder sb = new StringBuilder();\n+ sb.append(String.format(\"%.3f\", ((double)_ctimeRewrite.longValue())/1000000000)); //in sec\n+ return sb.toString();\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/RewriteCPlans.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/RewriteCPlans.java", "diff": "@@ -59,7 +59,7 @@ public class RewriteCPlans\nboolean oneappend = false;\nboolean twoappend = false;\nMatrixBlock lastResult = null;\n- if (curr.getOpcode().equalsIgnoreCase(\"tsmm\"))\n+ if (LineageCache.isReusable(curr))\n{\n// If the input to tsmm came from cbind, look for both the inputs in cache.\nLineageItem[] items = ((ComputationCPInstruction) curr).getLineageItems(ec);\n@@ -109,8 +109,13 @@ public class RewriteCPlans\nDMLScript.EXPLAIN = ExplainType.NONE;\ntry {\n+ long t0 = DMLScript.STATISTICS ? System.nanoTime() : 0;\nArrayList<Instruction> newInst = oneappend ? rewriteCbindTsmm(curr, ec, lrwec, lastResult) :\ntwoappend ? rewrite2CbindTsmm(curr, ec, lrwec, lastResult) : null;\n+ if (DMLScript.STATISTICS) {\n+ LineageCacheStatistics.incrementPRewriteTime(System.nanoTime() - t0);\n+ LineageCacheStatistics.incrementPRewrites();\n+ }\n//execute instructions\nBasicProgramBlock pb = getProgramBlock();\npb.setInstructions(newInst);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/utils/Statistics.java", "new_path": "src/main/java/org/tugraz/sysds/utils/Statistics.java", "diff": "@@ -941,8 +941,10 @@ public class Statistics\nif (DMLScript.LINEAGE && DMLScript.LINEAGE_REUSE) {\nsb.append(\"LineageCache hits (Mem/FS/Del): \" + LineageCacheStatistics.displayHits() + \".\\n\");\nsb.append(\"LineageCache writes (Mem/FS): \\t\" + LineageCacheStatistics.displayWtrites() + \".\\n\");\n+ sb.append(\"LineageCache Rewrites: \\t\" + LineageCacheStatistics.displayRewrites() + \".\\n\");\nsb.append(\"LineageCache FStimes (Rd/Wr): \\t\" + LineageCacheStatistics.displayTime() + \" sec.\\n\");\nsb.append(\"LineageCache costing time: \\t\" + LineageCacheStatistics.displayCostingTime() + \" sec.\\n\");\n+ sb.append(\"LineageCache rewrite time: \\t\" + LineageCacheStatistics.displayRewriteTime() + \" sec.\\n\");\n}\nif( ConfigurationManager.isCodegenEnabled() ) {\nsb.append(\"Codegen compile (DAG,CP,JC):\\t\" + getCodegenDAGCompile() + \"/\"\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -13,6 +13,7 @@ SYSTEMDS-10 Compiler Rework / Misc\n* 16 Remove instruction patching\n* 17 Refactoring of program block hierarchy OK\n* 18 Improve API for new dml-bodied builtin functions OK\n+ * 19 Break append instruction to cbind and rbind\nSYSTEMDS-20 New Data Model\n* 21 Finalize dense tensor blocks OK\n@@ -61,7 +62,7 @@ SYSTEMDS-70 Lineage Tracing and Reuse OK\n* 76 Generate runtime plan from lineage trace OK\n* 77 New builtin function for obtaining lineage OK\n* 78 Extended lineage tracing (parfor, funs, codegen) OK\n- * 79 Reuse cache configuration options (cmd line options)\n+ * 79 Reuse cache configuration options (cmd line options) OK\nSYSTEMDS-80 Improved distributed operations\n* 81 Avoid unnecessary replication on rmm\n@@ -84,6 +85,7 @@ SYSTEMDS-110 New Builtin Functions\n* 115 Builtin function for model debugging (slice finder) OK\n* 116 Builtin function for kmeans OK\n* 117 Builtin function for lm cross validation OK\n+ * 118 Builtin function for hyperparameter grid search with CVlm\nSYSTEMDS-120 Performance Features\n* 121 Avoid spark context creation on parfor result merge OK\n@@ -113,7 +115,9 @@ SYSTEMDS-160 Tensor Compiler/Runtime\nSYSTEMDS-170 Lineage full and partial reuse\n* 171 Initial version of partial rewrites OK\n* 172 Parfor integration (blocked waiting for results) OK\n- * 173 Improved cost estimates and ba+* support\n- * 174 Reuse rewrite for rbind-tsmm\n+ * 173 Improved cost estimates OK\n+ * 174 Reuse rewrite for rbind-tsmm OK\n+ * 175 Refactoring of lineage rewrite code\n* 175 Reuse rewrite for cbind/rbind-elementwise */+\n* 176 Reuse rewrite for aggregate\n+ * 177 Compiler assisted reuse (eg. CV, lmCG)\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/api/DMLOptions.java", "new_path": "src/main/java/org/tugraz/sysds/api/DMLOptions.java", "diff": "@@ -33,6 +33,7 @@ import org.apache.commons.cli.Options;\nimport org.apache.commons.cli.PosixParser;\nimport org.tugraz.sysds.common.Types.ExecMode;\nimport org.tugraz.sysds.hops.OptimizerUtils;\n+import org.tugraz.sysds.runtime.lineage.LineageCacheConfig.ReuseCacheType;\nimport org.tugraz.sysds.utils.Explain;\nimport org.tugraz.sysds.utils.Explain.ExplainType;\n@@ -61,7 +62,7 @@ public class DMLOptions {\npublic boolean help = false; // whether to print the usage option\npublic boolean lineage = false; // whether compute lineage trace\npublic boolean lineage_dedup = false; // whether deduplicate lineage items\n- public boolean lineage_reuse = false; // whether lineage-based reuse of intermediates\n+ public ReuseCacheType linReuseType = ReuseCacheType.NONE;\npublic final static DMLOptions defaultOptions = new DMLOptions(null);\n@@ -113,8 +114,12 @@ public class DMLOptions {\nif (lineageType != null){\nif (lineageType.equalsIgnoreCase(\"dedup\"))\ndmlOptions.lineage_dedup = lineageType.equalsIgnoreCase(\"dedup\");\n- else if (lineageType.equalsIgnoreCase(\"reuse\"))\n- dmlOptions.lineage_reuse = lineageType.equalsIgnoreCase(\"reuse\");\n+ else if (lineageType.equalsIgnoreCase(\"reuse_full\"))\n+ dmlOptions.linReuseType = ReuseCacheType.REUSE_FULL;\n+ else if (lineageType.equalsIgnoreCase(\"reuse_partial\"))\n+ dmlOptions.linReuseType = ReuseCacheType.REUSE_PARTIAL;\n+ else if (lineageType.equalsIgnoreCase(\"reuse_hybrid\"))\n+ dmlOptions.linReuseType = ReuseCacheType.REUSE_HYBRID;\nelse\nthrow new org.apache.commons.cli.ParseException(\"Invalid argument specified for -lineage option\");\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/api/DMLScript.java", "new_path": "src/main/java/org/tugraz/sysds/api/DMLScript.java", "diff": "@@ -65,7 +65,7 @@ import org.tugraz.sysds.runtime.controlprogram.parfor.util.IDHandler;\nimport org.tugraz.sysds.runtime.instructions.gpu.context.GPUContextPool;\nimport org.tugraz.sysds.runtime.io.IOUtilFunctions;\nimport org.tugraz.sysds.runtime.lineage.LineageCacheConfig;\n-import org.tugraz.sysds.runtime.lineage.LineageCacheConfig.CacheType;\n+import org.tugraz.sysds.runtime.lineage.LineageCacheConfig.ReuseCacheType;\nimport org.tugraz.sysds.runtime.util.LocalFileUtils;\nimport org.tugraz.sysds.runtime.util.HDFSTool;\nimport org.tugraz.sysds.utils.Explain;\n@@ -92,7 +92,7 @@ public class DMLScript\npublic static String GPU_MEMORY_ALLOCATOR = \"cuda\"; // GPU memory allocator to use\npublic static boolean LINEAGE = DMLOptions.defaultOptions.lineage; // whether compute lineage trace\npublic static boolean LINEAGE_DEDUP = DMLOptions.defaultOptions.lineage_dedup; // whether deduplicate lineage items\n- public static boolean LINEAGE_REUSE = DMLOptions.defaultOptions.lineage_reuse; // whether lineage-based reuse\n+ public static ReuseCacheType LINEAGE_REUSE = DMLOptions.defaultOptions.linReuseType; // whether lineage-based reuse\npublic static boolean USE_ACCELERATOR = DMLOptions.defaultOptions.gpu;\npublic static boolean FORCE_ACCELERATOR = DMLOptions.defaultOptions.forceGPU;\n@@ -194,7 +194,7 @@ public class DMLScript\nEXEC_MODE = dmlOptions.execMode;\nLINEAGE = dmlOptions.lineage;\nLINEAGE_DEDUP = dmlOptions.lineage_dedup;\n- LINEAGE_REUSE = dmlOptions.lineage_reuse;\n+ LINEAGE_REUSE = dmlOptions.linReuseType;\nString fnameOptConfig = dmlOptions.configFile;\nboolean isFile = dmlOptions.filePath != null;\n@@ -213,10 +213,7 @@ public class DMLScript\nreturn true;\n}\n- if( LINEAGE_REUSE ) {\n- //TODO proper cmd line configuration (SYSTEMDS-79)\n- LineageCacheConfig.setConfig(CacheType.FULL);\n- }\n+ LineageCacheConfig.setConfig(LINEAGE_REUSE);\nString dmlScriptStr = readDMLScript(isFile, fileOrScript);\nMap<String, String> argVals = dmlOptions.argVals;\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/MatrixAppendCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/MatrixAppendCPInstruction.java", "diff": "@@ -21,10 +21,13 @@ package org.tugraz.sysds.runtime.instructions.cp;\nimport org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\n+import org.tugraz.sysds.runtime.lineage.LineageItem;\n+import org.tugraz.sysds.runtime.lineage.LineageItemUtils;\n+import org.tugraz.sysds.runtime.lineage.LineageTraceable;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport org.tugraz.sysds.runtime.matrix.operators.Operator;\n-public final class MatrixAppendCPInstruction extends AppendCPInstruction {\n+public final class MatrixAppendCPInstruction extends AppendCPInstruction implements LineageTraceable {\nprotected MatrixAppendCPInstruction(Operator op, CPOperand in1, CPOperand in2, CPOperand in3, CPOperand out,\nAppendType type, String opcode, String istr) {\n@@ -51,4 +54,12 @@ public final class MatrixAppendCPInstruction extends AppendCPInstruction {\nec.setMatrixOutput(output.getName(), ret);\nec.releaseMatrixInput(input1.getName(), input2.getName());\n}\n+\n+ @Override\n+ public LineageItem[] getLineageItems(ExecutionContext ec) {\n+ //TODO: break append to cbind and rbind for full compilation chain\n+ String opcode = _type.toString().toLowerCase();\n+ return new LineageItem[]{new LineageItem(output.getName(),\n+ opcode, LineageItemUtils.getLineage(ec, input1, input2))};\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/Lineage.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/Lineage.java", "diff": "@@ -20,7 +20,7 @@ import org.tugraz.sysds.runtime.controlprogram.ForProgramBlock;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.instructions.Instruction;\nimport org.tugraz.sysds.runtime.instructions.cp.CPOperand;\n-import org.tugraz.sysds.runtime.lineage.LineageCacheConfig.CacheType;\n+import org.tugraz.sysds.runtime.lineage.LineageCacheConfig.ReuseCacheType;\nimport java.util.HashMap;\nimport java.util.Map;\n@@ -101,18 +101,18 @@ public class Lineage {\n}\npublic static void setLinReusePartial() {\n- LineageCacheConfig.setConfigTsmmCbind(CacheType.PARTIAL);\n+ LineageCacheConfig.setConfigTsmmCbind(ReuseCacheType.REUSE_PARTIAL);\n}\npublic static void setLinReuseFull() {\n- LineageCacheConfig.setConfigTsmmCbind(CacheType.FULL);\n+ LineageCacheConfig.setConfigTsmmCbind(ReuseCacheType.REUSE_FULL);\n}\npublic static void setLinReuseFullAndPartial() {\n- LineageCacheConfig.setConfigTsmmCbind(CacheType.HYBRID_FULL_PARTIAL);\n+ LineageCacheConfig.setConfigTsmmCbind(ReuseCacheType.REUSE_HYBRID);\n}\npublic static void setLinReuseNone() {\n- LineageCacheConfig.setConfigTsmmCbind(CacheType.NONE);\n+ LineageCacheConfig.setConfigTsmmCbind(ReuseCacheType.NONE);\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "diff": "@@ -28,6 +28,7 @@ import org.tugraz.sysds.runtime.instructions.Instruction;\nimport org.tugraz.sysds.runtime.instructions.cp.CPInstruction.CPType;\nimport org.tugraz.sysds.runtime.instructions.cp.ComputationCPInstruction;\nimport org.tugraz.sysds.runtime.instructions.cp.MMTSJCPInstruction;\n+import org.tugraz.sysds.runtime.lineage.LineageCacheConfig.ReuseCacheType;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport org.tugraz.sysds.runtime.util.LocalFileUtils;\n@@ -49,7 +50,7 @@ public class LineageCache {\n//--------------------- CACHE LOGIC METHODS ----------------------\npublic static boolean reuse(Instruction inst, ExecutionContext ec) {\n- if (!DMLScript.LINEAGE_REUSE)\n+ if (ReuseCacheType.isNone())\nreturn false;\nboolean reuse = false;\n@@ -86,7 +87,7 @@ public class LineageCache {\n}\npublic static void putValue(Instruction inst, ExecutionContext ec) {\n- if (!DMLScript.LINEAGE_REUSE)\n+ if (ReuseCacheType.isNone())\nreturn;\nif (inst instanceof ComputationCPInstruction && isReusable(inst) ) {\nLineageItem item = ((LineageTraceable) inst).getLineageItems(ec)[0];\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCacheConfig.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCacheConfig.java", "diff": "@@ -20,16 +20,21 @@ import org.tugraz.sysds.api.DMLScript;\nimport java.util.ArrayList;\npublic class LineageCacheConfig {\n- public enum CacheType {\n- FULL, // no rewrites\n- PARTIAL,\n- HYBRID_FULL_PARTIAL,\n+\n+ public enum ReuseCacheType {\n+ REUSE_FULL,\n+ REUSE_PARTIAL,\n+ REUSE_HYBRID,\nNONE;\npublic boolean isFullReuse() {\n- return this == FULL || this == HYBRID_FULL_PARTIAL;\n+ return this == REUSE_FULL || this == REUSE_HYBRID;\n}\npublic boolean isPartialReuse() {\n- return this == PARTIAL || this == HYBRID_FULL_PARTIAL;\n+ return this == REUSE_PARTIAL || this == REUSE_HYBRID;\n+ }\n+ public static boolean isNone() {\n+ return DMLScript.LINEAGE_REUSE == null\n+ || DMLScript.LINEAGE_REUSE == NONE;\n}\n}\n@@ -47,21 +52,21 @@ public class LineageCacheConfig {\nALL\n}\n- private static CacheType _cacheType = null;\n+ private static ReuseCacheType _cacheType = null;\nprivate static CachedItemHead _itemH = null;\nprivate static CachedItemTail _itemT = null;\n- public static void setConfigTsmmCbind(CacheType ct) {\n+ public static void setConfigTsmmCbind(ReuseCacheType ct) {\n_cacheType = ct;\n_itemH = CachedItemHead.TSMM;\n_itemT = CachedItemTail.CBIND;\n}\n- public static void setConfig(CacheType ct) {\n+ public static void setConfig(ReuseCacheType ct) {\n_cacheType = ct;\n}\n- public static void setConfig(CacheType ct, CachedItemHead ith, CachedItemTail itt) {\n+ public static void setConfig(ReuseCacheType ct, CachedItemHead ith, CachedItemTail itt) {\n_cacheType = ct;\n_itemH = ith;\n_itemT = itt;\n@@ -69,15 +74,15 @@ public class LineageCacheConfig {\npublic static void shutdownReuse() {\nDMLScript.LINEAGE = false;\n- DMLScript.LINEAGE_REUSE = false;\n+ DMLScript.LINEAGE_REUSE = ReuseCacheType.NONE;\n}\n- public static void restartReuse() {\n+ public static void restartReuse(ReuseCacheType rop) {\nDMLScript.LINEAGE = true;\n- DMLScript.LINEAGE_REUSE = true;\n+ DMLScript.LINEAGE_REUSE = rop;\n}\n- public static CacheType getCacheType() {\n+ public static ReuseCacheType getCacheType() {\nreturn _cacheType;\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageItem.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageItem.java", "diff": "@@ -151,7 +151,7 @@ public class LineageItem {\n} else\nret &= _data.equals(that._data);\n- if (_inputs != null)\n+ if (_inputs != null && ret && (_inputs.length == that._inputs.length))\nfor (int i = 0; i < _inputs.length; i++)\nret &= _inputs[i].equalsLI(that._inputs[i]);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/RewriteCPlans.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/RewriteCPlans.java", "diff": "@@ -42,6 +42,7 @@ import org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContextFactory;\nimport org.tugraz.sysds.runtime.instructions.Instruction;\nimport org.tugraz.sysds.runtime.instructions.cp.ComputationCPInstruction;\n+import org.tugraz.sysds.runtime.lineage.LineageCacheConfig.ReuseCacheType;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport org.tugraz.sysds.runtime.meta.MetaData;\nimport org.tugraz.sysds.utils.Explain;\n@@ -58,48 +59,17 @@ public class RewriteCPlans\n{\nboolean oneappend = false;\nboolean twoappend = false;\n+ boolean onerbind = false;\nMatrixBlock lastResult = null;\n- if (LineageCache.isReusable(curr))\n- {\n- // If the input to tsmm came from cbind, look for both the inputs in cache.\n- LineageItem[] items = ((ComputationCPInstruction) curr).getLineageItems(ec);\n- LineageItem item = items[0];\n- // TODO restructuring of rewrites to make them all\n- // independent of each other and this opening condition here\n- for (LineageItem source : item.getInputs())\n- if (source.getOpcode().equalsIgnoreCase(\"append\")) {\n- for (LineageItem input : source.getInputs()) {\n- // create tsmm lineage on top of the input of last append\n- LineageItem tmp = new LineageItem(\"toProbe\", curr.getOpcode(), new LineageItem[] {input});\n- if (LineageCache.probe(tmp)) {\n- oneappend = true; // at least one entry to reuse\n- if (lastResult == null)\n- lastResult = LineageCache.get(curr, tmp);\n- }\n- }\n- if (oneappend)\n- break; // no need to look for the next append\n+ MatrixBlock mmc = isTsmmCbind(curr, ec);\n+ if (mmc != null) {oneappend = true; lastResult = mmc;}\n+ MatrixBlock mm2c = isTsmm2Cbind(curr, ec);\n+ if (mm2c != null) {twoappend = true; lastResult = mm2c;}\n+ MatrixBlock mmr = isTsmmRbind(curr, ec);\n+ if (mmr != null) {onerbind = true; lastResult = mmr;}\n- // if not found in cache, look for two consecutive cbinds\n- LineageItem input = source.getInputs()[0];\n- if (input.getOpcode().equalsIgnoreCase(\"append\")) {\n- for (LineageItem L2appin : input.getInputs()) {\n- LineageItem tmp = new LineageItem(\"comb\", \"append\", new LineageItem[] {L2appin, source.getInputs()[1]});\n- LineageItem toProbe = new LineageItem(\"toProbe\", curr.getOpcode(), new LineageItem[] {tmp});\n- if (LineageCache.probe(toProbe)) {\n- twoappend = true;\n- if (lastResult == null)\n- lastResult = LineageCache.get(curr, toProbe);\n- }\n- }\n- }\n- }\n- }\n- else\n- return false;\n-\n- if (!oneappend && !twoappend)\n+ if (!oneappend && !twoappend && !onerbind)\nreturn false;\nExecutionContext lrwec = getExecutionContext();\n@@ -111,7 +81,8 @@ public class RewriteCPlans\ntry {\nlong t0 = DMLScript.STATISTICS ? System.nanoTime() : 0;\nArrayList<Instruction> newInst = oneappend ? rewriteCbindTsmm(curr, ec, lrwec, lastResult) :\n- twoappend ? rewrite2CbindTsmm(curr, ec, lrwec, lastResult) : null;\n+ twoappend ? rewrite2CbindTsmm(curr, ec, lrwec, lastResult) :\n+ onerbind ? rewriteRbindTsmm(curr, ec, lrwec, lastResult) : null;\nif (DMLScript.STATISTICS) {\nLineageCacheStatistics.incrementPRewriteTime(System.nanoTime() - t0);\nLineageCacheStatistics.incrementPRewrites();\n@@ -119,9 +90,10 @@ public class RewriteCPlans\n//execute instructions\nBasicProgramBlock pb = getProgramBlock();\npb.setInstructions(newInst);\n+ ReuseCacheType oldReuseOption = DMLScript.LINEAGE_REUSE;\nLineageCacheConfig.shutdownReuse();\npb.execute(lrwec);\n- LineageCacheConfig.restartReuse();\n+ LineageCacheConfig.restartReuse(oldReuseOption);\nec.setVariable(((ComputationCPInstruction)curr).output.getName(), lrwec.getVariable(LR_VAR));\n// add this to cache\nLineageCache.put(curr, ec);\n@@ -174,6 +146,33 @@ public class RewriteCPlans\nreturn genInst(lrwWrite, lrwec);\n}\n+ private static ArrayList<Instruction> rewriteRbindTsmm(Instruction curr, ExecutionContext ec, ExecutionContext lrwec, MatrixBlock lastResult)\n+ {\n+ // Create a transient read op over the last tsmm result\n+ MetaData md = new MetaData(lastResult.getDataCharacteristics());\n+ MatrixObject newmo = new MatrixObject(ValueType.FP64, \"lastResult\", md);\n+ newmo.acquireModify(lastResult);\n+ newmo.release();\n+ lrwec.setVariable(\"lastResult\", newmo);\n+ DataOp lastRes = HopRewriteUtils.createTransientRead(\"lastResult\", lastResult);\n+ // Create rightIndex op to find the last appended rows\n+ //TODO: support for block of rows\n+ MatrixObject mo = ec.getMatrixObject(((ComputationCPInstruction)curr).input1);\n+ lrwec.setVariable(\"oldMatrix\", mo);\n+ DataOp newMatrix = HopRewriteUtils.createTransientRead(\"oldMatrix\", mo);\n+ IndexingOp lastRow = HopRewriteUtils.createIndexingOp(newMatrix, new LiteralOp(mo.getNumRows()),\n+ new LiteralOp(mo.getNumRows()), new LiteralOp(1), new LiteralOp(mo.getNumColumns()));\n+ // tsmm(X + lastRow) = tsmm(X) + tsmm(lastRow)\n+ ReorgOp tlastRow = HopRewriteUtils.createTranspose(lastRow);\n+ AggBinaryOp tsmm_lr = HopRewriteUtils.createMatrixMultiply(tlastRow, lastRow);\n+ BinaryOp lrwHop = HopRewriteUtils.createBinary(lastRes, tsmm_lr, OpOp2.PLUS);\n+ DataOp lrwWrite = HopRewriteUtils.createTransientWrite(LR_VAR, lrwHop);\n+\n+ // generate runtime instructions\n+ LOG.debug(\"LINEAGE REWRITE rewriteRbindTsmm APPLIED\");\n+ return genInst(lrwWrite, lrwec);\n+ }\n+\nprivate static ArrayList<Instruction> rewrite2CbindTsmm(Instruction curr, ExecutionContext ec, ExecutionContext lrwec, MatrixBlock lastResult)\n{\n// Create a transient read op over the last tsmm result\n@@ -225,6 +224,85 @@ public class RewriteCPlans\nreturn newInst;\n}\n+ private static MatrixBlock isTsmmCbind(Instruction curr, ExecutionContext ec)\n+ {\n+ MatrixBlock lastResult = null;\n+ if (!LineageCache.isReusable(curr))\n+ return lastResult;\n+\n+ // If the input to tsmm came from cbind, look for both the inputs in cache.\n+ LineageItem[] items = ((ComputationCPInstruction) curr).getLineageItems(ec);\n+ LineageItem item = items[0];\n+\n+ // TODO restructuring of rewrites to make them all\n+ // independent of each other and this opening condition here\n+ for (LineageItem source : item.getInputs())\n+ if (source.getOpcode().equalsIgnoreCase(\"cbind\")) {\n+ for (LineageItem input : source.getInputs()) {\n+ // create tsmm lineage on top of the input of last append\n+ LineageItem tmp = new LineageItem(\"toProbe\", curr.getOpcode(), new LineageItem[] {input});\n+ if (LineageCache.probe(tmp)) {\n+ if (lastResult == null)\n+ lastResult = LineageCache.get(curr, tmp);\n+ }\n+ }\n+ }\n+ return lastResult;\n+ }\n+\n+ private static MatrixBlock isTsmmRbind(Instruction curr, ExecutionContext ec)\n+ {\n+ MatrixBlock lastResult = null;\n+ if (!LineageCache.isReusable(curr))\n+ return lastResult;\n+\n+ // If the input to tsmm came from rbind, look for both the inputs in cache.\n+ LineageItem[] items = ((ComputationCPInstruction) curr).getLineageItems(ec);\n+ LineageItem item = items[0];\n+\n+ // TODO restructuring of rewrites to make them all\n+ // independent of each other and this opening condition here\n+ for (LineageItem source : item.getInputs())\n+ if (source.getOpcode().equalsIgnoreCase(\"rbind\")) {\n+ for (LineageItem input : source.getInputs()) {\n+ // create tsmm lineage on top of the input of last append\n+ LineageItem tmp = new LineageItem(\"toProbe\", curr.getOpcode(), new LineageItem[] {input});\n+ if (LineageCache.probe(tmp)) {\n+ if (lastResult == null)\n+ lastResult = LineageCache.get(curr, tmp);\n+ }\n+ }\n+ }\n+ return lastResult;\n+ }\n+\n+ private static MatrixBlock isTsmm2Cbind (Instruction curr, ExecutionContext ec)\n+ {\n+ MatrixBlock lastResult = null;\n+ if (!LineageCache.isReusable(curr))\n+ return lastResult;\n+\n+ // If the input to tsmm came from cbind, look for both the inputs in cache.\n+ LineageItem[] items = ((ComputationCPInstruction) curr).getLineageItems(ec);\n+ LineageItem item = items[0];\n+ // look for two consecutive cbinds\n+ for (LineageItem source : item.getInputs())\n+ if (source.getOpcode().equalsIgnoreCase(\"cbind\")) {\n+ LineageItem input = source.getInputs()[0];\n+ if (input.getOpcode().equalsIgnoreCase(\"cbind\")) {\n+ for (LineageItem L2appin : input.getInputs()) {\n+ LineageItem tmp = new LineageItem(\"comb\", \"cbind\", new LineageItem[] {L2appin, source.getInputs()[1]});\n+ LineageItem toProbe = new LineageItem(\"toProbe\", curr.getOpcode(), new LineageItem[] {tmp});\n+ if (LineageCache.probe(toProbe)) {\n+ if (lastResult == null)\n+ lastResult = LineageCache.get(curr, toProbe);\n+ }\n+ }\n+ }\n+ }\n+ return lastResult;\n+ }\n+\nprivate static ExecutionContext getExecutionContext() {\nif( _lrEC == null )\n_lrEC = ExecutionContextFactory.createContext();\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/utils/Statistics.java", "new_path": "src/main/java/org/tugraz/sysds/utils/Statistics.java", "diff": "@@ -43,6 +43,7 @@ import org.tugraz.sysds.runtime.instructions.Instruction;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\nimport org.tugraz.sysds.runtime.instructions.cp.FunctionCallCPInstruction;\nimport org.tugraz.sysds.runtime.instructions.spark.SPInstruction;\n+import org.tugraz.sysds.runtime.lineage.LineageCacheConfig.ReuseCacheType;\nimport org.tugraz.sysds.runtime.lineage.LineageCacheStatistics;\nimport org.tugraz.sysds.runtime.matrix.data.LibMatrixDNN;\n@@ -938,7 +939,7 @@ public class Statistics\nsb.append(\"Functions recompiled:\\t\\t\" + getFunRecompiles() + \".\\n\");\nsb.append(\"Functions recompile time:\\t\" + String.format(\"%.3f\", ((double)getFunRecompileTime())/1000000000) + \" sec.\\n\");\n}\n- if (DMLScript.LINEAGE && DMLScript.LINEAGE_REUSE) {\n+ if (DMLScript.LINEAGE && !ReuseCacheType.isNone()) {\nsb.append(\"LineageCache hits (Mem/FS/Del): \" + LineageCacheStatistics.displayHits() + \".\\n\");\nsb.append(\"LineageCache writes (Mem/FS): \\t\" + LineageCacheStatistics.displayWtrites() + \".\\n\");\nsb.append(\"LineageCache Rewrites: \\t\" + LineageCacheStatistics.displayRewrites() + \".\\n\");\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/component/misc/CLIOptionsParserTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/component/misc/CLIOptionsParserTest.java", "diff": "@@ -30,6 +30,7 @@ import org.junit.Assert;\nimport org.junit.Test;\nimport org.tugraz.sysds.api.DMLOptions;\nimport org.tugraz.sysds.common.Types.ExecMode;\n+import org.tugraz.sysds.runtime.lineage.LineageCacheConfig.ReuseCacheType;\nimport org.tugraz.sysds.utils.Explain;\n@@ -131,7 +132,7 @@ public class CLIOptionsParserTest {\nDMLOptions o = DMLOptions.parseCLArguments(args);\nAssert.assertEquals(true, o.lineage);\nAssert.assertEquals(false, o.lineage_dedup);\n- Assert.assertEquals(false, o.lineage_reuse);\n+ Assert.assertEquals(ReuseCacheType.NONE, o.linReuseType);\n}\n@Test\n@@ -141,16 +142,35 @@ public class CLIOptionsParserTest {\nDMLOptions o = DMLOptions.parseCLArguments(args);\nAssert.assertEquals(true, o.lineage);\nAssert.assertEquals(true, o.lineage_dedup);\n- Assert.assertEquals(false, o.lineage_reuse);\n+ Assert.assertEquals(ReuseCacheType.NONE, o.linReuseType);\n}\n@Test\n- public void testLineageReuse() throws Exception {\n+ public void testLineageReuseF() throws Exception {\nString cl = \"systemds -f test.dml -lineage reuse\";\nString[] args = cl.split(\" \");\nDMLOptions o = DMLOptions.parseCLArguments(args);\nAssert.assertEquals(true, o.lineage);\n- Assert.assertEquals(true, o.lineage_reuse);\n+ Assert.assertEquals(ReuseCacheType.REUSE_FULL, o.linReuseType);\n+ Assert.assertEquals(false, o.lineage_dedup);\n+ }\n+\n+ @Test\n+ public void testLineageReuseP() throws Exception {\n+ String cl = \"systemds -f test.dml -lineage reuse\";\n+ String[] args = cl.split(\" \");\n+ DMLOptions o = DMLOptions.parseCLArguments(args);\n+ Assert.assertEquals(true, o.lineage);\n+ Assert.assertEquals(ReuseCacheType.REUSE_PARTIAL, o.linReuseType);\n+ Assert.assertEquals(false, o.lineage_dedup);\n+ }\n+ @Test\n+ public void testLineageReuseH() throws Exception {\n+ String cl = \"systemds -f test.dml -lineage reuse\";\n+ String[] args = cl.split(\" \");\n+ DMLOptions o = DMLOptions.parseCLArguments(args);\n+ Assert.assertEquals(true, o.lineage);\n+ Assert.assertEquals(ReuseCacheType.REUSE_HYBRID, o.linReuseType);\nAssert.assertEquals(false, o.lineage_dedup);\n}\n@@ -161,7 +181,37 @@ public class CLIOptionsParserTest {\nDMLOptions o = DMLOptions.parseCLArguments(args);\nAssert.assertEquals(true, o.lineage);\nAssert.assertEquals(true, o.lineage_dedup);\n- Assert.assertEquals(true, o.lineage_reuse);\n+ Assert.assertEquals(ReuseCacheType.REUSE_FULL, o.linReuseType);\n+ }\n+\n+ @Test\n+ public void testLineageDedupAndReuseF() throws Exception {\n+ String cl = \"systemds -f test.dml -lineage dedup reuse\";\n+ String[] args = cl.split(\" \");\n+ DMLOptions o = DMLOptions.parseCLArguments(args);\n+ Assert.assertEquals(true, o.lineage);\n+ Assert.assertEquals(true, o.lineage_dedup);\n+ Assert.assertEquals(ReuseCacheType.REUSE_FULL, o.linReuseType);\n+ }\n+\n+ @Test\n+ public void testLineageDedupAndReuseP() throws Exception {\n+ String cl = \"systemds -f test.dml -lineage dedup reuse\";\n+ String[] args = cl.split(\" \");\n+ DMLOptions o = DMLOptions.parseCLArguments(args);\n+ Assert.assertEquals(true, o.lineage);\n+ Assert.assertEquals(true, o.lineage_dedup);\n+ Assert.assertEquals(ReuseCacheType.REUSE_PARTIAL, o.linReuseType);\n+ }\n+\n+ @Test\n+ public void testLineageDedupAndReusuH() throws Exception {\n+ String cl = \"systemds -f test.dml -lineage dedup reuse\";\n+ String[] args = cl.split(\" \");\n+ DMLOptions o = DMLOptions.parseCLArguments(args);\n+ Assert.assertEquals(true, o.lineage);\n+ Assert.assertEquals(true, o.lineage_dedup);\n+ Assert.assertEquals(ReuseCacheType.REUSE_HYBRID, o.linReuseType);\n}\n@Test(expected = ParseException.class)\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/FullReusePerfTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/FullReusePerfTest.java", "diff": "@@ -77,7 +77,7 @@ public class FullReusePerfTest extends AutomatedTestBase\nproArgs.add(\"-stats\");\nproArgs.add(\"-lineage\");\n- proArgs.add(\"reuse\");\n+ proArgs.add(\"reuse_full\");\nproArgs.add(\"-explain\");\nproArgs.add(\"-args\");\nproArgs.add(input(\"X\"));\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageRewriteTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageRewriteTest.java", "diff": "@@ -32,29 +32,36 @@ public class LineageRewriteTest extends AutomatedTestBase {\nprotected static final String TEST_DIR = \"functions/lineage/\";\nprotected static final String TEST_NAME1 = \"RewriteTest3\";\nprotected static final String TEST_NAME2 = \"RewriteTest2\";\n+ protected static final String TEST_NAME3 = \"RewriteTest7\";\nprotected String TEST_CLASS_DIR = TEST_DIR + LineageRewriteTest.class.getSimpleName() + \"/\";\n- protected static final int numRecords = 1000;\n- protected static final int numFeatures = 100;\n+ protected static final int numRecords = 100;\n+ protected static final int numFeatures = 30;\n@Override\npublic void setUp() {\nTestUtils.clearAssertionInformation();\naddTestConfiguration(TEST_NAME1, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME1));\naddTestConfiguration(TEST_NAME2, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME2));\n+ addTestConfiguration(TEST_NAME3, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME3));\n}\n@Test\n- public void testRewrite1() {\n+ public void testtsmm2cbind() {\ntestRewrite(TEST_NAME1);\n}\n@Test\n- public void testRewrite2() {\n+ public void testtsmmcbind() {\ntestRewrite(TEST_NAME2);\n}\n+ @Test\n+ public void testtsmmrbind() {\n+ testRewrite(TEST_NAME3);\n+ }\n+\nprivate void testRewrite(String testname) {\ntry {\ngetAndLoadTestConfiguration(testname);\n@@ -75,10 +82,10 @@ public class LineageRewriteTest extends AutomatedTestBase {\nproArgs.clear();\nproArgs.add(\"-explain\");\n- proArgs.add(\"recompile_hops\");\n+ //proArgs.add(\"recompile_runtime\");\nproArgs.add(\"-stats\");\nproArgs.add(\"-lineage\");\n- proArgs.add(\"reuse\");\n+ proArgs.add(\"reuse_partial\");\nproArgs.add(\"-args\");\nproArgs.add(input(\"X\"));\nproArgs.add(output(\"Res\"));\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/lineage/RewriteTest7.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1);\n+\n+sum = 0;\n+tmp = X[1,];\n+R = matrix(0, 1, nrow(X));\n+\n+for (i in 2:nrow(X)) {\n+ Res1 = t(tmp) %*% tmp;\n+ tmp = rbind(tmp, X[i,]);\n+ while(FALSE) {}\n+ R[1,i] = sum(Res1);\n+ sum = sum + sum(Res1);\n+}\n+\n+write(R, $2, format=\"text\");\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/MatrixBuiltinNaryCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/MatrixBuiltinNaryCPInstruction.java", "diff": "@@ -23,10 +23,13 @@ import java.util.List;\nimport org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\n+import org.tugraz.sysds.runtime.lineage.LineageItem;\n+import org.tugraz.sysds.runtime.lineage.LineageItemUtils;\n+import org.tugraz.sysds.runtime.lineage.LineageTraceable;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport org.tugraz.sysds.runtime.matrix.operators.Operator;\n-public class MatrixBuiltinNaryCPInstruction extends BuiltinNaryCPInstruction {\n+public class MatrixBuiltinNaryCPInstruction extends BuiltinNaryCPInstruction implements LineageTraceable {\nprotected MatrixBuiltinNaryCPInstruction(Operator op, String opcode, String istr, CPOperand output, CPOperand[] inputs) {\nsuper(op, opcode, istr, output, inputs);\n@@ -63,4 +66,10 @@ public class MatrixBuiltinNaryCPInstruction extends BuiltinNaryCPInstruction {\noutput.getValueType(), outBlock.quickGetValue(0, 0)));\n}\n}\n+\n+ @Override\n+ public LineageItem[] getLineageItems(ExecutionContext ec) {\n+ return new LineageItem[]{new LineageItem(output.getName(),\n+ getOpcode(), LineageItemUtils.getLineage(ec, inputs))};\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageItemUtils.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageItemUtils.java", "diff": "@@ -32,6 +32,7 @@ import org.tugraz.sysds.hops.Hop;\nimport org.tugraz.sysds.hops.LiteralOp;\nimport org.tugraz.sysds.hops.Hop.DataGenMethod;\nimport org.tugraz.sysds.hops.Hop.DataOpTypes;\n+import org.tugraz.sysds.hops.Hop.OpOpN;\nimport org.tugraz.sysds.hops.rewrite.HopRewriteUtils;\nimport org.tugraz.sysds.lops.Lop;\nimport org.tugraz.sysds.lops.compile.Dag;\n@@ -58,6 +59,7 @@ import java.io.IOException;\nimport java.util.ArrayList;\nimport java.util.Arrays;\nimport java.util.HashMap;\n+import java.util.Map;\nimport java.util.stream.Collectors;\npublic class LineageItemUtils {\n@@ -129,7 +131,7 @@ public class LineageItemUtils {\n//recursively construct hops\nroot.resetVisitStatus();\n- HashMap<Long, Hop> operands = new HashMap<>();\n+ Map<Long, Hop> operands = new HashMap<>();\nrConstructHops(root, operands);\nHop out = HopRewriteUtils.createTransientWrite(\nvarname, operands.get(rootId));\n@@ -155,7 +157,7 @@ public class LineageItemUtils {\n.map(c -> ec.getLineage().getOrCreate(c)).toArray(LineageItem[]::new);\n}\n- private static void rConstructHops(LineageItem item, HashMap<Long, Hop> operands) {\n+ private static void rConstructHops(LineageItem item, Map<Long, Hop> operands) {\nif (item.isVisited())\nreturn;\n@@ -239,6 +241,12 @@ public class LineageItemUtils {\noperands.get(item.getInputs()[2].getId()), item.getOpcode()));\nbreak;\n}\n+ case BuiltinNary: {\n+ operands.put(item.getId(), HopRewriteUtils.createNary(\n+ OpOpN.valueOf(item.getOpcode().toUpperCase()),\n+ createNaryInputs(item, operands)));\n+ break;\n+ }\ncase MatrixIndexing: {\nif( \"rightIndex\".equals(item.getOpcode()) )\noperands.put(item.getId(), HopRewriteUtils.createIndexingOp(\n@@ -366,4 +374,12 @@ public class LineageItemUtils {\n}\ncurrent.setVisited();\n}\n+\n+ private static Hop[] createNaryInputs(LineageItem item, Map<Long, Hop> operands) {\n+ int len = item.getInputs().length;\n+ Hop[] ret = new Hop[len];\n+ for( int i=0; i<len; i++ )\n+ ret[i] = operands.get(item.getInputs()[i].getId());\n+ return ret;\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageTraceExecTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageTraceExecTest.java", "diff": "@@ -42,6 +42,7 @@ public class LineageTraceExecTest extends AutomatedTestBase {\nprotected static final String TEST_NAME3 = \"LineageTraceExec3\"; //read - matrix result\nprotected static final String TEST_NAME4 = \"LineageTraceExec4\"; //rand - matrix result - unspecified seed\nprotected static final String TEST_NAME5 = \"LineageTraceExec5\"; //rand - scalar result - unspecified seed\n+ protected static final String TEST_NAME6 = \"LineageTraceExec6\"; //nary rbind\nprotected String TEST_CLASS_DIR = TEST_DIR + LineageTraceExecTest.class.getSimpleName() + \"/\";\n@@ -60,6 +61,7 @@ public class LineageTraceExecTest extends AutomatedTestBase {\naddTestConfiguration( TEST_NAME3, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME3, new String[] {\"R\"}) );\naddTestConfiguration( TEST_NAME4, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME4, new String[] {\"R\"}) );\naddTestConfiguration( TEST_NAME5, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME5, new String[] {\"R\"}) );\n+ addTestConfiguration( TEST_NAME6, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME6, new String[] {\"R\"}) );\n}\n@Test\n@@ -87,6 +89,11 @@ public class LineageTraceExecTest extends AutomatedTestBase {\ntestLineageTraceExec(TEST_NAME5);\n}\n+ @Test\n+ public void testLineageTraceExec6() {\n+ testLineageTraceExec(TEST_NAME6);\n+ }\n+\nprivate void testLineageTraceExec(String testname) {\nSystem.out.println(\"------------ BEGIN \" + testname + \"------------\");\n@@ -103,7 +110,7 @@ public class LineageTraceExecTest extends AutomatedTestBase {\nprogramArgs = proArgs.toArray(new String[proArgs.size()]);\nfullDMLScriptName = getScript();\n- if( testname.equals(TEST_NAME3) ) {\n+ if( testname.equals(TEST_NAME3) || testname.equals(TEST_NAME6) ) {\ndouble[][] X = getRandomMatrix(numRecords, numFeatures, 0, 1, 0.8, -1);\nwriteInputMatrixWithMTD(\"X\", X, true);\n}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/lineage/LineageTraceExec6.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1);\n+\n+R = X;\n+for(i in 1:3)\n+ R = rbind(R,X,X,X);\n+\n+write(R, $2);\n" } ]

[ { "change_type": "MODIFY", "old_path": "scripts/builtin/lm.dml", "new_path": "scripts/builtin/lm.dml", "diff": "m_lm = function(Matrix[Double] X, Matrix[Double] y, Integer icpt = 0, Double reg = 1e-7, Double tol = 1e-7, Integer maxi = 0, Boolean verbose = TRUE)\nreturn (Matrix[Double] B) {\n- if (nrow (X) < 2000)\n+ if( ncol(X) <= 1024 )\nB = lmDS(X, y, icpt, reg, verbose)\nelse\nB = lmCG(X, y, icpt, reg, tol, maxi, verbose)\n" } ]

[ { "change_type": "MODIFY", "old_path": "scripts/builtin/steplm.dml", "new_path": "scripts/builtin/steplm.dml", "diff": "@@ -78,10 +78,12 @@ return(Matrix[Double] C, Matrix[Double] S) {\n# BEGIN STEPWISE LINEAR REGRESSION\nif (dir == \"forward\") {\ncontinue = TRUE\n- columns_fixed = matrix(0, rows = 1, cols = m_orig);\n- columns_fixed_ordered = matrix(0, rows = 1, cols = 1);\n+ columns_fixed = matrix(0, 1, m_orig);\n+ columns_fixed_ordered = matrix(0, 1, 1);\n+\n# X_global stores the best model found at each step\n- X_global = matrix(0, rows = n, cols = 1);\n+ X_global = matrix(0, n, 1);\n+\nif (icpt == 1 | icpt == 2) {\nbeta = mean(y);\nAIC_best = 2 + n * log(sum((beta - y) ^ 2) / n);\n@@ -89,39 +91,39 @@ return(Matrix[Double] C, Matrix[Double] S) {\nbeta = 0.0;\nAIC_best = n * log(sum(y ^ 2) / n);\n}\n- AICs = matrix(AIC_best, rows = 1, cols = m_orig);\n+ AICs = matrix(AIC_best, 1, m_orig);\nprint(\"Best AIC without any features: \" + AIC_best);\nboa_ncol = ncol(X_orig);\nif (icpt != 0) {\nboa_ncol = boa_ncol + 1\n}\n- beta_out_all = matrix(0, rows = boa_ncol, cols = m_orig * 1);\n+ beta_out_all = matrix(0, boa_ncol, m_orig * 1);\ny_ncol = 1;\ncolumn_best = 0;\n# First pass to examine single features\nfor (i in 1:m_orig, check = 0) {\n- columns_fixed_ordered_1 = matrix(i, rows = 1, cols = 1);\n+ columns_fixed_ordered_1 = as.matrix(i);\n[AIC_1, beta_out_i] = linear_regression(X_orig[, i], y, icpt);\nAICs[1, i] = AIC_1;\nAIC_cur = as.scalar(AICs[1, i]);\nif ((AIC_cur < AIC_best) & ((AIC_best - AIC_cur) > abs(thr * AIC_best))) {\ncolumn_best = i;\n- AIC_best = as.scalar(AICs[1, i]);\n+ AIC_best = AIC_cur;\n}\n- beta_out_all[1:nrow(beta_out_i), (i - 1) * y_ncol + 1:i * y_ncol] = beta_out_i[, 1:1];\n+ beta_out_all[1:nrow(beta_out_i), ((i - 1) * y_ncol + 1):(i * y_ncol)] = beta_out_i[, 1];\n}\n# beta best so far\nbeta_best = beta_out_all[, (column_best - 1) * y_ncol + 1:column_best * y_ncol];\nif (column_best == 0) {\nprint(\"AIC of an empty model is \" + AIC_best + \" and adding no feature achieves more than \" + (thr * 100) + \"% decrease in AIC!\");\n- Selected = matrix(0, rows = 1, cols = 1);\n+ Selected = matrix(0, 1, 1);\nif (icpt == 0) {\n- B = matrix(beta, rows = m_orig, cols = 1);\n+ B = matrix(beta, m_orig, 1);\n} else {\n- B_tmp = matrix(0, rows = m_orig + 1, cols = 1);\n+ B_tmp = matrix(0, m_orig + 1, 1);\nB_tmp[m_orig + 1,] = beta;\nB = B_tmp;\n}\n@@ -138,12 +140,12 @@ return(Matrix[Double] C, Matrix[Double] S) {\nwhile (continue) {\n# Subsequent passes over the features\n- beta_out_all_2 = matrix(0, rows = boa_ncol, cols = m_orig * 1);\n+ beta_out_all_2 = matrix(0, boa_ncol, m_orig * 1);\nfor (i in 1:m_orig, check = 0) {\nif (as.scalar(columns_fixed[1, i]) == 0) {\n# Construct the feature matrix\nX = cbind(X_global, X_orig[, i]);\n- tmp = matrix(0, rows = 1, cols = 1);\n+ tmp = as.matrix(0);\ntmp[1, 1] = i;\ncolumns_fixed_ordered_2 = append(columns_fixed_ordered, tmp);\n[AIC_2, beta_out_i] = linear_regression(X, y, icpt);\n@@ -183,7 +185,7 @@ return(Matrix[Double] C, Matrix[Double] S) {\n[AIC, beta_out] = linear_regression(X_global, y, icpt);\nSelected = columns_fixed_ordered;\nif (icpt != 0) {\n- Selected = cbind(Selected, matrix(boa_ncol, rows = 1, cols = 1))\n+ Selected = cbind(Selected, matrix(boa_ncol, 1, 1))\n}\nbeta_out = reorder_matrix(boa_ncol, beta_out, Selected);\nS = Selected;\n@@ -193,6 +195,7 @@ return(Matrix[Double] C, Matrix[Double] S) {\nstop(\"Currently only forward selection strategy is supported!\");\n}\n}\n+\n# Computes linear regression using lm and outputs AIC.\nlinear_regression = function(Matrix[Double] X, Matrix[Double] y, Integer icpt = 0)\nreturn(Double AIC, Matrix[Double] beta) {\n@@ -202,7 +205,7 @@ linear_regression = function(Matrix[Double] X, Matrix[Double] y, Integer icpt =\n# Introduce the intercept, shift and rescale the columns of X if needed\nif (icpt == 1 | icpt == 2) {\n# add the intercept column\n- ones_n = matrix(1, rows = n, cols = 1);\n+ ones_n = matrix(1, n, 1);\nX = cbind(X, ones_n);\nm = m - 1;\n}\n@@ -210,7 +213,7 @@ linear_regression = function(Matrix[Double] X, Matrix[Double] y, Integer icpt =\nm_ext = ncol(X);\n# BEGIN THE DIRECT SOLVE ALGORITHM (EXTERNAL CALL)\n- beta = lm(X = X, y = y);\n+ beta = lm(X = X, y = y, verbose=FALSE);\nm_ext = ncol(X);\n# COMPUTE AIC\ny_residual = y - X %*% beta;\n@@ -238,7 +241,7 @@ reorder_matrix = function(\nstop(\"Error: unable to re-order the matrix. Reason: B more than matrix X\");\n}\nif (num_empty_B > 0) {\n- pad_zeros = matrix(0, rows = num_empty_B, cols = 1);\n+ pad_zeros = matrix(0, num_empty_B, 1);\nB = rbind(B, pad_zeros);\nS = rbind(S, pad_zeros);\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinSTEPLmTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinSTEPLmTest.java", "diff": "@@ -33,8 +33,6 @@ public class BuiltinSTEPLmTest extends AutomatedTestBase\nprivate static final String TEST_CLASS_DIR = TEST_DIR + BuiltinSTEPLmTest.class.getSimpleName() + \"/\";\nprivate final static double eps = 1e-10;\n- private final static int rows = 10;\n- private final static int cols = 3;\nprivate final static double spSparse = 0.3;\nprivate final static double spDense = 0.7;\n@@ -45,25 +43,35 @@ public class BuiltinSTEPLmTest extends AutomatedTestBase\n@Test\npublic void testLmMatrixDenseCPlm() {\n- runSTEPLmTest(false, ExecType.CP);\n+ runSTEPLmTest(false, 10, 3, ExecType.CP);\n}\n@Test\npublic void testLmMatrixSparseCPlm() {\n- runSTEPLmTest(true, ExecType.CP);\n+ runSTEPLmTest(true, 10, 3, ExecType.CP);\n}\n@Test\npublic void testLmMatrixDenseSPlm() {\n- runSTEPLmTest(false, ExecType.SPARK);\n+ runSTEPLmTest(false, 10, 3, ExecType.SPARK);\n}\n@Test\npublic void testLmMatrixSparseSPlm() {\n- runSTEPLmTest(true, ExecType.SPARK);\n+ runSTEPLmTest(true, 10, 3, ExecType.SPARK);\n}\n- private void runSTEPLmTest(boolean sparse, ExecType instType) {\n+ @Test\n+ public void testLmMatrixDenseCPlm2() {\n+ runSTEPLmTest(false, 100, 3, ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testLmMatrixSparseCPlm2() {\n+ runSTEPLmTest(true, 100, 3, ExecType.CP);\n+ }\n+\n+ private void runSTEPLmTest(boolean sparse, int rows, int cols, ExecType instType) {\nExecMode platformOld = setExecMode(instType);\nString dml_test_name = TEST_NAME;\n@@ -88,6 +96,7 @@ public class BuiltinSTEPLmTest extends AutomatedTestBase\nrunRScript(true);\n//compare matrices\n+ //FIXME: currently only scenario w/o any features produce same results\nHashMap<CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(\"C\");\nHashMap<CellIndex, Double> dmfile1 = readDMLMatrixFromHDFS(\"S\");\nHashMap<CellIndex, Double> rfile = readRMatrixFromFS(\"C\");\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/builtin/steplm.R", "new_path": "src/test/scripts/functions/builtin/steplm.R", "diff": "@@ -20,12 +20,6 @@ args <- commandArgs(TRUE)\noptions(digits = 22)\nlibrary(\"Matrix\")\n-thr = 0.001;\n-intercept_status = 1;\n-X = as.matrix(readMM(paste(args[1], \"A.mtx\", sep = \"\")))\n-y = as.matrix(readMM(paste(args[1], \"B.mtx\", sep = \"\")))\n-\n-\nreorder_matrix <- function(\nncolX, B, S) {\n# This function assumes that B and S have same number of elements.\n@@ -76,15 +70,18 @@ reorder_matrix <- function(\ny = S0[b];\nP[x, y] = 1;\n}\n-\n}\n}\n-\nY = t(P) %*% B;\nreturn(Y)\n}\n+thr = 0.001;\n+intercept_status = 1;\n+X = as.matrix(readMM(paste(args[1], \"A.mtx\", sep = \"\")))\n+y = as.matrix(readMM(paste(args[1], \"B.mtx\", sep = \"\")))\n+\n# currently only the forward selection strategy in supported: start from one feature and iteratively add\n# features until AIC improves\ndir = 0;\n@@ -103,11 +100,11 @@ m_orig = ncol(X_orig);\nif (dir == 0) {\ncontinue = TRUE;\n- columns_fixed = matrix(0, nrow = 1, ncol = m_orig);\n- columns_fixed_ordered = matrix(0, nrow = 1, ncol = 1);\n+ columns_fixed = matrix(0, 1, m_orig);\n+ columns_fixed_ordered = matrix(0, 1, 1);\n# X_global stores the best model found at each step\n- X_global = matrix(0, nrow = n, ncol = 1);\n+ X_global = matrix(0, n, 1);\nif (intercept_status == 1 | intercept_status == 2) {\nbeta = mean(y);\n@@ -117,66 +114,40 @@ if (dir == 0) {\nAIC_best = n * log(sum(y ^ 2) / n);\n}\n- AICs = matrix(AIC_best, nrow = 1);\n- #print(\"Best AIC without any features: \" + AIC_best);\n- #print(AIC_best);\n+ AICs = matrix(t(AIC_best), 1, m_orig);\nboa_ncol = ncol(X_orig);\nif (intercept_status != 0) {\nboa_ncol = boa_ncol + 1\n}\n- beta_out_all = matrix(0, nrow = boa_ncol, ncol = m_orig * 1);\n-\n- y_ncol = ncol(y);\n+ beta_out_all = matrix(0, boa_ncol, m_orig * 1);\n+ y_ncol = 1;\ncolumn_best = 0;\n+\n# First pass to examine single features\nfor (i in 1:m_orig) {\n-\n- columns_fixed_ordered_1 = matrix(i, nrow = 1, ncol = 1);\n-\n- #[AIC_1, beta_out_i] = linear_regression(X_orig[, i], y, intercept_status);\n+ columns_fixed_ordered_1 = as.matrix(i);\nx = as.matrix(X_orig[, i]);\n- n = nrow(x);\n- m = ncol(x);\n-\n- # Introduce the intercept, shift and rescale the columns of X if needed\n- if (intercept_status == 1 | intercept_status == 2) {\n- # add the intercept column\n- ones_n <- matrix(1, nrow = n, ncol = 1);\n- x = cbind(X_orig[, i], ones_n);\n- m = m - 1;\n- }\n-\n- m_ext = ncol(x)\n-\n- # BEGIN THE DIRECT SOLVE ALGORITHM (EXTERNAL CALL)\n- beta_out_i = lm.fit(x, y)$coefficients\n-\n+ beta_out_i = as.matrix(lm.fit(x, y)$coefficients)\n# COMPUTE AIC\ny_residual = y - x %*% beta_out_i;\nss_res = sum(y_residual ^ 2);\n- eq_deg_of_freedom = m_ext;\n+ eq_deg_of_freedom = ncol(x);\nAIC_1 = (2 * eq_deg_of_freedom) + n * log(ss_res / n);\n-\n- #AICs[1, i] = AIC_1;\n- #AIC_cur =AICs[1, i];\n- if ((AIC_1 < AIC_best) & ((AIC_best - AIC_1) > abs(thr * AIC_best))) {\n+ AICs[1, i] = AIC_1;\n+ AIC_cur = AICs[1, i];\n+ if ((AIC_cur < AIC_best) & ((AIC_best - AIC_cur) > abs(thr * AIC_best))) {\ncolumn_best = i;\n- AIC_best = AIC_1;\n+ AIC_best = AIC_cur;\n}\n-\n- b = as.matrix(beta_out_i)\n-\n- beta_out_all[1:nrow(b), i:i] = b[1, 1];\n-\n-\n+ beta_out_all[1:nrow(beta_out_i), ((i - 1) * y_ncol + 1):(i * y_ncol)] = beta_out_i[,1];\n}\n# beta best so far\n- beta_best = beta_out_all[, (column_best - 1) * y_ncol + 1:column_best * y_ncol];\n+ beta_best = beta_out_all[, ((column_best - 1) * y_ncol + 1):(column_best * y_ncol)];\nif (column_best == 0) {\n@@ -190,15 +161,10 @@ if (dir == 0) {\n}\nbeta_out = B;\n-\n- print(\"estoy en el stop\");\n- print(beta_out);\n- print(Selected);\nwriteMM(as(beta_out, \"CsparseMatrix\"), paste(args[2], \"C\", sep = \"\"));\nwriteMM(as(Selected, \"CsparseMatrix\"), paste(args[2], \"S\", sep = \"\"));\nstop = 1;\n-\n}\ncolumns_fixed[1, column_best] = 1;\n@@ -208,7 +174,7 @@ if (dir == 0) {\nwhile (continue) {\n# Subsequent passes over the features\n- beta_out_all_2 = matrix(0, nrow = boa_ncol, ncol = m_orig * 1);\n+ beta_out_all_2 = matrix(0, boa_ncol, m_orig * 1);\nfor (i in 1:m_orig) {\nif (columns_fixed[1, i] == 0) {\n@@ -254,10 +220,8 @@ if (dir == 0) {\n}\n}\n-\n# have the best beta store in the matrix\n- beta_best = beta_out_all_2[, (column_best - 1) * y_ncol + 1:column_best * y_ncol];\n-\n+ beta_best = beta_out_all_2[, ((column_best - 1) * y_ncol + 1):(column_best * y_ncol)];\n# Append best found features (i.e., columns) to X_global\nif (is.null(columns_fixed[1, column_best])) {\n@@ -272,7 +236,6 @@ if (dir == 0) {\n} else {\nX_global = cbind(X_global, X_orig[, column_best]);\n}\n-\n} else {\ncontinue = FALSE;\n}\n@@ -306,7 +269,6 @@ if (dir == 0) {\neq_deg_of_freedom = m_ext;\nAIC = (2 * eq_deg_of_freedom) + n * log(ss_res / n);\n-\nSelected = columns_fixed_ordered;\nif (intercept_status != 0) {\nSelected = cbind(Selected, matrix(boa_ncol, nrow = 1, ncol = 1))\n@@ -314,12 +276,8 @@ if (dir == 0) {\nbeta_out = reorder_matrix(boa_ncol, beta_out, Selected);\n- print(beta_out);\n- print(Selected[1]);\n-\nwriteMM(as(beta_out, \"CsparseMatrix\"), paste(args[2], \"C\", sep = \"\"));\nwriteMM(as(Selected[1], \"CsparseMatrix\"), paste(args[2], \"S\", sep = \"\"));\n-\n}\n} else {\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -118,6 +118,6 @@ SYSTEMDS-160 Tensor Compiler/Runtime\n* 173 Improved cost estimates OK\n* 174 Reuse rewrite for rbind-tsmm OK\n* 175 Refactoring of lineage rewrite code\n- * 175 Reuse rewrite for cbind/rbind-elementwise */+\n- * 176 Reuse rewrite for aggregate\n- * 177 Compiler assisted reuse (eg. CV, lmCG)\n+ * 176 Reuse rewrite for cbind/rbind-elementwise */+\n+ * 177 Reuse rewrite for aggregate\n+ * 178 Compiler assisted reuse (eg. CV, lmCG)\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "diff": "@@ -61,8 +61,8 @@ public class LineageCache {\n//try to reuse full or partial intermediates\nif (LineageCacheConfig.getCacheType().isFullReuse())\nreuse = fullReuse(item, (ComputationCPInstruction)inst, ec);\n- else if (LineageCacheConfig.getCacheType().isPartialReuse())\n- reuse = RewriteCPlans.executeRewrites(inst, ec);\n+ if (LineageCacheConfig.getCacheType().isPartialReuse())\n+ reuse |= LineageRewriteReuse.executeRewrites(inst, ec);\n//create a placeholder if no reuse to avoid redundancy\n//(e.g., concurrent threads that try to start the computation)\n@@ -105,7 +105,9 @@ public class LineageCache {\nprivate static void putIntern(Instruction inst, LineageItem key, MatrixBlock value, double compcost) {\nif (_cache.containsKey(key))\n- throw new DMLRuntimeException(\"Redundant lineage caching detected: \"+inst);\n+ //can come here if reuse_partial option is enabled\n+ return;\n+ //throw new DMLRuntimeException(\"Redundant lineage caching detected: \"+inst);\n// Create a new entry.\nEntry newItem = new Entry(key, value, compcost);\n" }, { "change_type": "RENAME", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/RewriteCPlans.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageRewriteReuse.java", "diff": "@@ -48,72 +48,52 @@ import org.tugraz.sysds.runtime.meta.MetaData;\nimport org.tugraz.sysds.utils.Explain;\nimport org.tugraz.sysds.utils.Explain.ExplainType;\n-public class RewriteCPlans\n+public class LineageRewriteReuse\n{\nprivate static final String LR_VAR = \"__lrwrt\";\nprivate static BasicProgramBlock _lrPB = null;\nprivate static ExecutionContext _lrEC = null;\n- private static final Log LOG = LogFactory.getLog(RewriteCPlans.class.getName());\n+ private static final Log LOG = LogFactory.getLog(LineageRewriteReuse.class.getName());\npublic static boolean executeRewrites (Instruction curr, ExecutionContext ec)\n{\n- boolean oneappend = false;\n- boolean twoappend = false;\n- boolean onerbind = false;\n- MatrixBlock lastResult = null;\n-\n- MatrixBlock mmc = isTsmmCbind(curr, ec);\n- if (mmc != null) {oneappend = true; lastResult = mmc;}\n- MatrixBlock mm2c = isTsmm2Cbind(curr, ec);\n- if (mm2c != null) {twoappend = true; lastResult = mm2c;}\n- MatrixBlock mmr = isTsmmRbind(curr, ec);\n- if (mmr != null) {onerbind = true; lastResult = mmr;}\n-\n- if (!oneappend && !twoappend && !onerbind)\n- return false;\n-\nExecutionContext lrwec = getExecutionContext();\nExplainType et = DMLScript.EXPLAIN;\n// Disable explain not to print unnecessary logs\n// TODO extend recompiler to allow use without explain output\nDMLScript.EXPLAIN = ExplainType.NONE;\n- try {\n- long t0 = DMLScript.STATISTICS ? System.nanoTime() : 0;\n- ArrayList<Instruction> newInst = oneappend ? rewriteCbindTsmm(curr, ec, lrwec, lastResult) :\n- twoappend ? rewrite2CbindTsmm(curr, ec, lrwec, lastResult) :\n- onerbind ? rewriteRbindTsmm(curr, ec, lrwec, lastResult) : null;\n- if (DMLScript.STATISTICS) {\n- LineageCacheStatistics.incrementPRewriteTime(System.nanoTime() - t0);\n- LineageCacheStatistics.incrementPRewrites();\n- }\n- //execute instructions\n- BasicProgramBlock pb = getProgramBlock();\n- pb.setInstructions(newInst);\n- ReuseCacheType oldReuseOption = DMLScript.LINEAGE_REUSE;\n- LineageCacheConfig.shutdownReuse();\n- pb.execute(lrwec);\n- LineageCacheConfig.restartReuse(oldReuseOption);\n+ //check applicability and apply rewrite\n+ ArrayList<Instruction> newInst = rewriteTsmmCbind(curr, ec, lrwec); //tsmm(cbind(X, deltaX)) using tsmm(X)\n+ newInst = (newInst == null) ? rewriteTsmm2Cbind(curr, ec, lrwec) : newInst; //tsmm(cbind(cbind(X, deltaX), ones)) using tsmm(X)\n+ newInst = (newInst == null) ? rewriteTsmmRbind(curr, ec, lrwec) : newInst; //tsmm(rbind(X, deltaX)) using tsmm(X)\n+\n+ if (newInst == null)\n+ return false;\n+\n+ //execute instructions & write the o/p to symbol table\n+ executeInst(newInst, lrwec);\nec.setVariable(((ComputationCPInstruction)curr).output.getName(), lrwec.getVariable(LR_VAR));\n- // add this to cache\n+\n+ //put the result into the cache\nLineageCache.put(curr, ec);\n- }\n- catch (Exception e) {\n- throw new DMLRuntimeException(\"Error evaluating instruction: \" + curr.toString() , e);\n- }\nDMLScript.EXPLAIN = et;\nreturn true;\n}\n- private static ArrayList<Instruction> rewriteCbindTsmm(Instruction curr, ExecutionContext ec, ExecutionContext lrwec, MatrixBlock lastResult)\n+ /*--------------------------------REWRITE METHODS------------------------------*/\n+\n+ private static ArrayList<Instruction> rewriteTsmmCbind (Instruction curr, ExecutionContext ec, ExecutionContext lrwec)\n{\n- // Create a transient read op over the last tsmm result\n- MetaData md = new MetaData(lastResult.getDataCharacteristics());\n- MatrixObject newmo = new MatrixObject(ValueType.FP64, \"lastResult\", md);\n- newmo.acquireModify(lastResult);\n- newmo.release();\n- lrwec.setVariable(\"lastResult\", newmo);\n- DataOp lastRes = HopRewriteUtils.createTransientRead(\"lastResult\", lastResult);\n+ // Check the applicability of this rewrite.\n+ MatrixBlock cachedEntry = isTsmmCbind(curr, ec);\n+ if (cachedEntry == null)\n+ return null;\n+\n+ long t0 = DMLScript.STATISTICS ? System.nanoTime() : 0;\n+ // Create a transient read op over the cached tsmm result\n+ lrwec.setVariable(\"cachedEntry\", convMBtoMO(cachedEntry));\n+ DataOp lastRes = HopRewriteUtils.createTransientRead(\"cachedEntry\", cachedEntry);\n// Create rightIndex op to find the last matrix and the appended column\n// TODO: For now assumption is that a single column is being appended in a loop\n// Need to go down the lineage to find number of columns are being appended.\n@@ -141,20 +121,27 @@ public class RewriteCPlans\nBinaryOp lrwHop= HopRewriteUtils.createBinary(rowOne, rowTwo, OpOp2.RBIND);\nDataOp lrwWrite = HopRewriteUtils.createTransientWrite(LR_VAR, lrwHop);\n+ if (DMLScript.STATISTICS) {\n+ LineageCacheStatistics.incrementPRewriteTime(System.nanoTime() - t0);\n+ LineageCacheStatistics.incrementPRewrites();\n+ }\n+\n// generate runtime instructions\nLOG.debug(\"LINEAGE REWRITE rewriteCbindTsmm APPLIED\");\nreturn genInst(lrwWrite, lrwec);\n}\n- private static ArrayList<Instruction> rewriteRbindTsmm(Instruction curr, ExecutionContext ec, ExecutionContext lrwec, MatrixBlock lastResult)\n+ private static ArrayList<Instruction> rewriteTsmmRbind (Instruction curr, ExecutionContext ec, ExecutionContext lrwec)\n{\n+ // Check the applicability of this rewrite.\n+ MatrixBlock cachedEntry = isTsmmRbind(curr, ec);\n+ if (cachedEntry == null)\n+ return null;\n+\n+ long t0 = DMLScript.STATISTICS ? System.nanoTime() : 0;\n// Create a transient read op over the last tsmm result\n- MetaData md = new MetaData(lastResult.getDataCharacteristics());\n- MatrixObject newmo = new MatrixObject(ValueType.FP64, \"lastResult\", md);\n- newmo.acquireModify(lastResult);\n- newmo.release();\n- lrwec.setVariable(\"lastResult\", newmo);\n- DataOp lastRes = HopRewriteUtils.createTransientRead(\"lastResult\", lastResult);\n+ lrwec.setVariable(\"cachedEntry\", convMBtoMO(cachedEntry));\n+ DataOp lastRes = HopRewriteUtils.createTransientRead(\"cachedEntry\", cachedEntry);\n// Create rightIndex op to find the last appended rows\n//TODO: support for block of rows\nMatrixObject mo = ec.getMatrixObject(((ComputationCPInstruction)curr).input1);\n@@ -168,20 +155,28 @@ public class RewriteCPlans\nBinaryOp lrwHop = HopRewriteUtils.createBinary(lastRes, tsmm_lr, OpOp2.PLUS);\nDataOp lrwWrite = HopRewriteUtils.createTransientWrite(LR_VAR, lrwHop);\n+ if (DMLScript.STATISTICS) {\n+ LineageCacheStatistics.incrementPRewriteTime(System.nanoTime() - t0);\n+ LineageCacheStatistics.incrementPRewrites();\n+ }\n+\n// generate runtime instructions\nLOG.debug(\"LINEAGE REWRITE rewriteRbindTsmm APPLIED\");\nreturn genInst(lrwWrite, lrwec);\n}\n- private static ArrayList<Instruction> rewrite2CbindTsmm(Instruction curr, ExecutionContext ec, ExecutionContext lrwec, MatrixBlock lastResult)\n+ private static ArrayList<Instruction> rewriteTsmm2Cbind (Instruction curr, ExecutionContext ec, ExecutionContext lrwec)\n{\n+ // Check the applicability of this rewrite.\n+ MatrixBlock cachedEntry = isTsmm2Cbind(curr, ec);\n+ if (cachedEntry == null)\n+ return null;\n+\n+ long t0 = DMLScript.STATISTICS ? System.nanoTime() : 0;\n// Create a transient read op over the last tsmm result\n- MetaData md = new MetaData(lastResult.getDataCharacteristics());\n- MatrixObject newmo = new MatrixObject(ValueType.FP64, \"lastResult\", md);\n- newmo.acquireModify(lastResult);\n- newmo.release();\n- lrwec.setVariable(\"lastResult\", newmo);\n- DataOp lastRes = HopRewriteUtils.createTransientRead(\"lastResult\", lastResult);\n+ MatrixObject newmo = convMBtoMO(cachedEntry);\n+ lrwec.setVariable(\"cachedEntry\", newmo);\n+ DataOp lastRes = HopRewriteUtils.createTransientRead(\"cachedEntry\", cachedEntry);\nMatrixObject mo = ec.getMatrixObject(((ComputationCPInstruction)curr).input1);\nlrwec.setVariable(\"oldMatrix\", mo);\nDataOp newMatrix = HopRewriteUtils.createTransientRead(\"oldMatrix\", mo);\n@@ -212,76 +207,71 @@ public class RewriteCPlans\nNaryOp lrwHop = HopRewriteUtils.createNary(OpOpN.RBIND, rowOne, newCol, rowTwo);\nDataOp lrwWrite = HopRewriteUtils.createTransientWrite(LR_VAR, lrwHop);\n+ if (DMLScript.STATISTICS) {\n+ LineageCacheStatistics.incrementPRewriteTime(System.nanoTime() - t0);\n+ LineageCacheStatistics.incrementPRewrites();\n+ }\n+\n// generate runtime instructions\nLOG.debug(\"LINEAGE REWRITE rewrite2CbindTsmm APPLIED\");\nreturn genInst(lrwWrite, lrwec);\n}\n- private static ArrayList<Instruction> genInst(Hop hops, ExecutionContext ec) {\n- ArrayList<Instruction> newInst = Recompiler.recompileHopsDag(hops, ec.getVariables(), null, true, true, 0);\n- LOG.debug(\"EXPLAIN LINEAGE REWRITE \\nGENERIC (line \"+hops.getBeginLine()+\"):\\n\" + Explain.explain(hops,1));\n- LOG.debug(\"EXPLAIN LINEAGE REWRITE \\nGENERIC (line \"+hops.getBeginLine()+\"):\\n\" + Explain.explain(newInst,1));\n- return newInst;\n- }\n+ /*------------------------REWRITE APPLICABILITY CHECKS-------------------------*/\nprivate static MatrixBlock isTsmmCbind(Instruction curr, ExecutionContext ec)\n{\n- MatrixBlock lastResult = null;\n+ MatrixBlock cachedEntry = null;\nif (!LineageCache.isReusable(curr))\n- return lastResult;\n+ return cachedEntry;\n// If the input to tsmm came from cbind, look for both the inputs in cache.\nLineageItem[] items = ((ComputationCPInstruction) curr).getLineageItems(ec);\nLineageItem item = items[0];\n-\n- // TODO restructuring of rewrites to make them all\n- // independent of each other and this opening condition here\nfor (LineageItem source : item.getInputs())\nif (source.getOpcode().equalsIgnoreCase(\"cbind\")) {\nfor (LineageItem input : source.getInputs()) {\n// create tsmm lineage on top of the input of last append\nLineageItem tmp = new LineageItem(\"toProbe\", curr.getOpcode(), new LineageItem[] {input});\nif (LineageCache.probe(tmp)) {\n- if (lastResult == null)\n- lastResult = LineageCache.get(curr, tmp);\n+ if (cachedEntry == null)\n+ cachedEntry = LineageCache.get(curr, tmp);\n}\n}\n}\n- return lastResult;\n+ return cachedEntry;\n}\nprivate static MatrixBlock isTsmmRbind(Instruction curr, ExecutionContext ec)\n{\n- MatrixBlock lastResult = null;\n+ MatrixBlock cachedEntry = null;\nif (!LineageCache.isReusable(curr))\n- return lastResult;\n+ return cachedEntry;\n// If the input to tsmm came from rbind, look for both the inputs in cache.\nLineageItem[] items = ((ComputationCPInstruction) curr).getLineageItems(ec);\nLineageItem item = items[0];\n-\n- // TODO restructuring of rewrites to make them all\n- // independent of each other and this opening condition here\nfor (LineageItem source : item.getInputs())\nif (source.getOpcode().equalsIgnoreCase(\"rbind\")) {\nfor (LineageItem input : source.getInputs()) {\n// create tsmm lineage on top of the input of last append\nLineageItem tmp = new LineageItem(\"toProbe\", curr.getOpcode(), new LineageItem[] {input});\nif (LineageCache.probe(tmp)) {\n- if (lastResult == null)\n- lastResult = LineageCache.get(curr, tmp);\n+ if (cachedEntry == null)\n+ cachedEntry = LineageCache.get(curr, tmp);\n}\n}\n}\n- return lastResult;\n+ return cachedEntry;\n}\nprivate static MatrixBlock isTsmm2Cbind (Instruction curr, ExecutionContext ec)\n{\n- MatrixBlock lastResult = null;\n+ MatrixBlock cachedEntry = null;\nif (!LineageCache.isReusable(curr))\n- return lastResult;\n+ return cachedEntry;\n+ //TODO: support nary cbind\n// If the input to tsmm came from cbind, look for both the inputs in cache.\nLineageItem[] items = ((ComputationCPInstruction) curr).getLineageItems(ec);\nLineageItem item = items[0];\n@@ -294,13 +284,52 @@ public class RewriteCPlans\nLineageItem tmp = new LineageItem(\"comb\", \"cbind\", new LineageItem[] {L2appin, source.getInputs()[1]});\nLineageItem toProbe = new LineageItem(\"toProbe\", curr.getOpcode(), new LineageItem[] {tmp});\nif (LineageCache.probe(toProbe)) {\n- if (lastResult == null)\n- lastResult = LineageCache.get(curr, toProbe);\n+ if (cachedEntry == null)\n+ cachedEntry = LineageCache.get(curr, toProbe);\n+ }\n+ }\n}\n}\n+ return cachedEntry;\n}\n+\n+ /*----------------------INSTRUCTIONS GENERATION & EXECUTION-----------------------*/\n+\n+ private static ArrayList<Instruction> genInst(Hop hops, ExecutionContext ec) {\n+ ArrayList<Instruction> newInst = Recompiler.recompileHopsDag(hops, ec.getVariables(), null, true, true, 0);\n+ LOG.debug(\"EXPLAIN LINEAGE REWRITE \\nGENERIC (line \"+hops.getBeginLine()+\"):\\n\" + Explain.explain(hops,1));\n+ LOG.debug(\"EXPLAIN LINEAGE REWRITE \\nGENERIC (line \"+hops.getBeginLine()+\"):\\n\" + Explain.explain(newInst,1));\n+ return newInst;\n}\n- return lastResult;\n+\n+ private static void executeInst (ArrayList<Instruction> newInst, ExecutionContext lrwec)\n+ {\n+ // Disable explain not to print unnecessary logs\n+ // TODO extend recompiler to allow use without explain output\n+ DMLScript.EXPLAIN = ExplainType.NONE;\n+\n+ try {\n+ //execute instructions\n+ BasicProgramBlock pb = getProgramBlock();\n+ pb.setInstructions(newInst);\n+ ReuseCacheType oldReuseOption = DMLScript.LINEAGE_REUSE;\n+ LineageCacheConfig.shutdownReuse();\n+ pb.execute(lrwec);\n+ LineageCacheConfig.restartReuse(oldReuseOption);\n+ }\n+ catch (Exception e) {\n+ throw new DMLRuntimeException(\"Error executing lineage rewrites\" , e);\n+ }\n+ }\n+\n+ /*-------------------------------UTILITY METHODS----------------------------------*/\n+\n+ private static MatrixObject convMBtoMO (MatrixBlock cachedEntry) {\n+ MetaData md = new MetaData(cachedEntry.getDataCharacteristics());\n+ MatrixObject mo = new MatrixObject(ValueType.FP64, \"cachedEntry\", md);\n+ mo.acquireModify(cachedEntry);\n+ mo.release();\n+ return mo;\n}\nprivate static ExecutionContext getExecutionContext() {\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/api/DMLOptions.java", "new_path": "src/main/java/org/tugraz/sysds/api/DMLOptions.java", "diff": "@@ -121,7 +121,8 @@ public class DMLOptions {\nelse if (lineageType.equalsIgnoreCase(\"reuse_hybrid\"))\ndmlOptions.linReuseType = ReuseCacheType.REUSE_HYBRID;\nelse\n- throw new org.apache.commons.cli.ParseException(\"Invalid argument specified for -lineage option\");\n+ throw new org.apache.commons.cli.ParseException(\n+ \"Invalid argument specified for -lineage option: \" + lineageType);\n}\n}\n}\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/FullReuseTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/FullReuseTest.java", "diff": "@@ -20,6 +20,7 @@ import org.junit.Test;\nimport org.tugraz.sysds.hops.OptimizerUtils;\nimport org.tugraz.sysds.hops.recompile.Recompiler;\nimport org.tugraz.sysds.runtime.lineage.Lineage;\n+import org.tugraz.sysds.runtime.lineage.LineageCacheConfig.ReuseCacheType;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixValue;\nimport org.tugraz.sysds.test.AutomatedTestBase;\nimport org.tugraz.sysds.test.TestConfiguration;\n@@ -89,7 +90,7 @@ public class FullReuseTest extends AutomatedTestBase {\nproArgs.clear();\nproArgs.add(\"-stats\");\nproArgs.add(\"-lineage\");\n- proArgs.add(\"reuse\");\n+ proArgs.add(ReuseCacheType.REUSE_FULL.name().toLowerCase());\nproArgs.add(\"-args\");\nproArgs.add(output(\"X\"));\nprogramArgs = proArgs.toArray(new String[proArgs.size()]);\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -76,6 +76,7 @@ SYSTEMDS-100 Various Fixes\n* 105 Fix all application/function tests (various issues) OK\n* 106 Fix correctness of as.integer for negative numbers OK\n* 107 Fix correctness IPA check dimension-preserving OK\n+ * 108 Fix codegen optimizer (early costing abort) OK\nSYSTEMDS-110 New Builtin Functions\n* 111 Time builtin function for script-level measurements OK\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/codegen/opt/PlanSelectionFuseCostBasedV2.java", "new_path": "src/main/java/org/tugraz/sysds/hops/codegen/opt/PlanSelectionFuseCostBasedV2.java", "diff": "@@ -1000,6 +1000,8 @@ public class PlanSelectionFuseCostBasedV2 extends PlanSelection\nelse if( part.getExtConsumed().contains(current.getHopID()) ) {\ncosts += rGetPlanCosts(memo, current, visited, part, matPoints, plan,\ncomputeCosts, null, null, costBound - costs);\n+ if( costs >= costBound )\n+ return Double.POSITIVE_INFINITY;\n}\n//sanity check non-negative costs\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -117,8 +117,8 @@ SYSTEMDS-160 Tensor Compiler/Runtime\n* 171 Initial version of partial rewrites OK\n* 172 Parfor integration (blocked waiting for results) OK\n* 173 Improved cost estimates OK\n- * 174 Reuse rewrite for rbind-tsmm OK\n- * 175 Refactoring of lineage rewrite code\n+ * 174 Reuse rewrite for rbind/cbind-tsmm/ba+* OK\n+ * 175 Refactoring of lineage rewrite code OK\n* 176 Reuse rewrite for cbind/rbind-elementwise */+\n* 177 Reuse rewrite for aggregate\n* 178 Compiler assisted reuse (eg. CV, lmCG)\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "diff": "@@ -170,8 +170,7 @@ public class LineageCache {\npublic static boolean isReusable (Instruction inst) {\n// TODO: Move this to the new class LineageCacheConfig and extend\nreturn inst.getOpcode().equalsIgnoreCase(\"tsmm\")\n- || (LineageCacheConfig.getCacheType().isFullReuse()\n- && inst.getOpcode().equalsIgnoreCase(\"ba+*\"));\n+ || inst.getOpcode().equalsIgnoreCase(\"ba+*\");\n}\n//---------------- CACHE SPACE MANAGEMENT METHODS -----------------\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageRewriteReuse.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageRewriteReuse.java", "diff": "@@ -66,7 +66,12 @@ public class LineageRewriteReuse\n//check applicability and apply rewrite\nArrayList<Instruction> newInst = rewriteTsmmCbind(curr, ec, lrwec); //tsmm(cbind(X, deltaX)) using tsmm(X)\nnewInst = (newInst == null) ? rewriteTsmm2Cbind(curr, ec, lrwec) : newInst; //tsmm(cbind(cbind(X, deltaX), ones)) using tsmm(X)\n- newInst = (newInst == null) ? rewriteTsmmRbind(curr, ec, lrwec) : newInst; //tsmm(rbind(X, deltaX)) using tsmm(X)\n+ //tsmm(rbind(X, deltaX)) using C = tsmm(X) -> C + tsmm(deltaX)\n+ newInst = (newInst == null) ? rewriteTsmmRbind(curr, ec, lrwec) : newInst;\n+ //rbind(X,deltaX) %*% Y using C = X %*% Y -> rbind(C, deltaX %*% Y)\n+ newInst = (newInst == null) ? rewriteMatMulRbindLeft(curr, ec, lrwec) : newInst;\n+ //X %*% cbind(Y,deltaY)) using C = X %*% Y -> cbind(C, X %*% deltaY)\n+ newInst = (newInst == null) ? rewriteMatMulCbindRight(curr, ec, lrwec) : newInst;\nif (newInst == null)\nreturn false;\n@@ -217,6 +222,80 @@ public class LineageRewriteReuse\nreturn genInst(lrwWrite, lrwec);\n}\n+ private static ArrayList<Instruction> rewriteMatMulRbindLeft (Instruction curr, ExecutionContext ec, ExecutionContext lrwec)\n+ {\n+ // Check the applicability of this rewrite.\n+ MatrixBlock cachedEntry = isMatMulRbindLeft(curr, ec);\n+ if (cachedEntry == null)\n+ return null;\n+\n+ long t0 = DMLScript.STATISTICS ? System.nanoTime() : 0;\n+ // Create a transient read op over the last ba+* result\n+ lrwec.setVariable(\"cachedEntry\", convMBtoMO(cachedEntry));\n+ DataOp lastRes = HopRewriteUtils.createTransientRead(\"cachedEntry\", cachedEntry);\n+ // Create rightIndex op to find the last appended rows\n+ //TODO: support for block of rows\n+ MatrixObject moL = ec.getMatrixObject(((ComputationCPInstruction)curr).input1);\n+ lrwec.setVariable(\"leftMatrix\", moL);\n+ DataOp leftMatrix = HopRewriteUtils.createTransientRead(\"leftMatrix\", moL);\n+ MatrixObject moR = ec.getMatrixObject(((ComputationCPInstruction)curr).input2);\n+ lrwec.setVariable(\"rightMatrix\", moR);\n+ DataOp rightMatrix = HopRewriteUtils.createTransientRead(\"rightMatrix\", moR);\n+ //TODO avoid the indexing if possible (if deltaX found in cache)\n+ IndexingOp lastRow = HopRewriteUtils.createIndexingOp(leftMatrix, new LiteralOp(moL.getNumRows()),\n+ new LiteralOp(moL.getNumRows()), new LiteralOp(1), new LiteralOp(moL.getNumColumns()));\n+ // ba+*(X+lastRow, Y) = rbind(ba+*(X, Y), ba+*(lastRow, Y))\n+ AggBinaryOp rowTwo = HopRewriteUtils.createMatrixMultiply(lastRow, rightMatrix);\n+ BinaryOp lrwHop= HopRewriteUtils.createBinary(lastRes, rowTwo, OpOp2.RBIND);\n+ DataOp lrwWrite = HopRewriteUtils.createTransientWrite(LR_VAR, lrwHop);\n+\n+ if (DMLScript.STATISTICS) {\n+ LineageCacheStatistics.incrementPRewriteTime(System.nanoTime() - t0);\n+ LineageCacheStatistics.incrementPRewrites();\n+ }\n+\n+ // generate runtime instructions\n+ LOG.debug(\"LINEAGE REWRITE rewriteCbindTsmm APPLIED\");\n+ return genInst(lrwWrite, lrwec);\n+ }\n+\n+ private static ArrayList<Instruction> rewriteMatMulCbindRight (Instruction curr, ExecutionContext ec, ExecutionContext lrwec)\n+ {\n+ // Check the applicability of this rewrite.\n+ MatrixBlock cachedEntry = isMatMulCbindRight(curr, ec);\n+ if (cachedEntry == null)\n+ return null;\n+\n+ long t0 = DMLScript.STATISTICS ? System.nanoTime() : 0;\n+ // Create a transient read op over the last ba+* result\n+ lrwec.setVariable(\"cachedEntry\", convMBtoMO(cachedEntry));\n+ DataOp lastRes = HopRewriteUtils.createTransientRead(\"cachedEntry\", cachedEntry);\n+ // Create rightIndex op to find the last appended column\n+ //TODO: support for block of rows\n+ MatrixObject moL = ec.getMatrixObject(((ComputationCPInstruction)curr).input1);\n+ lrwec.setVariable(\"leftMatrix\", moL);\n+ DataOp leftMatrix = HopRewriteUtils.createTransientRead(\"leftMatrix\", moL);\n+ MatrixObject moR = ec.getMatrixObject(((ComputationCPInstruction)curr).input2);\n+ lrwec.setVariable(\"rightMatrix\", moR);\n+ DataOp rightMatrix = HopRewriteUtils.createTransientRead(\"rightMatrix\", moR);\n+ //TODO avoid the indexing if possible (if deltaY found in cache)\n+ IndexingOp lastCol = HopRewriteUtils.createIndexingOp(rightMatrix, new LiteralOp(1), new LiteralOp(moR.getNumRows()),\n+ new LiteralOp(moR.getNumColumns()), new LiteralOp(moR.getNumColumns()));\n+ // ba+*(X, Y+lastCol) = cbind(ba+*(X, Y), ba+*(X, lastCol))\n+ AggBinaryOp rowTwo = HopRewriteUtils.createMatrixMultiply(leftMatrix, lastCol);\n+ BinaryOp lrwHop= HopRewriteUtils.createBinary(lastRes, rowTwo, OpOp2.CBIND);\n+ DataOp lrwWrite = HopRewriteUtils.createTransientWrite(LR_VAR, lrwHop);\n+\n+ if (DMLScript.STATISTICS) {\n+ LineageCacheStatistics.incrementPRewriteTime(System.nanoTime() - t0);\n+ LineageCacheStatistics.incrementPRewrites();\n+ }\n+\n+ // generate runtime instructions\n+ LOG.debug(\"LINEAGE REWRITE rewriteCbindTsmm APPLIED\");\n+ return genInst(lrwWrite, lrwec);\n+ }\n+\n/*------------------------REWRITE APPLICABILITY CHECKS-------------------------*/\nprivate static MatrixBlock isTsmmCbind(Instruction curr, ExecutionContext ec)\n@@ -293,6 +372,50 @@ public class LineageRewriteReuse\nreturn cachedEntry;\n}\n+ private static MatrixBlock isMatMulRbindLeft(Instruction curr, ExecutionContext ec)\n+ {\n+ MatrixBlock cachedEntry = null;\n+ if (!LineageCache.isReusable(curr))\n+ return cachedEntry;\n+\n+ // If the left input to ba+* came from rbind, look for both the inputs in cache.\n+ LineageItem[] items = ((ComputationCPInstruction) curr).getLineageItems(ec);\n+ if (curr.getOpcode().equalsIgnoreCase(\"ba+*\")) {\n+ LineageItem left= items[0].getInputs()[0];\n+ LineageItem right = items[0].getInputs()[1];\n+ if (left.getOpcode().equalsIgnoreCase(\"rbind\")){\n+ LineageItem leftSource = left.getInputs()[0]; //left inpur of rbind = X\n+ // create ba+* lineage on top of the input of last append\n+ LineageItem tmp = new LineageItem(\"toProbe\", curr.getOpcode(), new LineageItem[] {leftSource, right});\n+ if (LineageCache.probe(tmp))\n+ cachedEntry = LineageCache.get(curr, tmp);\n+ }\n+ }\n+ return cachedEntry;\n+ }\n+\n+ private static MatrixBlock isMatMulCbindRight(Instruction curr, ExecutionContext ec)\n+ {\n+ MatrixBlock cachedEntry = null;\n+ if (!LineageCache.isReusable(curr))\n+ return cachedEntry;\n+\n+ // If the right input to ba+* came from cbind, look for both the inputs in cache.\n+ LineageItem[] items = ((ComputationCPInstruction) curr).getLineageItems(ec);\n+ if (curr.getOpcode().equalsIgnoreCase(\"ba+*\")) {\n+ LineageItem left = items[0].getInputs()[0];\n+ LineageItem right = items[0].getInputs()[1];\n+ if (right.getOpcode().equalsIgnoreCase(\"cbind\")) {\n+ LineageItem rightSource = right.getInputs()[0]; //left inpur of rbind = X\n+ // create ba+* lineage on top of the input of last append\n+ LineageItem tmp = new LineageItem(\"toProbe\", curr.getOpcode(), new LineageItem[] {left, rightSource});\n+ if (LineageCache.probe(tmp))\n+ cachedEntry = LineageCache.get(curr, tmp);\n+ }\n+ }\n+ return cachedEntry;\n+ }\n+\n/*----------------------INSTRUCTIONS GENERATION & EXECUTION-----------------------*/\nprivate static ArrayList<Instruction> genInst(Hop hops, ExecutionContext ec) {\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageRewriteTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageRewriteTest.java", "diff": "@@ -33,6 +33,8 @@ public class LineageRewriteTest extends AutomatedTestBase {\nprotected static final String TEST_NAME1 = \"RewriteTest3\";\nprotected static final String TEST_NAME2 = \"RewriteTest2\";\nprotected static final String TEST_NAME3 = \"RewriteTest7\";\n+ protected static final String TEST_NAME4 = \"RewriteTest8\";\n+ protected static final String TEST_NAME5 = \"RewriteTest9\";\nprotected String TEST_CLASS_DIR = TEST_DIR + LineageRewriteTest.class.getSimpleName() + \"/\";\n@@ -45,6 +47,8 @@ public class LineageRewriteTest extends AutomatedTestBase {\naddTestConfiguration(TEST_NAME1, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME1));\naddTestConfiguration(TEST_NAME2, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME2));\naddTestConfiguration(TEST_NAME3, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME3));\n+ addTestConfiguration(TEST_NAME4, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME4));\n+ addTestConfiguration(TEST_NAME5, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME5));\n}\n@Test\n@@ -62,6 +66,16 @@ public class LineageRewriteTest extends AutomatedTestBase {\ntestRewrite(TEST_NAME3);\n}\n+ @Test\n+ public void testmatmulrbindleft() {\n+ testRewrite(TEST_NAME4);\n+ }\n+\n+ @Test\n+ public void testmatmulcbindright() {\n+ testRewrite(TEST_NAME5);\n+ }\n+\nprivate void testRewrite(String testname) {\ntry {\ngetAndLoadTestConfiguration(testname);\n@@ -72,11 +86,14 @@ public class LineageRewriteTest extends AutomatedTestBase {\nproArgs.add(\"-lineage\");\nproArgs.add(\"-args\");\nproArgs.add(input(\"X\"));\n+ proArgs.add(input(\"Y\"));\nproArgs.add(output(\"Res\"));\nprogramArgs = proArgs.toArray(new String[proArgs.size()]);\nfullDMLScriptName = getScript();\ndouble[][] X = getRandomMatrix(numRecords, numFeatures, 0, 1, 0.8, -1);\n+ double[][] Y = getRandomMatrix(numFeatures, numRecords, 0, 1, 0.8, -1);\nwriteInputMatrixWithMTD(\"X\", X, true);\n+ writeInputMatrixWithMTD(\"Y\", Y, true);\nrunTest(true, EXCEPTION_NOT_EXPECTED, null, -1);\nHashMap<MatrixValue.CellIndex, Double> R_orig = readDMLMatrixFromHDFS(\"Res\");\n@@ -88,6 +105,7 @@ public class LineageRewriteTest extends AutomatedTestBase {\nproArgs.add(\"reuse_partial\");\nproArgs.add(\"-args\");\nproArgs.add(input(\"X\"));\n+ proArgs.add(input(\"Y\"));\nproArgs.add(output(\"Res\"));\nprogramArgs = proArgs.toArray(new String[proArgs.size()]);\nfullDMLScriptName = getScript();\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/lineage/RewriteTest2.dml", "new_path": "src/test/scripts/functions/lineage/RewriteTest2.dml", "diff": "@@ -30,4 +30,4 @@ for (i in 2:ncol(X)) {\nsum = sum + sum(Res1);\n}\n-write(R, $2, format=\"text\");\n+write(R, $3, format=\"text\");\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/lineage/RewriteTest3.dml", "new_path": "src/test/scripts/functions/lineage/RewriteTest3.dml", "diff": "@@ -33,4 +33,4 @@ for (i in 2:ncol(X)) {\nsum = sum + sum(Res1);\n}\n-write(R, $2, format=\"text\");\n+write(R, $3, format=\"text\");\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/lineage/RewriteTest7.dml", "new_path": "src/test/scripts/functions/lineage/RewriteTest7.dml", "diff": "@@ -30,4 +30,4 @@ for (i in 2:nrow(X)) {\nsum = sum + sum(Res1);\n}\n-write(R, $2, format=\"text\");\n+write(R, $3, format=\"text\");\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/lineage/RewriteTest8.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1);\n+Y = read($2);\n+\n+sum = 0;\n+tmp = X[1,];\n+R = matrix(0, 1, nrow(X));\n+\n+for (i in 2:nrow(X)) {\n+ Res1 = tmp %*% Y;\n+ tmp = rbind(tmp, X[i,]);\n+ while(FALSE) {}\n+ R[1,i] = sum(Res1);\n+ sum = sum + sum(Res1);\n+}\n+\n+write(R, $3, format=\"text\");\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/lineage/RewriteTest9.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1);\n+Y = read($2);\n+\n+sum = 0;\n+tmp = Y[,1];\n+R = matrix(0, 1, ncol(Y));\n+\n+for (i in 2:ncol(Y)) {\n+ Res1 = X %*% tmp;\n+ tmp = cbind(tmp, Y[,i]);\n+ while(FALSE) {}\n+ R[1,i] = sum(Res1);\n+ sum = sum + sum(Res1);\n+}\n+\n+write(R, $3, format=\"text\");\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/io/TensorReaderTextCell.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/io/TensorReaderTextCell.java", "diff": "@@ -69,18 +69,17 @@ public class TensorReaderTextCell extends TensorReader {\nret = new TensorBlock(schema, idims).allocateBlock();\ntry {\n- FastStringTokenizer st = new FastStringTokenizer(' ');\n-\nint[] ix = new int[dims.length];\nfor (InputSplit split : splits) {\nRecordReader<LongWritable, Text> reader = informat.getRecordReader(split, job, Reporter.NULL);\ntry {\nwhile (reader.next(key, value)) {\n- st.reset(value.toString());\n+ String[] parts = Arrays.stream(IOUtilFunctions.splitCSV(value.toString(), \" \"))\n+ .filter(s -> !s.isEmpty()).toArray(String[]::new);\nfor (int i = 0; i < ix.length; i++) {\n- ix[i] = st.nextInt() - 1;\n+ ix[i] = Integer.parseInt(parts[i]) - 1;\n}\n- ret.set(ix, st.nextToken());\n+ ret.set(ix, parts[ix.length]);\n}\n}\nfinally {\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/io/TensorWriterTextCell.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/io/TensorWriterTextCell.java", "diff": "@@ -20,6 +20,7 @@ package org.tugraz.sysds.runtime.io;\nimport org.apache.hadoop.fs.FileSystem;\nimport org.apache.hadoop.fs.Path;\nimport org.apache.hadoop.mapred.JobConf;\n+import org.tugraz.sysds.common.Types.ValueType;\nimport org.tugraz.sysds.conf.ConfigurationManager;\nimport org.tugraz.sysds.runtime.data.TensorBlock;\nimport org.tugraz.sysds.runtime.util.HDFSTool;\n@@ -63,11 +64,12 @@ public class TensorWriterTextCell extends TensorWriter {\nint[] ix = new int[dims.length];\nfor (long i = 0; i < src.getLength(); i++) {\nObject obj = src.get(ix);\n+ ValueType vt = src.isBasic() ? src.getValueType() : src.getSchema()[ix[1]];\nboolean skip;\n- if (!src.isBasic())\n- skip = UtilFunctions.objectToDouble(src.getSchema()[ix[1]], obj) == 0.0;\n+ if( vt == ValueType.STRING )\n+ skip = obj == null || ((String) obj).isEmpty();\nelse\n- skip = UtilFunctions.objectToDouble(src.getValueType(), obj) == 0.0;\n+ skip = UtilFunctions.objectToDouble(vt, obj) == 0.0;\nif (!skip) {\nfor (int j : ix)\nsb.append(j + 1).append(' ');\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/TestUtils.java", "new_path": "src/test/java/org/tugraz/sysds/test/TestUtils.java", "diff": "@@ -369,8 +369,7 @@ public class TestUtils\nAssert.assertEquals(tb1.getNumColumns(), tb2.getNumColumns());\nfor (int i = 0; i < tb1.getNumRows(); i++)\nfor (int j = 0; j < tb1.getNumColumns(); j++)\n- Assert.assertEquals(Double.valueOf(tb1.get(i, j)),\n- Double.valueOf(tb2.get(i, j)));\n+ Assert.assertEquals(tb1.get(new int[]{i, j}), tb2.get(new int[]{i, j}));\n}\npublic static TensorBlock createBasicTensor(ValueType vt, int rows, int cols, double sparsity) {\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/data/TensorTextCellTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/data/TensorTextCellTest.java", "diff": "@@ -55,7 +55,11 @@ public class TensorTextCellTest {\n@Test\npublic void testReadWriteTextCellBasicTensorString() {\n- testReadWriteTextCellBasicTensor(ValueType.STRING);\n+ TensorBlock tb1 = createBasicTensor(ValueType.STRING, 70, 30, 0.7);\n+ tb1.set(new int[]{0, 0}, \"\\\" f f \\\"\");\n+ tb1.set(new int[]{69, 29}, \"respect\");\n+ TensorBlock tb2 = writeAndReadBasicTensorTextCell(tb1);\n+ compareTensorBlocks(tb1, tb2);\n}\nprivate static void testReadWriteTextCellBasicTensor(ValueType vt) {\n@@ -91,7 +95,11 @@ public class TensorTextCellTest {\n@Test\npublic void testReadWriteTextCellDataTensorString() {\n- testReadWriteTextCellDataTensor(ValueType.STRING);\n+ TensorBlock tb1 = createDataTensor(ValueType.STRING, 70, 30, 0.7);\n+ tb1.set(new int[]{0, 0}, \"\\\" f f \\\"\");\n+ tb1.set(new int[]{69, 29}, \"respect\");\n+ TensorBlock tb2 = writeAndReadDataTensorTextCell(tb1);\n+ compareTensorBlocks(tb1, tb2);\n}\nprivate static void testReadWriteTextCellDataTensor(ValueType vt) {\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/BasicTensorBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/BasicTensorBlock.java", "diff": "@@ -451,10 +451,10 @@ public class BasicTensorBlock implements Serializable {\n}\n/**\n- * Copy a part of another tensor\n+ * Copy a part of another <code>BasicTensorBlock</code>\n* @param lower lower index of elements to copy (inclusive)\n* @param upper upper index of elements to copy (exclusive)\n- * @param src source tensor\n+ * @param src source <code>BasicTensorBlock</code>\n*/\npublic void copy(int[] lower, int[] upper, BasicTensorBlock src) {\n// TODO consider sparse\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DataTensorBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DataTensorBlock.java", "diff": "@@ -342,6 +342,12 @@ public class DataTensorBlock implements Serializable {\n}\n}\n+ /**\n+ * Copy a part of another <code>DataTensorBlock</code>\n+ * @param lower lower index of elements to copy (inclusive)\n+ * @param upper upper index of elements to copy (exclusive)\n+ * @param src source <code>DataTensorBlock</code>\n+ */\npublic void copy(int[] lower, int[] upper, DataTensorBlock src) {\nint[] subLower = lower.clone();\nif (upper[1] == 0) {\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/TensorBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/TensorBlock.java", "diff": "@@ -35,6 +35,14 @@ import java.io.ObjectInput;\nimport java.io.ObjectOutput;\nimport java.util.Arrays;\n+/**\n+ * A <code>TensorBlock</code> is the most top level representation of a tensor. There are two types of data representation\n+ * which can be used: Basic/Homogeneous and Data/Heterogeneous\n+ * Basic supports only one <code>ValueType</code>, while Data supports multiple <code>ValueType</code>s along the column\n+ * axis.\n+ * The format determines if the <code>TensorBlock</code> uses a <code>BasicTensorBlock</code> or a <code>DataTensorBlock</code>\n+ * for storing the data.\n+ */\npublic class TensorBlock implements CacheBlock, Externalizable {\nprivate static final long serialVersionUID = -8768054067319422277L;\n@@ -51,44 +59,81 @@ public class TensorBlock implements CacheBlock, Externalizable {\nprivate DataTensorBlock _dataTensor = null;\nprivate BasicTensorBlock _basicTensor = null;\n+ /**\n+ * Create a <code>TensorBlock</code> with [0,0] dimension and homogeneous representation (aka. basic).\n+ */\npublic TensorBlock() {\nthis(DEFAULT_DIMS, true);\n}\n+ /**\n+ * Create a <code>TensorBlock</code> with the given dimensions and the given data representation (basic/data).\n+ * @param dims dimensions\n+ * @param basic true -> basic, false -> data\n+ */\npublic TensorBlock(int[] dims, boolean basic) {\n_dims = dims;\n_basic = basic;\n}\n+ /**\n+ * Create a basic <code>TensorBlock</code> with the given <code>ValueType</code> and the given dimensions.\n+ * @param vt value type\n+ * @param dims dimensions\n+ */\npublic TensorBlock(ValueType vt, int[] dims) {\nthis(dims, true);\n_basicTensor = new BasicTensorBlock(vt, dims, false);\n}\n+ /**\n+ * Create a data <code>TensorBlock</code> with the given schema and the given dimensions.\n+ * @param schema schema of the columns\n+ * @param dims dimensions\n+ */\npublic TensorBlock(ValueType[] schema, int[] dims) {\nthis(dims, false);\n_dataTensor = new DataTensorBlock(schema, dims);\n}\n+ /**\n+ * Create a [1,1] basic FP64 <code>TensorBlock</code> containing the given value.\n+ * @param value value to put inside\n+ */\npublic TensorBlock(double value) {\n_dims = new int[]{1, 1};\n_basicTensor = new BasicTensorBlock(value);\n}\n+ /**\n+ * Wrap the given <code>BasicTensorBlock</code> inside a <code>TensorBlock</code>.\n+ * @param basicTensor basic tensor block\n+ */\npublic TensorBlock(BasicTensorBlock basicTensor) {\nthis(basicTensor._dims, true);\n_basicTensor = basicTensor;\n}\n+ /**\n+ * Wrap the given <code>DataTensorBlock</code> inside a <code>TensorBlock</code>.\n+ * @param dataTensor basic tensor block\n+ */\npublic TensorBlock(DataTensorBlock dataTensor) {\nthis(dataTensor._dims, false);\n_dataTensor = dataTensor;\n}\n+ /**\n+ * Copy constructor\n+ * @param that <code>TensorBlock</code> to copy\n+ */\npublic TensorBlock(TensorBlock that) {\ncopy(that);\n}\n+ /**\n+ * Reset all cells to 0.\n+ */\npublic void reset() {\nif (_basic) {\nif (_basicTensor == null)\n@@ -102,6 +147,10 @@ public class TensorBlock implements CacheBlock, Externalizable {\n}\n}\n+ /**\n+ * Reset data with new dimensions.\n+ * @param dims new dimensions\n+ */\npublic void reset(int[] dims) {\n_dims = dims;\nif (_basic) {\n@@ -127,6 +176,10 @@ public class TensorBlock implements CacheBlock, Externalizable {\nreturn _dataTensor != null && _dataTensor.isAllocated();\n}\n+ /**\n+ * If data is not yet allocated, allocate.\n+ * @return this <code>TensorBlock</code>\n+ */\npublic TensorBlock allocateBlock() {\nif (_basic) {\nif (_basicTensor == null)\n@@ -149,6 +202,10 @@ public class TensorBlock implements CacheBlock, Externalizable {\nreturn _dataTensor;\n}\n+ /**\n+ * Get the <code>ValueType</code> if this <code>TensorBlock</code> is homogeneous.\n+ * @return <code>ValueType</code> if homogeneous, null otherwise\n+ */\npublic ValueType getValueType() {\nif (_basic)\nreturn _basicTensor == null ? DEFAULT_VTYPE : _basicTensor.getValueType();\n@@ -156,6 +213,10 @@ public class TensorBlock implements CacheBlock, Externalizable {\nreturn null;\n}\n+ /**\n+ * Get the schema if this <code>TensorBlock</code> is heterogeneous.\n+ * @return value type if heterogeneous, null otherwise\n+ */\npublic ValueType[] getSchema() {\nif (_basic)\nreturn null;\n@@ -351,8 +412,7 @@ public class TensorBlock implements CacheBlock, Externalizable {\n}\n/**\n- * Set a cell to the value given as an `Object`. The type is inferred by either the `schema` or `valueType`, depending\n- * if the `TensorBlock` is a `BasicTensor` or `DataTensor`.\n+ * Set a cell to the value given as an `Object`.\n* @param ix indexes in each dimension, starting with 0\n* @param v value to set\n*/\n@@ -435,6 +495,12 @@ public class TensorBlock implements CacheBlock, Externalizable {\nreturn this;\n}\n+ /**\n+ * Copy a part of another <code>TensorBlock</code>\n+ * @param lower lower index of elements to copy (inclusive)\n+ * @param upper upper index of elements to copy (exclusive)\n+ * @param src source <code>TensorBlock</code>\n+ */\npublic TensorBlock copy(int[] lower, int[] upper, TensorBlock src) {\nif (_basic) {\nif (src._basic) {\n@@ -477,6 +543,11 @@ public class TensorBlock implements CacheBlock, Externalizable {\nreturn size;\n}\n+ /**\n+ * Get the exact serialized size of a <code>BasicTensorBlock</code> if written by\n+ * <code>TensorBlock.writeBlockData(DataOutput,BasicTensorBlock)</code>.\n+ * @param bt <code>BasicTensorBlock</code>\n+ */\npublic long getExactBlockDataSerializedSize(BasicTensorBlock bt) {\n// nnz, BlockType\nlong size = 8 + 1;\n@@ -540,6 +611,12 @@ public class TensorBlock implements CacheBlock, Externalizable {\n}\n}\n+ /**\n+ * Write a <code>BasicTensorBlock</code>.\n+ * @param out output stream\n+ * @param bt source <code>BasicTensorBlock</code>\n+ * @throws IOException if writing with the output stream fails\n+ */\npublic void writeBlockData(DataOutput out, BasicTensorBlock bt) throws IOException {\nout.writeLong(bt.getNonZeros()); // nnz\nif (bt.isEmpty(false)) {\n@@ -608,6 +685,12 @@ public class TensorBlock implements CacheBlock, Externalizable {\n}\n}\n+ /**\n+ * Read a <code>BasicTensorBlock</code>.\n+ * @param in input stream\n+ * @param bt destination <code>BasicTensorBlock</code>\n+ * @throws IOException if reading with the input stream fails\n+ */\nprotected void readBlockData(DataInput in, BasicTensorBlock bt) throws IOException {\nbt._nnz = in.readLong();\nswitch (BlockType.values()[in.readByte()]) {\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/ForProgramBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/controlprogram/ForProgramBlock.java", "diff": "@@ -31,8 +31,10 @@ import org.tugraz.sysds.runtime.DMLScriptException;\nimport org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject.UpdateType;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.instructions.Instruction;\n+import org.tugraz.sysds.runtime.instructions.cp.CPOperand;\nimport org.tugraz.sysds.runtime.instructions.cp.IntObject;\nimport org.tugraz.sysds.runtime.instructions.cp.ScalarObject;\n+import org.tugraz.sysds.runtime.lineage.Lineage;\nimport org.tugraz.sysds.runtime.lineage.LineagePath;\npublic class ForProgramBlock extends ProgramBlock\n@@ -134,6 +136,10 @@ public class ForProgramBlock extends ProgramBlock\n//set iteration variable\nec.setVariable(_iterPredVar, iterVar);\n+ if (DMLScript.LINEAGE) {\n+ Lineage li = ec.getLineage();\n+ li.set(_iterPredVar, li.getOrCreate(new CPOperand(iterVar)));\n+ }\n//execute all child blocks\nfor (int i = 0; i < _childBlocks.size(); i++) {\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/CPOperand.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/CPOperand.java", "diff": "@@ -57,6 +57,13 @@ public class CPOperand\n_isLiteral = literal;\n}\n+ public CPOperand(ScalarObject so) {\n+ _name = so.getStringValue();\n+ _valueType = so.getValueType();\n+ _dataType = DataType.SCALAR;\n+ _isLiteral = true;\n+ }\n+\npublic CPOperand(CPOperand variable){\n_name = variable._name;\n_valueType = variable._valueType;\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageItem.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageItem.java", "diff": "@@ -138,7 +138,7 @@ public class LineageItem {\n}\nprivate boolean equalsLI(LineageItem that) {\n- if (isVisited())\n+ if (isVisited() || this == that)\nreturn true;\nboolean ret = _opcode.equals(that._opcode);\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "diff": "@@ -25,6 +25,7 @@ import org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.instructions.CPInstructionParser;\nimport org.tugraz.sysds.runtime.instructions.Instruction;\n+import org.tugraz.sysds.runtime.instructions.cp.BinaryMatrixMatrixCPInstruction;\nimport org.tugraz.sysds.runtime.instructions.cp.CPInstruction.CPType;\nimport org.tugraz.sysds.runtime.instructions.cp.ComputationCPInstruction;\nimport org.tugraz.sysds.runtime.instructions.cp.MMTSJCPInstruction;\n@@ -171,7 +172,8 @@ public class LineageCache {\n// TODO: Move this to the new class LineageCacheConfig and extend\nreturn inst.getOpcode().equalsIgnoreCase(\"tsmm\")\n|| inst.getOpcode().equalsIgnoreCase(\"ba+*\")\n- || inst.getOpcode().equalsIgnoreCase(\"*\")\n+ || (inst.getOpcode().equalsIgnoreCase(\"*\") &&\n+ inst instanceof BinaryMatrixMatrixCPInstruction) //TODO support scalar\n|| inst.getOpcode().equalsIgnoreCase(\"rightIndex\");\n}\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -87,6 +87,7 @@ SYSTEMDS-110 New Builtin Functions\n* 116 Builtin function for kmeans OK\n* 117 Builtin function for lm cross validation OK\n* 118 Builtin function for hyperparameter grid search with CVlm\n+ * 119 Builtin functions for l2svm and msvm\nSYSTEMDS-120 Performance Features\n* 121 Avoid spark context creation on parfor result merge OK\n" }, { "change_type": "ADD", "old_path": null, "new_path": "scripts/builtin/l2svm.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Modifications Copyright 2019 Graz University of Technology\n+#\n+# Licensed to the Apache Software Foundation (ASF) under one\n+# or more contributor license agreements. See the NOTICE file\n+# distributed with this work for additional information\n+# regarding copyright ownership. The ASF licenses this file\n+# to you under the Apache License, Version 2.0 (the\n+# \"License\"); you may not use this file except in compliance\n+# with the License. You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing,\n+# software distributed under the License is distributed on an\n+# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+# KIND, either express or implied. See the License for the\n+# specific language governing permissions and limitations\n+# under the License.\n+#\n+#-------------------------------------------------------------\n+\n+\n+# Builtin function Implements binary-class SVM with squared slack variables\n+#\n+# INPUT PARAMETERS:\n+# ---------------------------------------------------------------------------------------------\n+# NAME TYPE DEFAULT MEANING\n+# ---------------------------------------------------------------------------------------------\n+# X Double --- matrix X of feature vectors\n+# Y Double --- matrix Y of class labels\n+# intercept Boolean False No Intercept ( If set to TRUE then a constant bias column is added to X)\n+# epsilon Double 0.001 Procedure terminates early if the reduction in objective function\n+# value is less than epsilon (tolerance) times the initial objective function value.\n+# lambda Double 1.0 Regularization parameter (lambda) for L2 regularization\n+# maxiterations Int 100 Maximum number of conjugate gradient iterations\n+# ---------------------------------------------------------------------------------------------\n+\n+\n+#Output(s)\n+# ---------------------------------------------------------------------------------------------\n+# NAME TYPE DEFAULT MEANING\n+# ---------------------------------------------------------------------------------------------\n+# model Double --- model matrix\n+\n+\n+m_l2svm = function(Matrix[Double] X, Matrix[Double] Y, Boolean intercept = FALSE,\n+Double epsilon = 0.001, Double lambda = 1, Integer maxiterations = 100, Boolean verbose = FALSE)\n+ return(Matrix[Double] model)\n+{\n+\n+ #check input parameter assertions\n+ if(nrow(X) < 2)\n+ stop(\"Stopping due to invalid inputs: Not possible to learn a binary class classifier without at least 2 rows\")\n+ if(epsilon < 0)\n+ stop(\"Stopping due to invalid argument: Tolerance (tol) must be non-negative\")\n+ if(lambda < 0)\n+ stop(\"Stopping due to invalid argument: Regularization constant (reg) must be non-negative\")\n+ if(maxiterations < 1)\n+ stop(\"Stopping due to invalid argument: Maximum iterations should be a positive integer\")\n+\n+ #check input lables and transform into -1/1\n+ check_min = min(Y)\n+ check_max = max(Y)\n+\n+ num_min = sum(Y == check_min)\n+ num_max = sum(Y == check_max)\n+\n+\n+ if(num_min + num_max != nrow(Y)) print(\"please check Y, it should contain only 2 labels\")\n+ else{\n+ if(check_min != -1 | check_max != +1)\n+ Y = 2/(check_max - check_min)*Y - (check_min + check_max)/(check_max - check_min)\n+ }\n+\n+ if(verbose) print('running L2-SVM ');\n+\n+ num_samples = nrow(X)\n+ dimensions = ncol(X)\n+\n+ if (intercept) {\n+ ones = matrix(1, rows=num_samples, cols=1)\n+ X = cbind(X, ones);\n+ }\n+\n+ num_rows_in_w = dimensions\n+ if(intercept){\n+ num_rows_in_w = num_rows_in_w + 1\n+ }\n+ w = matrix(0, rows=num_rows_in_w, cols=1)\n+\n+ g_old = t(X) %*% Y\n+ s = g_old\n+\n+ Xw = matrix(0, rows=nrow(X), cols=1)\n+\n+ iter = 0\n+ continue = 1\n+ while(continue == 1 & iter < maxiterations) {\n+ # minimizing primal obj along direction s\n+ step_sz = 0\n+ Xd = X %*% s\n+ wd = lambda * sum(w * s)\n+ dd = lambda * sum(s * s)\n+ continue1 = 1\n+ while(continue1 == 1){\n+ tmp_Xw = Xw + step_sz*Xd\n+ out = 1 - Y * (tmp_Xw)\n+ sv = (out > 0)\n+ out = out * sv\n+ g = wd + step_sz*dd - sum(out * Y * Xd)\n+ h = dd + sum(Xd * sv * Xd)\n+ step_sz = step_sz - g/h\n+ continue1 = (g*g/h >= epsilon)\n+ }\n+\n+ #update weights\n+ w = w + step_sz*s\n+ Xw = Xw + step_sz*Xd\n+\n+ out = 1 - Y * Xw\n+ sv = (out > 0)\n+ out = sv * out\n+ obj = 0.5 * sum(out * out) + lambda/2 * sum(w * w)\n+ g_new = t(X) %*% (out * Y) - lambda * w\n+\n+\n+ if(verbose) print(\"Iter, Obj \"+ iter + \", \"+obj)\n+\n+ tmp = sum(s * g_old)\n+ if(step_sz*tmp < epsilon*obj){\n+ continue = 0\n+ }\n+\n+ #non-linear CG step\n+ be = sum(g_new * g_new)/sum(g_old * g_old)\n+ s = be * s + g_new\n+ g_old = g_new\n+\n+ iter = iter + 1\n+ }\n+\n+ model = w\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "new_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "diff": "@@ -90,6 +90,8 @@ public enum Builtins {\nINVERSE(\"inv\", \"inverse\", false),\nIQM(\"interQuartileMean\", false),\nKMEANS(\"kmeans\", true),\n+ L2SVM( \"l2svm\", true),\n+ MULTISVM( \"multisvm\", true),\nLENGTH(\"length\", false),\nLINEAGE(\"lineage\", false),\nLIST(\"list\", false), //note: builtin and parbuiltin\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinL2SVMTest.java", "diff": "+/*\n+ * Copyright 2018 Graz University of Technology\n+ *\n+ * Licensed under the Apache License, Version 2.0 (the \"License\");\n+ * you may not use this file except in compliance with the License.\n+ * You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing, software\n+ * distributed under the License is distributed on an \"AS IS\" BASIS,\n+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+ * See the License for the specific language governing permissions and\n+ * limitations under the License.\n+ */\n+\n+\n+package org.tugraz.sysds.test.functions.builtin;\n+\n+import java.util.HashMap;\n+\n+import org.junit.Test;\n+import org.tugraz.sysds.api.DMLScript;\n+import org.tugraz.sysds.common.Types;\n+import org.tugraz.sysds.hops.OptimizerUtils;\n+import org.tugraz.sysds.lops.LopProperties;\n+import org.tugraz.sysds.runtime.matrix.data.MatrixValue;\n+import org.tugraz.sysds.test.AutomatedTestBase;\n+import org.tugraz.sysds.test.TestConfiguration;\n+import org.tugraz.sysds.test.TestUtils;\n+\n+\n+public class BuiltinL2SVMTest extends AutomatedTestBase {\n+ private final static String TEST_NAME = \"l2svm\";\n+ private final static String TEST_DIR = \"functions/builtin/\";\n+ private static final String TEST_CLASS_DIR = TEST_DIR + BuiltinL2SVMTest.class.getSimpleName() + \"/\";\n+\n+\n+ private final static double eps = 0.001;\n+ private final static int rows = 1000;\n+ private final static int colsX = 500;\n+ private final static double spSparse = 0.01;\n+ private final static double spDense = 0.7;\n+ private final static int max_iter = 10;\n+\n+\n+ @Override\n+ public void setUp() {\n+ TestUtils.clearAssertionInformation();\n+ addTestConfiguration(TEST_NAME,new TestConfiguration(TEST_CLASS_DIR, TEST_NAME,new String[]{\"C\"}));\n+ }\n+\n+ @Test\n+ public void testL2SVMDense() {\n+ runL2SVMTest(false, false, eps, 1.0, max_iter, LopProperties.ExecType.CP);\n+ }\n+ @Test\n+ public void testL2SVMSparse() {\n+ runL2SVMTest(true, false, eps, 1.0, max_iter, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testL2SVMIntercept() {\n+ runL2SVMTest(true,true, eps, 1.0, max_iter, LopProperties.ExecType.SPARK);\n+ }\n+\n+ @Test\n+ public void testL2SVMDenseIntercept() {\n+ runL2SVMTest(false,true, 1, 1.0, max_iter, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testL2SVMSparseLambda2() {\n+ runL2SVMTest(true,true, 1, 2.0, max_iter, LopProperties.ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testL2SVMSparseLambda100CP() {\n+ runL2SVMTest(true,true, 1, 100, max_iter, LopProperties.ExecType.CP);\n+ }\n+ @Test\n+ public void testL2SVMSparseLambda100Spark() {\n+ runL2SVMTest(true,true, 1, 100, max_iter, LopProperties.ExecType.SPARK);\n+ }\n+\n+ @Test\n+ public void testL2SVMIteration() {\n+ runL2SVMTest(true,true, 1, 2.0, 100, LopProperties.ExecType.CP);\n+ }\n+\n+ private void runL2SVMTest(boolean sparse, boolean intercept, double eps,\n+ double lambda, int run, LopProperties.ExecType instType)\n+ {\n+ Types.ExecMode platformOld = setExecMode(instType);\n+\n+ boolean oldFlag = OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION;\n+ boolean sparkConfigOld = DMLScript.USE_LOCAL_SPARK_CONFIG;\n+\n+ try\n+ {\n+ loadTestConfiguration(getTestConfiguration(TEST_NAME));\n+\n+ double sparsity = sparse ? spSparse : spDense;\n+\n+ String HOME = SCRIPT_DIR + TEST_DIR;\n+\n+ fullDMLScriptName = HOME + TEST_NAME + \".dml\";\n+ programArgs = new String[]{ \"-explain\", \"-stats\",\n+ \"-nvargs\", \"X=\" + input(\"X\"), \"Y=\" + input(\"Y\"), \"model=\" + output(\"model\"),\n+ \"inc=\" + String.valueOf(intercept).toUpperCase(),\"eps=\" + eps, \"lam=\" + lambda, \"max=\" + run};\n+\n+\n+ fullRScriptName = HOME + TEST_NAME + \".R\";\n+ rCmd = getRCmd(inputDir(), Boolean.toString(intercept), Double.toString(eps),\n+ Double.toString(lambda), Integer.toString(run), expectedDir());\n+\n+ //generate actual datasets\n+ double[][] X = getRandomMatrix(rows, colsX, 0, 100, sparsity, 10);\n+ double[][] Y= getRandomMatrix(rows, 1, -1, 1, 1, -1);\n+\n+ Y = TestUtils.round(Y);\n+\n+ writeInputMatrixWithMTD(\"X\", X, true);\n+ writeInputMatrixWithMTD(\"Y\", Y, true);\n+\n+ runTest(true, false, null, -1);\n+ runRScript(true);\n+\n+ HashMap<MatrixValue.CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(\"model\");\n+ HashMap<MatrixValue.CellIndex, Double> rfile = readRMatrixFromFS(\"model\");\n+ TestUtils.compareMatrices(dmlfile, rfile, eps, \"Stat-DML\", \"Stat-R\");\n+\n+ }\n+ finally {\n+ rtplatform = platformOld;\n+ DMLScript.USE_LOCAL_SPARK_CONFIG = sparkConfigOld;\n+ OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = oldFlag;\n+ OptimizerUtils.ALLOW_AUTO_VECTORIZATION = true;\n+ OptimizerUtils.ALLOW_OPERATOR_FUSION = true;\n+ }\n+ }\n+}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/builtin/l2svm.R", "diff": "+#-------------------------------------------------------------\n+#\n+# Licensed to the Apache Software Foundation (ASF) under one\n+# or more contributor license agreements. See the NOTICE file\n+# distributed with this work for additional information\n+# regarding copyright ownership. The ASF licenses this file\n+# to you under the Apache License, Version 2.0 (the\n+# \"License\"); you may not use this file except in compliance\n+# with the License. You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing,\n+# software distributed under the License is distributed on an\n+# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+# KIND, either express or implied. See the License for the\n+# specific language governing permissions and limitations\n+# under the License.\n+#\n+#-------------------------------------------------------------\n+\n+# JUnit test class: dml.test.integration.applications.L2SVMTest.java\n+# command line invocation assuming $L2SVM_HOME is set to the home of the R script\n+# Rscript $L2SVM_HOME/L2SVM.R $L2SVM_HOME/in/ 0.00000001 1 100 $L2SVM_HOME/expected/\n+\n+args <- commandArgs(TRUE)\n+library(\"Matrix\")\n+\n+X = readMM(paste(args[1], \"X.mtx\", sep=\"\"));\n+Y = readMM(paste(args[1], \"Y.mtx\", sep=\"\"));\n+\n+check_min = min(Y)\n+check_max = max(Y)\n+num_min = sum(Y == check_min)\n+num_max = sum(Y == check_max)\n+if(num_min + num_max != nrow(Y)){\n+ print(\"please check Y, it should contain only 2 labels\")\n+}else{\n+ if(check_min != -1 | check_max != +1)\n+ Y = 2/(check_max - check_min)*Y - (check_min + check_max)/(check_max - check_min)\n+}\n+\n+intercept = as.logical(args[2]);\n+epsilon = as.double(args[3]);\n+lambda = as.double(args[4]);\n+maxiterations = as.integer(args[5]);\n+\n+N = nrow(X)\n+D = ncol(X)\n+\n+if (intercept) {\n+ ones = matrix(1,N,1)\n+ X = cbind(X, ones);\n+}\n+\n+num_rows_in_w = D\n+if(intercept){\n+ num_rows_in_w = num_rows_in_w + 1\n+}\n+w = matrix(0, num_rows_in_w, 1)\n+\n+g_old = t(X) %*% Y\n+s = g_old\n+\n+Xw = matrix(0,nrow(X),1)\n+iter = 0\n+continue = TRUE\n+while(continue && iter < maxiterations){\n+ t = 0\n+ Xd = X %*% s\n+ wd = lambda * sum(w * s)\n+ dd = lambda * sum(s * s)\n+ continue1 = TRUE\n+ while(continue1){\n+ tmp_Xw = Xw + t*Xd\n+ out = 1 - Y * (tmp_Xw)\n+ sv = which(out > 0)\n+ g = wd + t*dd - sum(out[sv] * Y[sv] * Xd[sv])\n+ h = dd + sum(Xd[sv] * Xd[sv])\n+ t = t - g/h\n+ continue1 = (g*g/h >= epsilon)\n+ }\n+\n+ w = w + t*s\n+ Xw = Xw + t*Xd\n+\n+ out = 1 - Y * (X %*% w)\n+ sv = which(out > 0)\n+ obj = 0.5 * sum(out[sv] * out[sv]) + lambda/2 * sum(w * w)\n+ g_new = t(X[sv,]) %*% (out[sv] * Y[sv]) - lambda * w\n+\n+ #print(paste(\"OBJ : \", obj))\n+\n+ continue = (t*sum(s * g_old) >= epsilon*obj)\n+\n+ be = sum(g_new * g_new)/sum(g_old * g_old)\n+ s = be * s + g_new\n+ g_old = g_new\n+\n+ iter = iter + 1\n+}\n+\n+writeMM(as(w,\"CsparseMatrix\"), paste(args[6], \"model\", sep=\"\"));\n+\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/builtin/l2svm.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($X)\n+Y = read($Y)\n+model= l2svm(X=X, Y=Y, intercept = $inc, epsilon = $eps, lambda = $lam, maxiterations = $max )\n+write(model, $model)\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/DataGenOp.java", "new_path": "src/main/java/org/tugraz/sysds/hops/DataGenOp.java", "diff": "@@ -96,10 +96,7 @@ public class DataGenOp extends MultiThreadedHop\n_op = mthd;\n//ensure all parameters existing and consistent with data type\n- //TODO remove once this unnecessary parameter is cleaned up\n- if( !inputParameters.containsKey(DataExpression.RAND_TENSOR) )\n- inputParameters.put(DataExpression.RAND_TENSOR, new LiteralOp(false));\n- else if (HopRewriteUtils.isLiteralOfValue(inputParameters.get(DataExpression.RAND_TENSOR), true))\n+ if( inputParameters.containsKey(DataExpression.RAND_DIMS) )\nsetDataType(DataType.TENSOR);\nint index = 0;\n@@ -115,7 +112,7 @@ public class DataGenOp extends MultiThreadedHop\nHop sparsityOp = inputParameters.get(DataExpression.RAND_SPARSITY);\nif ( mthd == DataGenMethod.RAND && sparsityOp instanceof LiteralOp)\n- _sparsity = Double.valueOf(((LiteralOp)sparsityOp).getName());\n+ _sparsity = HopRewriteUtils.getDoubleValue((LiteralOp)sparsityOp);\n//generate base dir\nString scratch = ConfigurationManager.getScratchSpace();\n@@ -215,7 +212,7 @@ public class DataGenOp extends MultiThreadedHop\n@Override\nprotected double computeOutputMemEstimate( long dim1, long dim2, long nnz )\n{\n- double ret = 0;\n+ double ret;\nif ( _op == DataGenMethod.RAND && _sparsity != -1 ) {\nif( hasConstantValue(0.0) ) { //if empty block\n@@ -253,12 +250,18 @@ public class DataGenOp extends MultiThreadedHop\nif( (_op == DataGenMethod.RAND || _op == DataGenMethod.SINIT ) &&\nOptimizerUtils.ALLOW_WORSTCASE_SIZE_EXPRESSION_EVALUATION )\n{\n+ if (_paramIndexMap.containsKey(DataExpression.RAND_DIMS)) {\n+ // TODO size information for tensors\n+ return null;\n+ }\n+ else {\nlong dim1 = computeDimParameterInformation(getInput().get(_paramIndexMap.get(DataExpression.RAND_ROWS)), memo);\nlong dim2 = computeDimParameterInformation(getInput().get(_paramIndexMap.get(DataExpression.RAND_COLS)), memo);\nlong nnz = _sparsity >= 0 ? (long) (_sparsity * dim1 * dim2) : -1;\nif( dim1 >= 0 && dim2 >= 0 )\nreturn new MatrixCharacteristics(dim1, dim2, -1, nnz);\n}\n+ }\nelse if ( _op == DataGenMethod.SEQ )\n{\nHop from = getInput().get(_paramIndexMap.get(Statement.SEQ_FROM));\n@@ -454,7 +457,7 @@ public class DataGenOp extends MultiThreadedHop\n//sparsity awareness if requires\nif( ret && val != 0 ) {\nHop sparsity = getInput().get(_paramIndexMap.get(DataExpression.RAND_SPARSITY)); //sparsity\n- ret &= (sparsity == null || sparsity instanceof LiteralOp\n+ ret = (sparsity == null || sparsity instanceof LiteralOp\n&& HopRewriteUtils.getDoubleValueSafe((LiteralOp) sparsity) == 1);\n}\n@@ -522,8 +525,7 @@ public class DataGenOp extends MultiThreadedHop\nfor( Entry<String,Integer> e : _paramIndexMap.entrySet() ) {\nString key1 = e.getKey();\nint pos1 = e.getValue();\n- int pos2 = that2._paramIndexMap.containsKey(key1) ?\n- that2._paramIndexMap.get(key1) : -1;\n+ int pos2 = that2._paramIndexMap.getOrDefault(key1, -1);\nret &= ( pos2 >=0 && that2.getInput().get(pos2)!=null\n&& getInput().get(pos1) == that2.getInput().get(pos2) );\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/UnaryOp.java", "new_path": "src/main/java/org/tugraz/sysds/hops/UnaryOp.java", "diff": "@@ -254,7 +254,7 @@ public class UnaryOp extends MultiThreadedHop\n//special case single row block (no offsets needed)\nif( rlen > 0 && clen > 0 && rlen <= blen ) {\nLop offset = HopRewriteUtils.createDataGenOpByVal(new LiteralOp(1), new LiteralOp(clen),\n- new LiteralOp(\"1 1\"), DataType.MATRIX, ValueType.FP64, getCumulativeInitValue()).constructLops();\n+ null, DataType.MATRIX, ValueType.FP64, getCumulativeInitValue()).constructLops();\nreturn constructCumOffBinary(X, offset, aggtype, rlen, clen, blen);\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/recompile/Recompiler.java", "new_path": "src/main/java/org/tugraz/sysds/hops/recompile/Recompiler.java", "diff": "@@ -1327,6 +1327,8 @@ public class Recompiler\n|| d.getOp() == DataGenMethod.SAMPLE )\n{\nboolean initUnknown = !d.dimsKnown();\n+ // TODO refresh tensor size information\n+ if (params.containsKey(DataExpression.RAND_ROWS) && params.containsKey(DataExpression.RAND_COLS)) {\nint ix1 = params.get(DataExpression.RAND_ROWS);\nint ix2 = params.get(DataExpression.RAND_COLS);\n//update rows/cols by evaluating simple expression of literals, nrow, ncol, scalars, binaryops\n@@ -1336,6 +1338,7 @@ public class Recompiler\nif( !(initUnknown & d.dimsKnown()) )\nd.refreshSizeInformation();\n}\n+ }\nelse if ( d.getOp() == DataGenMethod.SEQ )\n{\nboolean initUnknown = !d.dimsKnown();\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/rewrite/HopRewriteUtils.java", "new_path": "src/main/java/org/tugraz/sysds/hops/rewrite/HopRewriteUtils.java", "diff": "@@ -307,7 +307,6 @@ public class HopRewriteUtils\nHashMap<String, Hop> params = new HashMap<>();\nparams.put(DataExpression.RAND_ROWS, rows);\nparams.put(DataExpression.RAND_COLS, cols);\n- params.put(DataExpression.RAND_DIMS, new LiteralOp(\"-1\")); //TODO\nparams.put(DataExpression.RAND_MIN, val);\nparams.put(DataExpression.RAND_MAX, val);\nparams.put(DataExpression.RAND_PDF, new LiteralOp(DataExpression.RAND_PDF_UNIFORM));\n@@ -337,11 +336,8 @@ public class HopRewriteUtils\npublic static DataGenOp copyDataGenOp( DataGenOp inputGen, double scale, double shift )\n{\nHashMap<String, Integer> params = inputGen.getParamIndexMap();\n- Hop rows = inputGen.getInput().get(params.get(DataExpression.RAND_ROWS));\n- Hop cols = inputGen.getInput().get(params.get(DataExpression.RAND_COLS));\nHop min = inputGen.getInput().get(params.get(DataExpression.RAND_MIN));\nHop max = inputGen.getInput().get(params.get(DataExpression.RAND_MAX));\n- Hop dims = inputGen.getInput().get(params.get(DataExpression.RAND_DIMS));\nHop pdf = inputGen.getInput().get(params.get(DataExpression.RAND_PDF));\nHop mean = inputGen.getInput().get(params.get(DataExpression.RAND_LAMBDA));\nHop sparsity = inputGen.getInput().get(params.get(DataExpression.RAND_SPARSITY));\n@@ -361,11 +357,18 @@ public class HopRewriteUtils\nHop smaxHop = new LiteralOp(smax);\nHashMap<String, Hop> params2 = new HashMap<>();\n+ if( !params.containsKey(DataExpression.RAND_DIMS) ) {\n+ Hop rows = inputGen.getInput().get(params.get(DataExpression.RAND_ROWS));\n+ Hop cols = inputGen.getInput().get(params.get(DataExpression.RAND_COLS));\nparams2.put(DataExpression.RAND_ROWS, rows);\nparams2.put(DataExpression.RAND_COLS, cols);\n+ }\n+ else {\n+ Hop dims = inputGen.getInput().get(params.get(DataExpression.RAND_DIMS));\n+ params2.put(DataExpression.RAND_DIMS, dims);\n+ }\nparams2.put(DataExpression.RAND_MIN, sminHop);\nparams2.put(DataExpression.RAND_MAX, smaxHop);\n- params2.put(DataExpression.RAND_DIMS, dims);\nparams2.put(DataExpression.RAND_PDF, pdf);\nparams2.put(DataExpression.RAND_LAMBDA, mean);\nparams2.put(DataExpression.RAND_SPARSITY, sparsity);\n@@ -393,7 +396,6 @@ public class HopRewriteUtils\nHashMap<String, Hop> params = new HashMap<>();\nparams.put(DataExpression.RAND_ROWS, rows);\nparams.put(DataExpression.RAND_COLS, cols);\n- params.put(DataExpression.RAND_DIMS, new LiteralOp(\"-1\")); //TODO\nparams.put(DataExpression.RAND_MIN, val);\nparams.put(DataExpression.RAND_MAX, val);\nparams.put(DataExpression.RAND_PDF, new LiteralOp(DataExpression.RAND_PDF_UNIFORM));\n@@ -428,7 +430,6 @@ public class HopRewriteUtils\nparams.put(DataExpression.RAND_COLS, cols);\nparams.put(DataExpression.RAND_MIN, val);\nparams.put(DataExpression.RAND_MAX, val);\n- params.put(DataExpression.RAND_DIMS, new LiteralOp(\"-1\")); //TODO\nparams.put(DataExpression.RAND_PDF, new LiteralOp(DataExpression.RAND_PDF_UNIFORM));\nparams.put(DataExpression.RAND_LAMBDA,new LiteralOp(-1.0));\nparams.put(DataExpression.RAND_SPARSITY, new LiteralOp(1.0));\n@@ -449,16 +450,18 @@ public class HopRewriteUtils\nHop val = new LiteralOp(value);\nHashMap<String, Hop> params = new HashMap<>();\n+ if (dt.isTensor())\n+ params.put(DataExpression.RAND_DIMS, dimsInput);\n+ else {\nparams.put(DataExpression.RAND_ROWS, rowInput);\nparams.put(DataExpression.RAND_COLS, colInput);\n- params.put(DataExpression.RAND_DIMS, dimsInput);\n+ }\nparams.put(DataExpression.RAND_MIN, val);\nparams.put(DataExpression.RAND_MAX, val);\nparams.put(DataExpression.RAND_PDF, new LiteralOp(DataExpression.RAND_PDF_UNIFORM));\nparams.put(DataExpression.RAND_LAMBDA, new LiteralOp(-1.0));\nparams.put(DataExpression.RAND_SPARSITY, new LiteralOp(1.0));\nparams.put(DataExpression.RAND_SEED, new LiteralOp(DataGenOp.UNSPECIFIED_SEED) );\n- params.put(DataExpression.RAND_TENSOR, new LiteralOp(false));\n//note internal refresh size information\nDataIdentifier di = new DataIdentifier(\"tmp\");\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/rewrite/RewriteAlgebraicSimplificationDynamic.java", "new_path": "src/main/java/org/tugraz/sysds/hops/rewrite/RewriteAlgebraicSimplificationDynamic.java", "diff": "@@ -214,7 +214,7 @@ public class RewriteAlgebraicSimplificationDynamic extends HopRewriteRule\n{\n//remove unnecessary right indexing\nHop hnew = HopRewriteUtils.createDataGenOpByVal( new LiteralOp(hi.getDim1()),\n- new LiteralOp(hi.getDim2()), new LiteralOp(\"1 1\"), DataType.MATRIX, ValueType.FP64, 0);\n+ new LiteralOp(hi.getDim2()), null, DataType.MATRIX, ValueType.FP64, 0);\nHopRewriteUtils.replaceChildReference(parent, hi, hnew, pos);\nHopRewriteUtils.cleanupUnreferenced(hi, input);\nhi = hnew;\n@@ -824,7 +824,7 @@ public class RewriteAlgebraicSimplificationDynamic extends HopRewriteRule\nelse //diagm2v TODO support tensor operation\nhnew = HopRewriteUtils.createDataGenOpByVal(\nHopRewriteUtils.createValueHop(input,true), new LiteralOp(1),\n- new LiteralOp(\"1 1\"), DataType.MATRIX, ValueType.FP64, 0);\n+ null, DataType.MATRIX, ValueType.FP64, 0);\n}\n}\nelse if( rhi.getOp() == ReOrgOp.RESHAPE )\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/lops/DataGen.java", "new_path": "src/main/java/org/tugraz/sysds/lops/DataGen.java", "diff": "@@ -132,7 +132,8 @@ public class DataGen extends Lop\n//sanity checks\nif ( method != DataGenMethod.RAND )\nthrow new LopsException(\"Invalid instruction generation for data generation method \" + method);\n- if( getInputs().size() != DataExpression.RAND_VALID_PARAM_NAMES.length ) {\n+ if( getInputs().size() != DataExpression.RAND_VALID_PARAM_NAMES.length - 2 && // tensor\n+ getInputs().size() != DataExpression.RAND_VALID_PARAM_NAMES.length - 1 ) { // matrix\nthrow new LopsException(printErrorLocation() + \"Invalid number of operands (\"\n+ getInputs().size() + \") for a Rand operation\");\n}\n@@ -145,17 +146,20 @@ public class DataGen extends Lop\nsb.append(RAND_OPCODE);\nsb.append(OPERAND_DELIMITOR);\n- Lop iLop = _inputParams.get(DataExpression.RAND_ROWS);\n+ Lop iLop = _inputParams.get(DataExpression.RAND_DIMS);\n+ if (iLop != null) {\nsb.append(iLop.prepScalarInputOperand(getExecType()));\nsb.append(OPERAND_DELIMITOR);\n-\n- iLop = _inputParams.get(DataExpression.RAND_COLS);\n+ }\n+ else {\n+ iLop = _inputParams.get(DataExpression.RAND_ROWS);\nsb.append(iLop.prepScalarInputOperand(getExecType()));\nsb.append(OPERAND_DELIMITOR);\n- iLop = _inputParams.get(DataExpression.RAND_DIMS);\n+ iLop = _inputParams.get(DataExpression.RAND_COLS);\nsb.append(iLop.prepScalarInputOperand(getExecType()));\nsb.append(OPERAND_DELIMITOR);\n+ }\nsb.append(getOutputParameters().getBlocksize());\nsb.append(OPERAND_DELIMITOR);\n@@ -219,9 +223,6 @@ public class DataGen extends Lop\niLop = _inputParams.get(DataExpression.RAND_COLS);\nString colsString = iLop.prepScalarLabel();\n- iLop = _inputParams.get(DataExpression.RAND_DIMS);\n- String dimsString = iLop.prepScalarLabel();\n-\nString blen = String.valueOf(getOutputParameters().getBlocksize());\niLop = _inputParams.get(DataExpression.RAND_MIN);\n@@ -243,8 +244,6 @@ public class DataGen extends Lop\nsb.append(OPERAND_DELIMITOR);\nsb.append(colsString);\nsb.append(OPERAND_DELIMITOR);\n- sb.append(dimsString);\n- sb.append(OPERAND_DELIMITOR);\nsb.append(blen);\nsb.append(OPERAND_DELIMITOR);\nsb.append(minString);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/parser/DataExpression.java", "new_path": "src/main/java/org/tugraz/sysds/parser/DataExpression.java", "diff": "@@ -53,7 +53,6 @@ import java.util.Map.Entry;\npublic class DataExpression extends DataIdentifier\n{\npublic static final String RAND_DIMS = \"dims\";\n- public static final String RAND_TENSOR = \"istensor\";\npublic static final String RAND_ROWS = \"rows\";\npublic static final String RAND_COLS = \"cols\";\n@@ -103,8 +102,7 @@ public class DataExpression extends DataIdentifier\npublic static final String DELIM_SPARSE = \"sparse\"; // applicable only for write\npublic static final String[] RAND_VALID_PARAM_NAMES =\n- {RAND_ROWS, RAND_COLS, RAND_DIMS, RAND_MIN, RAND_MAX, RAND_SPARSITY, RAND_SEED, RAND_PDF, RAND_LAMBDA,\n- RAND_TENSOR}; //FIXME: why is this istensor required at all\n+ {RAND_ROWS, RAND_COLS, RAND_DIMS, RAND_MIN, RAND_MAX, RAND_SPARSITY, RAND_SEED, RAND_PDF, RAND_LAMBDA};\npublic static final String[] RESHAPE_VALID_PARAM_NAMES =\n{ RAND_BY_ROW, RAND_DIMNAMES, RAND_DATA, RAND_ROWS, RAND_COLS, RAND_DIMS};\n@@ -497,26 +495,15 @@ public class DataExpression extends DataIdentifier\npublic void setMatrixDefault(){\nif (getVarParam(RAND_BY_ROW) == null)\naddVarParam(RAND_BY_ROW, new BooleanIdentifier(true, this));\n- if (getVarParam(RAND_DIMS) == null) {\n- StringIdentifier id = new StringIdentifier(\"1 1\", this);\n- addVarParam(RAND_DIMS, id);\n- }\n}\npublic void setTensorDefault(){\nif (getVarParam(RAND_BY_ROW) == null)\naddVarParam(RAND_BY_ROW, new BooleanIdentifier(true, this));\n- if (getVarParam(RAND_ROWS) == null) {\n- IntIdentifier id = new IntIdentifier(1L, this);\n- addVarParam(RAND_ROWS, id);\n- }\n- if (getVarParam(RAND_COLS) == null) {\n- IntIdentifier id = new IntIdentifier(1L, this);\n- addVarParam(RAND_COLS, id);\n- }\n}\npublic void setRandDefault() {\n+ if (getVarParam(RAND_DIMS) == null) {\nif( getVarParam(RAND_ROWS) == null ) {\nIntIdentifier id = new IntIdentifier(1L, this);\naddVarParam(RAND_ROWS, id);\n@@ -525,9 +512,6 @@ public class DataExpression extends DataIdentifier\nIntIdentifier id = new IntIdentifier(1L, this);\naddVarParam(RAND_COLS, id);\n}\n- if (getVarParam(RAND_DIMS) == null) {\n- StringIdentifier id = new StringIdentifier(\"1 1\", this);\n- addVarParam(RAND_DIMS, id);\n}\nif (getVarParam(RAND_MIN) == null) {\nDoubleIdentifier id = new DoubleIdentifier(0.0, this);\n@@ -553,10 +537,6 @@ public class DataExpression extends DataIdentifier\nDoubleIdentifier id = new DoubleIdentifier(1.0, this);\naddVarParam(RAND_LAMBDA, id);\n}\n- if (getVarParam(RAND_TENSOR) == null) {\n- BooleanIdentifier id = new BooleanIdentifier(false, this);\n- addVarParam(RAND_TENSOR, id);\n- }\n}\npublic void setOpCode(DataOp op) {\n@@ -677,9 +657,6 @@ public class DataExpression extends DataIdentifier\n// replace DataOp MATRIX with RAND -- Rand handles matrix generation for Scalar values\n// replace data parameter with min / max within Rand case below\n- // TODO either use dims or rows, cols depending on datatype\n- BooleanIdentifier id = new BooleanIdentifier(_opcode == DataOp.TENSOR, this);\n- addVarParam(RAND_TENSOR, id);\nthis.setOpCode(DataOp.RAND);\n}\n@@ -1221,12 +1198,7 @@ public class DataExpression extends DataIdentifier\nraiseValidateError(\"for Rand statement \" + RAND_SEED + \" has incorrect value type\", conditional);\n}\n- boolean isTensorOperation = false;\n- if (!(getVarParam(RAND_TENSOR) instanceof BooleanIdentifier)) {\n- raiseValidateError(\"for Rand statement \" + RAND_TENSOR + \" has incorrect value type\", conditional);\n- } else {\n- isTensorOperation = ((BooleanIdentifier)getVarParam(RAND_TENSOR)).getValue();\n- }\n+ boolean isTensorOperation = getVarParam(RAND_DIMS) != null;\nif ((getVarParam(RAND_MAX) instanceof StringIdentifier && !_strInit) ||\n(getVarParam(RAND_MAX) instanceof BooleanIdentifier && !isTensorOperation)) {\n@@ -1253,10 +1225,12 @@ public class DataExpression extends DataIdentifier\nlong rowsLong = -1L, colsLong = -1L;\n+ Expression rowsExpr = getVarParam(RAND_ROWS);\n+ Expression colsExpr = getVarParam(RAND_COLS);\n+ if (!isTensorOperation) {\n///////////////////////////////////////////////////////////////////\n// HANDLE ROWS\n///////////////////////////////////////////////////////////////////\n- Expression rowsExpr = getVarParam(RAND_ROWS);\nif( rowsExpr instanceof IntIdentifier ) {\nif( ((IntIdentifier) rowsExpr).getValue() < 0 ) {\nraiseValidateError(\"In rand statement, can only assign rows a long \" +\n@@ -1322,8 +1296,6 @@ public class DataExpression extends DataIdentifier\n///////////////////////////////////////////////////////////////////\n// HANDLE COLUMNS\n///////////////////////////////////////////////////////////////////\n-\n- Expression colsExpr = getVarParam(RAND_COLS);\nif( colsExpr instanceof IntIdentifier ) {\nif( ((IntIdentifier) colsExpr).getValue() < 0 ) {\nraiseValidateError(\"In rand statement, can only assign cols a long \" +\n@@ -1384,12 +1356,7 @@ public class DataExpression extends DataIdentifier\n// handle general expression\ncolsExpr.validateExpression(ids, currConstVars, conditional);\n}\n-\n- ///////////////////////////////////////////////////////////////////\n- // HANDLE DIMS\n- ///////////////////////////////////////////////////////////////////\n- // TODO handle dims for rand\n-\n+ }\n///////////////////////////////////////////////////////////////////\n// HANDLE MIN\n///////////////////////////////////////////////////////////////////\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/DataGenCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/DataGenCPInstruction.java", "diff": "@@ -176,7 +176,7 @@ public class DataGenCPInstruction extends UnaryCPInstruction {\nif ( opcode.equalsIgnoreCase(DataGen.RAND_OPCODE) ) {\nmethod = DataGenMethod.RAND;\n- InstructionUtils.checkNumFields ( s, 12 );\n+ InstructionUtils.checkNumFields ( s, 10, 11 );\n}\nelse if ( opcode.equalsIgnoreCase(DataGen.SEQ_OPCODE) ) {\nmethod = DataGenMethod.SEQ;\n@@ -199,21 +199,29 @@ public class DataGenCPInstruction extends UnaryCPInstruction {\nif ( method == DataGenMethod.RAND )\n{\n- CPOperand rows = new CPOperand(s[1]);\n- CPOperand cols = new CPOperand(s[2]);\n- CPOperand dims = new CPOperand(s[3]);\n- int blen = Integer.parseInt(s[4]);\n- double sparsity = !s[7].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n- Double.valueOf(s[7]) : -1;\n- long seed = !s[SEED_POSITION_RAND].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n- Long.valueOf(s[SEED_POSITION_RAND]) : -1;\n- String pdf = s[9];\n- String pdfParams = !s[10].contains( Lop.VARIABLE_NAME_PLACEHOLDER) ?\n- s[10] : null;\n- int k = Integer.parseInt(s[11]);\n+ int missing; // number of missing params (row & cols or dims)\n+ CPOperand rows = null, cols = null, dims = null;\n+ if (s.length == 12) {\n+ missing = 1;\n+ rows = new CPOperand(s[1]);\n+ cols = new CPOperand(s[2]);\n+ }\n+ else {\n+ missing = 2;\n+ dims = new CPOperand(s[1]);\n+ }\n+ int blen = Integer.parseInt(s[4 - missing]);\n+ double sparsity = !s[7 - missing].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ Double.parseDouble(s[7 - missing]) : -1;\n+ long seed = !s[SEED_POSITION_RAND - missing].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ Long.parseLong(s[SEED_POSITION_RAND - missing]) : -1;\n+ String pdf = s[9 - missing];\n+ String pdfParams = !s[10 - missing].contains( Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ s[10 - missing] : null;\n+ int k = Integer.parseInt(s[11 - missing]);\nreturn new DataGenCPInstruction(op, method, null, out, rows, cols, dims, blen,\n- s[5], s[6], sparsity, seed, pdf, pdfParams, k, opcode, str);\n+ s[5 - missing], s[6 - missing], sparsity, seed, pdf, pdfParams, k, opcode, str);\n}\nelse if ( method == DataGenMethod.SEQ)\n{\n@@ -253,9 +261,12 @@ public class DataGenCPInstruction extends UnaryCPInstruction {\n//process specific datagen operator\nif ( method == DataGenMethod.RAND ) {\n- long lrows = ec.getScalarInput(rows).getLongValue();\n- long lcols = ec.getScalarInput(cols).getLongValue();\n+ long lrows = -1, lcols = -1;\n+ if (dims == null) {\n+ lrows = ec.getScalarInput(rows).getLongValue();\n+ lcols = ec.getScalarInput(cols).getLongValue();\ncheckValidDimensions(lrows, lcols);\n+ }\n//generate pseudo-random seed (because not specified)\nlong lSeed = seed; //seed per invocation\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/StringInitCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/StringInitCPInstruction.java", "diff": "@@ -60,13 +60,12 @@ public class StringInitCPInstruction extends UnaryCPInstruction {\nthrow new DMLRuntimeException(\"Unsupported opcode: \"+opcode);\n//parse instruction\nString[] s = InstructionUtils.getInstructionPartsWithValueType ( str );\n- InstructionUtils.checkNumFields( s, 6 );\n+ InstructionUtils.checkNumFields( s, 5 );\nCPOperand out = new CPOperand(s[s.length-1]); // output is specified by the last operand\nlong rows = (s[1].contains( Lop.VARIABLE_NAME_PLACEHOLDER)?-1:Double.valueOf(s[1]).longValue());\nlong cols = (s[2].contains( Lop.VARIABLE_NAME_PLACEHOLDER)?-1:Double.valueOf(s[2]).longValue());\n- // Ignore dims\n- int blen = Integer.parseInt(s[4]);\n- String data = s[5];\n+ int blen = Integer.parseInt(s[3]);\n+ String data = s[4];\nreturn new StringInitCPInstruction(null, null, out, rows, cols, blen, data, opcode, str);\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/RandSPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/spark/RandSPInstruction.java", "diff": "@@ -189,7 +189,7 @@ public class RandSPInstruction extends UnarySPInstruction {\nDataGenMethod method = DataGenMethod.INVALID;\nif ( opcode.equalsIgnoreCase(DataGen.RAND_OPCODE) ) {\nmethod = DataGenMethod.RAND;\n- InstructionUtils.checkNumFields ( str, 12 );\n+ InstructionUtils.checkNumFields ( str, 10, 11 );\n}\nelse if ( opcode.equalsIgnoreCase(DataGen.SEQ_OPCODE) ) {\nmethod = DataGenMethod.SEQ;\n@@ -207,21 +207,29 @@ public class RandSPInstruction extends UnarySPInstruction {\nCPOperand out = new CPOperand(s[s.length-1]);\nif ( method == DataGenMethod.RAND ) {\n- CPOperand rows = new CPOperand(s[1]);\n- CPOperand cols = new CPOperand(s[2]);\n- CPOperand dims = new CPOperand(s[3]);\n- int blen = Integer.parseInt(s[4]);\n- double sparsity = !s[7].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n- Double.valueOf(s[7]).doubleValue() : -1;\n- long seed = !s[8].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n- Long.valueOf(s[8]).longValue() : -1;\n- String dir = s[9];\n- String pdf = s[10];\n- String pdfParams = !s[11].contains( Lop.VARIABLE_NAME_PLACEHOLDER) ?\n- s[11] : null;\n+ int missing; // number of missing params (row & cols or dims)\n+ CPOperand rows = null, cols = null, dims = null;\n+ if (s.length == 12) {\n+ missing = 1;\n+ rows = new CPOperand(s[1]);\n+ cols = new CPOperand(s[2]);\n+ }\n+ else {\n+ missing = 2;\n+ dims = new CPOperand(s[1]);\n+ }\n+ int blen = Integer.parseInt(s[4 - missing]);\n+ double sparsity = !s[7 - missing].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ Double.parseDouble(s[7 - missing]) : -1;\n+ long seed = !s[8 - missing].contains(Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ Long.parseLong(s[8 - missing]) : -1;\n+ String dir = s[9 - missing];\n+ String pdf = s[10 - missing];\n+ String pdfParams = !s[11 - missing].contains( Lop.VARIABLE_NAME_PLACEHOLDER) ?\n+ s[11 - missing] : null;\nreturn new RandSPInstruction(op, method, null, out, rows, cols, dims,\n- blen, s[5], s[6], sparsity, seed, dir, pdf, pdfParams, opcode, str);\n+ blen, s[5 - missing], s[6 - missing], sparsity, seed, dir, pdf, pdfParams, opcode, str);\n}\nelse if ( method == DataGenMethod.SEQ) {\nint blen = Integer.parseInt(s[3]);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageItemUtils.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageItemUtils.java", "diff": "@@ -176,9 +176,12 @@ public class LineageItemUtils {\nif (inst instanceof DataGenCPInstruction) {\nDataGenCPInstruction rand = (DataGenCPInstruction) inst;\nHashMap<String, Hop> params = new HashMap<>();\n+ if( rand.output.getDataType() == DataType.TENSOR)\n+ params.put(DataExpression.RAND_DIMS, new LiteralOp(rand.getDims()));\n+ else {\nparams.put(DataExpression.RAND_ROWS, new LiteralOp(rand.getRows()));\nparams.put(DataExpression.RAND_COLS, new LiteralOp(rand.getCols()));\n- params.put(DataExpression.RAND_DIMS, new LiteralOp(rand.getDims()));\n+ }\nparams.put(DataExpression.RAND_MIN, new LiteralOp(rand.getMinValue()));\nparams.put(DataExpression.RAND_MAX, new LiteralOp(rand.getMaxValue()));\nparams.put(DataExpression.RAND_PDF, new LiteralOp(rand.getPdf()));\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/data/TensorRandTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/data/TensorRandTest.java", "diff": "@@ -93,9 +93,8 @@ public class TensorRandTest extends AutomatedTestBase {\nString HOME = SCRIPT_DIR + TEST_DIR;\nfullDMLScriptName = HOME + testName + \".dml\";\n- StringBuilder dimensionsStringBuilder = new StringBuilder();\n- Arrays.stream(dimensions).forEach((dim) -> dimensionsStringBuilder.append(dim).append(\" \"));\n- String dimensionsString = dimensionsStringBuilder.toString();\n+ String dimensionsString = Arrays.toString(dimensions).replace(\",\", \"\");\n+ dimensionsString = dimensionsString.substring(1, dimensionsString.length() - 1);\nprogramArgs = new String[]{\"-explain\", \"-args\", dimensionsString, min, max, Integer.toString(seed)};\n" }, { "change_type": "MODIFY", "old_path": "src/test/scripts/functions/data/RandTensorTest.dml", "new_path": "src/test/scripts/functions/data/RandTensorTest.dml", "diff": "#\n#-------------------------------------------------------------\n-A = rand(dims=$1, min=$2, max=$3, seed=$4, istensor=TRUE)\n+A = rand(dims=$1, min=$2, max=$3, seed=$4)\nprint(toString(A))\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageDedupBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageDedupBlock.java", "diff": "package org.tugraz.sysds.runtime.lineage;\n-import org.tugraz.sysds.runtime.controlprogram.*;\n+import org.tugraz.sysds.runtime.controlprogram.BasicProgramBlock;\n+import org.tugraz.sysds.runtime.controlprogram.IfProgramBlock;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport java.util.ArrayList;\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageDedupUtils.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageDedupUtils.java", "diff": "package org.tugraz.sysds.runtime.lineage;\nimport org.tugraz.sysds.runtime.DMLRuntimeException;\n-import org.tugraz.sysds.runtime.controlprogram.*;\n+import org.tugraz.sysds.runtime.controlprogram.BasicProgramBlock;\n+import org.tugraz.sysds.runtime.controlprogram.ForProgramBlock;\n+import org.tugraz.sysds.runtime.controlprogram.IfProgramBlock;\n+import org.tugraz.sysds.runtime.controlprogram.ProgramBlock;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\npublic class LineageDedupUtils {\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/component/tensor/TensorSerializationTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/component/tensor/TensorSerializationTest.java", "diff": "@@ -25,9 +25,7 @@ import org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.caching.CacheDataInput;\nimport org.tugraz.sysds.runtime.controlprogram.caching.CacheDataOutput;\nimport org.tugraz.sysds.runtime.data.TensorBlock;\n-\n-import static org.tugraz.sysds.test.TestUtils.*;\n-\n+import org.tugraz.sysds.test.TestUtils;\npublic class TensorSerializationTest\n{\n@@ -82,15 +80,15 @@ public class TensorSerializationTest\n}\nprivate static void testSerializeBasicTensor(ValueType vt) {\n- TensorBlock tb1 = createBasicTensor(vt, 70, 30, 0.7);\n+ TensorBlock tb1 = TestUtils.createBasicTensor(vt, 70, 30, 0.7);\nTensorBlock tb2 = serializeAndDeserializeTensorBlock(tb1);\n- compareTensorBlocks(tb1, tb2);\n+ TestUtils.compareTensorBlocks(tb1, tb2);\n}\nprivate static void testSerializeDataTensor(ValueType vt) {\n- TensorBlock tb1 = createDataTensor(vt, 70, 30, 0.7);\n+ TensorBlock tb1 = TestUtils.createDataTensor(vt, 70, 30, 0.7);\nTensorBlock tb2 = serializeAndDeserializeTensorBlock(tb1);\n- compareTensorBlocks(tb1, tb2);\n+ TestUtils.compareTensorBlocks(tb1, tb2);\n}\nprivate static TensorBlock serializeAndDeserializeTensorBlock(TensorBlock tb1) {\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/data/TensorBinaryBlockParallelTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/data/TensorBinaryBlockParallelTest.java", "diff": "@@ -23,9 +23,7 @@ import org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.data.TensorBlock;\nimport org.tugraz.sysds.runtime.io.TensorReaderBinaryBlockParallel;\nimport org.tugraz.sysds.runtime.io.TensorWriterBinaryBlockParallel;\n-\n-import static org.tugraz.sysds.test.TestUtils.*;\n-\n+import org.tugraz.sysds.test.TestUtils;\npublic class TensorBinaryBlockParallelTest {\nstatic final String FILENAME = \"target/testTemp/functions/data/TensorBinaryBlockParallelTest/tensor\";\n@@ -61,9 +59,9 @@ public class TensorBinaryBlockParallelTest {\n}\nprivate static void testReadWriteBinaryBlockParallelBasicTensor(ValueType vt) {\n- TensorBlock tb1 = createBasicTensor(vt, 70, 3000, 0.7);\n+ TensorBlock tb1 = TestUtils.createBasicTensor(vt, 70, 3000, 0.7);\nTensorBlock tb2 = writeAndReadBasicTensorBinaryBlockParallel(tb1);\n- compareTensorBlocks(tb1, tb2);\n+ TestUtils.compareTensorBlocks(tb1, tb2);\n}\n@Test\n@@ -97,9 +95,9 @@ public class TensorBinaryBlockParallelTest {\n}\nprivate static void testReadWriteBinaryBlockParallelDataTensor(ValueType vt) {\n- TensorBlock tb1 = createDataTensor(vt, 70, 3000, 0.7);\n+ TensorBlock tb1 = TestUtils.createDataTensor(vt, 70, 3000, 0.7);\nTensorBlock tb2 = writeAndReadDataTensorBinaryBlockParallel(tb1);\n- compareTensorBlocks(tb1, tb2);\n+ TestUtils.compareTensorBlocks(tb1, tb2);\n}\nprivate static TensorBlock writeAndReadBasicTensorBinaryBlockParallel(TensorBlock tb1) {\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/data/TensorBinaryBlockTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/data/TensorBinaryBlockTest.java", "diff": "@@ -23,9 +23,7 @@ import org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.data.TensorBlock;\nimport org.tugraz.sysds.runtime.io.TensorReaderBinaryBlock;\nimport org.tugraz.sysds.runtime.io.TensorWriterBinaryBlock;\n-\n-import static org.tugraz.sysds.test.TestUtils.*;\n-\n+import org.tugraz.sysds.test.TestUtils;\npublic class TensorBinaryBlockTest {\nstatic final String FILENAME = \"target/testTemp/functions/data/TensorBinaryBlockTest/tensor\";\n@@ -61,9 +59,9 @@ public class TensorBinaryBlockTest {\n}\nprivate static void testReadWriteBinaryBlockBasicTensor(ValueType vt) {\n- TensorBlock tb1 = createBasicTensor(vt, 70, 3000, 0.7);\n+ TensorBlock tb1 = TestUtils.createBasicTensor(vt, 70, 3000, 0.7);\nTensorBlock tb2 = writeAndReadBasicTensorBinaryBlock(tb1);\n- compareTensorBlocks(tb1, tb2);\n+ TestUtils.compareTensorBlocks(tb1, tb2);\n}\n@Test\n@@ -97,9 +95,9 @@ public class TensorBinaryBlockTest {\n}\nprivate static void testReadWriteBinaryBlockDataTensor(ValueType vt) {\n- TensorBlock tb1 = createDataTensor(vt, 70, 3000, 0.7);\n+ TensorBlock tb1 = TestUtils.createDataTensor(vt, 70, 3000, 0.7);\nTensorBlock tb2 = writeAndReadDataTensorBinaryBlock(tb1);\n- compareTensorBlocks(tb1, tb2);\n+ TestUtils.compareTensorBlocks(tb1, tb2);\n}\nprivate static TensorBlock writeAndReadBasicTensorBinaryBlock(TensorBlock tb1) {\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/data/TensorTextCellParallelTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/data/TensorTextCellParallelTest.java", "diff": "@@ -23,9 +23,7 @@ import org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.data.TensorBlock;\nimport org.tugraz.sysds.runtime.io.TensorReaderTextCellParallel;\nimport org.tugraz.sysds.runtime.io.TensorWriterTextCellParallel;\n-\n-import static org.tugraz.sysds.test.TestUtils.*;\n-\n+import org.tugraz.sysds.test.TestUtils;\npublic class TensorTextCellParallelTest {\nstatic final String FILENAME = \"target/testTemp/functions/data/TensorTextCellParallelTest/tensor\";\n@@ -61,9 +59,9 @@ public class TensorTextCellParallelTest {\n}\nprivate static void testReadWriteTextCellParallelBasicTensor(ValueType vt) {\n- TensorBlock tb1 = createBasicTensor(vt, 70, 3000, 0.7);\n+ TensorBlock tb1 = TestUtils.createBasicTensor(vt, 70, 3000, 0.7);\nTensorBlock tb2 = writeAndReadBasicTensorTextCellParallel(tb1);\n- compareTensorBlocks(tb1, tb2);\n+ TestUtils.compareTensorBlocks(tb1, tb2);\n}\n@Test\n@@ -97,9 +95,9 @@ public class TensorTextCellParallelTest {\n}\nprivate static void testReadWriteTextCellParallelDataTensor(ValueType vt) {\n- TensorBlock tb1 = createDataTensor(vt, 70, 3000, 0.7);\n+ TensorBlock tb1 = TestUtils.createDataTensor(vt, 70, 3000, 0.7);\nTensorBlock tb2 = writeAndReadDataTensorTextCellParallel(tb1);\n- compareTensorBlocks(tb1, tb2);\n+ TestUtils.compareTensorBlocks(tb1, tb2);\n}\nprivate static TensorBlock writeAndReadBasicTensorTextCellParallel(TensorBlock tb1) {\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/data/TensorTextCellTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/data/TensorTextCellTest.java", "diff": "@@ -23,9 +23,7 @@ import org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.data.TensorBlock;\nimport org.tugraz.sysds.runtime.io.TensorReaderTextCell;\nimport org.tugraz.sysds.runtime.io.TensorWriterTextCell;\n-\n-import static org.tugraz.sysds.test.TestUtils.*;\n-\n+import org.tugraz.sysds.test.TestUtils;\npublic class TensorTextCellTest {\nstatic final String FILENAME = \"target/testTemp/functions/data/TensorTextCellTest/tensor\";\n@@ -57,17 +55,17 @@ public class TensorTextCellTest {\n@Test\npublic void testReadWriteTextCellBasicTensorString() {\n- TensorBlock tb1 = createBasicTensor(ValueType.STRING, 70, 30, 0.7);\n+ TensorBlock tb1 = TestUtils.createBasicTensor(ValueType.STRING, 70, 30, 0.7);\ntb1.set(new int[]{0, 0}, \"\\\" f f \\\"\");\ntb1.set(new int[]{69, 29}, \"respect\");\nTensorBlock tb2 = writeAndReadBasicTensorTextCell(tb1);\n- compareTensorBlocks(tb1, tb2);\n+ TestUtils.compareTensorBlocks(tb1, tb2);\n}\nprivate static void testReadWriteTextCellBasicTensor(ValueType vt) {\n- TensorBlock tb1 = createBasicTensor(vt, 70, 3000, 0.7);\n+ TensorBlock tb1 = TestUtils.createBasicTensor(vt, 70, 3000, 0.7);\nTensorBlock tb2 = writeAndReadBasicTensorTextCell(tb1);\n- compareTensorBlocks(tb1, tb2);\n+ TestUtils.compareTensorBlocks(tb1, tb2);\n}\n@Test\n@@ -97,17 +95,17 @@ public class TensorTextCellTest {\n@Test\npublic void testReadWriteTextCellDataTensorString() {\n- TensorBlock tb1 = createDataTensor(ValueType.STRING, 70, 30, 0.7);\n+ TensorBlock tb1 = TestUtils.createDataTensor(ValueType.STRING, 70, 30, 0.7);\ntb1.set(new int[]{0, 0}, \"\\\" f f \\\"\");\ntb1.set(new int[]{69, 29}, \"respect\");\nTensorBlock tb2 = writeAndReadDataTensorTextCell(tb1);\n- compareTensorBlocks(tb1, tb2);\n+ TestUtils.compareTensorBlocks(tb1, tb2);\n}\nprivate static void testReadWriteTextCellDataTensor(ValueType vt) {\n- TensorBlock tb1 = createDataTensor(vt, 70, 3000, 0.7);\n+ TensorBlock tb1 = TestUtils.createDataTensor(vt, 70, 3000, 0.7);\nTensorBlock tb2 = writeAndReadDataTensorTextCell(tb1);\n- compareTensorBlocks(tb1, tb2);\n+ TestUtils.compareTensorBlocks(tb1, tb2);\n}\nprivate static TensorBlock writeAndReadBasicTensorTextCell(TensorBlock tb1) {\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -77,6 +77,7 @@ SYSTEMDS-100 Various Fixes\n* 106 Fix correctness of as.integer for negative numbers OK\n* 107 Fix correctness IPA check dimension-preserving OK\n* 108 Fix codegen optimizer (early costing abort) OK\n+ * 109 Fix update-in-place w/ udf function calls OK\nSYSTEMDS-110 New Builtin Functions\n* 111 Time builtin function for script-level measurements OK\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/FunctionOp.java", "new_path": "src/main/java/org/tugraz/sysds/hops/FunctionOp.java", "diff": "package org.tugraz.sysds.hops;\nimport java.util.ArrayList;\n+import java.util.Arrays;\nimport java.util.List;\nimport org.tugraz.sysds.api.DMLScript;\n@@ -123,6 +124,11 @@ public class FunctionOp extends Hop\nreturn _outputNames;\n}\n+ public boolean containsOutput(String varname) {\n+ return Arrays.stream(getOutputVariableNames())\n+ .anyMatch(outName -> outName.equals(varname));\n+ }\n+\npublic FunctionType getFunctionType() {\nreturn _type;\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/rewrite/RewriteMarkLoopVariablesUpdateInPlace.java", "new_path": "src/main/java/org/tugraz/sysds/hops/rewrite/RewriteMarkLoopVariablesUpdateInPlace.java", "diff": "@@ -26,11 +26,12 @@ import java.util.List;\nimport org.tugraz.sysds.api.DMLScript;\nimport org.tugraz.sysds.common.Types.ExecMode;\nimport org.tugraz.sysds.hops.DataOp;\n+import org.tugraz.sysds.hops.FunctionOp;\nimport org.tugraz.sysds.hops.Hop;\n-import org.tugraz.sysds.hops.LeftIndexingOp;\n-import org.tugraz.sysds.hops.UnaryOp;\nimport org.tugraz.sysds.hops.Hop.DataOpTypes;\nimport org.tugraz.sysds.hops.Hop.OpOp1;\n+import org.tugraz.sysds.hops.LeftIndexingOp;\n+import org.tugraz.sysds.hops.UnaryOp;\nimport org.tugraz.sysds.parser.ForStatement;\nimport org.tugraz.sysds.parser.ForStatementBlock;\nimport org.tugraz.sysds.parser.IfStatement;\n@@ -128,7 +129,13 @@ public class RewriteMarkLoopVariablesUpdateInPlace extends StatementBlockRewrite\nreturn ret;\n}\n- private static boolean isApplicableForUpdateInPlace(Hop hop, String varname) {\n+ private static boolean isApplicableForUpdateInPlace(Hop hop, String varname)\n+ {\n+ // check erroneously marking a variable for update-in-place\n+ // that is written to by a function return value\n+ if(hop instanceof FunctionOp && ((FunctionOp)hop).containsOutput(varname))\n+ return false;\n+\n//NOTE: single-root-level validity check\nif( !hop.getName().equals(varname) )\nreturn true;\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/indexing/LeftIndexingUpdateInPlaceTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/indexing/LeftIndexingUpdateInPlaceTest.java", "diff": "@@ -34,7 +34,9 @@ import org.tugraz.sysds.test.TestUtils;\npublic class LeftIndexingUpdateInPlaceTest extends AutomatedTestBase\n{\nprivate final static String TEST_DIR = \"functions/indexing/\";\n- private final static String TEST_NAME = \"LeftIndexingUpdateInPlaceTest\";\n+ private final static String TEST_NAME1 = \"LeftIndexingUpdateInPlaceTest\";\n+ private final static String TEST_NAME2 = \"LeftIndexingUpdateInPlaceTestBugTest1\";\n+ private final static String TEST_NAME3 = \"LeftIndexingUpdateInPlaceTestBugTest2\";\nprivate final static String TEST_CLASS_DIR = TEST_DIR + LeftIndexingUpdateInPlaceTest.class.getSimpleName() + \"/\";\nprivate final static int rows1 = 1281;\n@@ -47,83 +49,88 @@ public class LeftIndexingUpdateInPlaceTest extends AutomatedTestBase\n@Override\npublic void setUp() {\n- addTestConfiguration(TEST_NAME, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME, new String[] {\"R\"}));\n+ addTestConfiguration(TEST_NAME1, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME1, new String[] {\"R\"}));\n+ addTestConfiguration(TEST_NAME2, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME2, new String[] {\"R\"}));\n+ addTestConfiguration(TEST_NAME3, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME3, new String[] {\"R\"}));\n}\n@Test\npublic void testSparseMatrixSparseMatrix() {\n- runLeftIndexingUpdateInPlaceTest(true, true, false, false);\n+ runLeftIndexingUpdateInPlaceTest(true, true, false, false, TEST_NAME1);\n}\n@Test\npublic void testSparseMatrixSparseVector() {\n- runLeftIndexingUpdateInPlaceTest(true, true, true, false);\n+ runLeftIndexingUpdateInPlaceTest(true, true, true, false, TEST_NAME1);\n}\n@Test\npublic void testSparseMatrixDenseMatrix() {\n- runLeftIndexingUpdateInPlaceTest(true, false, false, false);\n+ runLeftIndexingUpdateInPlaceTest(true, false, false, false, TEST_NAME1);\n}\n@Test\npublic void testSparseMatrixDenseVector() {\n- runLeftIndexingUpdateInPlaceTest(true, false, true, false);\n+ runLeftIndexingUpdateInPlaceTest(true, false, true, false, TEST_NAME1);\n}\n@Test\npublic void testDenseMatrixSparseMatrix() {\n- runLeftIndexingUpdateInPlaceTest(false, true, false, false);\n+ runLeftIndexingUpdateInPlaceTest(false, true, false, false, TEST_NAME1);\n}\n@Test\npublic void testDenseMatrixSparseVector() {\n- runLeftIndexingUpdateInPlaceTest(false, true, true, false);\n+ runLeftIndexingUpdateInPlaceTest(false, true, true, false, TEST_NAME1);\n}\n@Test\npublic void testDenseMatrixDenseMatrix() {\n- runLeftIndexingUpdateInPlaceTest(false, false, false, false);\n+ runLeftIndexingUpdateInPlaceTest(false, false, false, false, TEST_NAME1);\n}\n@Test\npublic void testDenseMatrixDenseVector() {\n- runLeftIndexingUpdateInPlaceTest(false, false, true, false);\n+ runLeftIndexingUpdateInPlaceTest(false, false, true, false, TEST_NAME1);\n}\n@Test\npublic void testSparseMatrixEmptyMatrix() {\n- runLeftIndexingUpdateInPlaceTest(true, true, false, true);\n+ runLeftIndexingUpdateInPlaceTest(true, true, false, true, TEST_NAME1);\n}\n@Test\npublic void testSparseMatrixEmptyVector() {\n- runLeftIndexingUpdateInPlaceTest(true, true, true, true);\n+ runLeftIndexingUpdateInPlaceTest(true, true, true, true, TEST_NAME1);\n}\n@Test\npublic void testDenseMatrixEmptyMatrix() {\n- runLeftIndexingUpdateInPlaceTest(false, true, false, true);\n+ runLeftIndexingUpdateInPlaceTest(false, true, false, true, TEST_NAME1);\n}\n@Test\npublic void testDenseMatrixEmptyVector() {\n- runLeftIndexingUpdateInPlaceTest(false, true, true, true);\n+ runLeftIndexingUpdateInPlaceTest(false, true, true, true, TEST_NAME1);\n}\n+ @Test\n+ public void testDenseMatrixDenseMatrixBugTest1() {\n+ runLeftIndexingUpdateInPlaceTest(false, false, false, false, TEST_NAME2);\n+ }\n- /**\n- *\n- * @param sparseM1\n- * @param sparseM2\n- * @param vectorM2\n- */\n- public void runLeftIndexingUpdateInPlaceTest(boolean sparseM1, boolean sparseM2, boolean vectorM2, boolean emptyM2)\n+ @Test\n+ public void testDenseMatrixDenseMatrixBugTest2() {\n+ runLeftIndexingUpdateInPlaceTest(false, false, false, false, TEST_NAME3);\n+ }\n+\n+ public void runLeftIndexingUpdateInPlaceTest(boolean sparseM1, boolean sparseM2, boolean vectorM2, boolean emptyM2, String testName)\n{\nExecMode oldRTP = rtplatform;\nrtplatform = ExecMode.HYBRID;\ntry {\n- TestConfiguration config = getTestConfiguration(TEST_NAME);\n+ TestConfiguration config = getTestConfiguration(testName);\nloadTestConfiguration(config);\ndouble spM1 = sparseM1 ? sparsity1 : sparsity2;\n@@ -131,10 +138,10 @@ public class LeftIndexingUpdateInPlaceTest extends AutomatedTestBase\nint colsM2 = vectorM2 ? cols3 : cols2;\nString HOME = SCRIPT_DIR + TEST_DIR;\n- fullDMLScriptName = HOME + TEST_NAME + \".dml\";\n+ fullDMLScriptName = HOME + testName + \".dml\";\nprogramArgs = new String[]{\"-explain\",\"-args\", input(\"A\"), input(\"B\"), output(\"R\")};\n- fullRScriptName = HOME + TEST_NAME + \".R\";\n+ fullRScriptName = HOME + testName + \".R\";\nrCmd = \"Rscript\" + \" \" + fullRScriptName + \" \" +\ninputDir() + \" \" + expectedDir();\n@@ -146,6 +153,7 @@ public class LeftIndexingUpdateInPlaceTest extends AutomatedTestBase\n//run dml and r script\nrunTest(true, false, null, 2); //2xrblk\n+ if(testName == TEST_NAME1) {\nrunRScript(true);\nHashMap<CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(\"R\");\n@@ -153,6 +161,7 @@ public class LeftIndexingUpdateInPlaceTest extends AutomatedTestBase\nTestUtils.compareMatrices(dmlfile, rfile, 0, \"DML\", \"R\");\ncheckDMLMetaDataFile(\"R\", new MatrixCharacteristics(rows1, cols1, 1, 1));\n}\n+ }\nfinally {\nrtplatform = oldRTP;\n}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/indexing/LeftIndexingUpdateInPlaceTestBugTest1.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+debug_print = function(String name, Matrix[Double] m) {\n+ print(name + \" ROWS=\" + nrow(m) + \" COLS=\" + ncol(m) + \"\\n\" + toString(m, decimal=10))\n+}\n+\n+bug_test = function(Matrix[Double] inputA) return (Matrix[Double] out) {\n+ out = inputA\n+ for ( i in 1:4 ) {\n+ debug_print(\"start inputA\", inputA[1:5, 1:5])\n+ out[,i] = inputA[,i] + 2 ^ i;\n+ debug_print(\"end inputA\", inputA[1:5, 1:5])\n+ }\n+}\n+\n+A = read($1)\n+B = read($2)\n+\n+for(i in 1:4) {\n+ A = bug_test(A)\n+ print(\"while loop iteration++\")\n+}\n+\n+debug_print(\"A\", A[1:5, 1:5])\n+debug_print(\"B\", B[1:5, 1:5])\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/indexing/LeftIndexingUpdateInPlaceTestBugTest2.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+debug_print = function(String name, Matrix[Double] m) {\n+ print(name + \" ROWS=\" + nrow(m) + \" COLS=\" + ncol(m) + \"\\n\" + toString(m, decimal=10))\n+}\n+\n+A = read($1)\n+B = read($2)\n+\n+# out will make A update-in-place from the second iteration of the outer loop\n+out = A\n+for(i in 1:4) {\n+ for ( i in 1:4 ) {\n+ debug_print(\"start A\", A[1:5, 1:5])\n+ out[,i] = A[,i] + 2 ^ i;\n+ debug_print(\"end A\", A[1:5, 1:5])\n+ }\n+ A = out\n+ print(\"while loop iteration++\")\n+}\n+\n+debug_print(\"A\", A[1:5, 1:5])\n+debug_print(\"B\", B[1:5, 1:5])\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/indexing/LeftIndexingUpdateInPlaceTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/indexing/LeftIndexingUpdateInPlaceTest.java", "diff": "@@ -34,9 +34,9 @@ import org.tugraz.sysds.test.TestUtils;\npublic class LeftIndexingUpdateInPlaceTest extends AutomatedTestBase\n{\nprivate final static String TEST_DIR = \"functions/indexing/\";\n- private final static String TEST_NAME1 = \"LeftIndexingUpdateInPlaceTest\";\n- private final static String TEST_NAME2 = \"LeftIndexingUpdateInPlaceTestBugTest1\";\n- private final static String TEST_NAME3 = \"LeftIndexingUpdateInPlaceTestBugTest2\";\n+ private final static String TEST_NAME1 = \"LeftIndexingUpdateInPlaceTest1\";\n+ private final static String TEST_NAME2 = \"LeftIndexingUpdateInPlaceTest2\";\n+ private final static String TEST_NAME3 = \"LeftIndexingUpdateInPlaceTest3\";\nprivate final static String TEST_CLASS_DIR = TEST_DIR + LeftIndexingUpdateInPlaceTest.class.getSimpleName() + \"/\";\nprivate final static int rows1 = 1281;\n@@ -153,14 +153,14 @@ public class LeftIndexingUpdateInPlaceTest extends AutomatedTestBase\n//run dml and r script\nrunTest(true, false, null, 2); //2xrblk\n- if(testName == TEST_NAME1) {\nrunRScript(true);\nHashMap<CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(\"R\");\nHashMap<CellIndex, Double> rfile = readRMatrixFromFS(\"R\");\nTestUtils.compareMatrices(dmlfile, rfile, 0, \"DML\", \"R\");\n- checkDMLMetaDataFile(\"R\", new MatrixCharacteristics(rows1, cols1, 1, 1));\n- }\n+ checkDMLMetaDataFile(\"R\", testName.equals(TEST_NAME1) ?\n+ new MatrixCharacteristics(rows1, cols1, 1, 1):\n+ new MatrixCharacteristics(1, 1, 1, 1));\n}\nfinally {\nrtplatform = oldRTP;\n" }, { "change_type": "RENAME", "old_path": "src/test/scripts/functions/indexing/LeftIndexingUpdateInPlaceTest.R", "new_path": "src/test/scripts/functions/indexing/LeftIndexingUpdateInPlaceTest1.R", "diff": "" }, { "change_type": "RENAME", "old_path": "src/test/scripts/functions/indexing/LeftIndexingUpdateInPlaceTest.dml", "new_path": "src/test/scripts/functions/indexing/LeftIndexingUpdateInPlaceTest1.dml", "diff": "" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/indexing/LeftIndexingUpdateInPlaceTest2.R", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+\n+args <- commandArgs(TRUE)\n+options(digits=22)\n+library(\"Matrix\")\n+\n+A = as.matrix(readMM(paste(args[1], \"A.mtx\", sep=\"\")))\n+\n+for(i in 1:4) {\n+ out = A\n+ for ( i in 1:4 ) {\n+ out[,i] = A[,i] + 2 ^ i;\n+ }\n+ A = out + A\n+}\n+\n+R = as.matrix(sum(A))\n+writeMM(as(R,\"CsparseMatrix\"), paste(args[2], \"R\", sep=\"\"))\n" }, { "change_type": "RENAME", "old_path": "src/test/scripts/functions/indexing/LeftIndexingUpdateInPlaceTestBugTest1.dml", "new_path": "src/test/scripts/functions/indexing/LeftIndexingUpdateInPlaceTest2.dml", "diff": "#\n#-------------------------------------------------------------\n-debug_print = function(String name, Matrix[Double] m) {\n- print(name + \" ROWS=\" + nrow(m) + \" COLS=\" + ncol(m) + \"\\n\" + toString(m, decimal=10))\n-}\n-\nbug_test = function(Matrix[Double] inputA) return (Matrix[Double] out) {\nout = inputA\nfor ( i in 1:4 ) {\n- debug_print(\"start inputA\", inputA[1:5, 1:5])\nout[,i] = inputA[,i] + 2 ^ i;\n- debug_print(\"end inputA\", inputA[1:5, 1:5])\n}\n+ out += inputA\n}\nA = read($1)\n-B = read($2)\nfor(i in 1:4) {\nA = bug_test(A)\n- print(\"while loop iteration++\")\n}\n-debug_print(\"A\", A[1:5, 1:5])\n-debug_print(\"B\", B[1:5, 1:5])\n+R = as.matrix(sum(A))\n+write(R, $3, format=\"text\")\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/indexing/LeftIndexingUpdateInPlaceTest3.R", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+\n+args <- commandArgs(TRUE)\n+options(digits=22)\n+library(\"Matrix\")\n+\n+A = as.matrix(readMM(paste(args[1], \"A.mtx\", sep=\"\")))\n+\n+for(i in 1:4) {\n+ out = A\n+ for ( i in 1:4 ) {\n+ out[,i] = A[,i] + 2 ^ i;\n+ }\n+ A = out + A\n+}\n+\n+R = as.matrix(sum(A))\n+writeMM(as(R,\"CsparseMatrix\"), paste(args[2], \"R\", sep=\"\"))\n" }, { "change_type": "RENAME", "old_path": "src/test/scripts/functions/indexing/LeftIndexingUpdateInPlaceTestBugTest2.dml", "new_path": "src/test/scripts/functions/indexing/LeftIndexingUpdateInPlaceTest3.dml", "diff": "#\n#-------------------------------------------------------------\n-debug_print = function(String name, Matrix[Double] m) {\n- print(name + \" ROWS=\" + nrow(m) + \" COLS=\" + ncol(m) + \"\\n\" + toString(m, decimal=10))\n-}\n-\nA = read($1)\n-B = read($2)\n-# out will make A update-in-place from the second iteration of the outer loop\n-out = A\nfor(i in 1:4) {\n+ out = A\nfor ( i in 1:4 ) {\n- debug_print(\"start A\", A[1:5, 1:5])\nout[,i] = A[,i] + 2 ^ i;\n- debug_print(\"end A\", A[1:5, 1:5])\n}\n- A = out\n- print(\"while loop iteration++\")\n+ A = out + A\n}\n-debug_print(\"A\", A[1:5, 1:5])\n-debug_print(\"B\", B[1:5, 1:5])\n+R = as.matrix(sum(A))\n+write(R, $3, format=\"text\")\n\\ No newline at end of file\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/parser/DMLTranslator.java", "new_path": "src/main/java/org/tugraz/sysds/parser/DMLTranslator.java", "diff": "package org.tugraz.sysds.parser;\nimport java.util.ArrayList;\n+import java.util.Arrays;\nimport java.util.HashMap;\nimport java.util.HashSet;\nimport java.util.Iterator;\n@@ -1527,6 +1528,10 @@ public class DMLTranslator\nreturn hops.get(((DataIdentifier) source).getName());\n}\ncatch ( Exception e ) {\n+ //print exception stacktrace for fatal exceptions w/o messages\n+ //to allow for error analysis other than ('no parse issue message')\n+ if( e.getMessage() == null )\n+ e.printStackTrace();\nthrow new ParseException(e.getMessage());\n}\n@@ -2052,10 +2057,12 @@ public class DMLTranslator\ntmpMatrix.add( 0, paramHops.get(DataExpression.RAND_DATA) );\ntmpMatrix.add( 1, paramHops.get(DataExpression.RAND_ROWS) );\ntmpMatrix.add( 2, paramHops.get(DataExpression.RAND_COLS) );\n- tmpMatrix.add( 3, paramHops.get(DataExpression.RAND_DIMS) );\n+ tmpMatrix.add( 3, !paramHops.containsKey(DataExpression.RAND_DIMS) ?\n+ new LiteralOp(\"-1\") : paramHops.get(DataExpression.RAND_DIMS));\ntmpMatrix.add( 4, paramHops.get(DataExpression.RAND_BY_ROW) );\n- currBuiltinOp = new ReorgOp(target.getName(), target.getDataType(), target.getValueType(),\n- ReOrgOp.RESHAPE, tmpMatrix);\n+ System.out.println(Arrays.toString(tmpMatrix.toArray()));\n+ currBuiltinOp = new ReorgOp(target.getName(), target.getDataType(),\n+ target.getValueType(), ReOrgOp.RESHAPE, tmpMatrix);\nbreak;\ndefault:\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/TestUtils.java", "new_path": "src/test/java/org/tugraz/sysds/test/TestUtils.java", "diff": "@@ -415,7 +415,7 @@ public class TestUtils\n}\n}\ncatch (IOException e) {\n- assertTrue(\"could not read from file \" + filePath, false);\n+ assertTrue(\"could not read from file \" + filePath+\": \"+e.getMessage(), false);\n}\nreturn expectedValues;\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -88,7 +88,7 @@ SYSTEMDS-110 New Builtin Functions\n* 116 Builtin function for kmeans OK\n* 117 Builtin function for lm cross validation OK\n* 118 Builtin function for hyperparameter grid search with CVlm\n- * 119 Builtin functions for l2svm and msvm\n+ * 119 Builtin functions for l2svm and msvm OK\nSYSTEMDS-120 Performance Features\n* 121 Avoid spark context creation on parfor result merge OK\n" }, { "change_type": "ADD", "old_path": null, "new_path": "scripts/builtin/msvm.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Modifications Copyright 2019 Graz University of Technology\n+#\n+# Licensed to the Apache Software Foundation (ASF) under one\n+# or more contributor license agreements. See the NOTICE file\n+# distributed with this work for additional information\n+# regarding copyright ownership. The ASF licenses this file\n+# to you under the Apache License, Version 2.0 (the\n+# \"License\"); you may not use this file except in compliance\n+# with the License. You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing,\n+# software distributed under the License is distributed on an\n+# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+# KIND, either express or implied. See the License for the\n+# specific language governing permissions and limitations\n+# under the License.\n+#\n+#-------------------------------------------------------------\n+\n+# Implements builtin multiclass SVM with squared slack variables,\n+# learns one-against-the-rest binary-class classifiers by making a function call to l2SVM\n+\n+# INPUT PARAMETERS:\n+# ---------------------------------------------------------------------------------------------\n+# NAME TYPE DEFAULT MEANING\n+# ---------------------------------------------------------------------------------------------\n+# X Double --- matrix X of feature vectors\n+# Y Double --- matrix Y of class labels\n+# intercept Boolean False No Intercept ( If set to TRUE then a constant bias column is added to X)\n+# num_classes integer 10 Number of classes\n+# epsilon Double 0.001 Procedure terminates early if the reduction in objective function\n+# value is less than epsilon (tolerance) times the initial objective function value.\n+# lambda Double 1.0 Regularization parameter (lambda) for L2 regularization\n+# maxiterations Int 100 Maximum number of conjugate gradient iterations\n+# ---------------------------------------------------------------------------------------------\n+\n+#Output(s)\n+# ---------------------------------------------------------------------------------------------\n+# NAME TYPE DEFAULT MEANING\n+# ---------------------------------------------------------------------------------------------\n+# model Double --- model matrix\n+\n+m_msvm = function(Matrix[Double] X, Matrix[Double] Y, Boolean intercept = FALSE, Integer num_classes =10,\n+ Double epsilon = 0.001, Double lambda = 1.0, Integer max_iterations = 100, Boolean verbose = FALSE)\n+ return(Matrix[Double] model)\n+{\n+ if(verbose)\n+ print(\"Built-in Multiclass-SVM started\")\n+\n+ num_samples = nrow(X)\n+ num_features = ncol(X)\n+\n+ num_rows_in_w = num_features\n+ if(intercept) {\n+ num_rows_in_w = num_rows_in_w + 1\n+ }\n+\n+ w = matrix(0, rows=num_rows_in_w, cols=num_classes)\n+\n+ parfor(iter_class in 1:num_classes) {\n+ Y_local = 2 * (Y == iter_class) - 1\n+ if(verbose) {\n+ print(\"iter class: \" + iter_class)\n+ print(\"y local: \" + toString(Y_local))\n+ }\n+ w[,iter_class] = l2svm(X=X, Y=Y_local, intercept=intercept,\n+ epsilon=epsilon, lambda=lambda, maxiterations=max_iterations)\n+ }\n+\n+ model = w\n+ if (verbose)\n+ print(\"model[\"+iter_class+\"]: \" + toString(model))\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "new_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "diff": "@@ -91,7 +91,7 @@ public enum Builtins {\nIQM(\"interQuartileMean\", false),\nKMEANS(\"kmeans\", true),\nL2SVM(\"l2svm\", true),\n- MULTISVM( \"multisvm\", true),\n+ MSVM(\"msvm\", true),\nLENGTH(\"length\", false),\nLINEAGE(\"lineage\", false),\nLIST(\"list\", false), //note: builtin and parbuiltin\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinMulticlassSVMTest.java", "diff": "+/*\n+ * Copyright 2018 Graz University of Technology\n+ *\n+ * Licensed under the Apache License, Version 2.0 (the \"License\");\n+ * you may not use this file except in compliance with the License.\n+ * You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing, software\n+ * distributed under the License is distributed on an \"AS IS\" BASIS,\n+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+ * See the License for the specific language governing permissions and\n+ * limitations under the License.\n+ */\n+\n+package org.tugraz.sysds.test.functions.builtin;\n+\n+import org.junit.Test;\n+import org.tugraz.sysds.api.DMLScript;\n+import org.tugraz.sysds.common.Types;\n+import org.tugraz.sysds.hops.OptimizerUtils;\n+import org.tugraz.sysds.lops.LopProperties;\n+import org.tugraz.sysds.runtime.matrix.data.MatrixValue;\n+import org.tugraz.sysds.test.AutomatedTestBase;\n+import org.tugraz.sysds.test.TestConfiguration;\n+import org.tugraz.sysds.test.TestUtils;\n+\n+import java.util.HashMap;\n+\n+public class BuiltinMulticlassSVMTest extends AutomatedTestBase\n+{\n+ private final static String TEST_NAME = \"multisvm\";\n+ private final static String TEST_DIR = \"functions/builtin/\";\n+ private static final String TEST_CLASS_DIR = TEST_DIR + BuiltinMulticlassSVMTest.class.getSimpleName() + \"/\";\n+\n+ private final static double eps = 0.001;\n+ private final static int rows = 1000;\n+ private final static int colsX = 500;\n+ private final static double spSparse = 0.01;\n+ private final static double spDense = 0.7;\n+ private final static int max_iter = 10;\n+ private final static int num_classes = 10;\n+\n+ @Override\n+ public void setUp() {\n+ TestUtils.clearAssertionInformation();\n+ addTestConfiguration(TEST_NAME,new TestConfiguration(TEST_CLASS_DIR, TEST_NAME,new String[]{\"C\"}));\n+ }\n+\n+ @Test\n+ public void testMSVMDense() {\n+ runMSVMTest(false, false, num_classes, eps, 1.0, max_iter, LopProperties.ExecType.CP);\n+ }\n+ @Test\n+ public void testMSVMSparse() {\n+ runMSVMTest(true, false, num_classes, eps, 1.0, max_iter, LopProperties.ExecType.CP);\n+ }\n+ @Test\n+ public void testMSVMInterceptSpark() {\n+ runMSVMTest(true,true, num_classes, eps, 1.0, max_iter, LopProperties.ExecType.SPARK);\n+ }\n+\n+ @Test\n+ public void testMSVMSparseLambda2() {\n+ runMSVMTest(true,true, num_classes, eps,2.0, max_iter, LopProperties.ExecType.CP);\n+ }\n+ @Test\n+ public void testMSVMSparseLambda100CP() {\n+ runMSVMTest(true,true, num_classes, 1, 100, max_iter, LopProperties.ExecType.CP);\n+ }\n+ @Test\n+ public void testMSVMSparseLambda100Spark() {\n+ runMSVMTest(true,true, num_classes, 1, 100, max_iter, LopProperties.ExecType.SPARK);\n+ }\n+ @Test\n+ public void testMSVMIteration() {\n+ runMSVMTest(true,true, num_classes, 1, 2.0, 100, LopProperties.ExecType.CP);\n+ }\n+ @Test\n+ public void testMSVMDenseIntercept() {\n+ runMSVMTest(false,true, num_classes, eps, 1.0, max_iter, LopProperties.ExecType.CP);\n+ }\n+ private void runMSVMTest(boolean sparse, boolean intercept, int classes, double eps,\n+ double lambda, int run, LopProperties.ExecType instType)\n+ {\n+ Types.ExecMode platformOld = setExecMode(instType);\n+\n+ boolean oldFlag = OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION;\n+ boolean sparkConfigOld = DMLScript.USE_LOCAL_SPARK_CONFIG;\n+\n+ try\n+ {\n+ loadTestConfiguration(getTestConfiguration(TEST_NAME));\n+\n+ double sparsity = sparse ? spSparse : spDense;\n+ String HOME = SCRIPT_DIR + TEST_DIR;\n+ fullDMLScriptName = HOME + TEST_NAME + \".dml\";\n+ programArgs = new String[]{ \"-explain\", \"-stats\",\n+ \"-nvargs\", \"X=\" + input(\"X\"), \"Y=\" + input(\"Y\"), \"model=\" + output(\"model\"),\n+ \"inc=\" + String.valueOf(intercept).toUpperCase(), \"num_classes=\" + classes, \"eps=\" + eps, \"lam=\" + lambda, \"max=\" + run};\n+\n+ fullRScriptName = HOME + TEST_NAME + \".R\";\n+ rCmd = getRCmd(inputDir(), Boolean.toString(intercept), Integer.toString(classes), Double.toString(eps),\n+ Double.toString(lambda), Integer.toString(run), expectedDir());\n+\n+ double[][] X = getRandomMatrix(rows, colsX, 0, 1, sparsity, -1);\n+ double[][] Y = getRandomMatrix(rows, 1, 0, num_classes, 1, -1);\n+ Y = TestUtils.round(Y);\n+\n+ writeInputMatrixWithMTD(\"X\", X, true);\n+ writeInputMatrixWithMTD(\"Y\", Y, true);\n+\n+ runTest(true, false, null, -1);\n+ runRScript(true);\n+\n+ HashMap<MatrixValue.CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(\"model\");\n+ HashMap<MatrixValue.CellIndex, Double> rfile = readRMatrixFromFS(\"model\");\n+ TestUtils.compareMatrices(dmlfile, rfile, eps, \"Stat-DML\", \"Stat-R\");\n+ }\n+ finally {\n+ rtplatform = platformOld;\n+ DMLScript.USE_LOCAL_SPARK_CONFIG = sparkConfigOld;\n+ OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = oldFlag;\n+ OptimizerUtils.ALLOW_AUTO_VECTORIZATION = true;\n+ OptimizerUtils.ALLOW_OPERATOR_FUSION = true;\n+ }\n+ }\n+}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/builtin/multisvm.R", "diff": "+#-------------------------------------------------------------\n+#\n+# Licensed to the Apache Software Foundation (ASF) under one\n+# or more contributor license agreements. See the NOTICE file\n+# distributed with this work for additional information\n+# regarding copyright ownership. The ASF licenses this file\n+# to you under the Apache License, Version 2.0 (the\n+# \"License\"); you may not use this file except in compliance\n+# with the License. You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing,\n+# software distributed under the License is distributed on an\n+# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+# KIND, either express or implied. See the License for the\n+# specific language governing permissions and limitations\n+# under the License.\n+#\n+#-------------------------------------------------------------\n+\n+args <- commandArgs(TRUE)\n+\n+library(\"Matrix\")\n+\n+X = as.matrix(readMM(paste(args[1], \"X.mtx\", sep=\"\")))\n+\n+check_X = sum(X)\n+if(check_X == 0){\n+ print(\"X has no non-zeros\")\n+}else{\n+ Y = as.matrix(readMM(paste(args[1], \"Y.mtx\", sep=\"\")))\n+ intercept = as.logical(args[2])\n+ num_classes = as.integer(args[3])\n+ epsilon = as.double(args[4])\n+ lambda = as.double(args[5])\n+ max_iterations = as.integer(args[6])\n+\n+ num_samples = nrow(X)\n+ num_features = ncol(X)\n+\n+ if (intercept == TRUE) {\n+ ones = matrix(1, num_samples, 1);\n+ X = cbind(X, ones);\n+ }\n+\n+ num_rows_in_w = num_features\n+ if(intercept == TRUE){\n+ num_rows_in_w = num_rows_in_w + 1\n+ }\n+ w = matrix(0, num_rows_in_w, num_classes)\n+\n+ for(iter_class in 1:num_classes){\n+ Y_local = 2 * (Y == iter_class) - 1\n+ w_class = matrix(0, num_features, 1)\n+ if (intercept == TRUE) {\n+ zero_matrix = matrix(0, 1, 1);\n+ w_class = t(cbind(t(w_class), zero_matrix));\n+ }\n+\n+ g_old = t(X) %*% Y_local\n+ s = g_old\n+\n+ Xw = matrix(0, nrow(X), 1)\n+ iter = 0\n+ continue = 1\n+ while(continue == 1) {\n+ # minimizing primal obj along direction s\n+ step_sz = 0\n+ Xd = X %*% s\n+ wd = lambda * sum(w_class * s)\n+ dd = lambda * sum(s * s)\n+ continue1 = 1\n+ while(continue1 == 1){\n+ tmp_Xw = Xw + step_sz*Xd\n+ out = 1 - Y_local * (tmp_Xw)\n+ sv = (out > 0)\n+ out = out * sv\n+ g = wd + step_sz*dd - sum(out * Y_local * Xd)\n+ h = dd + sum(Xd * sv * Xd)\n+ step_sz = step_sz - g/h\n+ if (g*g/h < 0.0000000001){\n+ continue1 = 0\n+ }\n+ }\n+\n+ #update weights\n+ w_class = w_class + step_sz*s\n+ Xw = Xw + step_sz*Xd\n+ #print(Xw)\n+ out = 1 - Y_local * Xw\n+ sv = (out > 0)\n+ out = sv * out\n+ obj = 0.5 * sum(out * out) + lambda/2 * sum(w_class * w_class)\n+ g_new = t(X) %*% (out * Y_local) - lambda * w_class\n+\n+ tmp = sum(s * g_old)\n+\n+ train_acc = sum(Y_local*(X%*%w_class) >= 0)/num_samples*100\n+ #print(paste(\"For class \", iter_class, \" iteration \", iter, \" training accuracy: \", train_acc, sep=\"\"))\n+\n+ if((step_sz*tmp < epsilon*obj) | (iter >= max_iterations-1)){\n+ continue = 0\n+ }\n+\n+ #non-linear CG step\n+ be = sum(g_new * g_new)/sum(g_old * g_old)\n+ s = be * s + g_new\n+ g_old = g_new\n+\n+ iter = iter + 1\n+ }\n+\n+ w[,iter_class] = w_class\n+ }\n+ #print(\"R model \"); print(w)\n+\n+ writeMM(as(w, \"CsparseMatrix\"), paste(args[7], \"model\", sep=\"\"))\n+}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/builtin/multisvm.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($X)\n+Y = read($Y)\n+model = msvm(X=X, Y=Y, intercept = $inc, num_classes= $num_classes,\n+ epsilon = $eps, lambda = $lam, max_iterations = $max )\n+write(model, $model)\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/BasicTensorBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/BasicTensorBlock.java", "diff": "@@ -163,6 +163,22 @@ public class BasicTensorBlock implements Serializable {\n}\n}\n+ /**\n+ * Recomputes and materializes the number of non-zero values\n+ * of the entire basic tensor block.\n+ *\n+ * @return number of non-zeros\n+ */\n+ public long recomputeNonZeros() {\n+ if( _sparse && _sparseBlock != null ) { //SPARSE\n+ throw new DMLRuntimeException(\"Sparse tensor block not supported\");\n+ }\n+ else if( !_sparse && _denseBlock != null ) { //DENSE\n+ _nnz = _denseBlock.countNonZeros();\n+ }\n+ return _nnz;\n+ }\n+\npublic boolean isAllocated() {\nreturn _sparse ? (_sparseBlock != null) : (_denseBlock != null);\n}\n@@ -260,9 +276,8 @@ public class BasicTensorBlock implements Serializable {\nret = true;\nif( _nnz==0 ) {\n//prevent under-estimation\n- //TODO recomputeNonZeros();\n- //TODO return false if _nnz != 0\nif(safe)\n+ recomputeNonZeros();\nret = (_nnz == 0);\n}\nreturn ret;\n@@ -322,18 +337,39 @@ public class BasicTensorBlock implements Serializable {\nthrow new NotImplementedException();\n}\nelse if (v != null) {\n- if (v instanceof Double)\n+ if (v instanceof Double) {\n+ double old = _denseBlock.get(ix);\n_denseBlock.set(ix, (Double) v);\n- else if (v instanceof Float)\n+ _nnz += (old == 0 ? 0 : -1) + ((Double) v == 0 ? 0 : 1);\n+ }\n+ else if (v instanceof Float) {\n+ double old = _denseBlock.get(ix);\n_denseBlock.set(ix, (Float) v);\n- else if (v instanceof Long)\n+ _nnz += (old == 0 ? 0 : -1) + ((Float) v == 0 ? 0 : 1);\n+ }\n+ else if (v instanceof Long) {\n+ long old = _denseBlock.getLong(ix);\n_denseBlock.set(ix, (Long) v);\n- else if (v instanceof Integer)\n+ _nnz += (old == 0 ? 0 : -1) + ((Long) v == 0 ? 0 : 1);\n+ }\n+ else if (v instanceof Integer) {\n+ long old = _denseBlock.getLong(ix);\n_denseBlock.set(ix, (Integer) v);\n- else if (v instanceof Boolean)\n+ _nnz += (old == 0 ? 0 : -1) + ((Integer) v == 0 ? 0 : 1);\n+ }\n+ else if (v instanceof Boolean) {\n+ long old = _denseBlock.getLong(ix);\n_denseBlock.set(ix, ((Boolean) v) ? 1.0 : 0.0);\n- else if (v instanceof String)\n+ _nnz += (old == 0 ? 0 : -1) + (!(Boolean) v ? 0 : 1);\n+ }\n+ else if (v instanceof String) {\n+ String old = _denseBlock.getString(ix);\n+ if (old != null && !old.isEmpty())\n+ _nnz--;\n_denseBlock.set(ix, (String) v);\n+ if (!((String) v).isEmpty())\n+ _nnz++;\n+ }\nelse\nthrow new DMLRuntimeException(\"BasicTensor.set(int[],Object) is not implemented for the given Object\");\n}\n@@ -346,7 +382,9 @@ public class BasicTensorBlock implements Serializable {\nthrow new NotImplementedException();\n}\nelse {\n+ double old = _denseBlock.get(r, c);\n_denseBlock.set(r, c, v);\n+ _nnz += (old == 0 ? 0 : -1) + (v == 0 ? 0 : 1);\n}\n}\n@@ -356,6 +394,10 @@ public class BasicTensorBlock implements Serializable {\n}\nelse {\n_denseBlock.set(v);\n+ if (v == 0)\n+ _nnz = 0;\n+ else\n+ _nnz = getLength();\n}\n}\n@@ -364,18 +406,30 @@ public class BasicTensorBlock implements Serializable {\nthrow new NotImplementedException();\n}\nelse {\n- if (v instanceof Double)\n+ if (v instanceof Double) {\n_denseBlock.set((Double) v);\n- else if (v instanceof Float)\n+ _nnz += ((Double) v == 0 ? 0 : 1);\n+ }\n+ else if (v instanceof Float) {\n_denseBlock.set((Float) v);\n- else if (v instanceof Long)\n+ _nnz += ((Float) v == 0 ? 0 : 1);\n+ }\n+ else if (v instanceof Long) {\n_denseBlock.set((Long) v);\n- else if (v instanceof Integer)\n+ _nnz += ((Long) v == 0 ? 0 : 1);\n+ }\n+ else if (v instanceof Integer) {\n_denseBlock.set((Integer) v);\n- else if (v instanceof Boolean)\n+ _nnz += ((Integer) v == 0 ? 0 : 1);\n+ }\n+ else if (v instanceof Boolean) {\n_denseBlock.set(((Boolean) v) ? 1.0 : 0.0);\n- else if (v instanceof String)\n+ _nnz += (!(Boolean) v ? 0 : 1);\n+ }\n+ else if (v instanceof String) {\n_denseBlock.set((String) v);\n+ _nnz += (((String) v).isEmpty() ? 0 : 1);\n+ }\nelse\nthrow new DMLRuntimeException(\"BasicTensor.set(Object) is not implemented for the given Object\");\n}\n@@ -387,8 +441,10 @@ public class BasicTensorBlock implements Serializable {\nelse {\nif (other.isSparse())\nthrow new NotImplementedException();\n- else\n+ else {\n_denseBlock.set(0, _dims[0], 0, _denseBlock.getCumODims(0), other.getDenseBlock());\n+ _nnz = other._nnz;\n+ }\n}\n}\n@@ -405,10 +461,12 @@ public class BasicTensorBlock implements Serializable {\n// TODO implement sparse set instead of converting to dense\nother.sparseToDense();\n_denseBlock.set(0, _dims[0], 0, _denseBlock.getCumODims(0), other.getDenseBlock());\n+ _nnz = other.getNonZeros();\n}\n}\nelse {\n_denseBlock.set(0, _dims[0], 0, _denseBlock.getCumODims(0), other.getDenseBlock());\n+ _nnz = other.getNonZeros();\n}\n}\n}\n@@ -443,7 +501,8 @@ public class BasicTensorBlock implements Serializable {\nif (that.isEmpty(false)) {\nif (_denseBlock != null)\n_denseBlock.reset(that._dims);\n- return;\n+ else\n+ _denseBlock = DenseBlockFactory.createDenseBlock(that._vt, that._dims);\n}\n//allocate and copy dense block\nallocateDenseBlock(false);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/data/DataTensorBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/data/DataTensorBlock.java", "diff": "@@ -331,9 +331,6 @@ public class DataTensorBlock implements Serializable {\nif (that._ixToCols[i] != null)\n_ixToCols[i] = that._ixToCols[i].clone();\nif (that.isAllocated()) {\n- if (that.isEmpty(false)) {\n- return;\n- }\nfor (int i = 0; i < _colsdata.length; i++) {\nif (that._colsdata[i] != null) {\n_colsdata[i] = new BasicTensorBlock(that._colsdata[i]);\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/DataGenCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/DataGenCPInstruction.java", "diff": "@@ -375,8 +375,8 @@ public class DataGenCPInstruction extends UnaryCPInstruction {\nif (runtimeSeed == null)\nruntimeSeed = (minValue == maxValue && sparsity == 1) ?\nDataGenOp.UNSPECIFIED_SEED : DataGenOp.generateRandomSeed();\n- int position = (method == DataGenMethod.RAND) ? SEED_POSITION_RAND + 1 :\n- (method == DataGenMethod.SAMPLE) ? SEED_POSITION_SAMPLE + 1 : 0;\n+ int position = (method == DataGenMethod.RAND) ? SEED_POSITION_RAND :\n+ (method == DataGenMethod.SAMPLE) ? SEED_POSITION_SAMPLE : 0;\ntmpInstStr = InstructionUtils.replaceOperand(\ntmpInstStr, position, String.valueOf(runtimeSeed));\n}\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -122,5 +122,5 @@ SYSTEMDS-160 Tensor Compiler/Runtime\n* 174 Reuse rewrite for rbind/cbind-tsmm/ba+* OK\n* 175 Refactoring of lineage rewrite code OK\n* 176 Reuse rewrite for cbind/rbind-elementwise */+\n- * 177 Reuse rewrite for aggregate\n+ * 177 Reuse rewrite for aggregate OK\n* 178 Compiler assisted reuse (eg. CV, lmCG)\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/ParameterizedBuiltinCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/ParameterizedBuiltinCPInstruction.java", "diff": "@@ -29,6 +29,8 @@ import java.util.List;\nimport java.util.stream.Collectors;\nimport java.util.stream.IntStream;\n+import org.tugraz.sysds.common.Types.DataType;\n+import org.tugraz.sysds.common.Types.ValueType;\nimport org.tugraz.sysds.lops.Lop;\nimport org.tugraz.sysds.parser.ParameterizedBuiltinFunctionExpression;\nimport org.tugraz.sysds.parser.Statement;\n@@ -43,6 +45,8 @@ import org.tugraz.sysds.runtime.data.TensorBlock;\nimport org.tugraz.sysds.runtime.functionobjects.ParameterizedBuiltin;\nimport org.tugraz.sysds.runtime.functionobjects.ValueFunction;\nimport org.tugraz.sysds.runtime.instructions.InstructionUtils;\n+import org.tugraz.sysds.runtime.lineage.LineageItem;\n+import org.tugraz.sysds.runtime.lineage.LineageItemUtils;\nimport org.tugraz.sysds.runtime.matrix.data.FrameBlock;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport org.tugraz.sysds.runtime.matrix.operators.Operator;\n@@ -392,4 +396,25 @@ public class ParameterizedBuiltinCPInstruction extends ComputationCPInstruction\n+ \"Use toString(X, rows=..., cols=...) if necessary.\");\n}\n}\n+\n+ @Override\n+ public LineageItem[] getLineageItems(ExecutionContext ec) {\n+ String opcode = getOpcode();\n+ if (opcode.equalsIgnoreCase(\"groupedagg\")) {\n+ CPOperand target = new CPOperand(params.get(Statement.GAGG_TARGET), ValueType.FP64, DataType.MATRIX);\n+ CPOperand groups = new CPOperand(params.get(Statement.GAGG_GROUPS), ValueType.FP64, DataType.MATRIX);\n+ String wt = params.containsKey(Statement.GAGG_WEIGHTS) ? params.get(Statement.GAGG_WEIGHTS) : String.valueOf(-1);\n+ CPOperand weights = new CPOperand(wt, ValueType.FP64, DataType.MATRIX);\n+ CPOperand fn = new CPOperand(params.get(Statement.GAGG_FN), ValueType.STRING, DataType.SCALAR, true);\n+ String ng = params.containsKey(Statement.GAGG_NUM_GROUPS) ? params.get(Statement.GAGG_NUM_GROUPS) : String.valueOf(-1);\n+ CPOperand ngroups = new CPOperand(ng , ValueType.INT64, DataType.SCALAR, true);\n+ return new LineageItem[]{new LineageItem(output.getName(),\n+ getOpcode(), LineageItemUtils.getLineage(ec, target, groups, weights, fn, ngroups))};\n+ }\n+ //TODO: generic interface to support all the ops\n+ else\n+ return new LineageItem[]{new LineageItem(output.getName(),\n+ getOpcode(), LineageItemUtils.getLineage(ec, input1,input2,input3))};\n+\n+ }\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCache.java", "diff": "@@ -20,6 +20,7 @@ import org.tugraz.sysds.api.DMLScript;\nimport org.tugraz.sysds.hops.OptimizerUtils;\nimport org.tugraz.sysds.hops.cost.CostEstimatorStaticRuntime;\nimport org.tugraz.sysds.lops.MMTSJ.MMTSJType;\n+import org.tugraz.sysds.parser.Statement;\nimport org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.caching.MatrixObject;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\n@@ -29,6 +30,7 @@ import org.tugraz.sysds.runtime.instructions.cp.BinaryMatrixMatrixCPInstruction;\nimport org.tugraz.sysds.runtime.instructions.cp.CPInstruction.CPType;\nimport org.tugraz.sysds.runtime.instructions.cp.ComputationCPInstruction;\nimport org.tugraz.sysds.runtime.instructions.cp.MMTSJCPInstruction;\n+import org.tugraz.sysds.runtime.instructions.cp.ParameterizedBuiltinCPInstruction;\nimport org.tugraz.sysds.runtime.lineage.LineageCacheConfig.ReuseCacheType;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport org.tugraz.sysds.runtime.util.LocalFileUtils;\n@@ -174,7 +176,8 @@ public class LineageCache {\n|| inst.getOpcode().equalsIgnoreCase(\"ba+*\")\n|| (inst.getOpcode().equalsIgnoreCase(\"*\") &&\ninst instanceof BinaryMatrixMatrixCPInstruction) //TODO support scalar\n- || inst.getOpcode().equalsIgnoreCase(\"rightIndex\");\n+ || inst.getOpcode().equalsIgnoreCase(\"rightIndex\")\n+ || inst.getOpcode().equalsIgnoreCase(\"groupedagg\");\n}\n//---------------- CACHE SPACE MANAGEMENT METHODS -----------------\n@@ -192,7 +195,8 @@ public class LineageCache {\nwhile ((valSize+_cachesize) > CACHELIMIT)\n{\ndouble reduction = _cache.get(_end._key).getValue().getInMemorySize();\n- if (_cache.get(_end._key)._compEst > getDiskSpillEstimate())\n+ if (_cache.get(_end._key)._compEst > getDiskSpillEstimate()\n+ && LineageCacheConfig.isSetSpill())\nspillToLocalFS(); // If re-computation is more expensive, spill data to disk.\nremoveEntry(reduction);\n@@ -295,6 +299,25 @@ public class LineageCache {\nbreak;\n}\n+ case ParameterizedBuiltin:\n+ {\n+ String opcode = ((ParameterizedBuiltinCPInstruction)inst).getOpcode();\n+ HashMap<String, String> params = ((ParameterizedBuiltinCPInstruction)inst).getParameterMap();\n+ long r1 = ec.getMatrixObject(params.get(Statement.GAGG_TARGET)).getNumRows();\n+ String fn = params.get(Statement.GAGG_FN);\n+ double xga = 0;\n+ if (opcode.equalsIgnoreCase(\"groupedagg\")) {\n+ if (fn.equalsIgnoreCase(\"sum\"))\n+ xga = 4;\n+ else if(fn.equalsIgnoreCase(\"count\"))\n+ xga = 1;\n+ //TODO: cm, variance\n+ }\n+ //TODO: support other PBuiltin ops\n+ nflops = 2 * r1+xga * r1;\n+ break;\n+ }\n+\ndefault:\nthrow new DMLRuntimeException(\"Lineage Cache: unsupported instruction: \"+inst.getOpcode());\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCacheConfig.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageCacheConfig.java", "diff": "@@ -38,8 +38,6 @@ public class LineageCacheConfig {\n}\n}\n- public ArrayList<String> _MMult = new ArrayList<>();\n-\npublic enum CachedItemHead {\nTSMM,\nALL\n@@ -52,10 +50,18 @@ public class LineageCacheConfig {\nALL\n}\n+ public ArrayList<String> _MMult = new ArrayList<>();\n+ public static boolean _allowSpill = true;\n+\nprivate static ReuseCacheType _cacheType = null;\nprivate static CachedItemHead _itemH = null;\nprivate static CachedItemTail _itemT = null;\n+ static {\n+ //setup static configuration parameters\n+ setSpill(false); //disable spilling of cache entries to disk\n+ }\n+\npublic static void setConfigTsmmCbind(ReuseCacheType ct) {\n_cacheType = ct;\n_itemH = CachedItemHead.TSMM;\n@@ -82,6 +88,14 @@ public class LineageCacheConfig {\nDMLScript.LINEAGE_REUSE = rop;\n}\n+ public static void setSpill(boolean toSpill) {\n+ _allowSpill = toSpill;\n+ }\n+\n+ public static boolean isSetSpill() {\n+ return _allowSpill;\n+ }\n+\npublic static ReuseCacheType getCacheType() {\nreturn _cacheType;\n}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageRewriteReuse.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageRewriteReuse.java", "diff": "@@ -18,6 +18,7 @@ package org.tugraz.sysds.runtime.lineage;\nimport java.util.ArrayList;\nimport java.util.HashMap;\n+import java.util.LinkedHashMap;\nimport java.util.Map;\nimport org.apache.commons.logging.Log;\n@@ -30,12 +31,15 @@ import org.tugraz.sysds.hops.DataOp;\nimport org.tugraz.sysds.hops.Hop;\nimport org.tugraz.sysds.hops.Hop.OpOp2;\nimport org.tugraz.sysds.hops.Hop.OpOpN;\n+import org.tugraz.sysds.hops.Hop.ParamBuiltinOp;\nimport org.tugraz.sysds.hops.IndexingOp;\nimport org.tugraz.sysds.hops.LiteralOp;\nimport org.tugraz.sysds.hops.NaryOp;\n+import org.tugraz.sysds.hops.ParameterizedBuiltinOp;\nimport org.tugraz.sysds.hops.ReorgOp;\nimport org.tugraz.sysds.hops.recompile.Recompiler;\nimport org.tugraz.sysds.hops.rewrite.HopRewriteUtils;\n+import org.tugraz.sysds.parser.Statement;\nimport org.tugraz.sysds.runtime.DMLRuntimeException;\nimport org.tugraz.sysds.runtime.controlprogram.BasicProgramBlock;\nimport org.tugraz.sysds.runtime.controlprogram.Program;\n@@ -44,6 +48,7 @@ import org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\nimport org.tugraz.sysds.runtime.controlprogram.context.ExecutionContextFactory;\nimport org.tugraz.sysds.runtime.instructions.Instruction;\nimport org.tugraz.sysds.runtime.instructions.cp.ComputationCPInstruction;\n+import org.tugraz.sysds.runtime.instructions.cp.ParameterizedBuiltinCPInstruction;\nimport org.tugraz.sysds.runtime.lineage.LineageCacheConfig.ReuseCacheType;\nimport org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\nimport org.tugraz.sysds.runtime.meta.MetaData;\n@@ -80,6 +85,8 @@ public class LineageRewriteReuse\nnewInst = (newInst == null) ? rewriteElementMulRbind(curr, ec, lrwec) : newInst;\n//cbind(X, deltaX) * cbind(Y, deltaY) -> cbind(C, deltaX * deltaY), where C = X * Y\nnewInst = (newInst == null) ? rewriteElementMulCbind(curr, ec, lrwec) : newInst;\n+ //aggregate(target=X+deltaX,...) = cbind(C, aggregate(target=deltaX,...)) where C = aggregate(target=X,...)\n+ newInst = (newInst == null) ? rewriteAggregateCbind(curr, ec, lrwec) : newInst;\nif (newInst == null)\nreturn false;\n@@ -449,6 +456,60 @@ public class LineageRewriteReuse\nreturn genInst(lrwWrite, lrwec);\n}\n+ private static ArrayList<Instruction> rewriteAggregateCbind (Instruction curr, ExecutionContext ec, ExecutionContext lrwec)\n+ {\n+ // Check the applicability of this rewrite.\n+ Map<String, MatrixBlock> inCache = new HashMap<>();\n+ if (!isAggCbind (curr, ec, inCache))\n+ return null;\n+\n+ long t0 = DMLScript.STATISTICS ? System.nanoTime() : 0;\n+ // Create a transient read op over the last * result\n+ MatrixBlock cachedEntry = inCache.get(\"lastMatrix\");\n+ lrwec.setVariable(\"cachedEntry\", convMBtoMO(cachedEntry));\n+ DataOp lastRes = HopRewriteUtils.createTransientRead(\"cachedEntry\", cachedEntry);\n+ //TODO: support for block of rows\n+ HashMap<String, String> params = ((ParameterizedBuiltinCPInstruction)curr).getParameterMap();\n+ MatrixObject mo = ec.getMatrixObject(params.get(Statement.GAGG_TARGET));\n+ lrwec.setVariable(\"oldMatrix\", mo);\n+ DataOp newMatrix = HopRewriteUtils.createTransientRead(\"oldMatrix\", mo);\n+ MatrixObject moG = ec.getMatrixObject(params.get(Statement.GAGG_GROUPS));\n+ lrwec.setVariable(\"groups\", moG);\n+ DataOp groups = HopRewriteUtils.createTransientRead(\"groups\", moG);\n+ String fn = params.get(Statement.GAGG_FN);\n+ int ngroups = (params.get(Statement.GAGG_NUM_GROUPS) != null) ?\n+ (int)Double.parseDouble(params.get(Statement.GAGG_NUM_GROUPS)) : -1;\n+ Hop lastCol;\n+ // Use deltaX from cache, or create rightIndex\n+ if (inCache.containsKey(\"deltaX\")) {\n+ MatrixBlock cachedRI = inCache.get(\"deltaX\");\n+ lrwec.setVariable(\"deltaX\", convMBtoMO(cachedRI));\n+ lastCol = HopRewriteUtils.createTransientRead(\"deltaX\", cachedRI);\n+ }\n+ else\n+ lastCol = HopRewriteUtils.createIndexingOp(newMatrix, new LiteralOp(1), new LiteralOp(mo.getNumRows()),\n+ new LiteralOp(mo.getNumColumns()), new LiteralOp(mo.getNumColumns()));\n+ // aggregate(target=X+lastCol,...) = cbind(aggregate(target=X,...), aggregate(target=lastCol,...))\n+ LinkedHashMap<String, Hop> args = new LinkedHashMap<>();\n+ args.put(\"target\", lastCol);\n+ args.put(\"groups\", groups);\n+ args.put(\"fn\", new LiteralOp(fn));\n+ if (ngroups != -1)\n+ args.put(\"ngroups\", new LiteralOp(ngroups));\n+ ParameterizedBuiltinOp rowTwo = HopRewriteUtils.createParameterizedBuiltinOp(newMatrix, args, ParamBuiltinOp.GROUPEDAGG);\n+ BinaryOp lrwHop= HopRewriteUtils.createBinary(lastRes, rowTwo, OpOp2.CBIND);\n+ DataOp lrwWrite = HopRewriteUtils.createTransientWrite(LR_VAR, lrwHop);\n+\n+ if (DMLScript.STATISTICS) {\n+ LineageCacheStatistics.incrementPRewriteTime(System.nanoTime() - t0);\n+ LineageCacheStatistics.incrementPRewrites();\n+ }\n+\n+ // generate runtime instructions\n+ LOG.debug(\"LINEAGE REWRITE rewriteElementMulCbind APPLIED\");\n+ return genInst(lrwWrite, lrwec);\n+ }\n+\n/*------------------------REWRITE APPLICABILITY CHECKS-------------------------*/\nprivate static boolean isTsmmCbind(Instruction curr, ExecutionContext ec, Map<String, MatrixBlock> inCache)\n@@ -630,6 +691,36 @@ public class LineageRewriteReuse\nreturn inCache.containsKey(\"lastMatrix\") ? true : false;\n}\n+ private static boolean isAggCbind (Instruction curr, ExecutionContext ec, Map<String, MatrixBlock> inCache)\n+ {\n+ if (!LineageCache.isReusable(curr)) {\n+ return false;\n+ }\n+\n+ // If the input to groupedagg came from cbind, look for both the inputs in cache.\n+ LineageItem[] items = ((ComputationCPInstruction) curr).getLineageItems(ec);\n+ if (curr.getOpcode().equalsIgnoreCase(\"groupedagg\")) {\n+ LineageItem target = items[0].getInputs()[0];\n+ LineageItem groups = items[0].getInputs()[1];\n+ LineageItem weights = items[0].getInputs()[2];\n+ LineageItem fn = items[0].getInputs()[3];\n+ LineageItem ngroups = items[0].getInputs()[4];\n+ if (target.getOpcode().equalsIgnoreCase(\"cbind\")) {\n+ // create groupedagg lineage on top of the input of last append\n+ LineageItem input1 = target.getInputs()[0];\n+ LineageItem tmp = new LineageItem(\"toProbe\", curr.getOpcode(),\n+ new LineageItem[] {input1, groups, weights, fn, ngroups});\n+ if (LineageCache.probe(tmp))\n+ inCache.put(\"lastMatrix\", LineageCache.get(tmp));\n+ // look for the appended column in cache\n+ if (LineageCache.probe(target.getInputs()[1]))\n+ inCache.put(\"deltaX\", LineageCache.get(target.getInputs()[1]));\n+ }\n+ }\n+ // return true only if the last tsmm is found\n+ return inCache.containsKey(\"lastMatrix\") ? true : false;\n+ }\n+\n/*----------------------INSTRUCTIONS GENERATION & EXECUTION-----------------------*/\nprivate static ArrayList<Instruction> genInst(Hop hops, ExecutionContext ec) {\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageRewriteTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/lineage/LineageRewriteTest.java", "diff": "@@ -37,6 +37,7 @@ public class LineageRewriteTest extends AutomatedTestBase {\nprotected static final String TEST_NAME5 = \"RewriteTest9\";\nprotected static final String TEST_NAME6 = \"RewriteTest10\";\nprotected static final String TEST_NAME7 = \"RewriteTest11\";\n+ protected static final String TEST_NAME8 = \"RewriteTest12\";\nprotected String TEST_CLASS_DIR = TEST_DIR + LineageRewriteTest.class.getSimpleName() + \"/\";\n@@ -53,44 +54,50 @@ public class LineageRewriteTest extends AutomatedTestBase {\naddTestConfiguration(TEST_NAME5, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME5));\naddTestConfiguration(TEST_NAME6, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME6));\naddTestConfiguration(TEST_NAME7, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME7));\n+ addTestConfiguration(TEST_NAME8, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME8));\n}\n@Test\npublic void testTsmm2Cbind() {\n- testRewrite(TEST_NAME1, false);\n+ testRewrite(TEST_NAME1, false, 0);\n}\n@Test\npublic void testTsmmCbind() {\n- testRewrite(TEST_NAME2, false);\n+ testRewrite(TEST_NAME2, false, 0);\n}\n@Test\npublic void testTsmmRbind() {\n- testRewrite(TEST_NAME3, false);\n+ testRewrite(TEST_NAME3, false, 0);\n}\n@Test\npublic void testMatmulRbindLeft() {\n- testRewrite(TEST_NAME4, false);\n+ testRewrite(TEST_NAME4, false, 0);\n}\n@Test\npublic void testMatmulCbindRight() {\n- testRewrite(TEST_NAME5, false);\n+ testRewrite(TEST_NAME5, false, 0);\n}\n@Test\npublic void testElementMulRbind() {\n- testRewrite(TEST_NAME6, true);\n+ testRewrite(TEST_NAME6, true, 0);\n}\n@Test\npublic void testElementMulCbind() {\n- testRewrite(TEST_NAME7, true);\n+ testRewrite(TEST_NAME7, true, 0);\n}\n- private void testRewrite(String testname, boolean elementwise) {\n+ @Test\n+ public void testaggregatecbind() {\n+ testRewrite(TEST_NAME8, false, 2);\n+ }\n+\n+ private void testRewrite(String testname, boolean elementwise, int classes) {\ntry {\ngetAndLoadTestConfiguration(testname);\nList<String> proArgs = new ArrayList<>();\n@@ -107,6 +114,13 @@ public class LineageRewriteTest extends AutomatedTestBase {\ndouble[][] X = getRandomMatrix(numRecords, numFeatures, 0, 1, 0.8, -1);\ndouble[][] Y = !elementwise ? getRandomMatrix(numFeatures, numRecords, 0, 1, 0.8, -1)\n: getRandomMatrix(numRecords, numFeatures, 0, 1, 0.8, -1);\n+ if (classes > 0) {\n+ Y = getRandomMatrix(numRecords, 1, 0, 1, 1, -1);\n+ for(int i=0; i<numRecords; i++){\n+ Y[i][0] = (int)(Y[i][0]*classes) + 1;\n+ Y[i][0] = (Y[i][0] > classes) ? classes : Y[i][0];\n+ }\n+ }\nwriteInputMatrixWithMTD(\"X\", X, true);\nwriteInputMatrixWithMTD(\"Y\", Y, true);\nrunTest(true, EXCEPTION_NOT_EXPECTED, null, -1);\n@@ -117,7 +131,7 @@ public class LineageRewriteTest extends AutomatedTestBase {\nproArgs.add(\"recompile_hops\");\nproArgs.add(\"-stats\");\nproArgs.add(\"-lineage\");\n- proArgs.add(\"reuse_partial\");\n+ proArgs.add(\"reuse_hybrid\");\nproArgs.add(\"-args\");\nproArgs.add(input(\"X\"));\nproArgs.add(input(\"Y\"));\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/lineage/RewriteTest12.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1);\n+Y = read($2);\n+\n+sum = 0;\n+ngroups = as.integer(max(Y));\n+tmp = X[,1];\n+R = matrix(0, 1, ncol(X));\n+\n+for (i in 2:ncol(X)) {\n+ Res1 = aggregate(target=tmp, groups=Y, fn=\"sum\", ngroups=ngroups);\n+ tmp = cbind(tmp, X[,i]);\n+ while(FALSE) {}\n+ R[1,i] = sum(Res1);\n+ sum = sum + sum(Res1);\n+}\n+\n+write(R, $3, format=\"text\");\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/rewrite/HopRewriteUtils.java", "new_path": "src/main/java/org/tugraz/sysds/hops/rewrite/HopRewriteUtils.java", "diff": "@@ -490,12 +490,12 @@ public class HopRewriteUtils\nHashMap<String, Hop> params = new HashMap<>();\nparams.put(DataExpression.RAND_ROWS, new LiteralOp(rows));\nparams.put(DataExpression.RAND_COLS, new LiteralOp(cols));\n- params.put(DataExpression.RAND_DIMS, new LiteralOp(\"-1\")); //TODO\nparams.put(DataExpression.RAND_MIN, str);\nparams.put(DataExpression.RAND_MAX, str);\nparams.put(DataExpression.RAND_SEED, new LiteralOp(DataGenOp.UNSPECIFIED_SEED));\n- Hop datagen = new DataGenOp(DataGenMethod.SINIT, new DataIdentifier(\"tmp\"), params);\n+ Hop datagen = new DataGenOp(DataGenMethod.SINIT,\n+ new DataIdentifier(\"tmp\", DataType.MATRIX), params);\ndatagen.setBlocksize(ConfigurationManager.getBlocksize());\ncopyLineNumbers(values.get(0), datagen);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/parser/DataIdentifier.java", "new_path": "src/main/java/org/tugraz/sysds/parser/DataIdentifier.java", "diff": "package org.tugraz.sysds.parser;\n+import org.tugraz.sysds.common.Types.DataType;\npublic class DataIdentifier extends Identifier\n{\n@@ -39,6 +40,11 @@ public class DataIdentifier extends Identifier\n_name = name;\n}\n+ public DataIdentifier(String name, DataType dt){\n+ this(name);\n+ _dataType = dt;\n+ }\n+\npublic DataIdentifier(){\n_name = null;\n}\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/api/DMLOptions.java", "new_path": "src/main/java/org/tugraz/sysds/api/DMLOptions.java", "diff": "@@ -114,7 +114,8 @@ public class DMLOptions {\nif (lineageType != null){\nif (lineageType.equalsIgnoreCase(\"dedup\"))\ndmlOptions.lineage_dedup = lineageType.equalsIgnoreCase(\"dedup\");\n- else if (lineageType.equalsIgnoreCase(\"reuse_full\"))\n+ else if (lineageType.equalsIgnoreCase(\"reuse_full\")\n+ || lineageType.equalsIgnoreCase(\"reuse\"))\ndmlOptions.linReuseType = ReuseCacheType.REUSE_FULL;\nelse if (lineageType.equalsIgnoreCase(\"reuse_partial\"))\ndmlOptions.linReuseType = ReuseCacheType.REUSE_PARTIAL;\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/component/misc/CLIOptionsParserTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/component/misc/CLIOptionsParserTest.java", "diff": "@@ -147,7 +147,7 @@ public class CLIOptionsParserTest {\n@Test\npublic void testLineageReuseF() throws Exception {\n- String cl = \"systemds -f test.dml -lineage reuse\";\n+ String cl = \"systemds -f test.dml -lineage reuse_full\";\nString[] args = cl.split(\" \");\nDMLOptions o = DMLOptions.parseCLArguments(args);\nAssert.assertEquals(true, o.lineage);\n@@ -157,7 +157,7 @@ public class CLIOptionsParserTest {\n@Test\npublic void testLineageReuseP() throws Exception {\n- String cl = \"systemds -f test.dml -lineage reuse\";\n+ String cl = \"systemds -f test.dml -lineage reuse_partial\";\nString[] args = cl.split(\" \");\nDMLOptions o = DMLOptions.parseCLArguments(args);\nAssert.assertEquals(true, o.lineage);\n@@ -166,7 +166,7 @@ public class CLIOptionsParserTest {\n}\n@Test\npublic void testLineageReuseH() throws Exception {\n- String cl = \"systemds -f test.dml -lineage reuse\";\n+ String cl = \"systemds -f test.dml -lineage reuse_hybrid\";\nString[] args = cl.split(\" \");\nDMLOptions o = DMLOptions.parseCLArguments(args);\nAssert.assertEquals(true, o.lineage);\n@@ -186,7 +186,7 @@ public class CLIOptionsParserTest {\n@Test\npublic void testLineageDedupAndReuseF() throws Exception {\n- String cl = \"systemds -f test.dml -lineage dedup reuse\";\n+ String cl = \"systemds -f test.dml -lineage dedup reuse_full\";\nString[] args = cl.split(\" \");\nDMLOptions o = DMLOptions.parseCLArguments(args);\nAssert.assertEquals(true, o.lineage);\n@@ -196,7 +196,7 @@ public class CLIOptionsParserTest {\n@Test\npublic void testLineageDedupAndReuseP() throws Exception {\n- String cl = \"systemds -f test.dml -lineage dedup reuse\";\n+ String cl = \"systemds -f test.dml -lineage dedup reuse_partial\";\nString[] args = cl.split(\" \");\nDMLOptions o = DMLOptions.parseCLArguments(args);\nAssert.assertEquals(true, o.lineage);\n@@ -206,7 +206,7 @@ public class CLIOptionsParserTest {\n@Test\npublic void testLineageDedupAndReusuH() throws Exception {\n- String cl = \"systemds -f test.dml -lineage dedup reuse\";\n+ String cl = \"systemds -f test.dml -lineage dedup reuse_hybrid\";\nString[] args = cl.split(\" \");\nDMLOptions o = DMLOptions.parseCLArguments(args);\nAssert.assertEquals(true, o.lineage);\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/ReorgOp.java", "new_path": "src/main/java/org/tugraz/sysds/hops/ReorgOp.java", "diff": "@@ -309,9 +309,9 @@ public class ReorgOp extends MultiThreadedHop\n// special cases where an input or output dimension is zero (i.e., 0x5 -> 1x0 is valid)\n// #nnz in output is exactly the same as in input\nif( dc.dimsKnown() ) {\n- if( rowsKnown() )\n+ if( rowsKnown() && getDim1()!=0 )\nret = new MatrixCharacteristics(getDim1(), dc.getRows()*dc.getCols()/getDim1(), -1, dc.getNonZeros());\n- else if( colsKnown() )\n+ else if( colsKnown() && getDim2()!=0 )\nret = new MatrixCharacteristics(dc.getRows()*dc.getCols()/getDim2(), getDim2(), -1, dc.getNonZeros());\nelse if( dimsKnown() )\nret = new MatrixCharacteristics(getDim1(), getDim2(), -1, -1);\n@@ -440,9 +440,9 @@ public class ReorgOp extends MultiThreadedHop\nrefreshColsParameterInformation(input3); //refresh cols\nsetNnz(input1.getNnz());\nif (!dimsKnown() && input1.dimsKnown()) { //reshape allows to infer dims, if input and 1 dim known\n- if (rowsKnown())\n+ if (rowsKnown() && getDim1()!=0)\nsetDim2(input1.getLength() / getDim1());\n- else if (colsKnown())\n+ else if (colsKnown() && getDim2()!=0)\nsetDim1(input1.getLength() / getDim2());\n}\n} else {\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/parser/DMLTranslator.java", "new_path": "src/main/java/org/tugraz/sysds/parser/DMLTranslator.java", "diff": "package org.tugraz.sysds.parser;\nimport java.util.ArrayList;\n-import java.util.Arrays;\nimport java.util.HashMap;\nimport java.util.HashSet;\nimport java.util.Iterator;\n@@ -2060,15 +2059,13 @@ public class DMLTranslator\ntmpMatrix.add( 3, !paramHops.containsKey(DataExpression.RAND_DIMS) ?\nnew LiteralOp(\"-1\") : paramHops.get(DataExpression.RAND_DIMS));\ntmpMatrix.add( 4, paramHops.get(DataExpression.RAND_BY_ROW) );\n- System.out.println(Arrays.toString(tmpMatrix.toArray()));\ncurrBuiltinOp = new ReorgOp(target.getName(), target.getDataType(),\ntarget.getValueType(), ReOrgOp.RESHAPE, tmpMatrix);\nbreak;\ndefault:\nthrow new ParseException(source.printErrorLocation() +\n- \"processDataExpression():: Unknown operation: \"\n- + source.getOpCode());\n+ \"processDataExpression():: Unknown operation: \" + source.getOpCode());\n}\n//set identifier meta data (incl dimensions and blocksizes)\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/misc/RewriteFuseBinaryOpChainTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/misc/RewriteFuseBinaryOpChainTest.java", "diff": "@@ -31,6 +31,7 @@ import org.tugraz.sysds.runtime.matrix.data.MatrixValue.CellIndex;\nimport org.tugraz.sysds.test.AutomatedTestBase;\nimport org.tugraz.sysds.test.TestConfiguration;\nimport org.tugraz.sysds.test.TestUtils;\n+import org.tugraz.sysds.utils.Statistics;\n/**\n* Regression test for function recompile-once issue with literal replacement.\n@@ -42,6 +43,7 @@ public class RewriteFuseBinaryOpChainTest extends AutomatedTestBase\nprivate static final String TEST_NAME2 = \"RewriteFuseBinaryOpChainTest2\"; //-* (X-s*Y)\nprivate static final String TEST_NAME3 = \"RewriteFuseBinaryOpChainTest3\"; //+* (s*Y+X)\nprivate static final String TEST_NAME4 = \"RewriteFuseBinaryOpChainTest4\"; //outer(X, s*Y, \"+\") not applied\n+ private static final String TEST_NAME5 = \"RewriteFuseBinaryOpChainTest5\"; //2 quantiles\nprivate static final String TEST_DIR = \"functions/misc/\";\nprivate static final String TEST_CLASS_DIR = TEST_DIR + RewriteFuseBinaryOpChainTest.class.getSimpleName() + \"/\";\n@@ -55,6 +57,7 @@ public class RewriteFuseBinaryOpChainTest extends AutomatedTestBase\naddTestConfiguration( TEST_NAME2, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME2, new String[] { \"R\" }) );\naddTestConfiguration( TEST_NAME3, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME3, new String[] { \"R\" }) );\naddTestConfiguration( TEST_NAME4, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME4, new String[] { \"R\" }) );\n+ addTestConfiguration( TEST_NAME5, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME5, new String[] { \"R\" }) );\n}\n@Test\n@@ -129,6 +132,16 @@ public class RewriteFuseBinaryOpChainTest extends AutomatedTestBase\ntestFuseBinaryChain( TEST_NAME4, true, ExecType.CP);\n}\n+ @Test\n+ public void testQuantilesNoRewriteCP() {\n+ testFuseBinaryChain(TEST_NAME5, false, ExecType.CP );\n+ }\n+\n+ @Test\n+ public void testQuantilesRewriteCP() {\n+ testFuseBinaryChain(TEST_NAME5, true, ExecType.CP);\n+ }\n+\nprivate void testFuseBinaryChain( String testname, boolean rewrites, ExecType instType )\n{\nExecMode platformOld = rtplatform;\n@@ -163,13 +176,16 @@ public class RewriteFuseBinaryOpChainTest extends AutomatedTestBase\nHashMap<CellIndex, Double> dmlfile = readDMLMatrixFromHDFS(\"S\");\nHashMap<CellIndex, Double> rfile = readRMatrixFromFS(\"S\");\nAssert.assertTrue(TestUtils.compareMatrices(dmlfile, rfile, eps, \"Stat-DML\", \"Stat-R\"));\n+\n+ if( testname.equals(TEST_NAME5) ) {\n+ //check for common subexpression elimination at lop level (independent of rewrites)\n+ Assert.assertTrue(Statistics.getCPHeavyHitterCount(\"qsort\") == 1);\n}\n- finally\n- {\n+ }\n+ finally {\nOptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = rewritesOld;\nrtplatform = platformOld;\nDMLScript.USE_LOCAL_SPARK_CONFIG = sparkConfigOld;\n}\n-\n}\n}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/misc/RewriteFuseBinaryOpChainTest5.R", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+args<-commandArgs(TRUE)\n+options(digits=22)\n+library(\"Matrix\")\n+\n+X = seq(1,100) + seq(100,1);\n+r1 = quantile(X, 0.05);\n+r2 = quantile(X, 0.95);\n+S = as.matrix(r1 + r2);\n+\n+writeMM(as(S, \"CsparseMatrix\"), paste(args[2], \"S\", sep=\"\"));\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/misc/RewriteFuseBinaryOpChainTest5.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = seq(1,100) + seq(100,1);\n+while(FALSE){}\n+\n+r1 = quantile(X, 0.05);\n+r2 = quantile(X, 0.95);\n+S = as.matrix(r1 + r2);\n+\n+while(FALSE){}\n+write(S,$1);\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -125,3 +125,6 @@ SYSTEMDS-160 Tensor Compiler/Runtime\n* 176 Reuse rewrite for cbind/rbind-elementwise */+\n* 177 Reuse rewrite for aggregate OK\n* 178 Compiler assisted reuse (eg. CV, lmCG)\n+\n+SYSTEMDS-180 New Builtin Functions II OK\n+ * 181 Builtin function for naive bayes OK\n\\ No newline at end of file\n" }, { "change_type": "ADD", "old_path": null, "new_path": "scripts/builtin/naivebayes.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Modifications Copyright 2019 Graz University of Technology\n+#\n+# Licensed to the Apache Software Foundation (ASF) under one\n+# or more contributor license agreements. See the NOTICE file\n+# distributed with this work for additional information\n+# regarding copyright ownership. The ASF licenses this file\n+# to you under the Apache License, Version 2.0 (the\n+# \"License\"); you may not use this file except in compliance\n+# with the License. You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing,\n+# software distributed under the License is distributed on an\n+# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+# KIND, either express or implied. See the License for the\n+# specific language governing permissions and limitations\n+# under the License.\n+#\n+#-------------------------------------------------------------\n+\n+m_naivebayes = function(Matrix[Double] D, Matrix[Double] C, Double laplace = 1, Boolean verbose = TRUE)\n+ return (Matrix[Double] prior, Matrix[Double] classConditionals)\n+{\n+ laplaceCorrection = laplace;\n+ numRows = nrow(D);\n+ numFeatures = ncol(D);\n+ numClasses = max(C);\n+\n+ # Compute conditionals\n+ # Compute the feature counts for each class\n+ classFeatureCounts = aggregate(target=D, groups=C, fn=\"sum\", ngroups=as.integer(numClasses));\n+\n+ # Compute the total feature count for each class\n+ # and add the number of features to this sum\n+ # for subsequent regularization (Laplace's rule)\n+ classSums = rowSums(classFeatureCounts) + numFeatures*laplaceCorrection;\n+\n+ # Compute class conditional probabilities\n+ classConditionals = (classFeatureCounts + laplaceCorrection) / classSums;\n+\n+ # Compute class priors\n+ classCounts = aggregate(target=C, groups=C, fn=\"count\", ngroups=as.integer(numClasses));\n+ prior = classCounts / numRows;\n+\n+ # Compute accuracy on training set\n+ if( verbose ) {\n+ logProbs = D %*% t(log(classConditionals)) + t(log(prior));\n+ acc = sum(rowIndexMax(logProbs) == C) / numRows * 100;\n+ print(\"Training Accuracy (%): \" + acc);\n+ }\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "new_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "diff": "@@ -113,6 +113,7 @@ public enum Builtins {\nNCOL(\"ncol\", false),\nNORMALIZE(\"normalize\", true),\nNROW(\"nrow\", false),\n+ NAIVEBAYES(\"naivebayes\", true, false),\nOUTER(\"outer\", false),\nOUTLIER(\"outlier\", true, false), //TODO parameterize opposite\nPPRED(\"ppred\", false),\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinNaiveBayesTest.java", "diff": "+/*\n+ * Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed under the Apache License, Version 2.0 (the \"License\");\n+ * you may not use this file except in compliance with the License.\n+ * You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing, software\n+ * distributed under the License is distributed on an \"AS IS\" BASIS,\n+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+ * See the License for the specific language governing permissions and\n+ * limitations under the License.\n+ */\n+\n+package org.tugraz.sysds.test.functions.builtin;\n+\n+import java.util.ArrayList;\n+import java.util.HashMap;\n+import java.util.List;\n+\n+import org.junit.Test;\n+import org.tugraz.sysds.runtime.matrix.data.MatrixValue.CellIndex;\n+import org.tugraz.sysds.test.AutomatedTestBase;\n+import org.tugraz.sysds.test.TestConfiguration;\n+import org.tugraz.sysds.test.TestUtils;\n+\n+public class BuiltinNaiveBayesTest extends AutomatedTestBase\n+{\n+ private final static String TEST_NAME = \"NaiveBayes\";\n+ private final static String TEST_DIR = \"functions/builtin/\";\n+ private final static String TEST_CLASS_DIR = TEST_DIR + BuiltinNaiveBayesTest.class.getSimpleName() + \"/\";\n+\n+ private final static int numClasses = 10;\n+\n+ @Override\n+ public void setUp() {\n+ addTestConfiguration(TEST_NAME,new TestConfiguration(TEST_CLASS_DIR, TEST_NAME,new String[]{\"B\"}));\n+ }\n+\n+ @Test\n+ public void testSmallDense() {\n+ testNaiveBayes(100, 50, 0.7);\n+ }\n+\n+ @Test\n+ public void testLargeDense() {\n+ testNaiveBayes(10000, 750, 0.7);\n+ }\n+\n+ @Test\n+ public void testSmallSparse() {\n+ testNaiveBayes(100, 50, 0.01);\n+ }\n+\n+ @Test\n+ public void testLargeSparse() {\n+ testNaiveBayes(10000, 750, 0.01);\n+ }\n+\n+ public void testNaiveBayes(int rows, int cols, double sparsity)\n+ {\n+ loadTestConfiguration(getTestConfiguration(TEST_NAME));\n+ String HOME = SCRIPT_DIR + TEST_DIR;\n+ fullDMLScriptName = HOME + TEST_NAME + \".dml\";\n+\n+ int classes = numClasses;\n+ double laplace_correction = 1;\n+\n+ List<String> proArgs = new ArrayList<>();\n+ proArgs.add(\"-stats\");\n+ proArgs.add(\"-args\");\n+ proArgs.add(input(\"X\"));\n+ proArgs.add(input(\"Y\"));\n+ proArgs.add(String.valueOf(classes));\n+ proArgs.add(String.valueOf(laplace_correction));\n+ proArgs.add(output(\"prior\"));\n+ proArgs.add(output(\"conditionals\"));\n+ programArgs = proArgs.toArray(new String[proArgs.size()]);\n+\n+ rCmd = getRCmd(inputDir(), Integer.toString(classes), Double.toString(laplace_correction), expectedDir());\n+\n+ double[][] X = getRandomMatrix(rows, cols, 0, 1, sparsity, -1);\n+ double[][] Y = getRandomMatrix(rows, 1, 0, 1, 1, -1);\n+ for(int i=0; i<rows; i++){\n+ Y[i][0] = (int)(Y[i][0]*classes) + 1;\n+ Y[i][0] = (Y[i][0] > classes) ? classes : Y[i][0];\n+ }\n+\n+ writeInputMatrixWithMTD(\"X\", X, true);\n+ writeInputMatrixWithMTD(\"Y\", Y, true);\n+\n+ runTest(true, EXCEPTION_NOT_EXPECTED, null, -1);\n+\n+ runRScript(true);\n+\n+ HashMap<CellIndex, Double> priorR = readRMatrixFromFS(\"prior\");\n+ HashMap<CellIndex, Double> priorSYSTEMDS= readDMLMatrixFromHDFS(\"prior\");\n+ HashMap<CellIndex, Double> conditionalsR = readRMatrixFromFS(\"conditionals\");\n+ HashMap<CellIndex, Double> conditionalsSYSTEMDS = readDMLMatrixFromHDFS(\"conditionals\");\n+ TestUtils.compareMatrices(priorR, priorSYSTEMDS, Math.pow(10, -12), \"priorR\", \"priorSYSTEMDS\");\n+ TestUtils.compareMatrices(conditionalsR, conditionalsSYSTEMDS, Math.pow(10.0, -12.0), \"conditionalsR\", \"conditionalsSYSTEMDS\");\n+ }\n+}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/builtin/NaiveBayes.R", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+args <- commandArgs(TRUE)\n+\n+library(\"Matrix\")\n+\n+D = as.matrix(readMM(paste(args[1], \"X.mtx\", sep=\"\")))\n+C = as.matrix(readMM(paste(args[1], \"Y.mtx\", sep=\"\")))\n+\n+# reading input args\n+numClasses = as.integer(args[2]);\n+laplace_correction = as.double(args[3]);\n+\n+numRows = nrow(D)\n+numFeatures = ncol(D)\n+\n+# Compute conditionals\n+\n+# Compute the feature counts for each class\n+classFeatureCounts = matrix(0, numClasses, numFeatures)\n+for (i in 1:numFeatures) {\n+ Col = D[,i]\n+ classFeatureCounts[,i] = aggregate(as.vector(Col), by=list(as.vector(C)), FUN=sum)[,2];\n+}\n+\n+# Compute the total feature count for each class\n+# and add the number of features to this sum\n+# for subsequent regularization (Laplace's rule)\n+classSums = rowSums(classFeatureCounts) + numFeatures*laplace_correction\n+\n+# Compute class conditional probabilities\n+ones = matrix(1, 1, numFeatures)\n+repClassSums = classSums %*% ones;\n+class_conditionals = (classFeatureCounts + laplace_correction) / repClassSums;\n+\n+# Compute class priors\n+class_counts = aggregate(as.vector(C), by=list(as.vector(C)), FUN=length)[,2]\n+class_prior = class_counts / numRows;\n+\n+# Compute accuracy on training set\n+ones = matrix(1, numRows, 1)\n+D_w_ones = cbind(D, ones)\n+model = cbind(class_conditionals, class_prior)\n+log_probs = D_w_ones %*% t(log(model))\n+pred = max.col(log_probs,ties.method=\"last\");\n+acc = sum(pred == C) / numRows * 100\n+\n+print(paste(\"Training Accuracy (%): \", acc, sep=\"\"))\n+\n+# write out the model\n+writeMM(as(class_prior, \"CsparseMatrix\"), paste(args[4], \"prior\", sep=\"\"));\n+writeMM(as(class_conditionals, \"CsparseMatrix\"), paste(args[4], \"conditionals\", sep=\"\"));\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/builtin/NaiveBayes.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1);\n+y = read($2);\n+[prior, conditionals] = naivebayes(D=X, C=y, laplace=$4);\n+write(prior, $5);\n+write(conditionals, $6);\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -128,3 +128,4 @@ SYSTEMDS-170 Lineage full and partial reuse\nSYSTEMDS-180 New Builtin Functions II OK\n* 181 Builtin function for naive bayes OK\n+ * 182 Builtin function for typeof (frame schema detection) OK\n\\ No newline at end of file\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "new_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "diff": "@@ -151,6 +151,7 @@ public enum Builtins {\nTAN(\"tan\", false),\nTANH(\"tanh\", false),\nTRACE(\"trace\", false),\n+ TYPEOF(\"typeOf\", false),\nVAR(\"var\", false),\nXOR(\"xor\", false),\nWINSORIZE(\"winsorize\", true, false), //TODO parameterize w/ prob, min/max val\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/Hop.java", "new_path": "src/main/java/org/tugraz/sysds/hops/Hop.java", "diff": "@@ -973,7 +973,7 @@ public abstract class Hop implements ParseInfo\nNOT, ABS, SIN, COS, TAN, ASIN, ACOS, ATAN, SINH, COSH, TANH, SIGN, SQRT, LOG, EXP,\nCAST_AS_SCALAR, CAST_AS_MATRIX, CAST_AS_FRAME, CAST_AS_DOUBLE, CAST_AS_INT, CAST_AS_BOOLEAN,\nPRINT, ASSERT, EIGEN, NROW, NCOL, LENGTH, ROUND, IQM, STOP, CEIL, FLOOR, MEDIAN, INVERSE, CHOLESKY,\n- SVD, EXISTS, LINEAGE,\n+ SVD, EXISTS, LINEAGE, TYPEOF,\n//cumulative sums, products, extreme values\nCUMSUM, CUMPROD, CUMMIN, CUMMAX, CUMSUMPROD,\n//fused ML-specific operators for performance\n@@ -1241,7 +1241,8 @@ public abstract class Hop implements ParseInfo\nHopsOpOp1LopsU.put(OpOp1.CAST_AS_SCALAR, org.tugraz.sysds.lops.Unary.OperationTypes.NOTSUPPORTED);\nHopsOpOp1LopsU.put(OpOp1.CAST_AS_MATRIX, org.tugraz.sysds.lops.Unary.OperationTypes.NOTSUPPORTED);\nHopsOpOp1LopsU.put(OpOp1.SPROP, org.tugraz.sysds.lops.Unary.OperationTypes.SPROP);\n- HopsOpOp1LopsU.put(OpOp1.SIGMOID, org.tugraz.sysds.lops.Unary.OperationTypes.SIGMOID);\n+ HopsOpOp1LopsU.put(OpOp1.SIGMOID, Unary.OperationTypes.SIGMOID);\n+ HopsOpOp1LopsU.put(OpOp1.TYPEOF, Unary.OperationTypes.TYPEOF);\nHopsOpOp1LopsU.put(OpOp1.LOG_NZ, org.tugraz.sysds.lops.Unary.OperationTypes.LOG_NZ);\nHopsOpOp1LopsU.put(OpOp1.CAST_AS_MATRIX, org.tugraz.sysds.lops.Unary.OperationTypes.CAST_AS_MATRIX);\nHopsOpOp1LopsU.put(OpOp1.CAST_AS_FRAME, org.tugraz.sysds.lops.Unary.OperationTypes.CAST_AS_FRAME);\n@@ -1281,6 +1282,7 @@ public abstract class Hop implements ParseInfo\nHopsOpOp1LopsUS.put(OpOp1.CEIL, org.tugraz.sysds.lops.UnaryCP.OperationTypes.CEIL);\nHopsOpOp1LopsUS.put(OpOp1.FLOOR, org.tugraz.sysds.lops.UnaryCP.OperationTypes.FLOOR);\nHopsOpOp1LopsUS.put(OpOp1.STOP, org.tugraz.sysds.lops.UnaryCP.OperationTypes.STOP);\n+ HopsOpOp1LopsUS.put(OpOp1.TYPEOF, UnaryCP.OperationTypes.TYPEOF);\n}\nprotected static final HashMap<OpOp3, Ternary.OperationType> HopsOpOp3Lops;\n@@ -1343,6 +1345,7 @@ public abstract class Hop implements ParseInfo\nHopsOpOp12String.put(OpOp1.INVERSE, \"inv\");\nHopsOpOp12String.put(OpOp1.SPROP, \"sprop\");\nHopsOpOp12String.put(OpOp1.SIGMOID, \"sigmoid\");\n+ HopsOpOp12String.put(OpOp1.TYPEOF, \"typeOf\");\nHopsStringOpOp1 = new HashMap<>();\nfor( Entry<OpOp1,String> e : HopsOpOp12String.entrySet() )\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/hops/UnaryOp.java", "new_path": "src/main/java/org/tugraz/sysds/hops/UnaryOp.java", "diff": "@@ -503,7 +503,7 @@ public class UnaryOp extends MultiThreadedHop\nsetRequiresRecompileIfNecessary();\n//ensure cp exec type for single-node operations\n- if( _op == OpOp1.PRINT || _op == OpOp1.ASSERT || _op == OpOp1.STOP\n+ if( _op == OpOp1.PRINT || _op == OpOp1.ASSERT || _op == OpOp1.STOP || _op == OpOp1.TYPEOF\n|| _op == OpOp1.INVERSE || _op == OpOp1.EIGEN || _op == OpOp1.CHOLESKY || _op == OpOp1.SVD\n|| getInput().get(0).getDataType() == DataType.LIST || isMetadataOperation() )\n{\n@@ -537,6 +537,10 @@ public class UnaryOp extends MultiThreadedHop\nsetDim1(input.getDim1());\nsetDim2(1);\n}\n+ else if(_op == OpOp1.TYPEOF) {\n+ setDim1(1);\n+ setDim2(input.getDim2());\n+ }\nelse //general case\n{\n// If output is a Matrix then this operation is of type (B = op(A))\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/lops/Unary.java", "new_path": "src/main/java/org/tugraz/sysds/lops/Unary.java", "diff": "/*\n+ * Modifications Copyright 2019 Graz University of Technology\n+ *\n* Licensed to the Apache Software Foundation (ASF) under one\n* or more contributor license agreements. See the NOTICE file\n* distributed with this work for additional information\n@@ -45,6 +47,7 @@ public class Unary extends Lop\nCUMSUM, CUMPROD, CUMMIN, CUMMAX, CUMSUMPROD,\nSPROP, SIGMOID, SUBTRACT_NZ, LOG_NZ,\nCAST_AS_MATRIX, CAST_AS_FRAME,\n+ TYPEOF,\nNOTSUPPORTED\n}\n@@ -272,6 +275,9 @@ public class Unary extends Lop\ncase SIGMOID:\nreturn \"sigmoid\";\n+ case TYPEOF:\n+ return \"typeOf\";\n+\ncase CAST_AS_MATRIX:\nreturn UnaryCP.CAST_AS_MATRIX_OPCODE;\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/lops/UnaryCP.java", "new_path": "src/main/java/org/tugraz/sysds/lops/UnaryCP.java", "diff": "@@ -31,7 +31,7 @@ public class UnaryCP extends Lop\npublic enum OperationTypes {\nNOT, ABS, SIN, COS, TAN, ASIN, ACOS, ATAN, SQRT, LOG, EXP, SINH, COSH, TANH,\nCAST_AS_SCALAR, CAST_AS_MATRIX, CAST_AS_FRAME, CAST_AS_DOUBLE, CAST_AS_INT, CAST_AS_BOOLEAN,\n- PRINT, ASSERT, NROW, NCOL, LENGTH, EXISTS, LINEAGE, ROUND, STOP, CEIL, FLOOR, CUMSUM, SOFTMAX\n+ PRINT, ASSERT, NROW, NCOL, LENGTH, EXISTS, LINEAGE, ROUND, STOP, CEIL, FLOOR, CUMSUM, SOFTMAX, TYPEOF\n}\npublic static final String CAST_AS_SCALAR_OPCODE = \"castdts\";\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/parser/BuiltinFunctionExpression.java", "new_path": "src/main/java/org/tugraz/sysds/parser/BuiltinFunctionExpression.java", "diff": "@@ -688,6 +688,14 @@ public class BuiltinFunctionExpression extends DataIdentifier\noutput.setBlocksize(id.getBlocksize());\noutput.setValueType(ValueType.FP64); //matrices always in double\nbreak;\n+ case TYPEOF:\n+ checkNumParameters(1);\n+ checkMatrixFrameParam(getFirstExpr());\n+ output.setDataType(DataType.FRAME);\n+ output.setDimensions(1, id.getDim2());\n+ output.setBlocksize (id.getBlocksize());\n+ output.setValueType(ValueType.STRING);\n+ break;\ncase CAST_AS_FRAME:\ncheckNumParameters(1);\ncheckMatrixScalarParam(getFirstExpr());\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/parser/DMLTranslator.java", "new_path": "src/main/java/org/tugraz/sysds/parser/DMLTranslator.java", "diff": "@@ -2447,6 +2447,7 @@ public class DMLTranslator\ncase CUMSUMPROD:\ncase CUMMIN:\ncase CUMMAX:\n+\ncurrBuiltinOp = new UnaryOp(target.getName(), target.getDataType(), target.getValueType(),\nOpOp1.valueOf(source.getOpCode().name()), expr);\nbreak;\n@@ -2565,6 +2566,7 @@ public class DMLTranslator\ncase INVERSE:\ncase CHOLESKY:\n+ case TYPEOF:\ncurrBuiltinOp = new UnaryOp(target.getName(), target.getDataType(), target.getValueType(),\nOpOp1.valueOf(source.getOpCode().name()), expr);\nbreak;\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/functionobjects/Builtin.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/functionobjects/Builtin.java", "diff": "@@ -49,7 +49,7 @@ public class Builtin extends ValueFunction\npublic enum BuiltinCode { SIN, COS, TAN, SINH, COSH, TANH, ASIN, ACOS, ATAN, LOG, LOG_NZ, MIN,\nMAX, ABS, SIGN, SQRT, EXP, PLOGP, PRINT, PRINTF, NROW, NCOL, LENGTH, LINEAGE, ROUND, MAXINDEX, MININDEX,\n- STOP, CEIL, FLOOR, CUMSUM, CUMPROD, CUMMIN, CUMMAX, CUMSUMPROD, INVERSE, SPROP, SIGMOID, EVAL, LIST }\n+ STOP, CEIL, FLOOR, CUMSUM, CUMPROD, CUMMIN, CUMMAX, CUMSUMPROD, INVERSE, SPROP, SIGMOID, EVAL, LIST, TYPEOF }\npublic BuiltinCode bFunc;\nprivate static final boolean FASTMATH = true;\n@@ -96,6 +96,7 @@ public class Builtin extends ValueFunction\nString2BuiltinCode.put( \"inverse\", BuiltinCode.INVERSE);\nString2BuiltinCode.put( \"sprop\", BuiltinCode.SPROP);\nString2BuiltinCode.put( \"sigmoid\", BuiltinCode.SIGMOID);\n+ String2BuiltinCode.put( \"typeOf\", BuiltinCode.TYPEOF);\n}\n// We should create one object for every builtin function that we support\n@@ -104,7 +105,7 @@ public class Builtin extends ValueFunction\nprivate static Builtin absObj = null, signObj = null, sqrtObj = null, expObj = null, plogpObj = null, printObj = null, printfObj;\nprivate static Builtin nrowObj = null, ncolObj = null, lengthObj = null, roundObj = null, ceilObj=null, floorObj=null;\nprivate static Builtin inverseObj=null, cumsumObj=null, cumprodObj=null, cumminObj=null, cummaxObj=null, cumsprodObj=null;\n- private static Builtin stopObj = null, spropObj = null, sigmoidObj = null;\n+ private static Builtin stopObj = null, spropObj = null, sigmoidObj = null, typeOfObj = null ;\nprivate Builtin(BuiltinCode bf) {\nbFunc = bf;\n@@ -288,6 +289,10 @@ public class Builtin extends ValueFunction\nsigmoidObj = new Builtin(BuiltinCode.SIGMOID);\nreturn sigmoidObj;\n+ case TYPEOF:\n+ if ( typeOfObj == null )\n+ typeOfObj = new Builtin(BuiltinCode.TYPEOF);\n+ return typeOfObj;\ndefault:\n// Unknown code --> return null\nreturn null;\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/CPInstructionParser.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/CPInstructionParser.java", "diff": "@@ -181,7 +181,7 @@ public class CPInstructionParser extends InstructionParser\nString2CPInstructionType.put( \"cholesky\",CPType.Unary);\nString2CPInstructionType.put( \"sprop\", CPType.Unary);\nString2CPInstructionType.put( \"sigmoid\", CPType.Unary);\n-\n+ String2CPInstructionType.put( \"typeOf\", CPType.Unary);\nString2CPInstructionType.put( \"printf\", CPType.BuiltinNary);\nString2CPInstructionType.put( \"cbind\", CPType.BuiltinNary);\nString2CPInstructionType.put( \"rbind\", CPType.BuiltinNary);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/UnaryCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/UnaryCPInstruction.java", "diff": "/*\n+ * Modifications Copyright 2019 Graz University of Technology\n+ *\n* Licensed to the Apache Software Foundation (ASF) under one\n* or more contributor license agreements. See the NOTICE file\n* distributed with this work for additional information\n@@ -76,8 +78,9 @@ public abstract class UnaryCPInstruction extends ComputationCPInstruction {\nelse if(in.getDataType() == DataType.MATRIX)\nreturn new UnaryMatrixCPInstruction(LibCommonsMath.isSupportedUnaryOperation(opcode) ?\nnull : InstructionUtils.parseUnaryOperator(opcode), in, out, opcode, str);\n+ else if(in.getDataType() == DataType.FRAME)\n+ return new UnaryFrameCPInstruction(InstructionUtils.parseUnaryOperator(opcode), in, out, opcode, str);\n}\n-\nreturn null;\n}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/UnaryFrameCPInstruction.java", "diff": "+/*\n+ * Modifications Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed to the Apache Software Foundation (ASF) under one\n+ * or more contributor license agreements. See the NOTICE file\n+ * distributed with this work for additional information\n+ * regarding copyright ownership. The ASF licenses this file\n+ * to you under the Apache License, Version 2.0 (the\n+ * \"License\"); you may not use this file except in compliance\n+ * with the License. You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing,\n+ * software distributed under the License is distributed on an\n+ * \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+ * KIND, either express or implied. See the License for the\n+ * specific language governing permissions and limitations\n+ * under the License.\n+ */\n+\n+package org.tugraz.sysds.runtime.instructions.cp;\n+\n+import org.tugraz.sysds.runtime.controlprogram.context.ExecutionContext;\n+import org.tugraz.sysds.runtime.matrix.data.FrameBlock;\n+import org.tugraz.sysds.runtime.matrix.operators.Operator;\n+\n+public class UnaryFrameCPInstruction extends UnaryCPInstruction {\n+ protected UnaryFrameCPInstruction(Operator op, CPOperand in, CPOperand out, String opcode, String instr) {\n+\n+ super(CPType.Unary, op, in, out, opcode, instr);\n+ }\n+\n+ @Override\n+ public void processInstruction(ExecutionContext ec) {\n+ FrameBlock inBlock = ec.getFrameInput(input1.getName());\n+ FrameBlock retBlock = inBlock.getSchemaTypeOf();\n+ ec.releaseFrameInput(input1.getName());\n+ ec.setFrameOutput(output.getName(), retBlock);\n+ }\n+}\n\\ No newline at end of file\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/matrix/data/FrameBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/matrix/data/FrameBlock.java", "diff": "/*\n+ * Modifications Copyright 2019 Graz University of Technology\n+ *\n* Licensed to the Apache Software Foundation (ASF) under one\n* or more contributor license agreements. See the NOTICE file\n* distributed with this work for additional information\n@@ -1180,6 +1182,13 @@ public class FrameBlock implements Writable, CacheBlock, Externalizable\nreturn result;\n}\n+ public FrameBlock getSchemaTypeOf() {\n+ FrameBlock fb = new FrameBlock(\n+ UtilFunctions.nCopies(getNumColumns(), ValueType.STRING));\n+ fb.appendRow(Arrays.stream(_schema)\n+ .map(vt -> vt.toString()).toArray(String[]::new));\n+ return fb;\n+ }\n///////\n// row iterators (over strings and boxed objects)\n@@ -1467,7 +1476,7 @@ public class FrameBlock implements Writable, CacheBlock, Externalizable\n}\n@Override\npublic void append(String value) {\n- append((value!=null)?Long.parseLong(value):null);\n+ append((value!=null)?Long.parseLong(value.trim()):null);\n}\n@Override\npublic void append(Long value) {\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/java/org/tugraz/sysds/test/functions/frame/TypeOfTest.java", "diff": "+/*\n+ * Modifications Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed to the Apache Software Foundation (ASF) under one\n+ * or more contributor license agreements. See the NOTICE file\n+ * distributed with this work for additional information\n+ * regarding copyright ownership. The ASF licenses this file\n+ * to you under the Apache License, Version 2.0 (the\n+ * \"License\"); you may not use this file except in compliance\n+ * with the License. You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing,\n+ * software distributed under the License is distributed on an\n+ * \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+ * KIND, either express or implied. See the License for the\n+ * specific language governing permissions and limitations\n+ * under the License.\n+ */\n+package org.tugraz.sysds.test.functions.frame;\n+\n+import org.junit.AfterClass;\n+import org.junit.Assert;\n+import org.junit.BeforeClass;\n+import org.junit.Test;\n+import org.tugraz.sysds.api.DMLScript;\n+import org.tugraz.sysds.common.Types;\n+import org.tugraz.sysds.lops.LopProperties.ExecType;\n+import org.tugraz.sysds.runtime.io.*;\n+import org.tugraz.sysds.runtime.matrix.data.FrameBlock;\n+import org.tugraz.sysds.runtime.matrix.data.InputInfo;\n+import org.tugraz.sysds.runtime.matrix.data.OutputInfo;\n+import org.tugraz.sysds.runtime.meta.MatrixCharacteristics;\n+import org.tugraz.sysds.runtime.util.HDFSTool;\n+import org.tugraz.sysds.runtime.util.UtilFunctions;\n+import org.tugraz.sysds.test.AutomatedTestBase;\n+import org.tugraz.sysds.test.TestConfiguration;\n+import org.tugraz.sysds.test.TestUtils;\n+\n+public class TypeOfTest extends AutomatedTestBase {\n+ private final static String TEST_NAME = \"TypeOf\";\n+ private final static String TEST_DIR = \"functions/frame/\";\n+ private static final String TEST_CLASS_DIR = TEST_DIR + TypeOfTest.class.getSimpleName() + \"/\";\n+\n+ private final static Types.ValueType[] schemaStrings = new Types.ValueType[]{Types.ValueType.STRING, Types.ValueType.STRING, Types.ValueType.STRING};\n+ private final static Types.ValueType[] schemaMixed = new Types.ValueType[]{Types.ValueType.STRING, Types.ValueType.FP64, Types.ValueType.INT64, Types.ValueType.BOOLEAN};\n+\n+ private final static int rows = 50;\n+\n+ @BeforeClass\n+ public static void init() {\n+ TestUtils.clearDirectory(TEST_DATA_DIR + TEST_CLASS_DIR);\n+ }\n+\n+ @AfterClass\n+ public static void cleanUp() {\n+ if (TEST_CACHE_ENABLED) {\n+ TestUtils.clearDirectory(TEST_DATA_DIR + TEST_CLASS_DIR);\n+ }\n+ }\n+\n+ private static void initFrameData(FrameBlock frame, double[][] data, Types.ValueType[] lschema) {\n+ Object[] row1 = new Object[lschema.length];\n+ for (int i = 0; i < rows; i++) {\n+ for (int j = 0; j < lschema.length; j++)\n+ data[i][j] = UtilFunctions.objectToDouble(lschema[j],\n+ row1[j] = UtilFunctions.doubleToObject(lschema[j], data[i][j]));\n+ frame.appendRow(row1);\n+ }\n+ }\n+\n+ @Override\n+ public void setUp() {\n+ TestUtils.clearAssertionInformation();\n+ addTestConfiguration(TEST_NAME, new TestConfiguration(TEST_CLASS_DIR, TEST_NAME, new String[]{\"B\"}));\n+ if (TEST_CACHE_ENABLED)\n+ setOutAndExpectedDeletionDisabled(true);\n+ }\n+\n+ @Test\n+ public void testTypeOfCP() {\n+ runtypeOfTest(schemaStrings, rows, schemaStrings.length, ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testTypeOfSpark() {\n+ runtypeOfTest(schemaStrings, rows, schemaStrings.length, ExecType.SPARK);\n+ }\n+\n+ @Test\n+ public void testTypeOfCPD2() {\n+ runtypeOfTest(schemaMixed, rows, schemaMixed.length, ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testTypeOfSparkD2() {\n+ runtypeOfTest(schemaMixed, rows, schemaMixed.length, ExecType.SPARK);\n+ }\n+\n+ private void runtypeOfTest(Types.ValueType[] schema, int rows, int cols, ExecType et) {\n+ if (et == ExecType.SPARK)\n+ DMLScript.USE_LOCAL_SPARK_CONFIG = true;\n+ try {\n+ getAndLoadTestConfiguration(TEST_NAME);\n+ String HOME = SCRIPT_DIR + TEST_DIR;\n+ fullDMLScriptName = HOME + TEST_NAME + \".dml\";\n+ programArgs = new String[]{\"-explain\", \"-args\", input(\"A\"), String.valueOf(rows), String.valueOf(cols), output(\"B\")};\n+ //data generation\n+ double[][] A = getRandomMatrix(rows, schema.length, -10, 10, 0.9, 2373);\n+ FrameBlock frame1 = new FrameBlock(schema);\n+ initFrameData(frame1, A, schema);\n+\n+ //write frame data to hdfs\n+ FrameWriter writer = FrameWriterFactory.createFrameWriter(OutputInfo.CSVOutputInfo);\n+ writer.writeFrameToHDFS(frame1, input(\"A\"), rows, schema.length);\n+ //write meta file\n+ HDFSTool.writeMetaDataFile(input(\"A.mtd\"), Types.ValueType.FP64, schema, Types.DataType.FRAME, new MatrixCharacteristics(rows, schema.length, 1000), OutputInfo.CSVOutputInfo);\n+\n+ //run testcase\n+ runTest(true, false, null, -1);\n+\n+ //read frame data from hdfs (not via readers to test physical schema)\n+ FrameReader reader = FrameReaderFactory.createFrameReader(InputInfo.BinaryBlockInputInfo);\n+ FrameBlock frame2 = ((FrameReaderBinaryBlock) reader).readFirstBlock(output(\"B\"));\n+\n+ //verify output schema\n+ for (int i = 0; i < schema.length; i++) {\n+ Assert.assertEquals(\"Wrong result: \" + frame2.getSchema()[i] + \".\",\n+ schema[i].toString(), frame2.get(0, i));\n+ }\n+ }\n+ catch (Exception ex) {\n+ ex.printStackTrace();\n+ throw new RuntimeException(ex);\n+ }\n+ }\n+}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/frame/TypeOf.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Modifications Copyright 2019 Graz University of Technology\n+#\n+# Licensed to the Apache Software Foundation (ASF) under one\n+# or more contributor license agreements. See the NOTICE file\n+# distributed with this work for additional information\n+# regarding copyright ownership. The ASF licenses this file\n+# to you under the Apache License, Version 2.0 (the\n+# \"License\"); you may not use this file except in compliance\n+# with the License. You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing,\n+# software distributed under the License is distributed on an\n+# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+# KIND, either express or implied. See the License for the\n+# specific language governing permissions and limitations\n+# under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1, rows=$2, cols=$3, data_type=\"frame\", format=\"csv\");\n+R = typeOf(X);\n+print(toString(R))\n+write(R, $4, format=\"binary\");\n+\n\\ No newline at end of file\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -13,7 +13,7 @@ SYSTEMDS-10 Compiler Rework / Misc\n* 16 Remove instruction patching\n* 17 Refactoring of program block hierarchy OK\n* 18 Improve API for new dml-bodied builtin functions OK\n- * 19 Break append instruction to cbind and rbind\n+ * 19 Cleanup pass cpvar+rmvar to mvvar instructions OK\nSYSTEMDS-20 New Data Model\n* 21 Finalize dense tensor blocks OK\n@@ -130,3 +130,7 @@ SYSTEMDS-170 Lineage full and partial reuse\nSYSTEMDS-180 New Builtin Functions II OK\n* 181 Builtin function for naive bayes OK\n* 182 Builtin function for typeof (frame schema detection) OK\n+\n+\n+Others:\n+ * Break append instruction to cbind and rbind\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/lops/compile/Dag.java", "new_path": "src/main/java/org/tugraz/sysds/lops/compile/Dag.java", "diff": "@@ -1004,7 +1004,8 @@ public class Dag<N extends Lop>\n* @return new list of potentially modified instructions\n*/\nprivate static ArrayList<Instruction> cleanupInstructions(List<Instruction> insts) {\n- //step 1: create mvvar instructions: assignvar s1 s2, rmvar s1 -> mvvar s1 s2\n+ //step 1: create mvvar instructions: assignvar s1 s2, rmvar s1 -> mvvar s1 s2,\n+ // cpvar m1 m2, rmvar m1 --> mvvar m1 m2\nList<Instruction> tmp1 = collapseAssignvarAndRmvarInstructions(insts);\n//step 2: create packed rmvar instructions: rmvar m1, rmvar m2 -> rmvar m1 m2\n@@ -1019,7 +1020,7 @@ public class Dag<N extends Lop>\nwhile( iter.hasNext() ) {\nInstruction inst = iter.next();\nif( iter.hasNext() && inst instanceof VariableCPInstruction\n- && ((VariableCPInstruction)inst).isAssignVariable() ) {\n+ && ((VariableCPInstruction)inst).isAssignOrCopyVariable() ) {\nVariableCPInstruction inst1 = (VariableCPInstruction) inst;\nInstruction inst2 = iter.next();\nif( inst2 instanceof VariableCPInstruction\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/VariableCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/VariableCPInstruction.java", "diff": "@@ -224,20 +224,25 @@ public class VariableCPInstruction extends CPInstruction implements LineageTrace\n}\npublic boolean isRemoveVariableNoFile() {\n- return (opcode == VariableOperationCode.RemoveVariable);\n+ return opcode == VariableOperationCode.RemoveVariable;\n}\npublic boolean isRemoveVariable() {\n- return (opcode == VariableOperationCode.RemoveVariable\n- || opcode == VariableOperationCode.RemoveVariableAndFile);\n+ return opcode == VariableOperationCode.RemoveVariable\n+ || opcode == VariableOperationCode.RemoveVariableAndFile;\n}\npublic boolean isAssignVariable() {\n- return (opcode == VariableOperationCode.AssignVariable);\n+ return opcode == VariableOperationCode.AssignVariable;\n+ }\n+\n+ public boolean isAssignOrCopyVariable() {\n+ return opcode == VariableOperationCode.AssignVariable\n+ || opcode == VariableOperationCode.CopyVariable;\n}\npublic boolean isCreateVariable() {\n- return (opcode == VariableOperationCode.CreateVariable);\n+ return opcode == VariableOperationCode.CreateVariable;\n}\npublic VariableOperationCode getVariableOpcode() {\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageMap.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/lineage/LineageMap.java", "diff": "@@ -211,7 +211,10 @@ public class LineageMap {\nprivate void processMoveLI(LineageItem li) {\nif (li.getName().equals(\"__pred\"))\nremoveLineageItem(li.getInputs()[0].getName());\n- else\n- addLineageItem(li);\n+ else {\n+ //remove from old and move to new key\n+ _traces.put(li.getName(),\n+ _traces.remove(li.getInputs()[0].getName()));\n+ }\n}\n}\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/SqlCPInstruction.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/cp/SqlCPInstruction.java", "diff": "@@ -92,7 +92,7 @@ public class SqlCPInstruction extends CPInstruction {\n}\n}\n- private void setCell(TensorBlock outBlock, ResultSet resultSet, ValueType valueType, int[] ix) throws SQLException {\n+ private static void setCell(TensorBlock outBlock, ResultSet resultSet, ValueType valueType, int[] ix) throws SQLException {\nint sqlCol = ix[1] + 1;\nswitch (valueType) {\ncase FP64: outBlock.set(ix, resultSet.getDouble(sqlCol)); break;\n@@ -105,7 +105,7 @@ public class SqlCPInstruction extends CPInstruction {\n}\n}\n- private ValueType[] getSchemaFromMetaData(ResultSetMetaData meta) throws SQLException {\n+ private static ValueType[] getSchemaFromMetaData(ResultSetMetaData meta) throws SQLException {\nValueType[] schema = new ValueType[meta.getColumnCount()];\nfor (int i = 0; i < meta.getColumnCount(); i++) {\nint type = meta.getColumnType(i + 1);\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/matrix/data/FrameBlock.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/matrix/data/FrameBlock.java", "diff": "@@ -1854,9 +1854,9 @@ public class FrameBlock implements Writable, CacheBlock, Externalizable\nrowTemp1[i] = \"STRING\";\nelse if (rowTemp1[i].equals(\"FP64\") || rowTemp2[i].equals(\"FP64\"))\nrowTemp1[i] = \"FP64\";\n- else if (rowTemp1[i].equals(\"FP32\") && new ArrayList<String> (Arrays.asList(\"INT64\", \"INT32\", \"CHARACTER\")).contains(rowTemp2[i].toString()) )\n+ else if (rowTemp1[i].equals(\"FP32\") && new ArrayList<> (Arrays.asList(\"INT64\", \"INT32\", \"CHARACTER\")).contains(rowTemp2[i].toString()) )\nrowTemp1[i] = \"FP32\";\n- else if (rowTemp1[i].equals(\"INT64\") && new ArrayList<String> (Arrays.asList(\"INT32\", \"CHARACTER\")).contains(rowTemp2[i].toString()))\n+ else if (rowTemp1[i].equals(\"INT64\") && new ArrayList<> (Arrays.asList(\"INT32\", \"CHARACTER\")).contains(rowTemp2[i].toString()))\nrowTemp1[i] = \"INT64\";\nelse if (rowTemp1[i].equals(\"INT32\") || rowTemp2[i].equals(\"CHARACTER\"))\nrowTemp1[i] = \"INT32\";\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/unary/frame/DetectSchemaTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/unary/frame/DetectSchemaTest.java", "diff": "@@ -142,7 +142,7 @@ public class DetectSchemaTest extends AutomatedTestBase {\n}\n}\n- private void initFrameDataString(FrameBlock frame1, double[][] data, Types.ValueType[] lschema) {\n+ private static void initFrameDataString(FrameBlock frame1, double[][] data, Types.ValueType[] lschema) {\nfor (int j = 0; j < lschema.length - 1; j++) {\nTypes.ValueType vt = lschema[j];\nswitch (vt) {\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -131,6 +131,11 @@ SYSTEMDS-180 New Builtin Functions II OK\n* 181 Builtin function for naive bayes OK\n* 182 Builtin function for typeof (frame schema detection) OK\n* 183 Builtin function detectSchema OK\n+ * 184 Builtin function GNMF\n+ * 185 Builtin function PNMF\n+ * 186 Builtin function ALS-DS\n+ * 187 Builtin function ALS-CG\n+ * 188 Builtin function ALS\nOthers:\n" }, { "change_type": "ADD", "old_path": null, "new_path": "scripts/builtin/gnmf.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Modifications Copyright 2019 Graz University of Technology\n+#\n+# Licensed to the Apache Software Foundation (ASF) under one\n+# or more contributor license agreements. See the NOTICE file\n+# distributed with this work for additional information\n+# regarding copyright ownership. The ASF licenses this file\n+# to you under the Apache License, Version 2.0 (the\n+# \"License\"); you may not use this file except in compliance\n+# with the License. You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing,\n+# software distributed under the License is distributed on an\n+# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+# KIND, either express or implied. See the License for the\n+# specific language governing permissions and limitations\n+# under the License.\n+#\n+#-------------------------------------------------------------\n+\n+# Implements Gaussian Nonnegative Matrix Factorization (GNMF)\n+#\n+# [Chao Liu, Hung-chih Yang, Jinliang Fan, Li-Wei He, Yi-Min Wang:\n+# Distributed nonnegative matrix factorization for web-scale dyadic\n+# data analysis on mapreduce. WWW 2010: 681-690]\n+\n+m_gnmf = function(Matrix[Double] X, Integer rnk, Double eps = 10^-8, Integer maxi = 0)\n+ return (Matrix[Double] W, Matrix[Double] H)\n+{\n+ #initialize W and H\n+ W = rand(rows=nrow(X), cols=rnk, min=-0.5, max=0.5);\n+ H = rand(rows=rnk, cols=ncol(X), min=-0.5, max=0.5);\n+\n+ i = 0;\n+\n+ while(i < maxi) {\n+ H = H * ((t(W) %*% X) / (((t(W) %*% W) %*% H)+eps));\n+ W = W * ((X %*% t(H)) / ((W %*% (H %*% t(H)))+eps));\n+ i = i + 1;\n+ }\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "new_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "diff": "@@ -83,6 +83,7 @@ public enum Builtins {\nEXP(\"exp\", false),\nEVAL(\"eval\", false),\nFLOOR(\"floor\", false),\n+ GNMF(\"gnmf\", true),\nIFELSE(\"ifelse\", false),\nIMG_MIRROR(\"img_mirror\", true),\nIMG_BRIGHTNESS(\"img_brightness\", true),\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinFactorizationTest.java", "diff": "+/*\n+ * Copyright 2019 Graz University of Technology\n+ *\n+ * Licensed under the Apache License, Version 2.0 (the \"License\");\n+ * you may not use this file except in compliance with the License.\n+ * You may obtain a copy of the License at\n+ *\n+ * http://www.apache.org/licenses/LICENSE-2.0\n+ *\n+ * Unless required by applicable law or agreed to in writing, software\n+ * distributed under the License is distributed on an \"AS IS\" BASIS,\n+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+ * See the License for the specific language governing permissions and\n+ * limitations under the License.\n+ */\n+\n+package org.tugraz.sysds.test.functions.builtin;\n+\n+import org.junit.Test;\n+import org.tugraz.sysds.api.DMLScript;\n+import org.tugraz.sysds.common.Types;\n+import org.tugraz.sysds.hops.OptimizerUtils;\n+import org.tugraz.sysds.lops.LopProperties.ExecType;\n+import org.tugraz.sysds.runtime.controlprogram.parfor.stat.InfrastructureAnalyzer;\n+import org.tugraz.sysds.runtime.instructions.InstructionUtils;\n+import org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\n+import org.tugraz.sysds.runtime.util.DataConverter;\n+import org.tugraz.sysds.test.AutomatedTestBase;\n+import org.tugraz.sysds.test.TestConfiguration;\n+import org.tugraz.sysds.test.TestUtils;\n+\n+public class BuiltinFactorizationTest extends AutomatedTestBase\n+{\n+ private final static String TEST_NAME1 = \"GNMF\";\n+ private final static String TEST_DIR = \"functions/builtin/\";\n+ private static final String TEST_CLASS_DIR = TEST_DIR + BuiltinFactorizationTest.class.getSimpleName() + \"/\";\n+\n+ private final static int rows = 3210;\n+ private final static int cols = 4012;\n+ private final static int rank = 50;\n+ private final static double sparsity = 0.01;\n+ private final static double max_iter = 10;\n+\n+ @Override\n+ public void setUp() {\n+ TestUtils.clearAssertionInformation();\n+ addTestConfiguration(TEST_NAME1,new TestConfiguration(TEST_CLASS_DIR, TEST_NAME1,new String[]{\"U\",\"V\"}));\n+ }\n+\n+ @Test\n+ public void testGNMFRewritesCP() {\n+ runFactorizationTest(TEST_NAME1, true, ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testGNMFNoRewritesCP() {\n+ runFactorizationTest(TEST_NAME1, false, ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testGNMFRewritesSpark() {\n+ runFactorizationTest(TEST_NAME1, true, ExecType.SPARK);\n+ }\n+\n+ @Test\n+ public void testGNMFNoRewritesSpark() {\n+ runFactorizationTest(TEST_NAME1, false, ExecType.SPARK);\n+ }\n+\n+ private void runFactorizationTest(String testname, boolean rewrites, ExecType instType)\n+ {\n+ Types.ExecMode platformOld = setExecMode(instType);\n+\n+ boolean oldFlag = OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION;\n+ boolean sparkConfigOld = DMLScript.USE_LOCAL_SPARK_CONFIG;\n+\n+ try\n+ {\n+ loadTestConfiguration(getTestConfiguration(testname));\n+\n+ String HOME = SCRIPT_DIR + TEST_DIR;\n+\n+ fullDMLScriptName = HOME + testname + \".dml\";\n+ programArgs = new String[]{ \"-explain\", \"-stats\",\n+ \"-args\", input(\"X\"), output(\"U\"), output(\"V\"),\n+ String.valueOf(rank), String.valueOf(max_iter)};\n+\n+ OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = rewrites;\n+\n+ //generate actual datasets X = W * (U %*% V)\n+ MatrixBlock U = MatrixBlock.randOperations(rows, rank, 1.0, -1, 1, \"uniform\", 3);\n+ MatrixBlock V = MatrixBlock.randOperations(rank, cols, 1.0, -1, 1, \"uniform\", 7);\n+ MatrixBlock X = U.aggregateBinaryOperations(U, V, new MatrixBlock(),\n+ InstructionUtils.getMatMultOperator(InfrastructureAnalyzer.getLocalParallelism()));\n+ MatrixBlock I = MatrixBlock.randOperations(rows, cols, sparsity, 1, 1, \"uniform\", 12);\n+ X = (MatrixBlock) X.binaryOperations(InstructionUtils.parseBinaryOperator(\"*\"), I, new MatrixBlock());\n+ double[][] Xa = DataConverter.convertToDoubleMatrix(X);\n+\n+ //write input incl meta data\n+ writeInputMatrixWithMTD(\"X\", Xa, true);\n+\n+ //run test case\n+ runTest(true, false, null, -1);\n+ }\n+ finally {\n+ rtplatform = platformOld;\n+ DMLScript.USE_LOCAL_SPARK_CONFIG = sparkConfigOld;\n+ OptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = oldFlag;\n+ OptimizerUtils.ALLOW_AUTO_VECTORIZATION = true;\n+ OptimizerUtils.ALLOW_OPERATOR_FUSION = true;\n+ }\n+ }\n+}\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/builtin/GNMF.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1)\n+[W, H] = gnmf(X=X, rnk=$4, maxi=$5)\n+write(W, $2)\n+write(H, $3)\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Tasks.txt", "new_path": "docs/Tasks.txt", "diff": "@@ -131,8 +131,8 @@ SYSTEMDS-180 New Builtin Functions II OK\n* 181 Builtin function for naive bayes OK\n* 182 Builtin function for typeof (frame schema detection) OK\n* 183 Builtin function detectSchema OK\n- * 184 Builtin function GNMF\n- * 185 Builtin function PNMF\n+ * 184 Builtin function GNMF OK\n+ * 185 Builtin function PNMF OK\n* 186 Builtin function ALS-DS\n* 187 Builtin function ALS-CG\n* 188 Builtin function ALS\n" }, { "change_type": "MODIFY", "old_path": "scripts/builtin/gnmf.dml", "new_path": "scripts/builtin/gnmf.dml", "diff": "# Distributed nonnegative matrix factorization for web-scale dyadic\n# data analysis on mapreduce. WWW 2010: 681-690]\n-m_gnmf = function(Matrix[Double] X, Integer rnk, Double eps = 10^-8, Integer maxi = 0)\n+m_gnmf = function(Matrix[Double] X, Integer rnk, Double eps = 10^-8, Integer maxi = 10)\nreturn (Matrix[Double] W, Matrix[Double] H)\n{\n#initialize W and H\n- W = rand(rows=nrow(X), cols=rnk, min=-0.5, max=0.5);\n- H = rand(rows=rnk, cols=ncol(X), min=-0.5, max=0.5);\n+ W = rand(rows=nrow(X), cols=rnk, min=-0.05, max=0.05);\n+ H = rand(rows=rnk, cols=ncol(X), min=-0.05, max=0.05);\ni = 0;\n-\nwhile(i < maxi) {\nH = H * ((t(W) %*% X) / (((t(W) %*% W) %*% H)+eps));\nW = W * ((X %*% t(H)) / ((W %*% (H %*% t(H)))+eps));\n" }, { "change_type": "ADD", "old_path": null, "new_path": "scripts/builtin/pnmf.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Modifications Copyright 2019 Graz University of Technology\n+#\n+# Licensed to the Apache Software Foundation (ASF) under one\n+# or more contributor license agreements. See the NOTICE file\n+# distributed with this work for additional information\n+# regarding copyright ownership. The ASF licenses this file\n+# to you under the Apache License, Version 2.0 (the\n+# \"License\"); you may not use this file except in compliance\n+# with the License. You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing,\n+# software distributed under the License is distributed on an\n+# \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY\n+# KIND, either express or implied. See the License for the\n+# specific language governing permissions and limitations\n+# under the License.\n+#\n+#-------------------------------------------------------------\n+\n+# Implements Poisson Nonnegative Matrix Factorization (PNMF)\n+#\n+# [Chao Liu, Hung-chih Yang, Jinliang Fan, Li-Wei He, Yi-Min Wang:\n+# Distributed nonnegative matrix factorization for web-scale dyadic\n+# data analysis on mapreduce. WWW 2010: 681-690]\n+\n+m_pnmf = function(Matrix[Double] X, Integer rnk, Double eps = 10^-8, Integer maxi = 10, Boolean verbose=TRUE)\n+ return (Matrix[Double] W, Matrix[Double] H)\n+{\n+ #initialize W and H\n+ W = rand(rows=nrow(X), cols=rnk, min=0, max=0.025);\n+ H = rand(rows=rnk, cols=ncol(X), min=0, max=0.025);\n+\n+ i = 0;\n+ while(i < maxi) {\n+ H = (H*(t(W)%*%(X/(W%*%H+eps)))) / t(colSums(W));\n+ W = (W*((X/(W%*%H+eps))%*%t(H))) / t(rowSums(H));\n+ i = i + 1;\n+ if( verbose ) {\n+ obj = sum(W%*%H) - sum(X*log(W%*%H+eps));\n+ print(\"iter=\" + i + \" obj=\" + obj);\n+ }\n+ }\n+}\n" }, { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "new_path": "src/main/java/org/tugraz/sysds/common/Builtins.java", "diff": "@@ -118,6 +118,7 @@ public enum Builtins {\nNAIVEBAYES(\"naivebayes\", true, false),\nOUTER(\"outer\", false),\nOUTLIER(\"outlier\", true, false), //TODO parameterize opposite\n+ PNMF(\"pnmf\", true),\nPPRED(\"ppred\", false),\nPROD(\"prod\", false),\nQR(\"qr\", false, ReturnType.MULTI_RETURN),\n" }, { "change_type": "MODIFY", "old_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinFactorizationTest.java", "new_path": "src/test/java/org/tugraz/sysds/test/functions/builtin/BuiltinFactorizationTest.java", "diff": "@@ -21,10 +21,6 @@ import org.tugraz.sysds.api.DMLScript;\nimport org.tugraz.sysds.common.Types;\nimport org.tugraz.sysds.hops.OptimizerUtils;\nimport org.tugraz.sysds.lops.LopProperties.ExecType;\n-import org.tugraz.sysds.runtime.controlprogram.parfor.stat.InfrastructureAnalyzer;\n-import org.tugraz.sysds.runtime.instructions.InstructionUtils;\n-import org.tugraz.sysds.runtime.matrix.data.MatrixBlock;\n-import org.tugraz.sysds.runtime.util.DataConverter;\nimport org.tugraz.sysds.test.AutomatedTestBase;\nimport org.tugraz.sysds.test.TestConfiguration;\nimport org.tugraz.sysds.test.TestUtils;\n@@ -32,6 +28,7 @@ import org.tugraz.sysds.test.TestUtils;\npublic class BuiltinFactorizationTest extends AutomatedTestBase\n{\nprivate final static String TEST_NAME1 = \"GNMF\";\n+ private final static String TEST_NAME2 = \"PNMF\";\nprivate final static String TEST_DIR = \"functions/builtin/\";\nprivate static final String TEST_CLASS_DIR = TEST_DIR + BuiltinFactorizationTest.class.getSimpleName() + \"/\";\n@@ -45,6 +42,7 @@ public class BuiltinFactorizationTest extends AutomatedTestBase\npublic void setUp() {\nTestUtils.clearAssertionInformation();\naddTestConfiguration(TEST_NAME1,new TestConfiguration(TEST_CLASS_DIR, TEST_NAME1,new String[]{\"U\",\"V\"}));\n+ addTestConfiguration(TEST_NAME2,new TestConfiguration(TEST_CLASS_DIR, TEST_NAME2,new String[]{\"U\",\"V\"}));\n}\n@Test\n@@ -67,6 +65,26 @@ public class BuiltinFactorizationTest extends AutomatedTestBase\nrunFactorizationTest(TEST_NAME1, false, ExecType.SPARK);\n}\n+ @Test\n+ public void testPNMFRewritesCP() {\n+ runFactorizationTest(TEST_NAME2, true, ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testPNMFNoRewritesCP() {\n+ runFactorizationTest(TEST_NAME2, false, ExecType.CP);\n+ }\n+\n+ @Test\n+ public void testPNMFRewritesSpark() {\n+ runFactorizationTest(TEST_NAME2, true, ExecType.SPARK);\n+ }\n+\n+ @Test\n+ public void testPNMFNoRewritesSpark() {\n+ runFactorizationTest(TEST_NAME2, false, ExecType.SPARK);\n+ }\n+\nprivate void runFactorizationTest(String testname, boolean rewrites, ExecType instType)\n{\nTypes.ExecMode platformOld = setExecMode(instType);\n@@ -82,21 +100,13 @@ public class BuiltinFactorizationTest extends AutomatedTestBase\nfullDMLScriptName = HOME + testname + \".dml\";\nprogramArgs = new String[]{ \"-explain\", \"-stats\",\n- \"-args\", input(\"X\"), output(\"U\"), output(\"V\"),\n+ \"-args\", input(\"X\"), output(\"W\"), output(\"H\"),\nString.valueOf(rank), String.valueOf(max_iter)};\nOptimizerUtils.ALLOW_ALGEBRAIC_SIMPLIFICATION = rewrites;\n- //generate actual datasets X = W * (U %*% V)\n- MatrixBlock U = MatrixBlock.randOperations(rows, rank, 1.0, -1, 1, \"uniform\", 3);\n- MatrixBlock V = MatrixBlock.randOperations(rank, cols, 1.0, -1, 1, \"uniform\", 7);\n- MatrixBlock X = U.aggregateBinaryOperations(U, V, new MatrixBlock(),\n- InstructionUtils.getMatMultOperator(InfrastructureAnalyzer.getLocalParallelism()));\n- MatrixBlock I = MatrixBlock.randOperations(rows, cols, sparsity, 1, 1, \"uniform\", 12);\n- X = (MatrixBlock) X.binaryOperations(InstructionUtils.parseBinaryOperator(\"*\"), I, new MatrixBlock());\n- double[][] Xa = DataConverter.convertToDoubleMatrix(X);\n-\n- //write input incl meta data\n+ //generate input and write incl meta data\n+ double[][] Xa = TestUtils.generateTestMatrix(rows, cols, 1, 10, sparsity, 7);\nwriteInputMatrixWithMTD(\"X\", Xa, true);\n//run test case\n" }, { "change_type": "ADD", "old_path": null, "new_path": "src/test/scripts/functions/builtin/PNMF.dml", "diff": "+#-------------------------------------------------------------\n+#\n+# Copyright 2019 Graz University of Technology\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+#-------------------------------------------------------------\n+\n+X = read($1)\n+[W, H] = pnmf(X=X, rnk=$4, maxi=$5)\n+write(W, $2)\n+write(H, $3)\n" } ]

[ { "change_type": "MODIFY", "old_path": "src/main/java/org/tugraz/sysds/runtime/instructions/SPInstructionParser.java", "new_path": "src/main/java/org/tugraz/sysds/runtime/instructions/SPInstructionParser.java", "diff": "@@ -420,7 +420,7 @@ public class SPInstructionParser extends InstructionParser\ncase Unary:\nparts = InstructionUtils.getInstructionPartsWithValueType(str);\n- CPOperand in = new CPOperand(\"\", Types.ValueType.UNKNOWN, Types.DataType.UNKNOWN);\n+ CPOperand in = new CPOperand(parts[1]);\nif(in.getDataType() == Types.DataType.MATRIX)\nreturn UnaryMatrixSPInstruction.parseInstruction(str);\nelse\n" } ]