.triton/dump/94361ae8a918b76700c87078e3d5a751/triton_.ttgir

#blocked = #triton_gpu.blocked<{sizePerThread = [8], threadsPerWarp = [32], warpsPerCTA = [4], order = [0], CTAsPerCGA = [1], CTASplitNum = [1], CTAOrder = [0]}>
#blocked1 = #triton_gpu.blocked<{sizePerThread = [4], threadsPerWarp = [32], warpsPerCTA = [4], order = [0], CTAsPerCGA = [1], CTASplitNum = [1], CTAOrder = [0]}>
module attributes {"triton_gpu.compute-capability" = 89 : i32, "triton_gpu.num-ctas" = 1 : i32, "triton_gpu.num-warps" = 4 : i32, "triton_gpu.threads-per-warp" = 32 : i32} {
  tt.func public @triton__0d1d2de(%arg0: !tt.ptr<bf16, 1> {tt.divisibility = 16 : i32}, %arg1: !tt.ptr<f32, 1> {tt.divisibility = 16 : i32}, %arg2: i32 {tt.divisibility = 16 : i32, tt.max_divisibility = 16 : i32}) attributes {noinline = false} {
    %cst = arith.constant dense<12865792> : tensor<1024xi32, #blocked>
    %cst_0 = arith.constant dense<12865792> : tensor<1024xi32, #blocked1>
    %c1024_i32 = arith.constant 1024 : i32
    %0 = tt.get_program_id x : i32
    %1 = arith.muli %0, %c1024_i32 : i32
    %2 = tt.make_range {end = 1024 : i32, start = 0 : i32} : tensor<1024xi32, #blocked>
    %3 = tt.make_range {end = 1024 : i32, start = 0 : i32} : tensor<1024xi32, #blocked1>
    %4 = tt.splat %1 : (i32) -> tensor<1024xi32, #blocked>
    %5 = tt.splat %1 : (i32) -> tensor<1024xi32, #blocked1>
    %6 = arith.addi %4, %2 : tensor<1024xi32, #blocked>
    %7 = arith.addi %5, %3 : tensor<1024xi32, #blocked1>
    %8 = arith.cmpi slt, %6, %cst : tensor<1024xi32, #blocked>
    %9 = arith.cmpi slt, %7, %cst_0 : tensor<1024xi32, #blocked1>
    %10 = tt.splat %arg0 : (!tt.ptr<bf16, 1>) -> tensor<1024x!tt.ptr<bf16, 1>, #blocked>
    %11 = tt.addptr %10, %6 : tensor<1024x!tt.ptr<bf16, 1>, #blocked>, tensor<1024xi32, #blocked>
    %12 = tt.load %11, %8 {cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<1024xbf16, #blocked>
    %13 = triton_gpu.convert_layout %12 : (tensor<1024xbf16, #blocked>) -> tensor<1024xbf16, #blocked1>
    %14 = arith.extf %13 : tensor<1024xbf16, #blocked1> to tensor<1024xf32, #blocked1>
    %15 = tt.splat %arg1 : (!tt.ptr<f32, 1>) -> tensor<1024x!tt.ptr<f32, 1>, #blocked1>
    %16 = tt.addptr %15, %7 : tensor<1024x!tt.ptr<f32, 1>, #blocked1>, tensor<1024xi32, #blocked1>
    tt.store %16, %14, %9 {cache = 1 : i32, evict = 1 : i32} : tensor<1024xf32, #blocked1>
    tt.return
  }
}