diff options
Diffstat (limited to 'mlir/test/Integration/GPU/ROCM/vecadd.mlir')
| -rw-r--r-- | mlir/test/Integration/GPU/ROCM/vecadd.mlir | 19 |
1 files changed, 11 insertions, 8 deletions
diff --git a/mlir/test/Integration/GPU/ROCM/vecadd.mlir b/mlir/test/Integration/GPU/ROCM/vecadd.mlir index 52a440eee1f6..41758256b25f 100644 --- a/mlir/test/Integration/GPU/ROCM/vecadd.mlir +++ b/mlir/test/Integration/GPU/ROCM/vecadd.mlir @@ -1,24 +1,24 @@ // RUN: mlir-opt %s \ // RUN: -convert-scf-to-cf \ // RUN: -gpu-kernel-outlining \ -// RUN: -pass-pipeline='gpu.module(strip-debuginfo,convert-gpu-to-rocdl,gpu-to-hsaco{chip=%chip})' \ -// RUN: -gpu-to-llvm \ +// RUN: -pass-pipeline='gpu.module(strip-debuginfo,convert-gpu-to-rocdl{use-bare-ptr-memref-call-conv=true},gpu-to-hsaco{chip=%chip})' \ +// RUN: -gpu-to-llvm=use-bare-pointers-for-kernels=true \ // RUN: | mlir-cpu-runner \ // RUN: --shared-libs=%linalg_test_lib_dir/libmlir_rocm_runtime%shlibext \ // RUN: --shared-libs=%linalg_test_lib_dir/libmlir_runner_utils%shlibext \ // RUN: --entry-point-result=void \ // RUN: | FileCheck %s -func.func @vecadd(%arg0 : memref<?xf32>, %arg1 : memref<?xf32>, %arg2 : memref<?xf32>) { +func.func @vecadd(%arg0 : memref<5xf32>, %arg1 : memref<5xf32>, %arg2 : memref<5xf32>) { %c0 = arith.constant 0 : index %c1 = arith.constant 1 : index - %block_dim = memref.dim %arg0, %c0 : memref<?xf32> + %block_dim = arith.constant 5 : index gpu.launch blocks(%bx, %by, %bz) in (%grid_x = %c1, %grid_y = %c1, %grid_z = %c1) threads(%tx, %ty, %tz) in (%block_x = %block_dim, %block_y = %c1, %block_z = %c1) { - %a = memref.load %arg0[%tx] : memref<?xf32> - %b = memref.load %arg1[%tx] : memref<?xf32> + %a = memref.load %arg0[%tx] : memref<5xf32> + %b = memref.load %arg1[%tx] : memref<5xf32> %c = arith.addf %a, %b : f32 - memref.store %c, %arg2[%tx] : memref<?xf32> + memref.store %c, %arg2[%tx] : memref<5xf32> gpu.terminator } return @@ -49,8 +49,11 @@ func.func @main() { %9 = call @mgpuMemGetDeviceMemRef1dFloat(%3) : (memref<?xf32>) -> (memref<?xf32>) %10 = call @mgpuMemGetDeviceMemRef1dFloat(%4) : (memref<?xf32>) -> (memref<?xf32>) %11 = call @mgpuMemGetDeviceMemRef1dFloat(%5) : (memref<?xf32>) -> (memref<?xf32>) + %12 = memref.cast %9 : memref<?xf32> to memref<5xf32> + %13 = memref.cast %10 : memref<?xf32> to memref<5xf32> + %14 = memref.cast %11 : memref<?xf32> to memref<5xf32> - call @vecadd(%9, %10, %11) : (memref<?xf32>, memref<?xf32>, memref<?xf32>) -> () + call @vecadd(%12, %13, %14) : (memref<5xf32>, memref<5xf32>, memref<5xf32>) -> () call @printMemrefF32(%8) : (memref<*xf32>) -> () return } |