Prefetch shared memory fix tests

Signed-off-by: Nirvedh Meshram <nirvedh@gmail.com>
iree-org · Jan 14, 2025 · a28d7eb · a28d7eb
1 parent acdbd18
commit a28d7eb
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 4 deletions.
diff --git a/compiler/src/iree/compiler/Codegen/LLVMGPU/test/ROCDL/config_igemm_tile_and_fuse.mlir b/compiler/src/iree/compiler/Codegen/LLVMGPU/test/ROCDL/config_igemm_tile_and_fuse.mlir
@@ -76,7 +76,7 @@ func.func @nhwc_conv_unaligned_mfma() {
 
 // CHECK-LABEL: func.func @nhwc_conv_unaligned_mfma
 //  CHECK-SAME:   #iree_codegen.translation_info<pipeline = LLVMGPUTileAndFuse workgroup_size = [256, 1, 1] subgroup_size = 64
-//  CHECK-SAME:   #iree_gpu.pipeline_options<prefetch_shared_memory = false, no_reduce_shared_memory_bank_conflicts = false
+//  CHECK-SAME:   #iree_gpu.pipeline_options<prefetch_shared_memory = true, no_reduce_shared_memory_bank_conflicts = false
 //  CHECK-SAME:   use_igemm_convolution = true
 
 //       CHECK:   linalg.conv_2d_nhwc_hwcf {{.*}}lowering_config = #iree_gpu.lowering_config
@@ -106,7 +106,7 @@ func.func @nchw_conv_unaligned_mfma() {
 
 // CHECK-LABEL: func.func @nchw_conv_unaligned_mfma
 //  CHECK-SAME:   #iree_codegen.translation_info<pipeline = LLVMGPUTileAndFuse workgroup_size = [256, 1, 1] subgroup_size = 64
-//  CHECK-SAME:   #iree_gpu.pipeline_options<prefetch_shared_memory = false, no_reduce_shared_memory_bank_conflicts = false
+//  CHECK-SAME:   #iree_gpu.pipeline_options<prefetch_shared_memory = true, no_reduce_shared_memory_bank_conflicts = false
 //  CHECK-SAME:   use_igemm_convolution = true
 
 //       CHECK:   linalg.conv_2d_nchw_fchw {{.*}}lowering_config = #iree_gpu.lowering_config

diff --git a/compiler/src/iree/compiler/Codegen/LLVMGPU/test/ROCDL/config_tile_and_fuse.mlir b/compiler/src/iree/compiler/Codegen/LLVMGPU/test/ROCDL/config_tile_and_fuse.mlir
@@ -277,7 +277,7 @@ func.func @unaligned_to_intrinsic_batched_matmul(%lhs : tensor<12x577x577xf32>,
 
 // CHECK-LABEL: func.func @unaligned_to_intrinsic_batched_matmul
 // CHECK-SAME:    #iree_codegen.translation_info<pipeline = LLVMGPUTileAndFuse workgroup_size = [64, 1, 1] subgroup_size = 64
-// CHECK-SAME:    {gpu_pipeline_options = #iree_gpu.pipeline_options<prefetch_shared_memory = false, no_reduce_shared_memory_bank_conflicts = false, use_igemm_convolution = false>}
+// CHECK-SAME:    {gpu_pipeline_options = #iree_gpu.pipeline_options<prefetch_shared_memory = true, no_reduce_shared_memory_bank_conflicts = false, use_igemm_convolution = false>}
 //      CHECK:    linalg.batch_matmul {{.*}}lowering_config = #iree_gpu.lowering_config
 //  CHECK-SAME:     padding = [1, 16, 16, 4]
 //  CHECK-SAME:     promote_operands = [0, 1, 2]
@@ -306,7 +306,7 @@ func.func @unaligned_to_intrinsic_batched_matmul_tiling_check(%lhs : tensor<12x5
 
 // CHECK-LABEL: func.func @unaligned_to_intrinsic_batched_matmul_tiling_check
 // CHECK-SAME:    #iree_codegen.translation_info<pipeline = LLVMGPUTileAndFuse workgroup_size = [256, 1, 1] subgroup_size = 64
-// CHECK-SAME:    {gpu_pipeline_options = #iree_gpu.pipeline_options<prefetch_shared_memory = false, no_reduce_shared_memory_bank_conflicts = false, use_igemm_convolution = false>}
+// CHECK-SAME:    {gpu_pipeline_options = #iree_gpu.pipeline_options<prefetch_shared_memory = true, no_reduce_shared_memory_bank_conflicts = false, use_igemm_convolution = false>}
 //      CHECK:    linalg.batch_matmul {{.*}}lowering_config = #iree_gpu.lowering_config
 //  CHECK-SAME:     padding = [1, 16, 512, 4]
 //  CHECK-SAME:     promote_operands = [0, 1, 2]