[Mlir-commits] [mlir] Revert "[mlir][test][gpu] Migrate CUDA tests to the TargetAttr compilation workflow (#65768) (PR #65848)
llvmlistbot at llvm.org
llvmlistbot at llvm.org
Sat Sep 9 04:15:21 PDT 2023
llvmbot wrote:
@llvm/pr-subscribers-mlir
<details>
<summary>Changes</summary>
This reverts commit d21b67293be15f8a89378e4785d70cc037866406.
--
Patch is 27.61 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/65848.diff
25 Files Affected:
- (modified) mlir/include/mlir/Conversion/GPUToNVVM/GPUToNVVMPass.h (+11-1)
- (modified) mlir/include/mlir/Conversion/Passes.td (+1)
- (modified) mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp (+10-1)
- (modified) mlir/lib/Dialect/SparseTensor/Pipelines/SparseTensorPipelines.cpp (+1-1)
- (modified) mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/dump-ptx.mlir (+2-1)
- (modified) mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-mma-2-4-f16.mlir (+3-2)
- (modified) mlir/test/Integration/Dialect/Vector/GPU/CUDA/test-reduction-distribute.mlir (+3-1)
- (modified) mlir/test/Integration/Dialect/Vector/GPU/CUDA/test-warp-distribute.mlir (+9-3)
- (modified) mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f16.mlir (+3-1)
- (modified) mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f32-bare-ptr.mlir (+3-1)
- (modified) mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f32.mlir (+3-1)
- (modified) mlir/test/Integration/GPU/CUDA/all-reduce-and.mlir (+6-2)
- (modified) mlir/test/Integration/GPU/CUDA/all-reduce-max.mlir (+3-1)
- (modified) mlir/test/Integration/GPU/CUDA/all-reduce-min.mlir (+3-1)
- (modified) mlir/test/Integration/GPU/CUDA/all-reduce-op.mlir (+3-1)
- (modified) mlir/test/Integration/GPU/CUDA/all-reduce-or.mlir (+3-1)
- (modified) mlir/test/Integration/GPU/CUDA/all-reduce-region.mlir (+3-1)
- (modified) mlir/test/Integration/GPU/CUDA/all-reduce-xor.mlir (+3-1)
- (modified) mlir/test/Integration/GPU/CUDA/async.mlir (+2-2)
- (modified) mlir/test/Integration/GPU/CUDA/gpu-to-cubin.mlir (+4-1)
- (modified) mlir/test/Integration/GPU/CUDA/multiple-all-reduce.mlir (+3-1)
- (modified) mlir/test/Integration/GPU/CUDA/printf.mlir (+2-1)
- (modified) mlir/test/Integration/GPU/CUDA/shuffle.mlir (+3-1)
- (modified) mlir/test/Integration/GPU/CUDA/two-modules.mlir (+3-1)
- (modified) mlir/test/lib/Dialect/GPU/TestLowerToNVVM.cpp (+42-34)
<pre>
diff --git a/mlir/include/mlir/Conversion/GPUToNVVM/GPUToNVVMPass.h b/mlir/include/mlir/Conversion/GPUToNVVM/GPUToNVVMPass.h
index e0f4c71051e506a..46f29c6dd8b9267 100644
--- a/mlir/include/mlir/Conversion/GPUToNVVM/GPUToNVVMPass.h
+++ b/mlir/include/mlir/Conversion/GPUToNVVM/GPUToNVVMPass.h
@@ -16,7 +16,9 @@ namespace mlir {
class LLVMTypeConverter;
class ConversionTarget;
class RewritePatternSet;
-class Pass;
+
+template <typename OpT>
+class OperationPass;
namespace gpu {
class GPUModuleOp;
@@ -43,6 +45,14 @@ void populateGpuSubgroupReduceOpLoweringPattern(LLVMTypeConverter &converter,
/// Collect a set of patterns to convert WMMA ops from GPU dialect to NVVM.
void populateGpuWMMAToNVVMConversionPatterns(LLVMTypeConverter &converter,
RewritePatternSet &patterns);
+
+/// Creates a pass that lowers GPU dialect operations to NVVM counterparts. The
+/// index bitwidth used for the lowering of the device side index computations
+/// is configurable.
+std::unique_ptr<OperationPass<gpu::GPUModuleOp>> createLowerGpuOpsToNVVMOpsPass(
+ unsigned indexBitwidth = kDeriveIndexBitwidthFromDataLayout,
+ bool hasRedux = false);
+
} // namespace mlir
#endif // MLIR_CONVERSION_GPUTONVVM_GPUTONVVMPASS_H_
diff --git a/mlir/include/mlir/Conversion/Passes.td b/mlir/include/mlir/Conversion/Passes.td
index 3218760931b8cb0..ed37abf85275bf3 100644
--- a/mlir/include/mlir/Conversion/Passes.td
+++ b/mlir/include/mlir/Conversion/Passes.td
@@ -486,6 +486,7 @@ def LowerHostCodeToLLVMPass : Pass<"lower-host-to-llvm", "ModuleOp"> {
def ConvertGpuOpsToNVVMOps : Pass<"convert-gpu-to-nvvm", "gpu::GPUModuleOp"> {
let summary = "Generate NVVM operations for gpu operations";
+ let constructor = "mlir::createLowerGpuOpsToNVVMOpsPass()";
let dependentDialects = [
"cf::ControlFlowDialect",
"memref::MemRefDialect",
diff --git a/mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp b/mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp
index 764b6a779b98c7a..06469dc82b3fc5d 100644
--- a/mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp
+++ b/mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp
@@ -210,7 +210,11 @@ struct GPULaneIdOpToNVVM : ConvertOpToLLVMPattern<gpu::LaneIdOp> {
/// code.
struct LowerGpuOpsToNVVMOpsPass
: public impl::ConvertGpuOpsToNVVMOpsBase<LowerGpuOpsToNVVMOpsPass> {
- using Base::Base;
+ LowerGpuOpsToNVVMOpsPass() = default;
+ LowerGpuOpsToNVVMOpsPass(unsigned indexBitwidth, bool hasRedux = false) {
+ this->indexBitwidth = indexBitwidth;
+ this->hasRedux = hasRedux;
+ }
void runOnOperation() override {
gpu::GPUModuleOp m = getOperation();
@@ -374,3 +378,8 @@ void mlir::populateGpuToNVVMConversionPatterns(LLVMTypeConverter &converter,
"__nv_tanh");
populateOpPatterns<math::TanOp>(converter, patterns, "__nv_tanf", "__nv_tan");
}
+
+std::unique_ptr<OperationPass<gpu::GPUModuleOp>>
+mlir::createLowerGpuOpsToNVVMOpsPass(unsigned indexBitwidth, bool hasRedux) {
+ return std::make_unique<LowerGpuOpsToNVVMOpsPass>(indexBitwidth, hasRedux);
+}
diff --git a/mlir/lib/Dialect/SparseTensor/Pipelines/SparseTensorPipelines.cpp b/mlir/lib/Dialect/SparseTensor/Pipelines/SparseTensorPipelines.cpp
index 24c4c4c43a93dea..a7fd5a25e68314e 100644
--- a/mlir/lib/Dialect/SparseTensor/Pipelines/SparseTensorPipelines.cpp
+++ b/mlir/lib/Dialect/SparseTensor/Pipelines/SparseTensorPipelines.cpp
@@ -52,7 +52,7 @@ void mlir::sparse_tensor::buildSparseCompiler(
pm.addPass(createSparseGPUCodegenPass());
pm.addNestedPass<gpu::GPUModuleOp>(createStripDebugInfoPass());
pm.addNestedPass<gpu::GPUModuleOp>(createConvertSCFToCFPass());
- pm.addNestedPass<gpu::GPUModuleOp>(createConvertGpuOpsToNVVMOps());
+ pm.addNestedPass<gpu::GPUModuleOp>(createLowerGpuOpsToNVVMOpsPass());
}
// TODO(springerm): Add sparse support to the BufferDeallocation pass and add
diff --git a/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/dump-ptx.mlir b/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/dump-ptx.mlir
index 0cb06b7bf1d2001..2c1ae3ee840d0f8 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/dump-ptx.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/dump-ptx.mlir
@@ -1,5 +1,6 @@
// RUN: mlir-opt %s \
-// RUN: | mlir-opt -test-lower-to-nvvm -debug-only=serialize-to-isa \
+// RUN: | mlir-opt -gpu-kernel-outlining \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,gpu-to-cubin{dump-ptx}))' \
// RUN: 2>&1 | FileCheck %s
// CHECK: Generated by LLVM NVPTX Back-End
diff --git a/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-mma-2-4-f16.mlir b/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-mma-2-4-f16.mlir
index 80972f244ec02d7..8eb90fd3ca9946e 100644
--- a/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-mma-2-4-f16.mlir
+++ b/mlir/test/Integration/Dialect/SparseTensor/GPU/CUDA/sparse-mma-2-4-f16.mlir
@@ -2,9 +2,10 @@
// NOTE: this test requires gpu-sm80
//
// RUN: mlir-opt \
-// RUN: --pass-pipeline="builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,convert-nvgpu-to-nvvm,affine-expand-index-ops,lower-affine,convert-arith-to-llvm),convert-vector-to-llvm,canonicalize,cse)" \
+// RUN: --pass-pipeline="builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,convert-nvgpu-to-nvvm,affine-expand-index-ops,lower-affine,convert-arith-to-llvm),convert-vector-to-llvm,canonicalize,cse,gpu.module(gpu-to-cubin{chip=sm_80 features=+ptx71}))" \
// RUN: %s \
-// RUN: | mlir-opt --test-lower-to-nvvm="cubin-chip=sm_80 cubin-features=+ptx71" \
+// RUN: | mlir-opt --convert-vector-to-scf --convert-scf-to-cf -convert-cf-to-llvm --convert-vector-to-llvm \
+// RUN: --convert-arith-to-llvm --gpu-to-llvm --reconcile-unrealized-casts \
// RUN: | mlir-cpu-runner \
// RUN: --shared-libs=%mlir_cuda_runtime \
// RUN: --shared-libs=%mlir_c_runner_utils \
diff --git a/mlir/test/Integration/Dialect/Vector/GPU/CUDA/test-reduction-distribute.mlir b/mlir/test/Integration/Dialect/Vector/GPU/CUDA/test-reduction-distribute.mlir
index 8c991493a2b0174..8571c5ca5f3dc26 100644
--- a/mlir/test/Integration/Dialect/Vector/GPU/CUDA/test-reduction-distribute.mlir
+++ b/mlir/test/Integration/Dialect/Vector/GPU/CUDA/test-reduction-distribute.mlir
@@ -1,7 +1,9 @@
// RUN: mlir-opt %s -test-vector-warp-distribute="hoist-uniform distribute-transfer-write propagate-distribution" -canonicalize |\
// RUN: mlir-opt -test-vector-warp-distribute=rewrite-warp-ops-to-scf-if |\
// RUN: mlir-opt -lower-affine -convert-vector-to-scf -convert-scf-to-cf -convert-vector-to-llvm \
-// RUN: -convert-arith-to-llvm -test-lower-to-nvvm | \
+// RUN: -convert-arith-to-llvm -gpu-kernel-outlining |\
+// RUN: mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,reconcile-unrealized-casts,gpu-to-cubin))' |\
+// RUN: mlir-opt -gpu-to-llvm -reconcile-unrealized-casts |\
// RUN: mlir-cpu-runner -e main -entry-point-result=void \
// RUN: -shared-libs=%mlir_cuda_runtime \
// RUN: -shared-libs=%mlir_c_runner_utils \
diff --git a/mlir/test/Integration/Dialect/Vector/GPU/CUDA/test-warp-distribute.mlir b/mlir/test/Integration/Dialect/Vector/GPU/CUDA/test-warp-distribute.mlir
index f26c18c4ae3dd28..c671c1843862f94 100644
--- a/mlir/test/Integration/Dialect/Vector/GPU/CUDA/test-warp-distribute.mlir
+++ b/mlir/test/Integration/Dialect/Vector/GPU/CUDA/test-warp-distribute.mlir
@@ -2,7 +2,9 @@
// everything on the same thread.
// RUN: mlir-opt %s -test-vector-warp-distribute=rewrite-warp-ops-to-scf-if -canonicalize | \
// RUN: mlir-opt -convert-vector-to-scf -convert-scf-to-cf -convert-cf-to-llvm -convert-vector-to-llvm -convert-arith-to-llvm \
-// RUN: -test-lower-to-nvvm | \
+// RUN: -gpu-kernel-outlining |\
+// RUN: mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,reconcile-unrealized-casts,gpu-to-cubin))' |\
+// RUN: mlir-opt -gpu-to-llvm -reconcile-unrealized-casts |\
// RUN: mlir-cpu-runner -e main -entry-point-result=void \
// RUN: -shared-libs=%mlir_cuda_runtime \
// RUN: -shared-libs=%mlir_c_runner_utils \
@@ -13,7 +15,9 @@
// RUN: mlir-opt %s -test-vector-warp-distribute="hoist-uniform distribute-transfer-write" \
// RUN: -test-vector-warp-distribute=rewrite-warp-ops-to-scf-if -canonicalize | \
// RUN: mlir-opt -convert-vector-to-scf -convert-scf-to-cf -convert-cf-to-llvm -convert-vector-to-llvm -convert-arith-to-llvm \
-// RUN: -test-lower-to-nvvm | \
+// RUN: -gpu-kernel-outlining |\
+// RUN: mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,reconcile-unrealized-casts,gpu-to-cubin))' |\
+// RUN: mlir-opt -gpu-to-llvm -reconcile-unrealized-casts |\
// RUN: mlir-cpu-runner -e main -entry-point-result=void \
// RUN: -shared-libs=%mlir_cuda_runtime \
// RUN: -shared-libs=%mlir_c_runner_utils \
@@ -23,7 +27,9 @@
// RUN: mlir-opt %s -test-vector-warp-distribute="hoist-uniform distribute-transfer-write propagate-distribution" \
// RUN: -test-vector-warp-distribute=rewrite-warp-ops-to-scf-if -canonicalize | \
// RUN: mlir-opt -convert-vector-to-scf -convert-scf-to-cf -convert-cf-to-llvm -convert-vector-to-llvm -convert-arith-to-llvm \
-// RUN: -test-lower-to-nvvm | \
+// RUN: -gpu-kernel-outlining |\
+// RUN: mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,reconcile-unrealized-casts,gpu-to-cubin))' |\
+// RUN: mlir-opt -gpu-to-llvm -reconcile-unrealized-casts |\
// RUN: mlir-cpu-runner -e main -entry-point-result=void \
// RUN: -shared-libs=%mlir_cuda_runtime \
// RUN: -shared-libs=%mlir_c_runner_utils \
diff --git a/mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f16.mlir b/mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f16.mlir
index 591bf1b4fd18231..535ba52d66f0062 100644
--- a/mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f16.mlir
+++ b/mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f16.mlir
@@ -1,5 +1,7 @@
// RUN: mlir-opt %s \
-// RUN: | mlir-opt -test-lower-to-nvvm="cubin-chip=sm_70" \
+// RUN: | mlir-opt -gpu-kernel-outlining \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,gpu-to-cubin{chip=sm_70}))' \
+// RUN: | mlir-opt --convert-scf-to-cf -gpu-to-llvm \
// RUN: | mlir-cpu-runner \
// RUN: --shared-libs=%mlir_cuda_runtime \
// RUN: --shared-libs=%mlir_runner_utils \
diff --git a/mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f32-bare-ptr.mlir b/mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f32-bare-ptr.mlir
index 51bd23f817b33f1..c4ca46521eeb4c7 100644
--- a/mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f32-bare-ptr.mlir
+++ b/mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f32-bare-ptr.mlir
@@ -3,7 +3,9 @@
// Similar to the wmma-matmul-f32 but but with the memref bare pointer lowering convention.
// This test also uses gpu.memcpy operations (instead of gpu.host_register).
// RUN: mlir-opt %s \
-// RUN: | mlir-opt -test-lower-to-nvvm="host-bare-ptr-calling-convention=1 kernel-bare-ptr-calling-convention=1 cubin-chip=sm_70" \
+// RUN: | mlir-opt -gpu-kernel-outlining \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm{use-bare-ptr-memref-call-conv=1},gpu-to-cubin{chip=sm_70}))' \
+// RUN: | mlir-opt --convert-scf-to-cf -gpu-to-llvm="use-bare-pointers-for-host=1 use-bare-pointers-for-kernels=1" \
// RUN: | mlir-cpu-runner \
// RUN: --shared-libs=%mlir_cuda_runtime \
// RUN: --entry-point-result=void \
diff --git a/mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f32.mlir b/mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f32.mlir
index 0307b3d504be9f6..ae410dce281b17a 100644
--- a/mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f32.mlir
+++ b/mlir/test/Integration/GPU/CUDA/TensorCore/wmma-matmul-f32.mlir
@@ -1,5 +1,7 @@
// RUN: mlir-opt %s \
-// RUN: | mlir-opt -test-lower-to-nvvm="cubin-chip=sm_70" \
+// RUN: | mlir-opt -gpu-kernel-outlining \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,gpu-to-cubin{chip=sm_70}))' \
+// RUN: | mlir-opt --convert-scf-to-cf -gpu-to-llvm \
// RUN: | mlir-cpu-runner \
// RUN: --shared-libs=%mlir_cuda_runtime \
// RUN: --shared-libs=%mlir_runner_utils \
diff --git a/mlir/test/Integration/GPU/CUDA/all-reduce-and.mlir b/mlir/test/Integration/GPU/CUDA/all-reduce-and.mlir
index b131b8682ddee06..f4324a14a36b64a 100644
--- a/mlir/test/Integration/GPU/CUDA/all-reduce-and.mlir
+++ b/mlir/test/Integration/GPU/CUDA/all-reduce-and.mlir
@@ -1,5 +1,7 @@
// RUN: mlir-opt %s \
-// RUN: | mlir-opt -test-lower-to-nvvm \
+// RUN: | mlir-opt -gpu-kernel-outlining \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,gpu-to-cubin))' \
+// RUN: | mlir-opt -gpu-to-llvm \
// RUN: | mlir-cpu-runner \
// RUN: --shared-libs=%mlir_cuda_runtime \
// RUN: --shared-libs=%mlir_runner_utils \
@@ -8,7 +10,9 @@
// Same as above but with the memref bare pointer lowering convention.
// RUN: mlir-opt %s \
-// RUN: | mlir-opt -test-lower-to-nvvm="kernel-bare-ptr-calling-convention=1" \
+// RUN: | mlir-opt -gpu-kernel-outlining \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm{use-bare-ptr-memref-call-conv=1},gpu-to-cubin))' \
+// RUN: | mlir-opt -gpu-to-llvm="use-bare-pointers-for-kernels=1" \
// RUN: | mlir-cpu-runner \
// RUN: --shared-libs=%mlir_cuda_runtime \
// RUN: --shared-libs=%mlir_runner_utils \
diff --git a/mlir/test/Integration/GPU/CUDA/all-reduce-max.mlir b/mlir/test/Integration/GPU/CUDA/all-reduce-max.mlir
index 155423db7e05049..0a8d38f14527914 100644
--- a/mlir/test/Integration/GPU/CUDA/all-reduce-max.mlir
+++ b/mlir/test/Integration/GPU/CUDA/all-reduce-max.mlir
@@ -1,5 +1,7 @@
// RUN: mlir-opt %s \
-// RUN: | mlir-opt -test-lower-to-nvvm \
+// RUN: | mlir-opt -gpu-kernel-outlining \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,gpu-to-cubin))' \
+// RUN: | mlir-opt -gpu-to-llvm \
// RUN: | mlir-cpu-runner \
// RUN: --shared-libs=%mlir_cuda_runtime \
// RUN: --shared-libs=%mlir_runner_utils \
diff --git a/mlir/test/Integration/GPU/CUDA/all-reduce-min.mlir b/mlir/test/Integration/GPU/CUDA/all-reduce-min.mlir
index e5047b6efa3bf25..bcd785d35291c66 100644
--- a/mlir/test/Integration/GPU/CUDA/all-reduce-min.mlir
+++ b/mlir/test/Integration/GPU/CUDA/all-reduce-min.mlir
@@ -1,5 +1,7 @@
// RUN: mlir-opt %s \
-// RUN: | mlir-opt -test-lower-to-nvvm \
+// RUN: | mlir-opt -gpu-kernel-outlining \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,gpu-to-cubin))' \
+// RUN: | mlir-opt -gpu-to-llvm \
// RUN: | mlir-cpu-runner \
// RUN: --shared-libs=%mlir_cuda_runtime \
// RUN: --shared-libs=%mlir_runner_utils \
diff --git a/mlir/test/Integration/GPU/CUDA/all-reduce-op.mlir b/mlir/test/Integration/GPU/CUDA/all-reduce-op.mlir
index 163e9fdba60c1a9..aa4b0e8820479dc 100644
--- a/mlir/test/Integration/GPU/CUDA/all-reduce-op.mlir
+++ b/mlir/test/Integration/GPU/CUDA/all-reduce-op.mlir
@@ -1,5 +1,7 @@
// RUN: mlir-opt %s \
-// RUN: | mlir-opt -test-lower-to-nvvm \
+// RUN: | mlir-opt -gpu-kernel-outlining \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,gpu-to-cubin))' \
+// RUN: | mlir-opt -gpu-to-llvm \
// RUN: | mlir-cpu-runner \
// RUN: --shared-libs=%mlir_cuda_runtime \
// RUN: --shared-libs=%mlir_runner_utils \
diff --git a/mlir/test/Integration/GPU/CUDA/all-reduce-or.mlir b/mlir/test/Integration/GPU/CUDA/all-reduce-or.mlir
index 381db2639c371f3..2e7d046c3921411 100644
--- a/mlir/test/Integration/GPU/CUDA/all-reduce-or.mlir
+++ b/mlir/test/Integration/GPU/CUDA/all-reduce-or.mlir
@@ -1,5 +1,7 @@
// RUN: mlir-opt %s \
-// RUN: | mlir-opt -test-lower-to-nvvm \
+// RUN: | mlir-opt -gpu-kernel-outlining \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,gpu-to-cubin))' \
+// RUN: | mlir-opt -gpu-to-llvm \
// RUN: | mlir-cpu-runner \
// RUN: --shared-libs=%mlir_cuda_runtime \
// RUN: --shared-libs=%mlir_runner_utils \
diff --git a/mlir/test/Integration/GPU/CUDA/all-reduce-region.mlir b/mlir/test/Integration/GPU/CUDA/all-reduce-region.mlir
index 23c6c117e67f36b..32cfa27c8988a2c 100644
--- a/mlir/test/Integration/GPU/CUDA/all-reduce-region.mlir
+++ b/mlir/test/Integration/GPU/CUDA/all-reduce-region.mlir
@@ -1,5 +1,7 @@
// RUN: mlir-opt %s \
-// RUN: | mlir-opt -test-lower-to-nvvm \
+// RUN: | mlir-opt -gpu-kernel-outlining \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,gpu-to-cubin))' \
+// RUN: | mlir-opt -gpu-to-llvm \
// RUN: | mlir-cpu-runner \
// RUN: --shared-libs=%mlir_cuda_runtime \
// RUN: --shared-libs=%mlir_runner_utils \
diff --git a/mlir/test/Integration/GPU/CUDA/all-reduce-xor.mlir b/mlir/test/Integration/GPU/CUDA/all-reduce-xor.mlir
index 3c5a100b5b90d57..30767b9495b6f20 100644
--- a/mlir/test/Integration/GPU/CUDA/all-reduce-xor.mlir
+++ b/mlir/test/Integration/GPU/CUDA/all-reduce-xor.mlir
@@ -1,5 +1,7 @@
// RUN: mlir-opt %s \
-// RUN: | mlir-opt -test-lower-to-nvvm \
+// RUN: | mlir-opt -gpu-kernel-outlining \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,gpu-to-cubin))' \
+// RUN: | mlir-opt -gpu-to-llvm \
// RUN: | mlir-cpu-runner \
// RUN: --shared-libs=%mlir_cuda_runtime \
// RUN: --shared-libs=%mlir_runner_utils \
diff --git a/mlir/test/Integration/GPU/CUDA/async.mlir b/mlir/test/Integration/GPU/CUDA/async.mlir
index d2a5127a34c3bdd..e6dd91ace974322 100644
--- a/mlir/test/Integration/GPU/CUDA/async.mlir
+++ b/mlir/test/Integration/GPU/CUDA/async.mlir
@@ -1,7 +1,7 @@
// RUN: mlir-opt %s \
// RUN: | mlir-opt -gpu-kernel-outlining \
-// RUN: | mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm),nvvm-attach-target)' \
-// RUN: | mlir-opt -gpu-async-region -gpu-to-llvm -gpu-module-to-binary \
+// RUN: | mlir-opt -pass-pipeline='builtin.module(gpu.module(strip-debuginfo,convert-gpu-to-nvvm,gpu-to-cubin))' \
+// RUN: | mlir-opt -gpu-async-region -gpu-to-llvm \
// RUN: | mlir-opt -async-to-async-runtime -async-runtime-ref-counting \
// RUN: | ...
<truncated>
</pre>
</details>
https://github.com/llvm/llvm-project/pull/65848
More information about the Mlir-commits
mailing list