[Mlir-commits] [mlir] [MLIR][NVGPU] Fix the cga_cluster.mlir test (PR #112191)

[llvmlistbot at llvm.org]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-mlir-gpu

@llvm/pr-subscribers-mlir

Author: Durgadoss R (durga4github)

<details>
<summary>Changes</summary>

This patch fixes the sm90 cluster test by:
* Fixing a typo in LowerGpuOpsToNVVMOps where one of the ClusterDim Op conversion pattern should actually be for the 
   ClusterDimBlocks Op. This addresses the compilation error for this test.
* The grid-size should be (4,4,1) instead of (2,2,1). This passes the scf-if check against the threshold of 3 below and actually 
   generates the required prints from the GPU.

---
Full diff: https://github.com/llvm/llvm-project/pull/112191.diff


2 Files Affected:

- (modified) mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp (+2-2) 
- (modified) mlir/test/Integration/GPU/CUDA/sm90/cga_cluster.mlir (+1-1) 


``````````diff
diff --git a/mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp b/mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp
index e83574b7342725..8638ae603a0bea 100644
--- a/mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp
+++ b/mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp
@@ -373,8 +373,8 @@ void mlir::populateGpuToNVVMConversionPatterns(
       NVVM::BlockInClusterIdYOp, NVVM::BlockInClusterIdZOp>>(
       converter, IndexKind::Other, IntrType::Id);
   patterns.add<gpu::index_lowering::OpLowering<
-      gpu::ClusterDimOp, NVVM::ClusterDimXOp, NVVM::ClusterDimYOp,
-      NVVM::ClusterDimZOp>>(converter, IndexKind::Other, IntrType::Dim);
+      gpu::ClusterDimBlocksOp, NVVM::ClusterDimBlocksXOp, NVVM::ClusterDimBlocksYOp,
+      NVVM::ClusterDimBlocksZOp>>(converter, IndexKind::Other, IntrType::Dim);
   patterns.add<gpu::index_lowering::OpLowering<
       gpu::BlockIdOp, NVVM::BlockIdXOp, NVVM::BlockIdYOp, NVVM::BlockIdZOp>>(
       converter, IndexKind::Grid, IntrType::Id);
diff --git a/mlir/test/Integration/GPU/CUDA/sm90/cga_cluster.mlir b/mlir/test/Integration/GPU/CUDA/sm90/cga_cluster.mlir
index 5c11d80178f727..c70c940564a264 100644
--- a/mlir/test/Integration/GPU/CUDA/sm90/cga_cluster.mlir
+++ b/mlir/test/Integration/GPU/CUDA/sm90/cga_cluster.mlir
@@ -18,7 +18,7 @@ module attributes {gpu.container_module} {
     return
   }
   gpu.module @gpumodule {
-    gpu.func @kernel_cluster() kernel attributes {gpu.known_block_size = array<i32: 1, 1, 1>, gpu.known_grid_size = array<i32: 2, 2, 1>} {
+    gpu.func @kernel_cluster() kernel attributes {gpu.known_block_size = array<i32: 1, 1, 1>, gpu.known_grid_size = array<i32: 4, 4, 1>} {
       %cidX = gpu.cluster_id  x
       %cidY = gpu.cluster_id  y
       %cidZ = gpu.cluster_id  z

``````````

</details>


https://github.com/llvm/llvm-project/pull/112191