[Mlir-commits] [mlir] [mlir][vector] fix: unroll vector.from_elements in gpu pipelines (PR #154774)

[Yang Bai]

https://github.com/yangtetris created https://github.com/llvm/llvm-project/pull/154774

### Problem

PR #142944 introduced a new canonicalization pattern which caused failures in the following GPU-related integration tests:

- mlir/test/Integration/GPU/CUDA/TensorCore/sm80/transform-mma-sync-matmul-f16-f16-accum.mlir
- mlir/test/Integration/GPU/CUDA/TensorCore/sm80/transform-mma-sync-matmul-f32.mlir

The issue occurs because the new canonicalization pattern can generate multi-dimensional `vector.from_elements` operations (rank > 1), but the GPU lowering pipelines were not equipped to handle these during the conversion to LLVM.

### Fix

This PR adds `vector::populateVectorFromElementsLoweringPatterns` to the GPU lowering passes that are integrated in `gpu-lower-to-nvvm-pipeline`:

- `GpuToLLVMConversionPass`: the general GPU-to-LLVM conversion pass.
- `LowerGpuOpsToNVVMOpsPass`: the NVVM-specific lowering pass.

>From 9b2fc699507eec6c0411c804e671e5f5b9d4b361 Mon Sep 17 00:00:00 2001
From: Yang Bai <yangb at nvidia.com>
Date: Thu, 21 Aug 2025 07:36:16 -0700
Subject: [PATCH] fix: add unroll vector.from_elements in gpu pipelines

---
 mlir/lib/Conversion/GPUCommon/GPUToLLVMConversion.cpp  | 3 +++
 mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp | 4 ++++
 2 files changed, 7 insertions(+)

diff --git a/mlir/lib/Conversion/GPUCommon/GPUToLLVMConversion.cpp b/mlir/lib/Conversion/GPUCommon/GPUToLLVMConversion.cpp
index 3cfbd898e49e2..e516118f75207 100644
--- a/mlir/lib/Conversion/GPUCommon/GPUToLLVMConversion.cpp
+++ b/mlir/lib/Conversion/GPUCommon/GPUToLLVMConversion.cpp
@@ -532,6 +532,9 @@ void GpuToLLVMConversionPass::runOnOperation() {
     // Vector transfer ops with rank > 1 should be lowered with VectorToSCF.
     vector::populateVectorTransferLoweringPatterns(patterns,
                                                    /*maxTransferRank=*/1);
+    // Transform N-D vector.from_elements to 1-D vector.from_elements before
+    // conversion.
+    vector::populateVectorFromElementsLoweringPatterns(patterns);
     if (failed(applyPatternsGreedily(getOperation(), std::move(patterns))))
       return signalPassFailure();
   }
diff --git a/mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp b/mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp
index 317bfc2970cf5..50ac1c60184eb 100644
--- a/mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp
+++ b/mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp
@@ -27,6 +27,7 @@
 #include "mlir/Dialect/Math/IR/Math.h"
 #include "mlir/Dialect/MemRef/IR/MemRef.h"
 #include "mlir/Dialect/NVGPU/IR/NVGPUDialect.h"
+#include "mlir/Dialect/Vector/Transforms/LoweringPatterns.h"
 #include "mlir/Transforms/DialectConversion.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 
@@ -369,6 +370,9 @@ struct LowerGpuOpsToNVVMOpsPass final
     {
       RewritePatternSet patterns(m.getContext());
       populateGpuRewritePatterns(patterns);
+      // Transform N-D vector.from_elements to 1-D vector.from_elements before
+      // conversion.
+      vector::populateVectorFromElementsLoweringPatterns(patterns);
       if (failed(applyPatternsGreedily(m, std::move(patterns))))
         return signalPassFailure();
     }