[Mlir-commits] [mlir] [mlir][XeGPU][Transform] Add vectorlinearize transform pass. (PR #158084)

[Adam Siemieniuk]

================
@@ -0,0 +1,101 @@
+//===- XeGPUVectorLinearize.cpp - Linearizes n-D vectors to 1-D vectors
+//-------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir/Dialect/SCF/IR/SCF.h"
+#include "mlir/Dialect/SCF/Transforms/Patterns.h"
+#include "mlir/Dialect/Vector/IR/VectorOps.h"
+#include "mlir/Dialect/Vector/Transforms/LoweringPatterns.h"
+#include "mlir/Dialect/Vector/Transforms/VectorRewritePatterns.h"
+#include "mlir/Dialect/XeGPU/IR/XeGPU.h"
+#include "mlir/Dialect/XeGPU/Transforms/Passes.h"
+#include "mlir/Pass/Pass.h"
+#include "mlir/Transforms/DialectConversion.h"
+#include "mlir/Transforms/GreedyPatternRewriteDriver.h"
+
+#include <optional>
+
+namespace mlir {
+namespace xegpu {
+#define GEN_PASS_DEF_XEGPUVECTORLINEARIZE
+#include "mlir/Dialect/XeGPU/Transforms/Passes.h.inc"
+} // namespace xegpu
+} // namespace mlir
+
+#define DEBUG_TYPE "xegpu-vector-linearize"
+
+using namespace mlir;
+
+namespace {
+struct XeGPUVectorLinearizePass final
+    : public xegpu::impl::XeGPUVectorLinearizeBase<XeGPUVectorLinearizePass> {
+  void runOnOperation() override {
+    // vector.broadcast and vector.gather requires progressive lowering
+    {
+      mlir::RewritePatternSet patterns(&getContext());
+      mlir::vector::populateVectorBroadcastLoweringPatterns(patterns);
+      mlir::vector::populateVectorGatherLoweringPatterns(patterns);
+      mlir::vector::populateVectorGatherToConditionalLoadPatterns(patterns);
+      // vector.transpose lowering
+      // Shuffle16x16 will fallback to Shuffle1D for non 16x16 sizes.
+      mlir::vector::populateVectorTransposeLoweringPatterns(
+          patterns, mlir::vector::VectorTransposeLowering::Shuffle16x16);
+      (void)mlir::applyPatternsGreedily(getOperation(), std::move(patterns));
+    }
+
+    // Unroll load store from <<MxN> to M <1xN> load/stores and then linearize
+    {
+      mlir::RewritePatternSet patterns(&getContext());
+      mlir::vector::UnrollVectorOptions vectorOptions;
+      vectorOptions.setNativeShapeFn(
+          [](mlir::Operation *op) -> std::optional<mlir::SmallVector<int64_t>> {
+            auto extractVectorType =
+                [](mlir::Operation *op) -> mlir::VectorType {
+              if (auto loadOp = mlir::dyn_cast<mlir::vector::LoadOp>(op))
+                return mlir::dyn_cast<mlir::VectorType>(
+                    loadOp.getResult().getType());
+              if (auto storeOp = mlir::dyn_cast<mlir::vector::StoreOp>(op))
+                return mlir::dyn_cast<mlir::VectorType>(
+                    storeOp.getValueToStore().getType());
----------------
adam-smnk wrote:

nit: you can use directly `op.getVectorType()` for both

https://github.com/llvm/llvm-project/pull/158084