[Mlir-commits] [mlir] [mlir] [XeGPU] Add XeGPU workgroup to subgroup pass (PR #139477)

[Adam Siemieniuk]

================
@@ -0,0 +1,387 @@
+//===- XeGPUWgToSg.cpp - XeGPU WorkGroup to Subgroup Pass -------===//
+//
+// 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/XeGPU/Transforms/Passes.h"
+
+#include "mlir/Dialect/Index/IR/IndexDialect.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
+#include "mlir/Dialect/Utils/IndexingUtils.h"
+#include "mlir/Dialect/XeGPU/IR/XeGPU.h"
+#include "mlir/Dialect/XeGPU/Transforms/Transforms.h"
+#include "mlir/Transforms/DialectConversion.h"
+#include "llvm/Support/Debug.h"
+#include <mlir/Dialect/GPU/IR/GPUDialect.h>
+#include <mlir/Dialect/Index/IR/IndexOps.h>
+#include <numeric>
+
+namespace mlir {
+namespace xegpu {
+#define GEN_PASS_DEF_XEGPUWGTOSG
+#include "mlir/Dialect/XeGPU/Transforms/Passes.h.inc"
+} // namespace xegpu
+} // namespace mlir
+
+#define DEBUG_TYPE "xegpu-wg-to-sg"
+#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
+#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
+
+using namespace mlir;
+
+namespace {
+
+// clang-format off
+/// This pattern transforms the CreateNdDescOp to create a subgroup descriptor
+/// from a workgroup descriptor. It replaces the offsets and sizes with
+/// appropriate values for the subgroup.
+/// It uses round-robin assignment to distribute the work to the subgroups.
+/// Following create_nd_desc operation:,
+///    %tdesc = xegpu.create_nd_tdesc %src[0, 0] : memref<24x24xf32>
+///       -> !xegpu.tensor_desc<24x24xf32, #xegpu.layout<sg_layout = [4, 4],
+///           sg_data = [2, 2], lane_layout = [2, 2], lane_data = [1, 1]>>
+/// is converted to 9 subgroup level operations based on the sg_layout & sg_data:
+///    %tdesc = xegpu.create_nd_tdesc %src[off1, off2] : memref<24x24xf32> ->
+///           !xegpu.tensor_desc<2x2xf32, #xegpu.layout<lane_layout = [2, 2], lane_data = [1, 1]>>
+///
+/// The sg_layout and sg_data attributes are dropped after the pass as they are no longer needed.
+///
+/// 24x24 matrix distribution example:
+/// sg_layout = [4, 4], sg_data = [2, 2]
+/// Each 8x8 matrix within the 24x24 matrix is called a distribution unit.
+/// dist_unit_shape = [8, 8] --> sg_layout[i] * sg_data[i]
+///
+/// +------------------------+
+/// | 8x8 | 8x8 | 8x8 |      <- 3 tiles across
+/// |-----+-----+-----|
+/// | 8x8 | 8x8 | 8x8 |      <- 3 tiles down
+/// |-----+-----+-----|
+/// | 8x8 | 8x8 | 8x8 |
+/// +------------------------+
+///
+/// Each 8x8 tile is further subdivided among subgroups:
+/// +------------------------+
+/// | 2x2 2x2 2x2 2x2 |  <- 4 subgroups across (each handles 2 columns)
+/// | 2x2 2x2 2x2 2x2 |  <- 4 subgroups down (each handles 2 rows)
+/// | 2x2 2x2 2x2 2x2 |
+/// | 2x2 2x2 2x2 2x2 |
+/// +------------------------+
+///
+/// Since the 24x24 matrix is divided into 8x8 distribution units, there will be 9
+/// distribution units (3x3) in total. Hence the 9 subgroup level operations.
+// clang-format on
+struct WgToSgCreateNdOp : public OpConversionPattern<xegpu::CreateNdDescOp> {
+  using OpConversionPattern<xegpu::CreateNdDescOp>::OpConversionPattern;
+
+  // Helper to extract mixed offsets into a Value array
+  SmallVector<Value> extractOffsets(ConversionPatternRewriter &rewriter,
+                                    xegpu::CreateNdDescOp op) const {
+    llvm::SmallVector<Value> offsets;
+    auto staticOffsets = op.getStaticOffsets();
+    auto dynamicOffsets = op.getOffsets();
+
+    for (size_t i = 0, j = 0; i != staticOffsets.size(); i++) {
+      if (ShapedType::isDynamic(staticOffsets[i]))
+        offsets.push_back(dynamicOffsets[j++]);
+      else
+        offsets.push_back(rewriter.create<arith::ConstantIndexOp>(
+            op.getLoc(), staticOffsets[i]));
+    }
+    return offsets;
+  }
+
+  // Convert linear subgroup ID to 2D coordinates
+  // TODO: Delinearize for nD
----------------
adam-smnk wrote:

Nothing wrong with that if it's helpful. AFAIK, the delinearization just adds affine maps and `affine.apply` which is nice for readability and (imo) less error prone than composing all the IR math by hand.
If it works fine, I'd still suggest using it as it solves the nD generalization.

https://github.com/llvm/llvm-project/pull/139477