[Mlir-commits] [mlir] [mlir][x86vector] Lower BF16 vector.contract to FMA using AVX2 BF16 packed ops. (PR #170267)

[Adam Siemieniuk]

================
@@ -0,0 +1,328 @@
+//===- VectorContractBF16ToFMA.cpp-----------------------------------------===//
+//
+// 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/MemRef/IR/MemRef.h"
+#include "mlir/Dialect/MemRef/Utils/MemRefUtils.h"
+#include "mlir/Dialect/Vector/IR/VectorOps.h"
+#include "mlir/Dialect/Vector/Utils/VectorUtils.h"
+#include "mlir/Dialect/X86Vector/Transforms.h"
+#include "mlir/Dialect/X86Vector/Utils/X86VectorUtils.h"
+#include "mlir/Dialect/X86Vector/X86VectorDialect.h"
+
+#include "mlir/IR/BuiltinAttributes.h"
+#include "mlir/IR/Dominance.h"
+#include "mlir/IR/PatternMatch.h"
+
+#include "mlir/Pass/Pass.h"
+#include "mlir/Transforms/GreedyPatternRewriteDriver.h"
+#include "llvm/Support/Casting.h"
+
+using namespace mlir;
+using namespace mlir::vector;
+using namespace mlir::x86vector;
+
+// Verifies that the LHS and RHS operands of a vector.contract are load or
+// vector.transfer_read operations on a memref source buffer, and checks
+// their bounds, dimensions, offsets, and strides.
+static bool validateVectorContractOperands(Value prodOp) {
+  Operation *defOp = prodOp.getDefiningOp();
+  if (!defOp)
+    return false;
+
+  // Verify that the transfer_read operation satisfies the following conditions:
+  // (1) It has no out-of-bounds dimensions.
+  // (2) The permutation map is non-identity.
----------------
adam-smnk wrote:

`is a minor identity`?
Maybe we can skip listing them in text. Now, the APIs speak for themselves.

https://github.com/llvm/llvm-project/pull/170267