09dfc5713d
The greedy rewriter is used in many different flows and it has a lot of
convenience (work list management, debugging actions, tracing, etc). But
it combines two kinds of greedy behavior 1) how ops are matched, 2)
folding wherever it can.
These are independent forms of greedy and leads to inefficiency. E.g.,
cases where one need to create different phases in lowering and is
required to applying patterns in specific order split across different
passes. Using the driver one ends up needlessly retrying folding/having
multiple rounds of folding attempts, where one final run would have
sufficed.
Of course folks can locally avoid this behavior by just building their
own, but this is also a common requested feature that folks keep on
working around locally in suboptimal ways.
For downstream users, there should be no behavioral change. Updating
from the deprecated should just be a find and replace (e.g., `find ./
-type f -exec sed -i
's|applyPatternsAndFoldGreedily|applyPatternsGreedily|g' {} \;` variety)
as the API arguments hasn't changed between the two.
103 lines
3.7 KiB
C++
103 lines
3.7 KiB
C++
//===- TestLinalgFusionTransforms.cpp - Test Linalg fusion patterns -------===//
|
|
//
|
|
// 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
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
//
|
|
// This file implements logic for testing Linalg fusion patterns.
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#include "mlir/Dialect/Affine/IR/AffineOps.h"
|
|
#include "mlir/Dialect/Func/IR/FuncOps.h"
|
|
#include "mlir/Dialect/Linalg/Transforms/Transforms.h"
|
|
#include "mlir/Dialect/SCF/Transforms/Patterns.h"
|
|
#include "mlir/Pass/Pass.h"
|
|
#include "mlir/Pass/PassManager.h"
|
|
#include "mlir/Transforms/GreedyPatternRewriteDriver.h"
|
|
#include "mlir/Transforms/Passes.h"
|
|
|
|
using namespace mlir;
|
|
using namespace mlir::linalg;
|
|
|
|
static LogicalResult fuseLinalgOpsGreedily(func::FuncOp f) {
|
|
OpBuilder b(f);
|
|
|
|
// Save original Linalg ops, we only want to make a pass over those.
|
|
SmallVector<LinalgOp, 8> linalgOps;
|
|
f.walk([&](LinalgOp op) {
|
|
// TODO: support multi-results.
|
|
if (op->getNumResults() <= 1)
|
|
linalgOps.push_back(op);
|
|
});
|
|
|
|
// Tile and Fuse for tensors inputs (TODO: all tensor operands).
|
|
bool changed = false;
|
|
for (LinalgOp linalgOp : llvm::reverse(linalgOps)) {
|
|
for (OpOperand &opOperand : linalgOp->getOpOperands()) {
|
|
if (isa<MemRefType>(opOperand.get().getType()))
|
|
continue;
|
|
if (isa<RankedTensorType>(opOperand.get().getType())) {
|
|
// Tile and Fuse tensor input.
|
|
if (opOperand.getOperandNumber() >= linalgOp.getNumDpsInputs())
|
|
continue;
|
|
auto info = fuseProducerOfTensor(b, opOperand);
|
|
if (failed(info))
|
|
continue;
|
|
auto *originalOp = info->originalProducer.getOperation();
|
|
auto *originalOpInLinalgOpsVector =
|
|
std::find(linalgOps.begin(), linalgOps.end(), originalOp);
|
|
*originalOpInLinalgOpsVector = info->fusedProducer;
|
|
// Don't mark for erasure in the tensor case, let DCE handle this.
|
|
changed = true;
|
|
}
|
|
}
|
|
}
|
|
|
|
return changed ? success() : failure();
|
|
}
|
|
|
|
namespace {
|
|
struct TestLinalgGreedyFusion
|
|
: public PassWrapper<TestLinalgGreedyFusion, OperationPass<func::FuncOp>> {
|
|
MLIR_DEFINE_EXPLICIT_INTERNAL_INLINE_TYPE_ID(TestLinalgGreedyFusion)
|
|
|
|
void getDependentDialects(DialectRegistry ®istry) const override {
|
|
registry.insert<affine::AffineDialect, linalg::LinalgDialect,
|
|
memref::MemRefDialect, scf::SCFDialect>();
|
|
}
|
|
StringRef getArgument() const final { return "test-linalg-greedy-fusion"; }
|
|
StringRef getDescription() const final {
|
|
return "Test Linalg fusion by applying a greedy test transformation.";
|
|
}
|
|
void runOnOperation() override {
|
|
MLIRContext *context = &getContext();
|
|
RewritePatternSet patterns =
|
|
linalg::getLinalgTilingCanonicalizationPatterns(context);
|
|
patterns.add<ExtractSliceOfPadTensorSwapPattern>(context);
|
|
scf::populateSCFForLoopCanonicalizationPatterns(patterns);
|
|
FrozenRewritePatternSet frozenPatterns(std::move(patterns));
|
|
OpPassManager pm(func::FuncOp::getOperationName());
|
|
pm.addPass(createLoopInvariantCodeMotionPass());
|
|
pm.addPass(createCanonicalizerPass());
|
|
pm.addPass(createCSEPass());
|
|
do {
|
|
(void)applyPatternsGreedily(getOperation(), frozenPatterns);
|
|
if (failed(runPipeline(pm, getOperation())))
|
|
this->signalPassFailure();
|
|
} while (succeeded(fuseLinalgOpsGreedily(getOperation())));
|
|
}
|
|
};
|
|
} // namespace
|
|
|
|
namespace mlir {
|
|
namespace test {
|
|
void registerTestLinalgGreedyFusion() {
|
|
PassRegistration<TestLinalgGreedyFusion>();
|
|
}
|
|
|
|
} // namespace test
|
|
} // namespace mlir
|