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.
125 lines
4.7 KiB
C++
125 lines
4.7 KiB
C++
//===- ArithToArmSME.cpp - Arith to ArmSME dialect conversion -------------===//
|
|
//
|
|
// 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/Conversion/ArithToArmSME/ArithToArmSME.h"
|
|
|
|
#include "mlir/Dialect/Arith/IR/Arith.h"
|
|
#include "mlir/Dialect/ArmSME/IR/ArmSME.h"
|
|
#include "mlir/Dialect/ArmSME/Utils/Utils.h"
|
|
#include "mlir/Pass/Pass.h"
|
|
#include "mlir/Transforms/GreedyPatternRewriteDriver.h"
|
|
|
|
namespace mlir {
|
|
#define GEN_PASS_DEF_ARITHTOARMSMECONVERSIONPASS
|
|
#include "mlir/Conversion/Passes.h.inc"
|
|
} // namespace mlir
|
|
|
|
#define DEBUG_TYPE "arith-to-arm-sme"
|
|
|
|
using namespace mlir;
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
// Conversion helpers
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
/// Returns true if 'val' is a splat of zero, false otherwise.
|
|
static bool isSplatZero(Type elemType, DenseElementsAttr val) {
|
|
if (llvm::isa<FloatType>(elemType))
|
|
return val && val.isSplat() && val.getSplatValue<APFloat>().isZero();
|
|
if (llvm::isa<IntegerType>(elemType))
|
|
return val && val.isSplat() && val.getSplatValue<APInt>().isZero();
|
|
return false;
|
|
}
|
|
|
|
namespace {
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
// ConstantOp
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
/// Conversion pattern for dense arith.constant.
|
|
struct ConstantOpToArmSMELowering : public OpRewritePattern<arith::ConstantOp> {
|
|
using OpRewritePattern<arith::ConstantOp>::OpRewritePattern;
|
|
|
|
LogicalResult matchAndRewrite(arith::ConstantOp constantOp,
|
|
PatternRewriter &rewriter) const final {
|
|
auto tileType = dyn_cast<VectorType>(constantOp.getType());
|
|
if (!tileType || !arm_sme::isValidSMETileVectorType(tileType))
|
|
return failure();
|
|
|
|
auto denseAttr = dyn_cast<DenseElementsAttr>(constantOp.getValueAttr());
|
|
if (!denseAttr || !denseAttr.isSplat())
|
|
return failure();
|
|
|
|
auto tileElementType = tileType.getElementType();
|
|
|
|
// Lower 'arith.constant dense<0>' to 'arm_sme.zero' op.
|
|
if (isSplatZero(tileElementType, denseAttr)) {
|
|
rewriter.replaceOpWithNewOp<arm_sme::ZeroOp>(constantOp, tileType);
|
|
return success();
|
|
}
|
|
|
|
// Lower non-zero constants to a loop of 'arm_sme.insert_tile_slice'
|
|
// ops that broadcast the constant to each tile slice.
|
|
auto loc = constantOp.getLoc();
|
|
|
|
// To fill a tile with a constant, we create a 1-D splat of the constant,
|
|
// then move that into each tile slice (the largest unit we can set at once,
|
|
// outside of operations like the outerproduct).
|
|
VectorType tileSliceType = VectorType::Builder(tileType).dropDim(0);
|
|
auto denseAttr1D = DenseElementsAttr::get(
|
|
tileSliceType, denseAttr.getSplatValue<Attribute>());
|
|
auto constantOp1D = rewriter.create<arith::ConstantOp>(loc, denseAttr1D);
|
|
|
|
auto initTile = rewriter.create<arm_sme::GetTileOp>(loc, tileType);
|
|
auto makeLoopBody = [&](OpBuilder &b, Location loc, Value tileSliceIndex,
|
|
Value currentTile) {
|
|
// Create 'arm_sme.insert_tile_slice' to write vector to tile
|
|
// slice.
|
|
auto nextTile = b.create<arm_sme::InsertTileSliceOp>(
|
|
loc, tileType, constantOp1D, currentTile, tileSliceIndex);
|
|
return nextTile.getResult();
|
|
};
|
|
auto forOp = mlir::arm_sme::createLoopOverTileSlices(
|
|
rewriter, loc, initTile, makeLoopBody);
|
|
rewriter.replaceOp(constantOp, forOp.getResult(0));
|
|
|
|
return success();
|
|
}
|
|
};
|
|
|
|
} // namespace
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
// Pattern population
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
void mlir::arith::populateArithToArmSMEConversionPatterns(
|
|
RewritePatternSet &patterns) {
|
|
patterns.add<ConstantOpToArmSMELowering>(patterns.getContext());
|
|
}
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
// Pass definition
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
namespace {
|
|
struct ArithToArmSMEConversionPass final
|
|
: impl::ArithToArmSMEConversionPassBase<ArithToArmSMEConversionPass> {
|
|
using impl::ArithToArmSMEConversionPassBase<
|
|
ArithToArmSMEConversionPass>::ArithToArmSMEConversionPassBase;
|
|
|
|
void runOnOperation() override {
|
|
RewritePatternSet patterns(&getContext());
|
|
arith::populateArithToArmSMEConversionPatterns(patterns);
|
|
if (failed(applyPatternsGreedily(getOperation(), std::move(patterns))))
|
|
return signalPassFailure();
|
|
}
|
|
};
|
|
} // namespace
|