StandardOps/Transforms/ExpandOps.cpp

//===- StdExpandDivs.cpp - Code to prepare Std for lowering Divs to LLVM  -===//
//
// 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 Std transformations to expand Divs operation to help for the
// lowering to LLVM. Currently implemented transformations are Ceil and Floor
// for Signed Integers.
//
//===----------------------------------------------------------------------===//

#include "PassDetail.h"
#include "mlir/Dialect/StandardOps/IR/Ops.h"
#include "mlir/Dialect/StandardOps/Transforms/Passes.h"
#include "mlir/IR/PatternMatch.h"

using namespace mlir;

namespace {

/// Converts `atomic_rmw` that cannot be lowered to a simple atomic op with
/// AtomicRMWOpLowering pattern, e.g. with "minf" or "maxf" attributes, to
/// `generic_atomic_rmw` with the expanded code.
///
/// %x = atomic_rmw "maxf" %fval, %F[%i] : (f32, memref<10xf32>) -> f32
///
/// will be lowered to
///
/// %x = std.generic_atomic_rmw %F[%i] : memref<10xf32> {
/// ^bb0(%current: f32):
///   %cmp = cmpf "ogt", %current, %fval : f32
///   %new_value = select %cmp, %current, %fval : f32
///   atomic_yield %new_value : f32
/// }
struct AtomicRMWOpConverter : public OpRewritePattern<AtomicRMWOp> {
public:
  using OpRewritePattern::OpRewritePattern;

  LogicalResult matchAndRewrite(AtomicRMWOp op,
                                PatternRewriter &rewriter) const final {
    CmpFPredicate predicate;
    switch (op.kind()) {
    case AtomicRMWKind::maxf:
      predicate = CmpFPredicate::OGT;
      break;
    case AtomicRMWKind::minf:
      predicate = CmpFPredicate::OLT;
      break;
    default:
      return failure();
    }

    auto loc = op.getLoc();
    auto genericOp =
        rewriter.create<GenericAtomicRMWOp>(loc, op.memref(), op.indices());
    OpBuilder bodyBuilder =
        OpBuilder::atBlockEnd(genericOp.getBody(), rewriter.getListener());

    Value lhs = genericOp.getCurrentValue();
    Value rhs = op.value();
    Value cmp = bodyBuilder.create<CmpFOp>(loc, predicate, lhs, rhs);
    Value select = bodyBuilder.create<SelectOp>(loc, cmp, lhs, rhs);
    bodyBuilder.create<AtomicYieldOp>(loc, select);

    rewriter.replaceOp(op, genericOp.getResult());
    return success();
  }
};

/// Converts `memref_reshape` that has a target shape of a statically-known
/// size to `memref_reinterpret_cast`.
struct MemRefReshapeOpConverter : public OpRewritePattern<MemRefReshapeOp> {
public:
  using OpRewritePattern::OpRewritePattern;

  LogicalResult matchAndRewrite(MemRefReshapeOp op,
                                PatternRewriter &rewriter) const final {
    auto shapeType = op.shape().getType().cast<MemRefType>();
    if (!shapeType.hasStaticShape())
      return failure();

    int64_t rank = shapeType.cast<MemRefType>().getDimSize(0);
    SmallVector<OpFoldResult, 4> sizes, strides;
    sizes.resize(rank);
    strides.resize(rank);

    Location loc = op.getLoc();
    Value stride = rewriter.create<ConstantIndexOp>(loc, 1);
    for (int i = rank - 1; i >= 0; --i) {
      Value index = rewriter.create<ConstantIndexOp>(loc, i);
      Value size = rewriter.create<LoadOp>(loc, op.shape(), index);
      if (!size.getType().isa<IndexType>())
        size = rewriter.create<IndexCastOp>(loc, size, rewriter.getIndexType());
      sizes[i] = size;
      strides[i] = stride;
      if (i > 0)
        stride = rewriter.create<MulIOp>(loc, stride, size);
    }
    rewriter.replaceOpWithNewOp<MemRefReinterpretCastOp>(
        op, op.getType(), op.source(), /*offset=*/rewriter.getIndexAttr(0),
        sizes, strides);
    return success();
  }
};

/// Expands SignedCeilDivIOP (n, m) into
///   1) x = (m > 0) ? -1 : 1
///   2) (n*m>0) ? ((n+x) / m) + 1 : - (-n / m)
struct SignedCeilDivIOpConverter : public OpRewritePattern<SignedCeilDivIOp> {
public:
  using OpRewritePattern::OpRewritePattern;
  LogicalResult matchAndRewrite(SignedCeilDivIOp op,
                                PatternRewriter &rewriter) const final {
    Location loc = op.getLoc();
    SignedCeilDivIOp signedCeilDivIOp = cast<SignedCeilDivIOp>(op);
    Type type = signedCeilDivIOp.getType();
    Value a = signedCeilDivIOp.lhs();
    Value b = signedCeilDivIOp.rhs();
    Value plusOne =
        rewriter.create<ConstantOp>(loc, rewriter.getIntegerAttr(type, 1));
    Value zero =
        rewriter.create<ConstantOp>(loc, rewriter.getIntegerAttr(type, 0));
    Value minusOne =
        rewriter.create<ConstantOp>(loc, rewriter.getIntegerAttr(type, -1));
    // Compute x = (b>0) ? -1 : 1.
    Value compare = rewriter.create<CmpIOp>(loc, CmpIPredicate::sgt, b, zero);
    Value x = rewriter.create<SelectOp>(loc, compare, minusOne, plusOne);
    // Compute positive res: 1 + ((x+a)/b).
    Value xPlusA = rewriter.create<AddIOp>(loc, x, a);
    Value xPlusADivB = rewriter.create<SignedDivIOp>(loc, xPlusA, b);
    Value posRes = rewriter.create<AddIOp>(loc, plusOne, xPlusADivB);
    // Compute negative res: - ((-a)/b).
    Value minusA = rewriter.create<SubIOp>(loc, zero, a);
    Value minusADivB = rewriter.create<SignedDivIOp>(loc, minusA, b);
    Value negRes = rewriter.create<SubIOp>(loc, zero, minusADivB);
    // Result is (a*b>0) ? pos result : neg result.
    // Note, we want to avoid using a*b because of possible overflow.
    // The case that matters are a>0, a==0, a<0, b>0 and b<0. We do
    // not particuliarly care if a*b<0 is true or false when b is zero
    // as this will result in an illegal divide. So `a*b<0` can be reformulated
    // as `(a<0 && b<0) || (a>0 && b>0)' or `(a<0 && b<0) || (a>0 && b>=0)'.
    // We pick the first expression here.
    Value aNeg = rewriter.create<CmpIOp>(loc, CmpIPredicate::slt, a, zero);
    Value aPos = rewriter.create<CmpIOp>(loc, CmpIPredicate::sgt, a, zero);
    Value bNeg = rewriter.create<CmpIOp>(loc, CmpIPredicate::slt, b, zero);
    Value bPos = rewriter.create<CmpIOp>(loc, CmpIPredicate::sgt, b, zero);
    Value firstTerm = rewriter.create<AndOp>(loc, aNeg, bNeg);
    Value secondTerm = rewriter.create<AndOp>(loc, aPos, bPos);
    Value compareRes = rewriter.create<OrOp>(loc, firstTerm, secondTerm);
    Value res = rewriter.create<SelectOp>(loc, compareRes, posRes, negRes);
    // Perform substitution and return success.
    rewriter.replaceOp(op, {res});
    return success();
  }
};

/// Expands SignedFloorDivIOP (n, m) into
///   1)  x = (m<0) ? 1 : -1
///   2)  return (n*m<0) ? - ((-n+x) / m) -1 : n / m
struct SignedFloorDivIOpConverter : public OpRewritePattern<SignedFloorDivIOp> {
public:
  using OpRewritePattern::OpRewritePattern;
  LogicalResult matchAndRewrite(SignedFloorDivIOp op,
                                PatternRewriter &rewriter) const final {
    Location loc = op.getLoc();
    SignedFloorDivIOp signedFloorDivIOp = cast<SignedFloorDivIOp>(op);
    Type type = signedFloorDivIOp.getType();
    Value a = signedFloorDivIOp.lhs();
    Value b = signedFloorDivIOp.rhs();
    Value plusOne =
        rewriter.create<ConstantOp>(loc, rewriter.getIntegerAttr(type, 1));
    Value zero =
        rewriter.create<ConstantOp>(loc, rewriter.getIntegerAttr(type, 0));
    Value minusOne =
        rewriter.create<ConstantOp>(loc, rewriter.getIntegerAttr(type, -1));
    // Compute x = (b<0) ? 1 : -1.
    Value compare = rewriter.create<CmpIOp>(loc, CmpIPredicate::slt, b, zero);
    Value x = rewriter.create<SelectOp>(loc, compare, plusOne, minusOne);
    // Compute negative res: -1 - ((x-a)/b).
    Value xMinusA = rewriter.create<SubIOp>(loc, x, a);
    Value xMinusADivB = rewriter.create<SignedDivIOp>(loc, xMinusA, b);
    Value negRes = rewriter.create<SubIOp>(loc, minusOne, xMinusADivB);
    // Compute positive res: a/b.
    Value posRes = rewriter.create<SignedDivIOp>(loc, a, b);
    // Result is (a*b<0) ? negative result : positive result.
    // Note, we want to avoid using a*b because of possible overflow.
    // The case that matters are a>0, a==0, a<0, b>0 and b<0. We do
    // not particuliarly care if a*b<0 is true or false when b is zero
    // as this will result in an illegal divide. So `a*b<0` can be reformulated
    // as `(a>0 && b<0) || (a>0 && b<0)' or `(a>0 && b<0) || (a>0 && b<=0)'.
    // We pick the first expression here.
    Value aNeg = rewriter.create<CmpIOp>(loc, CmpIPredicate::slt, a, zero);
    Value aPos = rewriter.create<CmpIOp>(loc, CmpIPredicate::sgt, a, zero);
    Value bNeg = rewriter.create<CmpIOp>(loc, CmpIPredicate::slt, b, zero);
    Value bPos = rewriter.create<CmpIOp>(loc, CmpIPredicate::sgt, b, zero);
    Value firstTerm = rewriter.create<AndOp>(loc, aNeg, bPos);
    Value secondTerm = rewriter.create<AndOp>(loc, aPos, bNeg);
    Value compareRes = rewriter.create<OrOp>(loc, firstTerm, secondTerm);
    Value res = rewriter.create<SelectOp>(loc, compareRes, negRes, posRes);
    // Perform substitution and return success.
    rewriter.replaceOp(op, {res});
    return success();
  }
};

struct StdExpandOpsPass : public StdExpandOpsBase<StdExpandOpsPass> {
  void runOnFunction() override {
    MLIRContext &ctx = getContext();

    OwningRewritePatternList patterns;
    populateStdExpandOpsPatterns(&ctx, patterns);

    ConversionTarget target(getContext());

    target.addLegalDialect<StandardOpsDialect>();
    target.addDynamicallyLegalOp<AtomicRMWOp>([](AtomicRMWOp op) {
      return op.kind() != AtomicRMWKind::maxf &&
             op.kind() != AtomicRMWKind::minf;
    });
    target.addDynamicallyLegalOp<MemRefReshapeOp>([](MemRefReshapeOp op) {
      return !op.shape().getType().cast<MemRefType>().hasStaticShape();
    });
    target.addIllegalOp<SignedCeilDivIOp>();
    target.addIllegalOp<SignedFloorDivIOp>();
    if (failed(
            applyPartialConversion(getFunction(), target, std::move(patterns))))
      signalPassFailure();
  }
};

} // namespace

void mlir::populateStdExpandOpsPatterns(MLIRContext *context,
                                        OwningRewritePatternList &patterns) {
  patterns.insert<AtomicRMWOpConverter, MemRefReshapeOpConverter,
                  SignedCeilDivIOpConverter, SignedFloorDivIOpConverter>(
      context);
}

std::unique_ptr<Pass> mlir::createStdExpandOpsPass() {
  return std::make_unique<StdExpandOpsPass>();
}