1 //===- StdExpandDivs.cpp - Code to prepare Std for lowering Divs to LLVM -===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This file Std transformations to expand Divs operation to help for the 10 // lowering to LLVM. Currently implemented transformations are Ceil and Floor 11 // for Signed Integers. 12 // 13 //===----------------------------------------------------------------------===// 14 15 #include "PassDetail.h" 16 17 #include "mlir/Dialect/Arithmetic/IR/Arithmetic.h" 18 #include "mlir/Dialect/Arithmetic/Transforms/Passes.h" 19 #include "mlir/Dialect/MemRef/IR/MemRef.h" 20 #include "mlir/Dialect/MemRef/Transforms/Passes.h" 21 #include "mlir/Dialect/StandardOps/IR/Ops.h" 22 #include "mlir/IR/TypeUtilities.h" 23 #include "mlir/Transforms/DialectConversion.h" 24 25 using namespace mlir; 26 27 namespace { 28 29 /// Converts `atomic_rmw` that cannot be lowered to a simple atomic op with 30 /// AtomicRMWOpLowering pattern, e.g. with "minf" or "maxf" attributes, to 31 /// `memref.generic_atomic_rmw` with the expanded code. 32 /// 33 /// %x = atomic_rmw "maxf" %fval, %F[%i] : (f32, memref<10xf32>) -> f32 34 /// 35 /// will be lowered to 36 /// 37 /// %x = memref.generic_atomic_rmw %F[%i] : memref<10xf32> { 38 /// ^bb0(%current: f32): 39 /// %cmp = arith.cmpf "ogt", %current, %fval : f32 40 /// %new_value = select %cmp, %current, %fval : f32 41 /// memref.atomic_yield %new_value : f32 42 /// } 43 struct AtomicRMWOpConverter : public OpRewritePattern<memref::AtomicRMWOp> { 44 public: 45 using OpRewritePattern::OpRewritePattern; 46 47 LogicalResult matchAndRewrite(memref::AtomicRMWOp op, 48 PatternRewriter &rewriter) const final { 49 arith::CmpFPredicate predicate; 50 switch (op.kind()) { 51 case arith::AtomicRMWKind::maxf: 52 predicate = arith::CmpFPredicate::OGT; 53 break; 54 case arith::AtomicRMWKind::minf: 55 predicate = arith::CmpFPredicate::OLT; 56 break; 57 default: 58 return failure(); 59 } 60 61 auto loc = op.getLoc(); 62 auto genericOp = rewriter.create<memref::GenericAtomicRMWOp>( 63 loc, op.memref(), op.indices()); 64 OpBuilder bodyBuilder = 65 OpBuilder::atBlockEnd(genericOp.getBody(), rewriter.getListener()); 66 67 Value lhs = genericOp.getCurrentValue(); 68 Value rhs = op.value(); 69 Value cmp = bodyBuilder.create<arith::CmpFOp>(loc, predicate, lhs, rhs); 70 Value select = bodyBuilder.create<arith::SelectOp>(loc, cmp, lhs, rhs); 71 bodyBuilder.create<memref::AtomicYieldOp>(loc, select); 72 73 rewriter.replaceOp(op, genericOp.getResult()); 74 return success(); 75 } 76 }; 77 78 /// Converts `memref.reshape` that has a target shape of a statically-known 79 /// size to `memref.reinterpret_cast`. 80 struct MemRefReshapeOpConverter : public OpRewritePattern<memref::ReshapeOp> { 81 public: 82 using OpRewritePattern::OpRewritePattern; 83 84 LogicalResult matchAndRewrite(memref::ReshapeOp op, 85 PatternRewriter &rewriter) const final { 86 auto shapeType = op.shape().getType().cast<MemRefType>(); 87 if (!shapeType.hasStaticShape()) 88 return failure(); 89 90 int64_t rank = shapeType.cast<MemRefType>().getDimSize(0); 91 SmallVector<OpFoldResult, 4> sizes, strides; 92 sizes.resize(rank); 93 strides.resize(rank); 94 95 Location loc = op.getLoc(); 96 Value stride = rewriter.create<arith::ConstantIndexOp>(loc, 1); 97 for (int i = rank - 1; i >= 0; --i) { 98 Value size; 99 // Load dynamic sizes from the shape input, use constants for static dims. 100 if (op.getType().isDynamicDim(i)) { 101 Value index = rewriter.create<arith::ConstantIndexOp>(loc, i); 102 size = rewriter.create<memref::LoadOp>(loc, op.shape(), index); 103 if (!size.getType().isa<IndexType>()) 104 size = rewriter.create<arith::IndexCastOp>( 105 loc, rewriter.getIndexType(), size); 106 sizes[i] = size; 107 } else { 108 sizes[i] = rewriter.getIndexAttr(op.getType().getDimSize(i)); 109 size = 110 rewriter.create<arith::ConstantOp>(loc, sizes[i].get<Attribute>()); 111 } 112 strides[i] = stride; 113 if (i > 0) 114 stride = rewriter.create<arith::MulIOp>(loc, stride, size); 115 } 116 rewriter.replaceOpWithNewOp<memref::ReinterpretCastOp>( 117 op, op.getType(), op.source(), /*offset=*/rewriter.getIndexAttr(0), 118 sizes, strides); 119 return success(); 120 } 121 }; 122 123 struct ExpandOpsPass : public ExpandOpsBase<ExpandOpsPass> { 124 void runOnOperation() override { 125 MLIRContext &ctx = getContext(); 126 127 RewritePatternSet patterns(&ctx); 128 memref::populateExpandOpsPatterns(patterns); 129 ConversionTarget target(ctx); 130 131 target.addLegalDialect<arith::ArithmeticDialect, memref::MemRefDialect, 132 StandardOpsDialect>(); 133 target.addDynamicallyLegalOp<memref::AtomicRMWOp>( 134 [](memref::AtomicRMWOp op) { 135 return op.kind() != arith::AtomicRMWKind::maxf && 136 op.kind() != arith::AtomicRMWKind::minf; 137 }); 138 target.addDynamicallyLegalOp<memref::ReshapeOp>([](memref::ReshapeOp op) { 139 return !op.shape().getType().cast<MemRefType>().hasStaticShape(); 140 }); 141 if (failed(applyPartialConversion(getOperation(), target, 142 std::move(patterns)))) 143 signalPassFailure(); 144 } 145 }; 146 147 } // namespace 148 149 void mlir::memref::populateExpandOpsPatterns(RewritePatternSet &patterns) { 150 patterns.add<AtomicRMWOpConverter, MemRefReshapeOpConverter>( 151 patterns.getContext()); 152 } 153 154 std::unique_ptr<Pass> mlir::memref::createExpandOpsPass() { 155 return std::make_unique<ExpandOpsPass>(); 156 } 157