[Mlir-commits] [mlir] [mlir][emitc] Add op modelling C expressions (PR #71631)
Simon Camphausen
llvmlistbot at llvm.org
Thu Nov 9 06:21:44 PST 2023
================
@@ -0,0 +1,117 @@
+//===- Transforms.cpp - Patterns and transforms for the EmitC dialect------===//
+//
+// 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/EmitC/Transforms/Transforms.h"
+#include "mlir/Dialect/EmitC/IR/EmitC.h"
+#include "mlir/IR/IRMapping.h"
+#include "mlir/IR/PatternMatch.h"
+#include "llvm/Support/Debug.h"
+
+namespace mlir {
+namespace emitc {
+
+ExpressionOp createExpression(Operation *op, OpBuilder &builder) {
+ assert(ExpressionOp::isCExpression(*op) && "Expected a C expression");
+
+ // Create an expression yielding the value returned by op.
+ assert(op->getNumResults() == 1 && "Expected exactly one result");
+ Value result = op->getResult(0);
+ Type resultType = result.getType();
+ Location loc = op->getLoc();
+
+ builder.setInsertionPointAfter(op);
+ auto expressionOp = builder.create<emitc::ExpressionOp>(loc, resultType);
+
+ // Replace all op's uses with the new expression's result.
+ result.replaceAllUsesWith(expressionOp.getResult());
+
+ // Create an op to yield op's value.
+ Region ®ion = expressionOp.getRegion();
+ Block &block = region.emplaceBlock();
+ builder.setInsertionPointToEnd(&block);
+ auto yieldOp = builder.create<emitc::YieldOp>(loc, result);
+
+ // Move op into the new expression.
+ op->moveBefore(yieldOp);
+
+ return expressionOp;
+}
+
+} // namespace emitc
+} // namespace mlir
+
+using namespace mlir;
+using namespace mlir::emitc;
+
+namespace {
+
+struct FoldExpressionOp : public OpRewritePattern<ExpressionOp> {
+ using OpRewritePattern<ExpressionOp>::OpRewritePattern;
+ LogicalResult matchAndRewrite(ExpressionOp expressionOp,
+ PatternRewriter &rewriter) const override {
+ bool anythingFolded = false;
+ for (Operation &op : llvm::make_early_inc_range(
+ expressionOp.getBody()->without_terminator())) {
+ // Don't fold expressions whose result value has its address taken.
+ auto applyOp = dyn_cast<emitc::ApplyOp>(op);
+ if (applyOp && applyOp.getApplicableOperator() == "&")
+ continue;
+
+ for (Value operand : op.getOperands()) {
+ auto usedExpression =
+ dyn_cast_or_null<ExpressionOp>(operand.getDefiningOp());
+
+ if (!usedExpression)
+ continue;
+
+ // Don't fold expressions with multiple users: assume any
+ // re-materialization was done separately.
+ if (!usedExpression.getResult().hasOneUse())
+ continue;
+
+ // Don't fold expressions with side effects.
+ if (usedExpression.hasSideEffects())
+ continue;
+
+ // Fold the used expression into this expression by cloning all
+ // instructions in the used expression just before the operation using
+ // its value.
+ rewriter.setInsertionPoint(&op);
+ IRMapping mapper;
+ for (Operation &opToClone :
+ usedExpression.getBody()->without_terminator()) {
+ Operation *clone = rewriter.clone(opToClone, mapper);
+ mapper.map(&opToClone, clone);
+ }
+
+ auto usedYield =
+ cast<YieldOp>(usedExpression.getBody()->getTerminator());
+ Operation *expressionRoot = usedYield.getResult().getDefiningOp();
+ assert(expressionRoot && "Used expression has no root operation");
----------------
simon-camp wrote:
```suggestion
Operation *expressionRoot = usedExpression.getRootOp();
```
https://github.com/llvm/llvm-project/pull/71631
More information about the Mlir-commits
mailing list