[Mlir-commits] [mlir] [mlir][EmitC]Expand the MemRefToEmitC pass - Lowering `AllocOp` (PR #148257)
Gil Rapaport
llvmlistbot at llvm.org
Fri Jul 25 08:37:02 PDT 2025
================
@@ -77,6 +81,75 @@ struct ConvertAlloca final : public OpConversionPattern<memref::AllocaOp> {
}
};
+struct ConvertAlloc final : public OpConversionPattern<memref::AllocOp> {
+ using OpConversionPattern::OpConversionPattern;
+ LogicalResult
+ matchAndRewrite(memref::AllocOp allocOp, OpAdaptor operands,
+ ConversionPatternRewriter &rewriter) const override {
+ mlir::Location loc = allocOp.getLoc();
+ MemRefType memrefType = allocOp.getType();
+ if (!memrefType.hasStaticShape()) {
+ // TODO: Handle Dynamic shapes in the future. If the size
+ // of the allocation is the result of some function, we could
+ // potentially evaluate the function and use the result in the call to
+ // allocate.
+ return rewriter.notifyMatchFailure(
+ loc, "cannot transform alloc with dynamic shape");
+ }
+
+ mlir::Type sizeTType = mlir::emitc::SizeTType::get(rewriter.getContext());
+ Type elementType = memrefType.getElementType();
+ mlir::emitc::CallOpaqueOp sizeofElementOp =
+ rewriter.create<mlir::emitc::CallOpaqueOp>(
+ loc, sizeTType, rewriter.getStringAttr("sizeof"),
+ mlir::ValueRange{},
+ mlir::ArrayAttr::get(rewriter.getContext(),
+ {mlir::TypeAttr::get(elementType)}));
+ mlir::Value sizeofElement = sizeofElementOp.getResult(0);
+
+ long val = memrefType.getShape().front();
----------------
aniragil wrote:
This only works for 1D memrefs (please also give `val` a more meaningful name).
Relying on the shape for allocation is only correct for identity-stride memrefs. The `MemrefToLLVM` pass enforces this constraint when lowering `AllocOp`, so we should probably do the same.
https://github.com/llvm/llvm-project/pull/148257
More information about the Mlir-commits
mailing list