[Mlir-commits] [mlir] [MLIR][Bufferization] BufferResultsToOutParams: Add an option to eliminate AllocOp and avoid Copy (PR #90011)

[llvmlistbot at llvm.org]

================
@@ -118,10 +119,14 @@ static LogicalResult updateReturnOps(func::FuncOp func,
         keepAsReturnOperands.push_back(operand);
     }
     OpBuilder builder(op);
-    for (auto t : llvm::zip(copyIntoOutParams, appendedEntryArgs)) {
-      if (failed(
-              memCpyFn(builder, op.getLoc(), std::get<0>(t), std::get<1>(t))))
-        return WalkResult::interrupt();
+    for (auto [orig, arg] : llvm::zip(copyIntoOutParams, appendedEntryArgs)) {
+      if (avoidBufferResultAllocAndCopy && isa<memref::AllocOp>(orig.getDefiningOp())) {
----------------
Menooker wrote:

Thanks for the comments. This added option in the pass was supposed to work on statically shaped allocated memref. I will update the code to filter out dynamic-shaped cases.

https://github.com/llvm/llvm-project/pull/90011