[mlir][async] Erase op later to preserve insertion point

[matthias-springer]

Delay the erasure of an op, so that the insertion point of the rewriter remains valid.

This commit is in preparation of the One-Shot Dialect Conversion refactoring. (The current implementation works with the current dialect conversion driver because op erasure is delayed.)

[llvmbot]

@llvm/pr-subscribers-mlir-async

@llvm/pr-subscribers-mlir

Author: Matthias Springer (matthias-springer)

Changes

Delay the erasure of an op, so that the insertion point of the rewriter remains valid.

This commit is in preparation of the One-Shot Dialect Conversion refactoring. (The current implementation works with the current dialect conversion driver because op erasure is delayed.)

---

Full diff: https://github.com/llvm/llvm-project/pull/146516.diff

1 Files Affected:

• (modified) mlir/lib/Dialect/Async/Transforms/AsyncToAsyncRuntime.cpp (+1-1)

diff --git a/mlir/lib/Dialect/Async/Transforms/AsyncToAsyncRuntime.cpp b/mlir/lib/Dialect/Async/Transforms/AsyncToAsyncRuntime.cpp index c289faaf9983c..8601bb5aaada9 100644 --- a/mlir/lib/Dialect/Async/Transforms/AsyncToAsyncRuntime.cpp +++ b/mlir/lib/Dialect/Async/Transforms/AsyncToAsyncRuntime.cpp @@ -723,8 +723,8 @@ class YieldOpLowering : public OpConversionPattern<async::YieldOp> { // Switch the coroutine completion token to available state. rewriter.create<RuntimeSetAvailableOp>(loc, *coro.asyncToken); - rewriter.eraseOp(op); rewriter.create<cf::BranchOp>(loc, coro.cleanup); + rewriter.eraseOp(op); return success(); }