[Mlir-commits] [mlir] [Affine] Parallelize loops with MemoryEffects::Free ops. (PR #172388)

[Oleksandr Alex Zinenko]

================
@@ -152,6 +171,7 @@ bool mlir::affine::isLoopMemoryParallel(AffineForOp forOp) {
         loadAndStoreOps.push_back(op);
     } else if (!isa<AffineForOp, AffineYieldOp, AffineIfOp>(op) &&
                !hasSingleEffect<MemoryEffects::Allocate>(op) &&
+               !isDeallocationOfLocallyDefined(op, forOp) &&
----------------
ftynse wrote:

For C-style `free`, most likely, but I'd rather keep it conservative given that this semantics doesn't say anything about the underlying implementation.

https://github.com/llvm/llvm-project/pull/172388