[Mlir-commits] [mlir] [mlir] Don't hoist transfers from potentially zero trip loops (PR #112752)

[llvmlistbot at llvm.org]

================
@@ -41,7 +44,8 @@ namespace linalg {
 ///
 /// WARNING: This hoisting does not model parallelism and is generally incorrect
 /// when used on distributed loops with memref semantics!
-void hoistRedundantVectorTransfers(Operation *root);
+void hoistRedundantVectorTransfers(Operation *root,
+                                   bool verifyNonZeroTrip = false);
----------------
Max191 wrote:

Maybe there could still be some issues if the trip count is >0 for all threads, but not the same for all threads. I can't think of a specific example at the moment, but it is something that is not considered by this logic, so I left a comment noting that for now.

https://github.com/llvm/llvm-project/pull/112752