[mlir][linalg] Fix memory leak in pack-dynamic-inner-tile.mlir

[matthias-springer]

This fixes the test when running with ASAN.

[llvmbot]

@llvm/pr-subscribers-mlir-linalg

@llvm/pr-subscribers-mlir

Author: Matthias Springer (matthias-springer)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/137589.diff

1 Files Affected:

• (modified) mlir/test/Integration/Dialect/Linalg/CPU/pack-dynamic-inner-tile.mlir (+6-2)

diff --git a/mlir/test/Integration/Dialect/Linalg/CPU/pack-dynamic-inner-tile.mlir b/mlir/test/Integration/Dialect/Linalg/CPU/pack-dynamic-inner-tile.mlir
index 7b410ccee9633..187d7b862604b 100644
--- a/mlir/test/Integration/Dialect/Linalg/CPU/pack-dynamic-inner-tile.mlir
+++ b/mlir/test/Integration/Dialect/Linalg/CPU/pack-dynamic-inner-tile.mlir
@@ -125,9 +125,13 @@ module @transforms attributes { transform.with_named_sequence } {
     %bufferize = transform.bufferization.one_shot_bufferize %module
       {bufferize_function_boundaries=true} : (!transform.any_op) -> !transform.any_op
 
-    // 5. Canonicalize
+    // 5. Deallocate buffers.
     %func_op_bufferized = transform.structured.match ops{["func.func"]} in %bufferize : (!transform.any_op) -> !transform.op<"func.func">
-    transform.apply_patterns to %func_op_bufferized {
+    %func_op_deallocated = transform.apply_registered_pass "buffer-deallocation-pipeline" to %func_op_bufferized
+      : (!transform.op<"func.func">) -> !transform.op<"func.func">
+
+    // 6. Canonicalize
+    transform.apply_patterns to %func_op_deallocated {
       transform.apply_patterns.canonicalization
     } : !transform.op<"func.func">
 

[banach-space]

Thanks for the fix, LGTM!

This fixes the test when running with ASAN

How do you enable ASAN? LLVM_USE_SANITIZER=memory?

[matthias-springer]

-DLLVM_USE_SANITIZER="Address"

Btw, the unpack test also has a leak, but the buffer deallocation pass inserts the dealloc inside of vector.mask, which is invalid.

[banach-space]

Btw, the unpack test also has a leak, but the buffer deallocation pass inserts the dealloc inside of vector.mask, which is invalid.

I can look into that if you want, sounds like a familiar issue.

[matthias-springer]

Feel free to take a took. But I think we're going to have to extend the buffer deallocation pass in some way (to make sure that we pick the correct insertion point for the dealloc op)