================
@@ -57,7 +61,7 @@ def tma_load(
     b_tma.load(b2, mbar_group[0], coords=[64, 0], predicate=p)
 
 
- at NVDSL.mlir_func
+ at NVDSL.mlir_func(dump_only)
----------------
castigli wrote:
done.
https://github.com/llvm/llvm-project/pull/156830