Fix a bug to increment stride tile correctly (#503)

* Fix a bug to increment stride tile correctly * Update regular_tile_access_iterator_tensor_op.h Co-authored-by: peisun1115 <[email protected]> Co-authored-by: Haicheng Wu <[email protected]>
NVIDIA · Jun 4, 2022 · a29dfb1 · a29dfb1
1 parent 0abaac8
commit a29dfb1
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/include/cutlass/transform/threadblock/regular_tile_access_iterator_tensor_op.h b/include/cutlass/transform/threadblock/regular_tile_access_iterator_tensor_op.h
@@ -614,7 +614,7 @@ class RegularTileAccessIterator<Shape_, Element_,
     add_pointer_offset(coord.contiguous() * sections_per_stage_ * stride_ *
                            ThreadMap::kElementsPerAccess / sections_ +
                        coord.strided() * Shape::kStrided * stride_ *
-                           Layout::kElementsPerAccess);
+                           Layout::kElementsPerAccess / Layout::kFactor);
   }
 };