fixed sweep logic

abd9c245 · Jing Zhang · a5aa963c · abd9c245
Commit abd9c245 authored Dec 08, 2021 by Jing Zhang
Hide whitespace changes
Inline Side-by-side

Showing with 1 addition and 1 deletion

composable_kernel/include/tensor_operation/threadwise_tensor_slice_transfer.hpp ...ude/tensor_operation/threadwise_tensor_slice_transfer.hpp +1 -1

No files found.
--- a/composable_kernel/include/tensor_operation/threadwise_tensor_slice_transfer.hpp
+++ b/composable_kernel/include/tensor_operation/threadwise_tensor_slice_transfer.hpp
@@ -166,7 +166,7 @@ struct ThreadwiseTensorSliceTransfer_v1r3
                    index_t tmp = ordered_access_idx[I0];

                    static_for<1, i, 1>{}([&](auto j) {
-                        tmp = tmp * ordered_access_lengths[j - 1] + ordered_access_idx[j];
+                        tmp = tmp * ordered_access_lengths[j] + ordered_access_idx[j];
                    });

                    forward_sweep_(i) = tmp % 2 == 0;