You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
See reproducer in https://godbolt.org/z/M6a7zxxdK
Given the fact that stride >= width it should be safe to vectorize the load from l_out[(i + 1) * stride + j] and the store to l_out[i*stride + j] since they won't overlap.
Replacing stride with width works in this case.
The text was updated successfully, but these errors were encountered:
See reproducer in https://godbolt.org/z/M6a7zxxdK
Given the fact that
stride >= width
it should be safe to vectorize the load froml_out[(i + 1) * stride + j]
and the store tol_out[i*stride + j]
since they won't overlap.Replacing
stride
withwidth
works in this case.The text was updated successfully, but these errors were encountered: