Is this a lookup overhead thing or a memcpy based overhead regression? In the case of the latter it seems like this may result in an unexpected mutation of the

Is this a lookup overhead thing or a memcpy based overhead regression? In the case of the latter it seems like this may result in an unexpected mutation of the source data?