nir: add vec7 and vec15, vectorize loads with a hole between them, vectorize load_smem_amd
This will enable aggressive vectorization of scalar loads on AMD hw.
The commits adding vec7 and vec15 were removed.
Edited by Marek Olšák
Due to an influx of spam, we have had to impose restrictions on new accounts. Please see this wiki page for instructions on how to get full permissions. Sorry for the inconvenience.