Binary Vector Search at 350GB/S Using ARM Neon
1–3 of 3 posts
Re: Binary Vector Search at 350GB/S Using ARM Neon
#2re: optimization for 1024b vectors — do you pad shorter ones, or fallback to a more general kernel?
Re: Binary Vector Search at 350GB/S Using ARM Neon
#3re: optimization for 1024b vectors — do you pad shorter ones, or fallback to a more general kernel?
We do a projection of the original vectors so that it matches one of our optimized kernel. This generally gives us better recall vs. simple padding since all bits are utilized.