Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
p0nce
on Nov 30, 2022
|
parent
|
context
|
favorite
| on:
AVX 512 will be the future
Well this makes the instruction useless unless all memory pointed to makes no page fault.
janwas
on Dec 1, 2022
[–]
Agreed. This is why it is even more important for the application to pad its arrays to the vector size. Then we might also skip the masking entirely.
p0nce
on Dec 1, 2022
|
parent
[–]
Yes. It also makes the masked instruction more portable to arm or non-x86, which is great. It also avoids the side channel?
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search: