- Instead of permuting the scales individually, multiply unpermuted scales, then permute the result (doesn't seem to improve performance).
- Get rid of masked loads so that the compiler can use
vperm*
directly on memory operands (promising). - Somehow use one
vpdpbusds
instead of two (doesn't seem to be possible). - Somehow use the accumulator in
vpdpbusds
instead of a separate subtraction at the very end (also doesn't seem to be possible).
-
Notifications
You must be signed in to change notification settings - Fork 0
dfyz/llama-avx-512
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
About
Expermenting with quantized AVX-512 dot product for llama.cpp
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published