CUDA: refactor and optimize IQ MMVQ (#8215)

Browse source

* CUDA: refactor and optimize IQ MMVQ

* uint -> uint32_t

* __dp4a -> ggml_cuda_dp4a

* remove MIN_CC_DP4A checks

* change default

* try CI fix

...

This commit is contained in:

Johannes Gäßler 2024-07-01 20:39:06 +02:00 • committed by GitHub

parent dae57a1ebc

commit cb5fad4c6c

No known key found for this signature in database

GPG key ID: B5690EEEBB952194

8 changed files with 406 additions and 487 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Download patch file Download diff file Expand all files Collapse all files

682

ggml/src/ggml-cuda/vecdotq.cuh

View file

File diff suppressed because it is too large Load diff