Iwan Kawrakow 595d2ae32d iq6_k: slightly better Zen4 iqk_mul_mat ...

We now arrive at pp-512 = 147 t/s for LLaMA-3.1-8B.
TG-128 is 9.5 t/s. This is better than last commit,
but still kind of slow compared to Q6_K.

My last commit message is wrong: also iq3_k needs a fix
for overflow.

2024-08-09 16:00:31 +02:00

..

cmake

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

include

iq6_k: WIP (nothing works)

2024-08-09 16:00:31 +02:00

src

iq6_k: slightly better Zen4 iqk_mul_mat

2024-08-09 16:00:31 +02:00

.gitignore

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

CMakeLists.txt

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00