Repository navigation

ggml-cpu: AVX2/AVX-VNNI kernels for PQ2_0 + Q8_K activations + tiled GEMM (x86 without AVX-512) - #206

Merged

bri-prism merged 3 commits into

PrismML-Eng:prismfrom

alenizy361:cpu-pq2-avx2-gemm

Sep 21, 2026

Merged

ggml-cpu: AVX2/AVX-VNNI kernels for PQ2_0 + Q8_K activations + tiled GEMM (x86 without AVX-512)#206
bri-prism merged 3 commits into
PrismML-Eng:prismfrom
alenizy361:cpu-pq2-avx2-gemm

Commits on Sep 19, 2026
- ggml-cpu: AVX2/AVX-VNNI kernels for PQ2_0 + Q8_K activations + tiled GEMM (x86 without AVX-512)
  
  alenizy361
  and
  claude
  committed
  164c337
Commits on Sep 20, 2026
- ggml-cpu, llama-quant: fix PQ2_0 non-x86 link break and 128-column shape contract
  
  alenizy361
  and
  claude
  committed
  10df298
Commits on Sep 21, 2026
- ggml-cpu: keep PQ2_0 rows that are not multiples of 256 on the Q8_0 activation path
  
  alenizy361
  and
  claude
  committed
  d2c9ddd