IQ2_XXS: much faster CPU prompt processing#515

Merged

ikawrakow merged 6 commits intomainfrom

ik/iq2_xxs_gemm

Jun 11, 2025

Commits on Jun 10, 2025

Much faster iq2_xxs GEMM
Iwan Kawrakow
committed
iq2_xxs: q8_2_x4 GEMM
Iwan Kawrakow
committed
iq2_xxs: use template for q8_2_x4 GEMM
Iwan Kawrakow
committed
Fix AVX2
Iwan Kawrakow
committed

Commits on Jun 11, 2025

Cleanup
Iwan Kawrakow
committed
NEON is not working yet, so still use Q8_K GEMM
Iwan Kawrakow
committed