ggml-cpu: add 2x2 register-blocked AVX2 kernel for ggml_vec_dot_q8_0_q8_0 by codspeed-hq[bot] · Pull Request #45 · AvalancheHQ/llama.cpp

Performance Gate Passed

✅ 28 untouched benchmarks

_{Comparing codspeed-optim-2-2-register-blocked-avx2-kernel-for-ggml-vec-dot-1785415701799 (93f4a9a) with master (46819c9)}