[MOD-16688] x86 distance kernels: multi-accumulator unrolling + SIMD for smaller dimensions#984

Open

GuyAv46 wants to merge 3 commits into

mainfrom

perf-x86-multi-accumulators-small-dims

Commits on Jul 2, 2026

Multi-accumulator x86 SIMD kernels + lower small-dim SIMD thresholds

GuyAv46
and
claude
committed
Resolve residual block handling at compile time via wider dispatch modulus

GuyAv46
and
claude
committed

Commits on Jul 5, 2026

Use 2 accumulators where 4 showed no benefit (FP64 + all AVX-512)

GuyAv46
and
claude
committed