Fix regression in assembleAndSum PQ decoder performance#379

Merged

jkni merged 2 commits intomainfrom

128-bit-assemble-and-sum

Dec 31, 2024

Commits on Dec 24, 2024

Add implementations of decoding PQ similarities in SimdOps for 128-bit max vector width
jkni
committed
Use scalar implementation for assembleAndSum if preferred vector width is 128-bits
jkni
committed