Commit Graph

1 Commits

Author SHA1 Message Date
Madan mohan Manokar
3dceb69212 core: vectorize JacobiSVD with double-accumulating SIMD and AVX512 dispatch
Add VBLAS::dotD and VBLAS::givensD to vectorize JacobiSVDImpl_'s dot-product,
Givens rotation and norm loops with double-precision accumulation matching the
scalar path.

Dispatch lapack for both AVX512_SKX and AVX512_ICL.
2026-08-14 15:56:30 +00:00