mirror of
https://github.com/amd/blis.git
synced 2026-07-18 17:47:28 +00:00
825244052d
Removed function calls to 6x48, 6x32, 6x16, 6x8, 6x4 fringe kernels. These are now handled in the main kernel directly Introduced new fringe kernels 6x3, 6x2 in the n-direction Removed fringe kernel calls to AVX2 code path Access pattern inside the kernel changed so that the C-matrix stores are along the cache line Various instruction level optimizations These kerels are enabled also on the Zen6 code path Tiny path modified to take these newer kernels on zen5 and zen6 Added Gtests for these new kernels AMD-Internal: CPUPL-8424, SWLCSG-3272 Co-authored-by: Rohan Rayan rohrayan@amd.com