Small performance improvement for avx-512 on Skylake-SP - #191

Open

rdolbeau wants to merge 3 commits into

FFTW:masterfrom

rdolbeau:avx512skx

Commits on Jul 14, 2024

Replace Scatter/Gather in AVX512 DP by a sequence of instructions to assemble/disassemble the vector in 128 bits chunks.
rdolbeau
committed
e4f897f
In AVX-512 LDu/STu, handle pair of single as a double.
rdolbeau
committed
c6cd1cd
new AVX512 code requires AVX512DQ, add the compiler flags in configure.ac
rdolbeau
committed
fd08d11