mirror of https://github.com/oxen-io/lokinet
the AVX2 codepaths now appear to be 32-bit clean.
old hard-coded inline asm is still included if requested. -rick nb: is a vector of eight floats not the same layout as a simple linear array of same? (Aside from the alignment requirements) netbsd-family build fixes, also - the AVX2 codepaths are _compiler-specific_, they use features _exclusive_ to gcc and clangpull/18/head
parent
7f809eb53b
commit
e3a94101b4
Loading…
Reference in New Issue