We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
8a5b838
optimize AVX some more, but still slower than SSE with MSVC
