[Bug 218203] Implement AVX2 accelerated Fletcher algorithms

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=218203

--- Comment #4 from Adam Stylinski <kungfujesus06 at gmail.com> ---
Created attachment 181326
  --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=181326&action=edit
SSE4 implementation

Also not as impactful, but definitely measurable is an SSE4 variant.  This will
actually run on Atoms and a much larger number of other pre-Haswell CPUs.  

The ZFS On Linux version does 8 padds per loop iteration because they are
avoiding the pmovzxdq instruction.  This is probably to be compatible with
SSSE3 and not require SSE4.

-- 
You are receiving this mail because:
You are the assignee for the bug.