md/raid6 algorithms: xor_syndrome() for SSE2
authorMarkus Stockhausen <stockhausen@collogia.de>
Mon, 15 Dec 2014 01:57:05 +0000 (12:57 +1100)
committerNeilBrown <neilb@suse.de>
Tue, 21 Apr 2015 22:00:42 +0000 (08:00 +1000)
commita582564b24bec0443b5c5ff43ee6d1258f8bd658
tree44b3c4918a5242c3d420fbeb8c061834dca0c100
parent9a5ce91d053961b7cc8fa56bd083819a9fc92734
md/raid6 algorithms: xor_syndrome() for SSE2

The second and (last) optimized XOR syndrome calculation. This version
supports right and left side optimization. All CPUs with architecture
older than Haswell will benefit from it.

It should be noted that SSE2 movntdq kills performance for memory areas
that are read and written simultaneously in chunks smaller than cache
line size. So use movdqa instead for P/Q writes in sse21 and sse22 XOR
functions.

Signed-off-by: Markus Stockhausen <stockhausen@collogia.de>
Signed-off-by: NeilBrown <neilb@suse.de>
lib/raid6/sse2.c