tree d4407d39343af61e369396a27ec9aa5443f7e48e
parent 754cb46fa0b799769e2646654d65ef7dc5b61977
author Michael Munday <mike.munday@ibm.com> 1524668148 -0400
committer Michael Munday <mike.munday@ibm.com> 1524783825 +0000

internal/chacha20: add s390x SIMD implementation

Based on the SIMD algorithm described in:
ChaCha, a variant of Salsa20 by Daniel J. Bernstein
https://cr.yp.to/chacha/chacha-20080128.pdf

Requires the vector facility (vx).

name            old speed     new speed      delta
ChaCha20/32     178MB/s ± 0%   174MB/s ± 0%    -2.10%  (p=0.000 n=9+10)
ChaCha20/63     341MB/s ± 0%   337MB/s ± 0%    -1.16%  (p=0.000 n=10+10)
ChaCha20/64     367MB/s ± 0%   335MB/s ± 0%    -8.73%  (p=0.000 n=10+10)
ChaCha20/256    404MB/s ± 0%  1448MB/s ± 0%  +258.61%  (p=0.000 n=9+10)
ChaCha20/1024   410MB/s ± 0%  1568MB/s ± 0%  +282.73%  (p=0.000 n=9+10)
ChaCha20/1350   393MB/s ± 0%  1389MB/s ± 0%  +253.58%  (p=0.000 n=10+10)
ChaCha20/65536  414MB/s ± 0%  1634MB/s ± 0%  +294.79%  (p=0.000 n=10+10)

Change-Id: I9a600fb5ae8ee3f3b81ae6b01cff139c1272d684
Reviewed-on: https://go-review.googlesource.com/35842
Run-TryBot: Michael Munday <mike.munday@ibm.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>