tree 295974d2a56176a4105341928d8fee06f72aa7af
parent f8f265b9cfd57970b2bc8b3dd8531cedaf57ccc1
author erifan01 <eric.fang@arm.com> 1526451907 +0000
committer Cherry Zhang <cherryyz@google.com> 1555944316 +0000

math/big: optimize mulAddVWW on arm64 for better performance

Unroll the cycle 4 times to reduce load overhead.

Benchmarks:
name                old time/op    new time/op    delta
MulAddVWW/1-8         15.9ns ± 0%    11.9ns ± 0%  -24.92%  (p=0.000 n=8+8)
MulAddVWW/2-8         16.1ns ± 0%    13.9ns ± 1%  -13.82%  (p=0.000 n=8+8)
MulAddVWW/3-8         18.9ns ± 0%    17.3ns ± 0%   -8.47%  (p=0.000 n=8+8)
MulAddVWW/4-8         21.7ns ± 0%    19.5ns ± 0%  -10.14%  (p=0.000 n=8+8)
MulAddVWW/5-8         25.1ns ± 0%    22.5ns ± 0%  -10.27%  (p=0.000 n=8+8)
MulAddVWW/10-8        41.6ns ± 0%    40.0ns ± 0%   -3.79%  (p=0.000 n=8+8)
MulAddVWW/100-8        368ns ± 0%     363ns ± 0%   -1.36%  (p=0.000 n=8+8)
MulAddVWW/1000-8      3.52µs ± 0%    3.52µs ± 0%   -0.14%  (p=0.000 n=8+8)
MulAddVWW/10000-8     35.1µs ± 0%    35.1µs ± 0%   -0.01%  (p=0.000 n=7+6)
MulAddVWW/100000-8     351µs ± 0%     351µs ± 0%   +0.15%  (p=0.038 n=8+8)

Change-Id: I052a4db286ac6e4f3293289c7e9a82027da0405e
Reviewed-on: https://go-review.googlesource.com/c/go/+/155780
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>