]> Cypherpunks repositories - gostls13.git/commit
math/big: optimize subVV function for loong64
authorHuang Qiqi <huangqiqi@loongson.cn>
Tue, 11 Jun 2024 11:06:29 +0000 (19:06 +0800)
committerabner chenc <chenguoqi@loongson.cn>
Tue, 15 Apr 2025 11:56:52 +0000 (04:56 -0700)
commit396a48bea6f2dc07ef103da298eec05cb6167892
treedb7044be32332ee8a47c7298181e4b45e31579e5
parent72fa8adbdc773343aca62d45c482074c860918b5
math/big: optimize subVV function for loong64

Benchmark results on Loongson 3C5000 (which is an LA464 implementation):

goos: linux
goarch: loong64
pkg: math/big
cpu: Loongson-3C5000 @ 2200.00MHz
             │ test/old_3c5000_subvv.log │      test/new_3c5000_subvv.log      │
             │          sec/op           │   sec/op     vs base                │
SubVV/1                     10.920n ± 0%   7.657n ± 0%  -29.88% (p=0.000 n=20)
SubVV/2                     14.100n ± 0%   8.841n ± 0%  -37.30% (p=0.000 n=20)
SubVV/3                      16.38n ± 0%   11.06n ± 0%  -32.48% (p=0.000 n=20)
SubVV/4                      18.65n ± 0%   12.85n ± 0%  -31.10% (p=0.000 n=20)
SubVV/5                      20.93n ± 0%   14.79n ± 0%  -29.34% (p=0.000 n=20)
SubVV/10                     32.30n ± 0%   22.29n ± 0%  -30.99% (p=0.000 n=20)
SubVV/100                    244.3n ± 0%   149.2n ± 0%  -38.93% (p=0.000 n=20)
SubVV/1000                   2.292µ ± 0%   1.378µ ± 0%  -39.88% (p=0.000 n=20)
SubVV/10000                  26.26µ ± 0%   25.64µ ± 0%   -2.33% (p=0.000 n=20)
SubVV/100000                 341.3µ ± 0%   238.0µ ± 0%  -30.26% (p=0.000 n=20)
geomean                      209.1n        144.5n       -30.86%

Change-Id: I3863c2c6728f1b0f8fecbf77de13254299c5b1cb
Reviewed-on: https://go-review.googlesource.com/c/go/+/659877
Reviewed-by: abner chenc <chenguoqi@loongson.cn>
Reviewed-by: Carlos Amedee <carlos@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
src/math/big/arith_loong64.s