]> Cypherpunks repositories - gostls13.git/commit
crypto/subtle: implement xorBytes in hardware on loong64
authorXiaolin Zhao <zhaoxiaolin@loongson.cn>
Tue, 4 Jun 2024 05:03:06 +0000 (13:03 +0800)
committerGopher Robot <gobot@golang.org>
Wed, 11 Sep 2024 19:40:18 +0000 (19:40 +0000)
commit69827b5c8ddf93be65bfc8b17d331bb3ff7b704c
tree112bf2bcdca91a5fa4a5d3330b5e6ae8d6292131
parent1dfb33e8612d20f41cf4e034d9d0838abf75e04b
crypto/subtle: implement xorBytes in hardware on loong64

goos: linux
goarch: loong64
pkg: crypto/subtle
cpu: Loongson-3A6000 @ 2500.00MHz
                    │  bench.old   │              bench.new              │
                    │    sec/op    │   sec/op     vs base                │
XORBytes/8Bytes       11.250n ± 0%   6.403n ± 0%  -43.08% (p=0.000 n=20)
XORBytes/128Bytes      24.61n ± 0%   12.21n ± 0%  -50.39% (p=0.000 n=20)
XORBytes/2048Bytes     216.7n ± 0%   108.3n ± 0%  -50.02% (p=0.000 n=20)
XORBytes/32768Bytes    3.657µ ± 0%   1.683µ ± 0%  -53.98% (p=0.000 n=20)
geomean                121.7n        61.44n       -49.52%

                    │  bench.old   │               bench.new                │
                    │     B/s      │      B/s       vs base                 │
XORBytes/8Bytes       678.1Mi ± 0%   1191.5Mi ± 0%   +75.72% (p=0.000 n=20)
XORBytes/128Bytes     4.844Gi ± 0%    9.766Gi ± 0%  +101.63% (p=0.000 n=20)
XORBytes/2048Bytes    8.801Gi ± 0%   17.619Gi ± 0%  +100.18% (p=0.000 n=20)
XORBytes/32768Bytes   8.346Gi ± 0%   18.137Gi ± 0%  +117.32% (p=0.000 n=20)
geomean               3.918Gi         7.763Gi        +98.14%

goos: linux
goarch: loong64
pkg: crypto/subtle
cpu: Loongson-3A5000 @ 2500.00MHz
                    │  bench.old   │              bench.new              │
                    │    sec/op    │   sec/op     vs base                │
XORBytes/8Bytes       16.420n ± 0%   8.806n ± 0%  -46.37% (p=0.000 n=20)
XORBytes/128Bytes      35.84n ± 0%   16.42n ± 0%  -54.19% (p=0.000 n=20)
XORBytes/2048Bytes     332.0n ± 0%   160.5n ± 0%  -51.64% (p=0.000 n=20)
XORBytes/32768Bytes    4.944µ ± 0%   2.474µ ± 0%  -49.96% (p=0.000 n=20)
geomean                176.3n        87.05n       -50.62%

                    │  bench.old   │               bench.new                │
                    │     B/s      │      B/s       vs base                 │
XORBytes/8Bytes       464.7Mi ± 0%    866.4Mi ± 0%   +86.45% (p=0.000 n=20)
XORBytes/128Bytes     3.326Gi ± 0%    7.261Gi ± 0%  +118.31% (p=0.000 n=20)
XORBytes/2048Bytes    5.745Gi ± 0%   11.880Gi ± 0%  +106.80% (p=0.000 n=20)
XORBytes/32768Bytes   6.172Gi ± 0%   12.334Gi ± 0%   +99.83% (p=0.000 n=20)
geomean               2.705Gi         5.477Gi       +102.52%

Change-Id: Id404f9023a57025f78b6922659cfa8870881d646
Reviewed-on: https://go-review.googlesource.com/c/go/+/590175
Reviewed-by: abner chenc <chenguoqi@loongson.cn>
Reviewed-by: Roland Shoemaker <roland@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Tim King <taking@google.com>
Reviewed-by: Tim King <taking@google.com>
src/crypto/subtle/xor_generic.go
src/crypto/subtle/xor_loong64.go [new file with mode: 0644]
src/crypto/subtle/xor_loong64.s [new file with mode: 0644]