Xiaolin Zhao
|
b369b723c8
|
crypto/internal/poly1305: implement function update in assembly on loong64
The performance improvements on Loongson-3A5000 and Loongson-3A6000 are as follows:
goos: linux
goarch: loong64
pkg: golang.org/x/crypto/internal/poly1305
cpu: Loongson-3A5000 @ 2500.00MHz
| bench.old | bench.new |
| sec/op | sec/op vs base |
64 122.8n ± 0% 100.0n ± 0% -18.57% (p=0.000 n=10)
1K 1152.0n ± 0% 732.2n ± 0% -36.44% (p=0.000 n=10)
2M 2.356m ± 0% 1.443m ± 0% -38.74% (p=0.000 n=10)
64Unaligned 122.7n ± 0% 101.5n ± 0% -17.24% (p=0.000 n=10)
1KUnaligned 1152.0n ± 0% 745.4n ± 0% -35.30% (p=0.000 n=10)
2MUnaligned 2.336m ± 0% 1.473m ± 0% -36.94% (p=0.000 n=10)
Write64 77.92n ± 0% 54.88n ± 0% -29.57% (p=0.000 n=10)
Write1K 1106.0n ± 0% 683.3n ± 0% -38.22% (p=0.000 n=10)
Write2M 2.356m ± 0% 1.444m ± 0% -38.72% (p=0.000 n=10)
Write64Unaligned 77.87n ± 0% 55.69n ± 0% -28.49% (p=0.000 n=10)
Write1KUnaligned 1106.0n ± 0% 708.1n ± 0% -35.97% (p=0.000 n=10)
Write2MUnaligned 2.335m ± 0% 1.471m ± 0% -37.01% (p=0.000 n=10)
geomean 6.373µ 4.272µ -32.96%
| bench.old | bench.new |
| B/s | B/s vs base |
64 497.1Mi ± 0% 610.3Mi ± 0% +22.78% (p=0.000 n=10)
1K 847.6Mi ± 0% 1333.7Mi ± 0% +57.35% (p=0.000 n=10)
2M 849.0Mi ± 0% 1385.9Mi ± 0% +63.24% (p=0.000 n=10)
64Unaligned 497.4Mi ± 0% 600.9Mi ± 0% +20.81% (p=0.000 n=10)
1KUnaligned 847.6Mi ± 0% 1310.1Mi ± 0% +54.57% (p=0.000 n=10)
2MUnaligned 856.3Mi ± 0% 1357.9Mi ± 0% +58.58% (p=0.000 n=10)
Write64 783.3Mi ± 0% 1112.2Mi ± 0% +41.99% (p=0.000 n=10)
Write1K 882.8Mi ± 0% 1429.1Mi ± 0% +61.88% (p=0.000 n=10)
Write2M 849.0Mi ± 0% 1385.4Mi ± 0% +63.18% (p=0.000 n=10)
Write64Unaligned 783.8Mi ± 0% 1096.1Mi ± 0% +39.85% (p=0.000 n=10)
Write1KUnaligned 882.8Mi ± 0% 1379.0Mi ± 0% +56.20% (p=0.000 n=10)
Write2MUnaligned 856.5Mi ± 0% 1359.9Mi ± 0% +58.76% (p=0.000 n=10)
geomean 772.2Mi 1.125Gi +49.18%
goos: linux
goarch: loong64
pkg: golang.org/x/crypto/internal/poly1305
cpu: Loongson-3A6000-HV @ 2500.00MHz
| bench.old | bench.new |
| sec/op | sec/op vs base |
64 92.06n ± 0% 71.55n ± 0% -22.28% (p=0.000 n=10)
1K 998.4n ± 0% 607.7n ± 0% -39.13% (p=0.000 n=10)
2M 1.976m ± 0% 1.165m ± 0% -41.07% (p=0.000 n=10)
64Unaligned 92.05n ± 0% 71.55n ± 0% -22.27% (p=0.000 n=10)
1KUnaligned 998.3n ± 0% 607.6n ± 0% -39.13% (p=0.000 n=10)
2MUnaligned 1.975m ± 0% 1.222m ± 0% -38.11% (p=0.000 n=10)
Write64 65.24n ± 0% 45.23n ± 0% -30.67% (p=0.000 n=10)
Write1K 970.8n ± 0% 577.6n ± 0% -40.51% (p=0.000 n=10)
Write2M 1.965m ± 0% 1.163m ± 0% -40.81% (p=0.000 n=10)
Write64Unaligned 65.24n ± 0% 45.24n ± 0% -30.66% (p=0.000 n=10)
Write1KUnaligned 970.8n ± 0% 577.6n ± 0% -40.50% (p=0.000 n=10)
Write2MUnaligned 1.965m ± 0% 1.222m ± 0% -37.81% (p=0.000 n=10)
geomean 5.317µ 3.426µ -35.58%
| bench.old | bench.new |
| B/s | B/s vs base |
64 663.0Mi ± 0% 853.1Mi ± 0% +28.67% (p=0.000 n=10)
1K 978.1Mi ± 0% 1606.9Mi ± 0% +64.28% (p=0.000 n=10)
2M 1012.0Mi ± 0% 1717.4Mi ± 0% +69.70% (p=0.000 n=10)
64Unaligned 663.1Mi ± 0% 853.1Mi ± 0% +28.65% (p=0.000 n=10)
1KUnaligned 978.2Mi ± 0% 1607.1Mi ± 0% +64.29% (p=0.000 n=10)
2MUnaligned 1012.6Mi ± 0% 1636.2Mi ± 0% +61.58% (p=0.000 n=10)
Write64 935.5Mi ± 0% 1349.3Mi ± 0% +44.23% (p=0.000 n=10)
Write1K 1005.9Mi ± 0% 1690.9Mi ± 0% +68.09% (p=0.000 n=10)
Write2M 1017.7Mi ± 0% 1719.5Mi ± 0% +68.95% (p=0.000 n=10)
Write64Unaligned 935.6Mi ± 0% 1349.3Mi ± 0% +44.22% (p=0.000 n=10)
Write1KUnaligned 1006.0Mi ± 0% 1690.9Mi ± 0% +68.08% (p=0.000 n=10)
Write2MUnaligned 1017.7Mi ± 0% 1636.4Mi ± 0% +60.80% (p=0.000 n=10)
geomean 925.6Mi 1.403Gi +55.22%
Change-Id: If05a8bfc868b3e6f903ff169eed7a894af741f9b
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/638455
Reviewed-by: David Chase <drchase@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Reviewed-by: abner chenc <chenguoqi@loongson.cn>
|
2025-03-11 17:59:26 -07:00 |
|
Paul E. Murphy
|
6c21748958
|
internal/poly1305: extend ppc64le support to ppc64
The cipher needs to load the stream in LE order. Use the byte
reversing loads on BE.
Also, remove the unused variable poly1305Mask in the PPC64
asm file too.
Change-Id: Ie90fe7bb0ea7a3bcb76583e0cf9c1e4133499541
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/614298
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Archana Ravindar <aravinda@redhat.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: David Chase <drchase@google.com>
|
2024-10-04 20:59:56 +00:00 |
|
Dmitri Shuralyov
|
8779cbd1c9
|
all: update go directive to 1.18
Done with:
go get go@1.18
go mod tidy
go fix ./...
Using go1.21.3.
Also update avo to v0.5.0 in the curve25519/internal/field/_asm module.
It's newer and produces no diff in the generated code.
For golang/go#60268.
Change-Id: I9bd771ee8561595d7f68aaca76df6e3e33d35013
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/534141
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Damien Neil <dneil@google.com>
Auto-Submit: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
|
2023-10-11 22:02:22 +00:00 |
|
Filippo Valsorda
|
c084706c22
|
poly1305: deprecate public package
Fixes golang/go#36646
Change-Id: Ic19dd2171c84472fc9d3f44803224b87fc5c0417
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/345649
Trust: Filippo Valsorda <filippo@golang.org>
Trust: Katie Hockman <katie@golang.org>
Run-TryBot: Filippo Valsorda <filippo@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Roland Shoemaker <roland@golang.org>
Reviewed-by: Katie Hockman <katie@golang.org>
|
2021-09-15 21:47:49 +00:00 |
|