Files
crypto/argon2/blamka_loong64.go
T
Xiaolin Zhao 388684e50b argon2: add loong64 SIMD implementation
The performance gains on Loongson 3A6000 and 3A5000 are as follows:

goos: linux
goarch: loong64
pkg: golang.org/x/crypto/argon2
cpu: Loongson-3A6000-HV @ 2500.00MHz
                                             |  bench.old   |              bench.new              |
                                             |    sec/op    |   sec/op     vs base                |
Argon2i/_Time:_3_Memory:_32_MB,_Threads:_1     131.23m ± 0%   67.56m ± 1%  -48.52% (p=0.000 n=10)
Argon2i/_Time:_4_Memory:_32_MB,_Threads:_1     171.28m ± 2%   90.20m ± 0%  -47.34% (p=0.000 n=10)
Argon2i/_Time:_5_Memory:_32_MB,_Threads:_1      213.3m ± 0%   112.6m ± 0%  -47.21% (p=0.000 n=10)
Argon2i/_Time:_3_Memory:_64_MB,_Threads:_4      269.5m ± 0%   147.2m ± 0%  -45.37% (p=0.000 n=10)
Argon2i/_Time:_4_Memory:_64_MB,_Threads:_4      357.7m ± 0%   195.4m ± 0%  -45.36% (p=0.000 n=10)
Argon2i/_Time:_5_Memory:_64_MB,_Threads:_4      449.8m ± 0%   243.8m ± 0%  -45.79% (p=0.000 n=10)
Argon2d/_Time:_3,_Memory:_32_MB,_Threads:_1    126.56m ± 0%   67.43m ± 0%  -46.72% (p=0.000 n=10)
Argon2d/_Time:_4,_Memory:_32_MB,_Threads:_1    168.57m ± 0%   90.04m ± 0%  -46.58% (p=0.000 n=10)
Argon2d/_Time:_5,_Memory:_32_MB,_Threads:_1     210.5m ± 0%   112.7m ± 0%  -46.45% (p=0.000 n=10)
Argon2d/_Time:_3,_Memory:_64_MB,_Threads:_4     264.8m ± 0%   145.0m ± 1%  -45.23% (p=0.000 n=10)
Argon2d/_Time:_4,_Memory:_64_MB,_Threads:_4     353.8m ± 0%   193.7m ± 0%  -45.26% (p=0.000 n=10)
Argon2d/_Time:_5,_Memory:_64_MB,_Threads:_4     444.4m ± 0%   242.3m ± 0%  -45.49% (p=0.000 n=10)
Argon2id/_Time:_3,_Memory:_32_MB,_Threads:_1   126.89m ± 0%   66.62m ± 0%  -47.50% (p=0.000 n=10)
Argon2id/_Time:_4,_Memory:_32_MB,_Threads:_1   169.02m ± 0%   89.07m ± 0%  -47.30% (p=0.000 n=10)
Argon2id/_Time:_5,_Memory:_32_MB,_Threads:_1    210.7m ± 0%   111.0m ± 0%  -47.34% (p=0.000 n=10)
Argon2id/_Time:_3,_Memory:_64_MB,_Threads:_4    267.6m ± 1%   145.8m ± 0%  -45.51% (p=0.000 n=10)
Argon2id/_Time:_4,_Memory:_64_MB,_Threads:_4    355.1m ± 0%   194.1m ± 0%  -45.34% (p=0.000 n=10)
Argon2id/_Time:_5,_Memory:_64_MB,_Threads:_4    443.6m ± 0%   242.5m ± 0%  -45.33% (p=0.000 n=10)
geomean                                         240.8m        129.3m       -46.32%

goos: linux
goarch: loong64
pkg: golang.org/x/crypto/argon2
cpu: Loongson-3A5000 @ 2500.00MHz
                                             |  bench.old  |              bench.new              |
                                             |   sec/op    |   sec/op     vs base                |
Argon2i/_Time:_3_Memory:_32_MB,_Threads:_1     209.9m ± 1%   109.7m ± 2%  -47.75% (p=0.000 n=10)
Argon2i/_Time:_4_Memory:_32_MB,_Threads:_1     278.1m ± 0%   143.7m ± 0%  -48.34% (p=0.000 n=10)
Argon2i/_Time:_5_Memory:_32_MB,_Threads:_1     346.7m ± 0%   178.1m ± 0%  -48.63% (p=0.000 n=10)
Argon2i/_Time:_3_Memory:_64_MB,_Threads:_4     455.3m ± 0%   240.8m ± 0%  -47.12% (p=0.000 n=10)
Argon2i/_Time:_4_Memory:_64_MB,_Threads:_4     604.6m ± 0%   317.7m ± 0%  -47.45% (p=0.000 n=10)
Argon2i/_Time:_5_Memory:_64_MB,_Threads:_4     754.8m ± 0%   395.4m ± 0%  -47.61% (p=0.000 n=10)
Argon2d/_Time:_3,_Memory:_32_MB,_Threads:_1    206.9m ± 1%   107.6m ± 0%  -48.00% (p=0.000 n=10)
Argon2d/_Time:_4,_Memory:_32_MB,_Threads:_1    274.3m ± 0%   141.8m ± 1%  -48.32% (p=0.000 n=10)
Argon2d/_Time:_5,_Memory:_32_MB,_Threads:_1    342.4m ± 0%   175.6m ± 0%  -48.71% (p=0.000 n=10)
Argon2d/_Time:_3,_Memory:_64_MB,_Threads:_4    450.2m ± 0%   237.9m ± 0%  -47.15% (p=0.000 n=10)
Argon2d/_Time:_4,_Memory:_64_MB,_Threads:_4    597.7m ± 0%   314.0m ± 0%  -47.46% (p=0.000 n=10)
Argon2d/_Time:_5,_Memory:_64_MB,_Threads:_4    745.8m ± 0%   390.7m ± 1%  -47.61% (p=0.000 n=10)
Argon2id/_Time:_3,_Memory:_32_MB,_Threads:_1   207.6m ± 0%   107.9m ± 0%  -48.05% (p=0.000 n=10)
Argon2id/_Time:_4,_Memory:_32_MB,_Threads:_1   275.0m ± 0%   142.0m ± 0%  -48.34% (p=0.000 n=10)
Argon2id/_Time:_5,_Memory:_32_MB,_Threads:_1   342.9m ± 1%   176.0m ± 0%  -48.66% (p=0.000 n=10)
Argon2id/_Time:_3,_Memory:_64_MB,_Threads:_4   450.6m ± 1%   238.5m ± 0%  -47.07% (p=0.000 n=10)
Argon2id/_Time:_4,_Memory:_64_MB,_Threads:_4   598.5m ± 1%   314.6m ± 0%  -47.44% (p=0.000 n=10)
Argon2id/_Time:_5,_Memory:_64_MB,_Threads:_4   746.4m ± 0%   391.0m ± 0%  -47.61% (p=0.000 n=10)
geomean                                        398.6m        207.9m       -47.86%

Change-Id: Iaa9d134d68dd2f0972fc5768d7e66f7b1ff0ebd3
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/657795
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: abner chenc <chenguoqi@loongson.cn>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Reviewed-by: Carlos Amedee <carlos@golang.org>
2025-04-13 20:23:35 -07:00

60 lines
1.3 KiB
Go

// Copyright 2025 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//go:build loong64 && gc && !purego
package argon2
import "golang.org/x/sys/cpu"
//go:noescape
func mixBlocks1VX(out, in1, in2 *block)
//go:noescape
func mixBlocks2VX(out, in1, in2, t *block)
//go:noescape
func xorBlocksVX(out, in1, in2, t *block)
//go:noescape
func blamkaVX(b *block)
func processBlockVX(out, in1, in2 *block, xor bool) {
var t block
mixBlocks1VX(&t, in1, in2)
if cpu.Loong64.HasLSX {
blamkaVX(&t)
} else {
for i := 0; i < blockLength; i += 16 {
blamkaGeneric(
&t[i+0], &t[i+1], &t[i+2], &t[i+3],
&t[i+4], &t[i+5], &t[i+6], &t[i+7],
&t[i+8], &t[i+9], &t[i+10], &t[i+11],
&t[i+12], &t[i+13], &t[i+14], &t[i+15],
)
}
for i := 0; i < blockLength/8; i += 2 {
blamkaGeneric(
&t[i], &t[i+1], &t[16+i], &t[16+i+1],
&t[32+i], &t[32+i+1], &t[48+i], &t[48+i+1],
&t[64+i], &t[64+i+1], &t[80+i], &t[80+i+1],
&t[96+i], &t[96+i+1], &t[112+i], &t[112+i+1],
)
}
}
if xor {
xorBlocksVX(out, in1, in2, &t)
} else {
mixBlocks2VX(out, in1, in2, &t)
}
}
func processBlock(out, in1, in2 *block) {
processBlockVX(out, in1, in2, false)
}
func processBlockXOR(out, in1, in2 *block) {
processBlockVX(out, in1, in2, true)
}