Commit Graph
6 Commits
Author SHA1 Message Date
Lynn Boger 1a865804d5 x/crypto/internal/poly1305: improve sum_ppc64le.s
This contains a few minor improvements to sum_ppc64le.s
which result in up to 10% performance improvement for
some of the benchmarks in this directory.

- ADDZE followed by ADD can be combined into ADDE
- PCALIGN added to the loop
- Eliminate a few unnecessary register moves

goos: linux
goarch: ppc64le
pkg: golang.org/x/crypto/internal/poly1305
cpu: POWER10
                 │ poly.orig.out │              poly.out              │
                 │    sec/op     │   sec/op     vs base               │
64                   40.34n ± 0%   38.13n ± 0%   -5.47% (p=0.002 n=6)
1K                   482.2n ± 0%   444.6n ± 0%   -7.81% (p=0.002 n=6)
2M                   978.4µ ± 0%   879.3µ ± 0%  -10.12% (p=0.002 n=6)
64Unaligned          40.35n ± 0%   38.16n ± 0%   -5.42% (p=0.002 n=6)
1KUnaligned          482.0n ± 0%   444.2n ± 0%   -7.84% (p=0.002 n=6)
2MUnaligned          978.4µ ± 0%   879.4µ ± 0%  -10.12% (p=0.002 n=6)
Write64              32.69n ± 0%   30.71n ± 0%   -6.04% (p=0.002 n=6)
Write1K              472.4n ± 0%   436.5n ± 0%   -7.60% (p=0.002 n=6)
Write2M              978.3µ ± 0%   879.4µ ± 0%  -10.11% (p=0.002 n=6)
Write64Unaligned     32.67n ± 0%   30.71n ± 0%   -6.00% (p=0.002 n=6)
Write1KUnaligned     472.6n ± 0%   436.4n ± 0%   -7.66% (p=0.002 n=6)
Write2MUnaligned     978.5µ ± 0%   879.6µ ± 0%  -10.10% (p=0.002 n=6)
geomean              2.569µ        2.367µ        -7.87%

Change-Id: I63314e7252ef10fb2d157f623c4bc2e31a63ae32
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/558775
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Paul Murphy <murp@ibm.com>
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
2024-02-15 13:30:21 +00:00
Lynn Boger ed6320f186 internal/chacha20: fix minor naming issue
This corrects a compile error due to a mismatch between function
description and the assembler code that first appeared when running
TryBots on the change to vendor this code. This fixes that problem
and adds and minor change based on a comment in CL 195959.

Change-Id: I42cbd3f75e281913f12f74850deaedf14eb21989
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/205659
Reviewed-by: Filippo Valsorda <filippo@golang.org>
2019-11-06 20:26:28 +00:00
Lynn Boger af544f31c8 internal/chacha20: improve chacha20 performance on ppc64le
This improves the performance of the asm implementation for
chacha20 on ppc64le by updating to the vsx implementation
provided in cryptogams. The previous implementation was found to
not perform as well as possible on power9. This implementation
improves performance on both power8 and power9.

Power9 improvement with this change as compared to current:

name               old time/op    new time/op     delta
ChaCha20/32-64        361ns ± 0%      225ns ± 0%  -37.67%  (p=1.000 n=1+1)
ChaCha20/63-64        364ns ± 0%      229ns ± 0%  -37.09%  (p=1.000 n=1+1)
ChaCha20/64-64        364ns ± 0%      231ns ± 0%  -36.54%  (p=1.000 n=1+1)
ChaCha20/256-64       332ns ± 0%      199ns ± 0%  -40.06%  (p=1.000 n=1+1)
ChaCha20/1024-64     1.24µs ± 0%     0.70µs ± 0%  -43.23%  (p=1.000 n=1+1)
ChaCha20/1350-64     1.89µs ± 0%     1.03µs ± 0%  -45.35%  (p=1.000 n=1+1)
ChaCha20/65536-64    77.0µs ± 0%     42.5µs ± 0%  -44.83%  (p=1.000 n=1+1)

This is discussed in issue golang/go#25051.

A few asm instructions vmrgew and vmrgow were just added in Go 1.14
so have been encoded using WORD at this point.

Change-Id: I2b192a63cf46b0b20195e60e4412c43c5dd14ad8
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/195959
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Carlos Eduardo Seo <cseo@linux.vnet.ibm.com>
2019-10-10 18:54:27 +00:00
Lynn Boger f99c8df09e poly1305: improve performance with asm for ppc64le
This adds an asm implementation for poly1305 on ppc64le, based on
the amd64 asm implementation using the mac interface.

The improvements on a power8 based on the poly1305 benchmarks are:

    name              old time/op   new time/op    delta
    64                  172ns ± 0%      78ns ± 0%   -54.77%  (p=1.000 n=1+1)
    1K                 1.47µs ± 0%    0.59µs ± 0%   -59.69%  (p=1.000 n=1+1)
    2M                 2.84ms ± 0%    1.12ms ± 0%   -60.47%  (p=1.000 n=1+1)
    64Unaligned         172ns ± 0%      78ns ± 0%   -54.59%  (p=1.000 n=1+1)
    1KUnaligned        1.47µs ± 0%    0.59µs ± 0%   -59.69%  (p=1.000 n=1+1)
    2MUnaligned        2.84ms ± 0%    1.13ms ± 0%   -60.23%  (p=1.000 n=1+1)
    Write64             100ns ± 0%      46ns ± 0%   -53.80%  (p=1.000 n=1+1)
    Write1K            1.40µs ± 0%    0.56µs ± 0%   -59.90%  (p=1.000 n=1+1)
    Write2M            2.84ms ± 0%    1.12ms ± 0%   -60.46%  (p=1.000 n=1+1)
    Write64Unaligned    100ns ± 0%      46ns ± 0%   -53.60%  (p=1.000 n=1+1)
    Write1KUnaligned   1.40µs ± 0%    0.56µs ± 0%   -59.90%  (p=1.000 n=1+1)
    Write2MUnaligned   2.84ms ± 0%    1.13ms ± 0%   -60.22%  (p=1.000 n=1+1)

Change-Id: I77cc9bb3645a6b1a6edc414b5651dc37ae4a7410
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/173421
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Munday <mike.munday@ibm.com>
2019-06-05 12:30:33 +00:00
Lynn Boger fbbbdbdd6f sha3: revert alignment optimization on ppc64
A previous fix to improve performance based on alignment
to sha3 was made in both ppc64le and ppc64, when it should
have only applied to ppc64le.  This changes the build tags
so it is only done for ppc64le.

Fixes golang/go#15392

Change-Id: Idf32a0370f3c76fc2b54a2897a668acbae5d43c5
Reviewed-on: https://go-review.googlesource.com/22323
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2016-04-20 21:40:36 +00:00
Lynn Boger c197bcf24c golang/x/crypto/sha3: use better alignment
Change the build tags for xor.go and xor_unaligned.go
to perform better on ppc64/ppc64le.

This improves performance for these sha3 benchmarks:

benchmark                        old MB/s     new MB/s     speedup
BenchmarkSha3_512_MTU            68.42        76.03        1.11x
BenchmarkSha3_384_MTU            91.12        105.38       1.16x
BenchmarkSha3_256_MTU            110.19       132.63       1.20x
BenchmarkSha3_224_MTU            114.46       139.12       1.22x
BenchmarkShake128_MTU            129.77       163.14       1.26x
BenchmarkShake256_MTU            125.01       151.93       1.22x
BenchmarkShake256_16x            110.60       129.44       1.17x
BenchmarkShake256_1MiB           128.51       154.23       1.20x
BenchmarkSha3_512_1MiB           74.18        82.39        1.11x

Fixes golang/go#14847

Change-Id: I59ba763b4eaf3e567399b132ad820fb2dcbbf129
Reviewed-on: https://go-review.googlesource.com/20821
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
2016-03-17 20:21:02 +00:00