ac174400f4
Improvements are made in the assembler code which improves time and
space by 9-10%.
1. ANDCC, followed by SLD is combined and replaced by CLRLSLDI.
2. MOVWZ can use an indexed load and eliminate an ADD instruction in some cases.
Example: ADD R7,R10,R7 followed by MOVWZ 0(R7),R5 can be replaced with just MOVWZ (R7)(R10),R5.
3. Optimizations for the block after the "short" label includes the same MOVWZ use of indexed load, as well as other improvements.
The gain from code changes can be seen as follows, generated by
benchstat:
goos: linux
goarch: ppc64le
pkg: hash/crc32
cpu: POWER10
| oldCrc.out | newCrc.out |
| sec/op | sec/op vs base |
CRC32/poly=IEEE/size=15/align=0-12 50.19n ± 1% 39.85n ± 0% -20.59% (p=0.002 n=6)
CRC32/poly=IEEE/size=15/align=1-12 50.18n ± 1% 39.87n ± 0% -20.54% (p=0.002 n=6)
CRC32/poly=IEEE/size=40/align=0-12 40.25n ± 0% 36.95n ± 0% -8.19% (p=0.002 n=6)
CRC32/poly=IEEE/size=40/align=1-12 40.31n ± 0% 36.95n ± 0% -8.36% (p=0.002 n=6)
CRC32/poly=IEEE/size=512/align=0-12 38.03n ± 0% 38.17n ± 0% +0.37% (p=0.002 n=6)
CRC32/poly=IEEE/size=512/align=1-12 89.19n ± 1% 73.65n ± 0% -17.43% (p=0.002 n=6)
CRC32/poly=IEEE/size=1kB/align=0-12 50.73n ± 7% 50.14n ± 0% -1.18% (p=0.002 n=6)
CRC32/poly=IEEE/size=1kB/align=1-12 101.00n ± 37% 81.58n ± 0% -19.23% (p=0.002 n=6)
CRC32/poly=IEEE/size=4kB/align=0-12 98.30n ± 45% 93.05n ± 0% -5.34% (p=0.043 n=6)
CRC32/poly=IEEE/size=4kB/align=1-12 140.8n ± 0% 125.8n ± 0% -10.65% (p=0.002 n=6)
CRC32/poly=IEEE/size=32kB/align=0-12 525.8n ± 0% 528.5n ± 0% +0.52% (p=0.011 n=6)
CRC32/poly=IEEE/size=32kB/align=1-12 584.4n ± 1% 576.3n ± 0% -1.39% (p=0.002 n=6)
geomean 90.51n 81.74n -9.69%
| oldCrc.out | newCrc.out |
| B/s | B/s vs base |
CRC32/poly=IEEE/size=15/align=0-12 285.0Mi ± 1% 359.0Mi ± 0% +25.94% (p=0.002 n=6)
CRC32/poly=IEEE/size=15/align=1-12 285.1Mi ± 1% 358.8Mi ± 0% +25.86% (p=0.002 n=6)
CRC32/poly=IEEE/size=40/align=0-12 947.8Mi ± 0% 1032.3Mi ± 0% +8.91% (p=0.002 n=6)
CRC32/poly=IEEE/size=40/align=1-12 946.2Mi ± 0% 1032.5Mi ± 0% +9.12% (p=0.002 n=6)
CRC32/poly=IEEE/size=512/align=0-12 12.54Gi ± 0% 12.49Gi ± 0% -0.37% (p=0.002 n=6)
CRC32/poly=IEEE/size=512/align=1-12 5.346Gi ± 1% 6.475Gi ± 0% +21.12% (p=0.002 n=6)
CRC32/poly=IEEE/size=1kB/align=0-12 18.80Gi ± 7% 19.02Gi ± 0% +1.20% (p=0.002 n=6)
CRC32/poly=IEEE/size=1kB/align=1-12 9.454Gi ± 27% 11.690Gi ± 0% +23.66% (p=0.002 n=6)
CRC32/poly=IEEE/size=4kB/align=0-12 38.86Gi ± 31% 41.00Gi ± 0% +5.49% (p=0.041 n=6)
CRC32/poly=IEEE/size=4kB/align=1-12 27.10Gi ± 0% 30.32Gi ± 0% +11.89% (p=0.002 n=6)
CRC32/poly=IEEE/size=32kB/align=0-12 58.05Gi ± 0% 57.74Gi ± 0% -0.53% (p=0.009 n=6)
CRC32/poly=IEEE/size=32kB/align=1-12 52.22Gi ± 1% 52.95Gi ± 0% +1.41% (p=0.002 n=6)
geomean 6.074Gi 6.724Gi +10.70%
Change-Id: I378c0e84e798656384a8009f4ac48b51614489b2
Cq-Include-Trybots: luci.golang.try:gotip-linux-ppc64_power10,gotip-linux-ppc64_power8,gotip-linux-ppc64le_power8,gotip-linux-ppc64le_power9,gotip-linux-ppc64le_power10
Reviewed-on: https://go-review.googlesource.com/c/go/+/582395
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Eli Bendersky <eliben@google.com>
|
||
|---|---|---|
| .github | ||
| api | ||
| doc | ||
| lib/time | ||
| misc | ||
| src | ||
| test | ||
| .gitattributes | ||
| .gitignore | ||
| codereview.cfg | ||
| CONTRIBUTING.md | ||
| go.env | ||
| LICENSE | ||
| PATENTS | ||
| README.md | ||
| SECURITY.md | ||
The Go Programming Language
Go is an open source programming language that makes it easy to build simple, reliable, and efficient software.
Gopher image by Renee French, licensed under Creative Commons 4.0 Attributions license.
Our canonical Git repository is located at https://go.googlesource.com/go. There is a mirror of the repository at https://github.com/golang/go.
Unless otherwise noted, the Go source files are distributed under the BSD-style license found in the LICENSE file.
Download and Install
Binary Distributions
Official binary distributions are available at https://go.dev/dl/.
After downloading a binary release, visit https://go.dev/doc/install for installation instructions.
Install From Source
If a binary distribution is not available for your combination of operating system and architecture, visit https://go.dev/doc/install/source for source installation instructions.
Contributing
Go is the work of thousands of contributors. We appreciate your help!
To contribute, please read the contribution guidelines at https://go.dev/doc/contribute.
Note that the Go project uses the issue tracker for bug reports and proposals only. See https://go.dev/wiki/Questions for a list of places to ask questions about the Go language.