Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

zstd: Shorter and faster asm for decSymbol.newState #896

Merged
merged 2 commits into from
Dec 9, 2023
Merged

zstd: Shorter and faster asm for decSymbol.newState #896

merged 2 commits into from
Dec 9, 2023
+68 −107

Commits on Dec 9, 2023

  1. zstd: Shorter asm for decSymbol.newState 

    The asm needs to compute decSymbol.newState, which is

	uint16(state >> 16),

or, equivalently (except for types),

	uint32(state) >> 16.

This can be accomplished by a MOVL+SHRL, the former of which is elided
by avo, so we get a single instruction for both the BMI2 and non-BMI2
cases.

Benchmarks show no difference on a new BMI2-supporting machine, but on
an older i7, decompression throughput is a tiny bit faster:

	goos: linux
	goarch: amd64
	pkg: github.com/klauspost/compress/zstd
	cpu: Intel(R) Core(TM) i7-3770K CPU @ 3.50GHz
	                                       │     old      │                shift                │
	                                       │     B/s      │     B/s       vs base               │
	Decoder_DecodeAll/kppkn.gtb.zst-8        441.4Mi ± 2%   450.4Mi ± 0%  +2.03% (p=0.000 n=10)
	Decoder_DecodeAll/geo.protodata.zst-8    1.148Gi ± 1%   1.152Gi ± 0%  +0.34% (p=0.009 n=10)
	Decoder_DecodeAll/plrabn12.txt.zst-8     347.9Mi ± 0%   356.6Mi ± 1%  +2.48% (p=0.000 n=10)
	Decoder_DecodeAll/lcet10.txt.zst-8       417.4Mi ± 0%   427.3Mi ± 0%  +2.37% (p=0.000 n=10)
	Decoder_DecodeAll/asyoulik.txt.zst-8     347.1Mi ± 0%   352.7Mi ± 1%  +1.62% (p=0.003 n=10)
	Decoder_DecodeAll/alice29.txt.zst-8      346.3Mi ± 1%   352.6Mi ± 0%  +1.83% (p=0.000 n=10)
	Decoder_DecodeAll/html_x_4.zst-8         1.440Gi ± 0%   1.445Gi ± 0%  +0.29% (p=0.019 n=10)
	Decoder_DecodeAll/paper-100k.pdf.zst-8   4.191Gi ± 0%   4.210Gi ± 0%  +0.45% (p=0.007 n=10)
	Decoder_DecodeAll/fireworks.jpeg.zst-8   8.891Gi ± 0%   8.849Gi ± 0%  -0.47% (p=0.000 n=10)
	Decoder_DecodeAll/urls.10K.zst-8         589.6Mi ± 0%   600.2Mi ± 0%  +1.80% (p=0.001 n=10)
	Decoder_DecodeAll/html.zst-8             926.1Mi ± 1%   937.9Mi ± 0%  +1.27% (p=0.000 n=10)
	Decoder_DecodeAll/comp-data.bin.zst-8    389.6Mi ± 0%   395.1Mi ± 0%  +1.40% (p=0.000 n=10)
	geomean                                  832.6Mi        843.3Mi       +1.28%
    @greatroar
    greatroar committed Dec 9, 2023
    Configuration menu
    Copy the full SHA
    3818b77 View commit details
    Browse the repository at this point in the history

  2. zstd: Remove unused parameter in asm generator

    @greatroar
    greatroar committed Dec 9, 2023
    Configuration menu
    Copy the full SHA
    0eb409c View commit details
    Browse the repository at this point in the history