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

huff0: Assembler improvements #736

Merged
merged 1 commit into from
Jan 9, 2023
Merged

huff0: Assembler improvements #736

merged 1 commit into from
Jan 9, 2023

Commits on Jan 8, 2023

  1. huff0: Assembler improvements 

    Main changes:

* Compute out[id * dstEvery + i] statically. This shaves four
  instructions off the main loops. (It also frees up a register.)

* Track "exhausted" by addition instead or OR. This gets rid of an
  additional instruction. The variable is now also zeroed inside the
  loop as a dependency hint.

Benchmark results show small speedups on some datasets:

name                                           old speed      new speed      delta
Decompress1XTable/digits-8                      350MB/s ± 0%   350MB/s ± 1%    ~     (p=0.764 n=10+9)
Decompress1XTable/gettysburg-8                  270MB/s ± 1%   268MB/s ± 1%  -0.72%  (p=0.001 n=10+10)
Decompress1XTable/twain-8                       329MB/s ± 1%   328MB/s ± 0%    ~     (p=0.035 n=10+9)
Decompress1XTable/low-ent.10k-8                 387MB/s ± 1%   386MB/s ± 0%    ~     (p=0.027 n=10+8)
Decompress1XTable/superlow-ent-10k-8            377MB/s ± 0%   375MB/s ± 0%  -0.48%  (p=0.000 n=10+10)
Decompress1XTable/crash2-8                     17.0MB/s ± 0%  16.9MB/s ± 0%  -0.36%  (p=0.004 n=9+10)
Decompress1XTable/endzerobits-8                53.3MB/s ± 0%  53.0MB/s ± 0%  -0.55%  (p=0.000 n=10+9)
Decompress1XTable/endnonzero-8                 11.3MB/s ± 0%  11.3MB/s ± 1%    ~     (p=0.060 n=10+10)
Decompress1XTable/case1-8                      22.0MB/s ± 0%  21.9MB/s ± 1%    ~     (p=0.015 n=9+9)
Decompress1XTable/case2-8                      18.1MB/s ± 1%  18.1MB/s ± 1%    ~     (p=0.202 n=10+9)
Decompress1XTable/case3-8                      19.1MB/s ± 1%  19.2MB/s ± 1%    ~     (p=0.056 n=9+10)
Decompress1XTable/pngdata.001-8                 374MB/s ± 0%   374MB/s ± 0%    ~     (p=0.148 n=10+10)
Decompress1XTable/normcount2-8                 54.4MB/s ± 1%  54.4MB/s ± 1%    ~     (p=0.617 n=10+10)
Decompress1XNoTable/digits/100-8                280MB/s ± 0%   280MB/s ± 1%    ~     (p=0.951 n=9+10)
Decompress1XNoTable/digits/10000-8              366MB/s ± 1%   367MB/s ± 0%    ~     (p=0.090 n=10+9)
Decompress1XNoTable/digits/262143-8             348MB/s ± 1%   349MB/s ± 0%    ~     (p=0.043 n=10+10)
Decompress1XNoTable/gettysburg/100-8            276MB/s ± 0%   277MB/s ± 1%  +0.44%  (p=0.009 n=10+10)
Decompress1XNoTable/gettysburg/10000-8          363MB/s ± 1%   363MB/s ± 0%    ~     (p=0.041 n=10+7)
Decompress1XNoTable/gettysburg/262143-8         349MB/s ± 1%   350MB/s ± 0%    ~     (p=0.123 n=10+10)
Decompress1XNoTable/twain/100-8                 267MB/s ± 0%   268MB/s ± 0%    ~     (p=0.052 n=10+10)
Decompress1XNoTable/twain/10000-8               357MB/s ± 3%   363MB/s ± 0%  +1.74%  (p=0.000 n=10+10)
Decompress1XNoTable/twain/262143-8              320MB/s ± 2%   329MB/s ± 0%  +3.09%  (p=0.000 n=10+10)
Decompress1XNoTable/low-ent.10k/100-8           183MB/s ± 1%   184MB/s ± 0%    ~     (p=0.211 n=9+10)
Decompress1XNoTable/low-ent.10k/10000-8         377MB/s ± 3%   385MB/s ± 1%  +2.14%  (p=0.000 n=10+10)
Decompress1XNoTable/low-ent.10k/262143-8        386MB/s ± 1%   389MB/s ± 1%  +0.84%  (p=0.005 n=10+10)
Decompress1XNoTable/superlow-ent-10k/262143-8   382MB/s ± 2%   389MB/s ± 1%  +1.89%  (p=0.001 n=10+10)
Decompress1XNoTable/crash2/100-8                276MB/s ± 2%   278MB/s ± 0%    ~     (p=0.180 n=10+8)
Decompress1XNoTable/crash2/10000-8              373MB/s ± 1%   374MB/s ± 1%    ~     (p=0.315 n=10+10)
Decompress1XNoTable/crash2/262143-8             373MB/s ± 1%   375MB/s ± 0%    ~     (p=0.165 n=10+8)
Decompress1XNoTable/endzerobits/100-8           184MB/s ± 0%   184MB/s ± 1%    ~     (p=0.845 n=9+9)
Decompress1XNoTable/endzerobits/10000-8         384MB/s ± 1%   386MB/s ± 0%  +0.61%  (p=0.007 n=10+10)
Decompress1XNoTable/endzerobits/262143-8        387MB/s ± 2%   389MB/s ± 0%    ~     (p=0.963 n=9+8)
Decompress1XNoTable/endnonzero/100-8            181MB/s ± 2%   183MB/s ± 0%    ~     (p=0.017 n=9+10)
Decompress1XNoTable/endnonzero/10000-8          385MB/s ± 0%   382MB/s ± 1%  -0.88%  (p=0.001 n=8+10)
Decompress1XNoTable/endnonzero/262143-8         387MB/s ± 1%   385MB/s ± 2%    ~     (p=0.143 n=10+10)
Decompress1XNoTable/case1/100-8                 278MB/s ± 2%   282MB/s ± 1%    ~     (p=0.013 n=10+9)
Decompress1XNoTable/case1/10000-8               373MB/s ± 1%   373MB/s ± 0%    ~     (p=0.274 n=10+8)
Decompress1XNoTable/case1/262143-8              374MB/s ± 1%   374MB/s ± 0%    ~     (p=0.589 n=10+9)
Decompress1XNoTable/case2/100-8                 274MB/s ± 0%   274MB/s ± 0%  -0.26%  (p=0.002 n=10+9)
Decompress1XNoTable/case2/10000-8               378MB/s ± 0%   377MB/s ± 0%    ~     (p=0.093 n=10+10)
Decompress1XNoTable/case2/262143-8              377MB/s ± 1%   376MB/s ± 1%    ~     (p=0.225 n=10+10)
Decompress1XNoTable/case3/100-8                 266MB/s ± 0%   265MB/s ± 0%  -0.20%  (p=0.007 n=10+9)
Decompress1XNoTable/case3/10000-8               371MB/s ± 0%   372MB/s ± 0%    ~     (p=0.211 n=10+9)
Decompress1XNoTable/case3/262143-8              373MB/s ± 0%   374MB/s ± 0%    ~     (p=0.073 n=10+10)
Decompress1XNoTable/pngdata.001/100-8           239MB/s ± 0%   239MB/s ± 0%    ~     (p=0.889 n=9+10)
Decompress1XNoTable/pngdata.001/10000-8         384MB/s ± 0%   384MB/s ± 0%    ~     (p=0.228 n=10+8)
Decompress1XNoTable/pngdata.001/262143-8        377MB/s ± 0%   379MB/s ± 0%  +0.56%  (p=0.000 n=10+10)
Decompress1XNoTable/normcount2/100-8            281MB/s ± 1%   282MB/s ± 1%    ~     (p=0.015 n=10+10)
Decompress1XNoTable/normcount2/10000-8          368MB/s ± 0%   370MB/s ± 0%  +0.37%  (p=0.004 n=10+10)
Decompress1XNoTable/normcount2/262143-8         371MB/s ± 0%   371MB/s ± 0%    ~     (p=0.034 n=8+10)
Decompress4XNoTable/digits/100-8                200MB/s ± 1%   201MB/s ± 0%    ~     (p=0.274 n=8+10)
Decompress4XNoTable/digits/10000-8              603MB/s ± 0%   622MB/s ± 1%  +3.20%  (p=0.000 n=8+10)
Decompress4XNoTable/digits/262143-8             578MB/s ± 0%   595MB/s ± 1%  +2.87%  (p=0.000 n=8+10)
Decompress4XNoTable/gettysburg/100-8            260MB/s ± 0%   260MB/s ± 1%    ~     (p=0.011 n=8+10)
Decompress4XNoTable/gettysburg/10000-8          643MB/s ± 0%   657MB/s ± 1%  +2.19%  (p=0.000 n=10+9)
Decompress4XNoTable/gettysburg/262143-8         572MB/s ± 0%   589MB/s ± 0%  +2.93%  (p=0.000 n=8+10)
Decompress4XNoTable/twain/100-8                 206MB/s ± 1%   206MB/s ± 1%    ~     (p=0.436 n=10+10)
Decompress4XNoTable/twain/10000-8               639MB/s ± 1%   653MB/s ± 1%  +2.25%  (p=0.000 n=10+10)
Decompress4XNoTable/twain/262143-8              516MB/s ± 0%   522MB/s ± 1%  +1.09%  (p=0.004 n=10+10)
Decompress4XNoTable/low-ent.10k/100-8           207MB/s ± 1%   207MB/s ± 0%    ~     (p=1.000 n=10+9)
Decompress4XNoTable/low-ent.10k/10000-8         631MB/s ± 0%   653MB/s ± 0%  +3.42%  (p=0.000 n=10+9)
Decompress4XNoTable/low-ent.10k/262143-8        685MB/s ± 1%   696MB/s ± 0%  +1.61%  (p=0.000 n=10+10)
Decompress4XNoTable/superlow-ent-10k/262143-8   684MB/s ± 1%   695MB/s ± 1%  +1.51%  (p=0.000 n=9+10)
Decompress4XNoTable/case1/100-8                 208MB/s ± 1%   207MB/s ± 0%    ~     (p=0.353 n=10+10)
Decompress4XNoTable/case1/10000-8               601MB/s ± 0%   621MB/s ± 1%  +3.22%  (p=0.000 n=10+10)
Decompress4XNoTable/case1/262143-8              613MB/s ± 1%   632MB/s ± 0%  +3.14%  (p=0.000 n=10+10)
Decompress4XNoTable/case2/100-8                 210MB/s ± 2%   208MB/s ± 2%    ~     (p=0.315 n=10+9)
Decompress4XNoTable/case2/10000-8               618MB/s ± 0%   636MB/s ± 0%  +2.95%  (p=0.000 n=10+10)
Decompress4XNoTable/case2/262143-8              635MB/s ± 0%   651MB/s ± 0%  +2.56%  (p=0.000 n=7+10)
Decompress4XNoTable/case3/100-8                 199MB/s ± 1%   200MB/s ± 1%    ~     (p=0.055 n=10+10)
Decompress4XNoTable/case3/10000-8               615MB/s ± 0%   633MB/s ± 1%  +2.94%  (p=0.000 n=10+10)
Decompress4XNoTable/case3/262143-8              620MB/s ± 0%   639MB/s ± 1%  +3.00%  (p=0.000 n=10+10)
Decompress4XNoTable/pngdata.001/100-8           212MB/s ± 0%   211MB/s ± 1%    ~     (p=0.211 n=10+9)
Decompress4XNoTable/pngdata.001/10000-8         649MB/s ± 0%   667MB/s ± 1%  +2.76%  (p=0.000 n=10+10)
Decompress4XNoTable/pngdata.001/262143-8        646MB/s ± 0%   660MB/s ± 0%  +2.28%  (p=0.000 n=9+10)
Decompress4XNoTable/normcount2/100-8            261MB/s ± 1%   262MB/s ± 1%    ~     (p=0.031 n=9+9)
Decompress4XNoTable/normcount2/10000-8          589MB/s ± 1%   613MB/s ± 0%  +3.99%  (p=0.000 n=10+9)
Decompress4XNoTable/normcount2/262143-8         585MB/s ± 3%   617MB/s ± 1%  +5.57%  (p=0.000 n=10+10)
Decompress4XNoTableTableLog8/digits-8           579MB/s ± 2%   610MB/s ± 0%  +5.33%  (p=0.000 n=10+10)
Decompress4XTable/digits-8                      584MB/s ± 1%   607MB/s ± 1%  +3.89%  (p=0.000 n=10+10)
Decompress4XTable/gettysburg-8                  370MB/s ± 0%   373MB/s ± 1%  +0.67%  (p=0.009 n=10+10)
Decompress4XTable/twain-8                       512MB/s ± 2%   523MB/s ± 1%  +2.08%  (p=0.000 n=9+10)
Decompress4XTable/low-ent.10k-8                 656MB/s ± 1%   677MB/s ± 1%  +3.21%  (p=0.000 n=10+10)
Decompress4XTable/superlow-ent-10k-8            603MB/s ± 4%   626MB/s ± 1%  +3.91%  (p=0.000 n=9+10)
Decompress4XTable/case1-8                      21.1MB/s ± 0%  21.0MB/s ± 0%  -0.55%  (p=0.000 n=9+9)
Decompress4XTable/case2-8                      17.6MB/s ± 0%  17.6MB/s ± 1%    ~     (p=0.736 n=9+10)
Decompress4XTable/case3-8                      18.7MB/s ± 1%  18.7MB/s ± 1%    ~     (p=0.642 n=10+10)
Decompress4XTable/pngdata.001-8                 648MB/s ± 0%   657MB/s ± 0%  +1.50%  (p=0.000 n=10+8)
Decompress4XTable/normcount2-8                 49.7MB/s ± 1%  49.7MB/s ± 1%    ~     (p=0.839 n=10+10)
[Geo mean]                                      271MB/s        274MB/s       +0.96%
    @greatroar
    greatroar committed Jan 8, 2023
    Configuration menu
    Copy the full SHA
    3683ed4 View commit details
    Browse the repository at this point in the history