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: Faster CRC checking/skipping #696

Merged
merged 2 commits into from
Nov 25, 2022
Merged

zstd: Faster CRC checking/skipping #696

merged 2 commits into from
Nov 25, 2022

Commits on Nov 23, 2022

  1. zstd: Check ignoreChecksum option once 

    Decoder.nextBlock no longer calls the checksum finalizer if not
necessary.
    @greatroar
    greatroar committed Nov 23, 2022
    Configuration menu
    Copy the full SHA
    eb034a2 View commit details
    Browse the repository at this point in the history

Commits on Nov 25, 2022

  1. zstd: Inline blockDec.checkCRC 

    checkCRC is now a uint32 with a boolean indicating whether it needs
checking. This makes the comparison shorter and avoids some allocations.

Benchmark results on older i7:

name                                                  old speed      new speed      delta
Decoder_DecoderSmall/kppkn.gtb.zst/buffered-8          458MB/s ± 0%   458MB/s ± 0%     ~     (p=0.139 n=9+8)
Decoder_DecoderSmall/kppkn.gtb.zst/unbuffered-8        568MB/s ± 2%   566MB/s ± 4%     ~     (p=0.971 n=10+10)
Decoder_DecoderSmall/geo.protodata.zst/buffered-8     1.19GB/s ± 0%  1.20GB/s ± 1%     ~     (p=0.051 n=9+10)
Decoder_DecoderSmall/geo.protodata.zst/unbuffered-8    832MB/s ± 2%   839MB/s ± 2%     ~     (p=0.156 n=10+9)
Decoder_DecoderSmall/plrabn12.txt.zst/buffered-8       352MB/s ± 0%   352MB/s ± 0%     ~     (p=0.148 n=10+10)
Decoder_DecoderSmall/plrabn12.txt.zst/unbuffered-8     534MB/s ± 7%   542MB/s ± 3%     ~     (p=0.315 n=10+10)
Decoder_DecoderSmall/lcet10.txt.zst/buffered-8         422MB/s ± 0%   422MB/s ± 0%     ~     (p=0.408 n=8+10)
Decoder_DecoderSmall/lcet10.txt.zst/unbuffered-8       606MB/s ± 9%   618MB/s ± 7%     ~     (p=0.393 n=10+10)
Decoder_DecoderSmall/asyoulik.txt.zst/buffered-8       362MB/s ± 5%   367MB/s ± 0%   +1.44%  (p=0.026 n=9+10)
Decoder_DecoderSmall/asyoulik.txt.zst/unbuffered-8     449MB/s ± 3%   463MB/s ± 4%   +3.28%  (p=0.003 n=10+9)
Decoder_DecoderSmall/alice29.txt.zst/buffered-8        361MB/s ± 0%   361MB/s ± 0%     ~     (p=0.888 n=8+9)
Decoder_DecoderSmall/alice29.txt.zst/unbuffered-8      569MB/s ± 8%   577MB/s ± 2%     ~     (p=0.541 n=9+8)
Decoder_DecoderSmall/html_x_4.zst/buffered-8          2.46GB/s ± 1%  2.46GB/s ± 0%     ~     (p=0.720 n=10+9)
Decoder_DecoderSmall/html_x_4.zst/unbuffered-8        1.67GB/s ± 6%  1.65GB/s ± 4%     ~     (p=0.315 n=10+9)
Decoder_DecoderSmall/paper-100k.pdf.zst/buffered-8    3.87GB/s ± 0%  3.87GB/s ± 0%     ~     (p=0.497 n=10+9)
Decoder_DecoderSmall/paper-100k.pdf.zst/unbuffered-8  1.84GB/s ± 7%  1.83GB/s ± 2%     ~     (p=0.912 n=10+10)
Decoder_DecoderSmall/fireworks.jpeg.zst/buffered-8    8.63GB/s ± 1%  8.61GB/s ± 0%     ~     (p=0.063 n=10+10)
Decoder_DecoderSmall/fireworks.jpeg.zst/unbuffered-8  3.34GB/s ± 3%  3.32GB/s ± 5%     ~     (p=0.579 n=10+10)
Decoder_DecoderSmall/urls.10K.zst/buffered-8           586MB/s ± 1%   584MB/s ± 0%   -0.46%  (p=0.005 n=9+7)
Decoder_DecoderSmall/urls.10K.zst/unbuffered-8         877MB/s ± 2%   879MB/s ± 2%     ~     (p=0.796 n=10+10)
Decoder_DecoderSmall/html.zst/buffered-8               961MB/s ± 0%   963MB/s ± 0%   +0.19%  (p=0.004 n=9+9)
Decoder_DecoderSmall/html.zst/unbuffered-8             709MB/s ± 1%   715MB/s ± 3%     ~     (p=0.203 n=8+10)
Decoder_DecoderSmall/comp-data.bin.zst/buffered-8      407MB/s ± 0%   405MB/s ± 0%   -0.47%  (p=0.000 n=10+8)
Decoder_DecoderSmall/comp-data.bin.zst/unbuffered-8    152MB/s ± 3%   151MB/s ± 4%     ~     (p=0.853 n=10+10)

name                                                  old alloc/op   new alloc/op   delta
Decoder_DecoderSmall/kppkn.gtb.zst/buffered-8            54.0B ± 0%     54.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/kppkn.gtb.zst/unbuffered-8         4.57kB ± 3%    4.57kB ± 3%     ~     (p=0.985 n=10+10)
Decoder_DecoderSmall/geo.protodata.zst/buffered-8        49.0B ± 0%     49.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/geo.protodata.zst/unbuffered-8     4.47kB ± 0%    4.47kB ± 1%     ~     (p=0.481 n=10+10)
Decoder_DecoderSmall/plrabn12.txt.zst/buffered-8         68.6B ± 2%     68.0B ± 0%     ~     (p=0.173 n=10+9)
Decoder_DecoderSmall/plrabn12.txt.zst/unbuffered-8      4.64kB ± 5%    4.62kB ± 4%     ~     (p=1.000 n=10+10)
Decoder_DecoderSmall/lcet10.txt.zst/buffered-8           63.0B ± 0%     63.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/lcet10.txt.zst/unbuffered-8        4.65kB ± 3%    4.56kB ± 3%   -1.94%  (p=0.014 n=10+10)
Decoder_DecoderSmall/asyoulik.txt.zst/buffered-8         53.0B ± 0%     53.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/asyoulik.txt.zst/unbuffered-8      4.49kB ± 1%    4.45kB ± 2%   -0.93%  (p=0.006 n=9+9)
Decoder_DecoderSmall/alice29.txt.zst/buffered-8          77.9B ± 1%     77.5B ± 1%     ~     (p=0.176 n=9+8)
Decoder_DecoderSmall/alice29.txt.zst/unbuffered-8       4.52kB ± 2%    4.48kB ± 1%   -0.94%  (p=0.003 n=10+9)
Decoder_DecoderSmall/html_x_4.zst/buffered-8             53.0B ± 6%     56.0B ± 0%     ~     (p=0.059 n=10+8)
Decoder_DecoderSmall/html_x_4.zst/unbuffered-8          4.46kB ± 0%    4.48kB ± 2%     ~     (p=1.000 n=8+10)
Decoder_DecoderSmall/paper-100k.pdf.zst/buffered-8       48.0B ± 0%     48.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/paper-100k.pdf.zst/unbuffered-8    4.46kB ± 0%    4.43kB ± 0%   -0.72%  (p=0.000 n=10+10)
Decoder_DecoderSmall/fireworks.jpeg.zst/buffered-8       48.0B ± 0%     48.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/fireworks.jpeg.zst/unbuffered-8    4.07kB ± 0%    4.04kB ± 0%   -0.79%  (p=0.000 n=9+10)
Decoder_DecoderSmall/urls.10K.zst/buffered-8             66.0B ± 0%     66.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/urls.10K.zst/unbuffered-8          4.60kB ± 3%    4.53kB ± 5%     ~     (p=0.052 n=10+10)
Decoder_DecoderSmall/html.zst/buffered-8                 49.0B ± 0%     49.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/html.zst/unbuffered-8              4.47kB ± 0%    4.46kB ± 1%     ~     (p=0.165 n=8+10)
Decoder_DecoderSmall/comp-data.bin.zst/buffered-8        48.0B ± 0%     48.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/comp-data.bin.zst/unbuffered-8     4.44kB ± 0%    4.44kB ± 0%     ~     (p=0.171 n=10+10)

name                                                  old allocs/op  new allocs/op  delta
Decoder_DecoderSmall/kppkn.gtb.zst/buffered-8             1.00 ± 0%      1.00 ± 0%     ~     (all equal)
Decoder_DecoderSmall/kppkn.gtb.zst/unbuffered-8           40.0 ± 0%      32.0 ± 0%  -20.00%  (p=0.000 n=10+10)
Decoder_DecoderSmall/geo.protodata.zst/buffered-8         1.00 ± 0%      1.00 ± 0%     ~     (all equal)
Decoder_DecoderSmall/geo.protodata.zst/unbuffered-8       40.0 ± 0%      32.0 ± 0%  -20.00%  (p=0.000 n=10+10)
Decoder_DecoderSmall/plrabn12.txt.zst/buffered-8          1.00 ± 0%      1.00 ± 0%     ~     (all equal)
Decoder_DecoderSmall/plrabn12.txt.zst/unbuffered-8        40.0 ± 0%      32.0 ± 0%  -20.00%  (p=0.000 n=10+10)
Decoder_DecoderSmall/lcet10.txt.zst/buffered-8            1.00 ± 0%      1.00 ± 0%     ~     (all equal)
Decoder_DecoderSmall/lcet10.txt.zst/unbuffered-8          40.0 ± 0%      32.0 ± 0%  -20.00%  (p=0.000 n=10+10)
Decoder_DecoderSmall/asyoulik.txt.zst/buffered-8          1.00 ± 0%      1.00 ± 0%     ~     (all equal)
Decoder_DecoderSmall/asyoulik.txt.zst/unbuffered-8        40.0 ± 0%      32.0 ± 0%  -20.00%  (p=0.000 n=10+10)
Decoder_DecoderSmall/alice29.txt.zst/buffered-8           1.00 ± 0%      1.00 ± 0%     ~     (all equal)
Decoder_DecoderSmall/alice29.txt.zst/unbuffered-8         40.0 ± 0%      32.0 ± 0%  -20.00%  (p=0.000 n=10+10)
Decoder_DecoderSmall/html_x_4.zst/buffered-8              1.00 ± 0%      1.00 ± 0%     ~     (all equal)
Decoder_DecoderSmall/html_x_4.zst/unbuffered-8            40.0 ± 0%      32.0 ± 0%  -20.00%  (p=0.000 n=10+10)
Decoder_DecoderSmall/paper-100k.pdf.zst/buffered-8        1.00 ± 0%      1.00 ± 0%     ~     (all equal)
Decoder_DecoderSmall/paper-100k.pdf.zst/unbuffered-8      40.0 ± 0%      32.0 ± 0%  -20.00%  (p=0.000 n=10+10)
Decoder_DecoderSmall/fireworks.jpeg.zst/buffered-8        1.00 ± 0%      1.00 ± 0%     ~     (all equal)
Decoder_DecoderSmall/fireworks.jpeg.zst/unbuffered-8      32.0 ± 0%      24.0 ± 0%  -25.00%  (p=0.000 n=10+10)
Decoder_DecoderSmall/urls.10K.zst/buffered-8              1.00 ± 0%      1.00 ± 0%     ~     (all equal)
Decoder_DecoderSmall/urls.10K.zst/unbuffered-8            40.0 ± 0%      32.0 ± 0%  -20.00%  (p=0.000 n=10+10)
Decoder_DecoderSmall/html.zst/buffered-8                  1.00 ± 0%      1.00 ± 0%     ~     (all equal)
Decoder_DecoderSmall/html.zst/unbuffered-8                40.0 ± 0%      32.0 ± 0%  -20.00%  (p=0.000 n=10+10)
Decoder_DecoderSmall/comp-data.bin.zst/buffered-8         1.00 ± 0%      1.00 ± 0%     ~     (all equal)
Decoder_DecoderSmall/comp-data.bin.zst/unbuffered-8       32.0 ± 0%      32.0 ± 0%     ~     (all equal)

On a newer i5:

name                                                   old speed      new speed      delta
Decoder_DecoderSmall/kppkn.gtb.zst/buffered-16          730MB/s ± 2%   729MB/s ± 2%     ~     (p=0.971 n=10+10)
Decoder_DecoderSmall/kppkn.gtb.zst/unbuffered-16       1.16GB/s ± 5%  1.17GB/s ± 5%     ~     (p=0.720 n=10+9)
Decoder_DecoderSmall/geo.protodata.zst/buffered-16     2.45GB/s ± 2%  2.45GB/s ± 4%     ~     (p=0.853 n=10+10)
Decoder_DecoderSmall/geo.protodata.zst/unbuffered-16   2.47GB/s ± 1%  2.47GB/s ± 1%     ~     (p=0.631 n=10+10)
Decoder_DecoderSmall/plrabn12.txt.zst/buffered-16       520MB/s ± 1%   528MB/s ± 2%   +1.58%  (p=0.015 n=10+10)
Decoder_DecoderSmall/plrabn12.txt.zst/unbuffered-16    1.15GB/s ±12%  1.18GB/s ± 5%     ~     (p=0.393 n=10+10)
Decoder_DecoderSmall/lcet10.txt.zst/buffered-16         615MB/s ± 2%   627MB/s ± 2%   +1.80%  (p=0.001 n=10+10)
Decoder_DecoderSmall/lcet10.txt.zst/unbuffered-16      1.19GB/s ± 4%  1.20GB/s ± 5%     ~     (p=0.684 n=10+10)
Decoder_DecoderSmall/asyoulik.txt.zst/buffered-16       565MB/s ± 2%   572MB/s ± 2%     ~     (p=0.075 n=10+10)
Decoder_DecoderSmall/asyoulik.txt.zst/unbuffered-16    1.10GB/s ±10%  1.11GB/s ± 9%     ~     (p=0.481 n=10+10)
Decoder_DecoderSmall/alice29.txt.zst/buffered-16        560MB/s ± 1%   567MB/s ± 2%   +1.22%  (p=0.008 n=10+9)
Decoder_DecoderSmall/alice29.txt.zst/unbuffered-16      928MB/s ± 3%   919MB/s ± 3%     ~     (p=0.218 n=10+10)
Decoder_DecoderSmall/html_x_4.zst/buffered-16          4.12GB/s ± 2%  4.19GB/s ± 2%   +1.75%  (p=0.008 n=9+10)
Decoder_DecoderSmall/html_x_4.zst/unbuffered-16        4.79GB/s ± 2%  4.77GB/s ± 2%     ~     (p=0.542 n=10+10)
Decoder_DecoderSmall/paper-100k.pdf.zst/buffered-16    6.02GB/s ± 2%  5.96GB/s ± 2%     ~     (p=0.123 n=10+10)
Decoder_DecoderSmall/paper-100k.pdf.zst/unbuffered-16  5.78GB/s ± 1%  5.81GB/s ± 1%     ~     (p=0.123 n=10+10)
Decoder_DecoderSmall/fireworks.jpeg.zst/buffered-16    11.5GB/s ± 2%  11.3GB/s ± 2%   -1.77%  (p=0.005 n=10+10)
Decoder_DecoderSmall/fireworks.jpeg.zst/unbuffered-16  9.83GB/s ± 1%  9.80GB/s ± 1%     ~     (p=0.353 n=10+10)
Decoder_DecoderSmall/urls.10K.zst/buffered-16           845MB/s ± 1%   845MB/s ± 2%     ~     (p=0.796 n=10+10)
Decoder_DecoderSmall/urls.10K.zst/unbuffered-16        1.46GB/s ± 2%  1.46GB/s ± 3%     ~     (p=0.971 n=10+10)
Decoder_DecoderSmall/html.zst/buffered-16              1.80GB/s ± 2%  1.84GB/s ± 2%   +2.11%  (p=0.006 n=9+8)
Decoder_DecoderSmall/html.zst/unbuffered-16            2.16GB/s ± 1%  2.18GB/s ± 1%   +0.97%  (p=0.019 n=10+10)
Decoder_DecoderSmall/comp-data.bin.zst/buffered-16      702MB/s ± 3%   708MB/s ± 1%     ~     (p=0.165 n=10+10)
Decoder_DecoderSmall/comp-data.bin.zst/unbuffered-16    609MB/s ± 1%   608MB/s ± 1%     ~     (p=0.150 n=9+10)

name                                                   old alloc/op   new alloc/op   delta
Decoder_DecoderSmall/kppkn.gtb.zst/buffered-16            55.0B ± 0%     55.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/kppkn.gtb.zst/unbuffered-16         4.51kB ± 1%    4.47kB ± 1%   -0.93%  (p=0.001 n=10+9)
Decoder_DecoderSmall/geo.protodata.zst/buffered-16        49.0B ± 0%     49.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/geo.protodata.zst/unbuffered-16     4.48kB ± 1%    4.44kB ± 0%   -0.87%  (p=0.000 n=10+10)
Decoder_DecoderSmall/plrabn12.txt.zst/buffered-16         74.5B ± 1%     74.2B ± 2%     ~     (p=0.450 n=10+10)
Decoder_DecoderSmall/plrabn12.txt.zst/unbuffered-16      4.56kB ± 2%    4.53kB ± 2%     ~     (p=0.306 n=10+10)
Decoder_DecoderSmall/lcet10.txt.zst/buffered-16           67.5B ± 1%     67.0B ± 0%   -0.74%  (p=0.033 n=10+10)
Decoder_DecoderSmall/lcet10.txt.zst/unbuffered-16        4.55kB ± 1%    4.50kB ± 1%   -1.14%  (p=0.000 n=9+10)
Decoder_DecoderSmall/asyoulik.txt.zst/buffered-16         54.0B ± 0%     54.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/asyoulik.txt.zst/unbuffered-16      4.47kB ± 0%    4.44kB ± 0%   -0.66%  (p=0.000 n=9+10)
Decoder_DecoderSmall/alice29.txt.zst/buffered-16          74.6B ± 6%     75.3B ± 7%     ~     (p=0.883 n=10+10)
Decoder_DecoderSmall/alice29.txt.zst/unbuffered-16       4.48kB ± 0%    4.45kB ± 1%   -0.57%  (p=0.005 n=7+9)
Decoder_DecoderSmall/html_x_4.zst/buffered-16             54.0B ± 0%     52.4B ± 5%   -2.96%  (p=0.046 n=8+10)
Decoder_DecoderSmall/html_x_4.zst/unbuffered-16          4.48kB ± 0%    4.46kB ± 0%   -0.43%  (p=0.016 n=10+9)
Decoder_DecoderSmall/paper-100k.pdf.zst/buffered-16       48.0B ± 0%     48.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/paper-100k.pdf.zst/unbuffered-16    4.46kB ± 0%    4.43kB ± 0%   -0.69%  (p=0.000 n=10+10)
Decoder_DecoderSmall/fireworks.jpeg.zst/buffered-16       48.0B ± 0%     48.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/fireworks.jpeg.zst/unbuffered-16    4.07kB ± 0%    4.04kB ± 0%   -0.80%  (p=0.000 n=7+9)
Decoder_DecoderSmall/urls.10K.zst/buffered-16             71.4B ± 1%     71.6B ± 1%     ~     (p=0.656 n=10+10)
Decoder_DecoderSmall/urls.10K.zst/unbuffered-16          4.55kB ± 2%    4.53kB ± 2%     ~     (p=0.529 n=10+10)
Decoder_DecoderSmall/html.zst/buffered-16                 49.0B ± 0%     49.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/html.zst/unbuffered-16              4.48kB ± 0%    4.44kB ± 0%   -0.91%  (p=0.000 n=10+9)
Decoder_DecoderSmall/comp-data.bin.zst/buffered-16        48.0B ± 0%     48.0B ± 0%     ~     (all equal)
Decoder_DecoderSmall/comp-data.bin.zst/unbuffered-16     4.44kB ± 0%    4.44kB ± 0%     ~     (p=0.897 n=10+10)

allocs/op is the same as on the previous machine.
    @greatroar
    greatroar committed Nov 25, 2022
    Configuration menu
    Copy the full SHA
    2f19fd3 View commit details
    Browse the repository at this point in the history