huff0: Add amd64 assembly for low tablelogs #518
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Solid improvement on large payload, but regression on small. ``` benchmark old MB/s new MB/s speedup BenchmarkDecompress4XNoTable/digits/100-32 296.37 154.65 0.52x BenchmarkDecompress4XNoTable/digits/10000-32 693.64 916.51 1.32x BenchmarkDecompress4XNoTable/digits/262143-32 631.10 876.36 1.39x BenchmarkDecompress4XNoTable/gettysburg/100-32 343.16 204.11 0.59x BenchmarkDecompress4XNoTable/twain/100-32 297.23 162.42 0.55x BenchmarkDecompress4XNoTable/low-ent.10k/100-32 267.55 149.23 0.56x BenchmarkDecompress4XNoTable/low-ent.10k/10000-32 752.34 912.03 1.21x BenchmarkDecompress4XNoTable/low-ent.10k/262143-32 778.37 1023.79 1.32x BenchmarkDecompress4XNoTable/superlow-ent-10k/262143-32 777.72 1020.92 1.31x BenchmarkDecompress4XNoTable/case1/100-32 307.09 158.09 0.51x BenchmarkDecompress4XNoTable/case1/10000-32 693.03 899.10 1.30x BenchmarkDecompress4XNoTable/case1/262143-32 696.06 954.08 1.37x BenchmarkDecompress4XNoTable/case2/100-32 292.50 159.58 0.55x BenchmarkDecompress4XNoTable/case2/10000-32 713.88 924.82 1.30x BenchmarkDecompress4XNoTable/case2/262143-32 724.83 975.00 1.35x BenchmarkDecompress4XNoTable/case3/100-32 298.72 158.89 0.53x BenchmarkDecompress4XNoTable/case3/10000-32 704.81 926.74 1.31x BenchmarkDecompress4XNoTable/case3/262143-32 708.32 969.43 1.37x BenchmarkDecompress4XNoTable/pngdata.001/100-32 285.21 153.79 0.54x BenchmarkDecompress4XNoTable/normcount2/100-32 335.74 205.54 0.61x BenchmarkDecompress4XNoTable/normcount2/10000-32 677.11 910.41 1.34x BenchmarkDecompress4XNoTable/normcount2/262143-32 682.78 942.87 1.38x BenchmarkDecompress4XNoTableTableLog8/digits-32 678.59 932.54 1.37x ```
klauspost
force-pushed
the
huff0-add-8bit-asm
branch
from
1e1c59a
to
23de804
Compare
|
@WojciechMula Added asm for low tablelogs. Should be the final cases of 4X up to speed. |
Great! TBH I wasn't sure if this 8-bit specialisation is worth rewritting --- it never appeared in |
|
You are right, though it is a low hanging fruit, so we might as well take it. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Solid improvement on large payload, but regression on small, so we fall back to Go for those.