Skip to content

benchmarks.md

Latest commit

 

History

History
121 lines (105 loc) · 6.79 KB

benchmarks.md

File metadata and controls

121 lines (105 loc) · 6.79 KB
Raw

benchmarks

generated from bench/results/2026-10-18-091846-566808-nyx.json.

run date: 2026-10-18T09:18:46.507560+00:00 method: warm-cache repeated local run; archive generation excluded from timed samples.

environment

  • platform: Linux-7.2.9-2-cachyos-x86_64-with-glibc2.44
  • cpu: 12th Gen Intel(R) Core(TM) i5-1235U
  • logical processors: 12
  • memory: 8028516352 bytes
  • python: 3.14.8 (main, Oct 13 2026, 10:33:10) [GCC 16.2.1 20260810]
  • rustc: rustc 1.98.1 (48a229cea 2026-09-01)
  • source revision: None
  • working tree dirty: none
  • oxar: 0.1.0
  • stdlib zlib runtime: 1.3.1.zlib-ng
  • stdlib lzma module: /usr/lib/python3.14/lzma.py
  • stdlib bz2 module: /usr/lib/python3.14/bz2.py
  • stdlib zstd module: /usr/lib/python3.14/compression/zstd/__init__.py

workload

  • entries: 250 files of 4096 bytes
  • codec input: 262144 bytes
  • repetitions: 7
  • zip size: 550475 bytes
  • tar.gz size: 523166 bytes
  • input sha256: f0dfcf0a7a654c0fd09b1bd84632cef79783ad071a0f16698bfc1285eb407aec

results

timings are seconds. iqr is the interquartile range across raw samples. archive generation is excluded; extraction target creation is included. the run is warm-cache and uses the same host for each paired operation.

operation median (s) iqr (s) output bytes
oxar.zip.list 0.001386 0.000091
stdlib.zip.list 0.000487 0.000016
oxar.zip.read_middle 0.000304 0.000092 4096
stdlib.zip.read_middle 0.000544 0.000040 4096
oxar.zip.extract 0.005376 0.000260
stdlib.zip.extract 0.006121 0.000178
oxar.tar_gz.list 0.002263 0.000112
stdlib.tar_gz.list 0.005376 0.000038
oxar.tar_gz.read_middle 0.000993 0.000071 4096
stdlib.tar_gz.read_middle 0.006578 0.000039 4096
oxar.tar_gz.extract 0.005504 0.000519
stdlib.tar_gz.extract 0.015795 0.000506
oxar.codec.gzip.compressible.compress 0.000095 0.000011 1338
stdlib.codec.gzip.compressible.compress 0.000069 0.000014 1338
oxar.codec.gzip.compressible.decompress 0.000043 0.000006 262144
stdlib.codec.gzip.compressible.decompress 0.000034 0.000011 262144
oxar.codec.bzip2.compressible.compress 0.028184 0.000584 132
stdlib.codec.bzip2.compressible.compress 0.030015 0.000455 132
oxar.codec.bzip2.compressible.decompress 0.001563 0.000200 262144
stdlib.codec.bzip2.compressible.decompress 0.001115 0.000022 262144
oxar.codec.xz.compressible.compress 0.011231 0.000296 200
stdlib.codec.xz.compressible.compress 0.004444 0.000330 200
oxar.codec.xz.compressible.decompress 0.000126 0.000009 262144
stdlib.codec.xz.compressible.decompress 0.000137 0.000016 262144
oxar.codec.zstd.compressible.compress 0.000034 0.000003 60
stdlib.codec.zstd.compressible.compress 0.000034 0.000009 60
oxar.codec.zstd.compressible.decompress 0.000098 0.000003 262144
stdlib.codec.zstd.compressible.decompress 0.000112 0.000005 262144
oxar.codec.gzip.random.compress 0.004356 0.000088 262244
stdlib.codec.gzip.random.compress 0.003570 0.000038 262244
oxar.codec.gzip.random.decompress 0.000035 0.000007 262144
stdlib.codec.gzip.random.decompress 0.000036 0.000005 262144
oxar.codec.bzip2.random.compress 0.019558 0.000161 263702
stdlib.codec.bzip2.random.compress 0.019876 0.000488 263702
oxar.codec.bzip2.random.decompress 0.010774 0.000115 262144
stdlib.codec.bzip2.random.decompress 0.010239 0.000048 262144
oxar.codec.xz.random.compress 0.051843 0.003745 262216
stdlib.codec.xz.random.compress 0.047256 0.002699 262216
oxar.codec.xz.random.decompress 0.000042 0.000015 262144
stdlib.codec.xz.random.decompress 0.000038 0.000007 262144
oxar.codec.zstd.random.compress 0.000067 0.000038 262159
stdlib.codec.zstd.random.compress 0.000082 0.000038 262159
oxar.codec.zstd.random.decompress 0.000019 0.000006 262144
stdlib.codec.zstd.random.decompress 0.000025 0.000005 262144

paired winners

paired cases: 22. oxar wins 13, stdlib wins 9, ties 0.

winner is the implementation with the lower median time. the factor is the slower median divided by the faster median; for example, 2.0x means the winner's median was half the other median. these are descriptive medians, not a statistical significance test.

operation winner median factor oxar median (s) stdlib median (s)
codec.bzip2.compressible.compress oxar 1.06x 0.028184 0.030015
codec.bzip2.compressible.decompress stdlib 1.40x 0.001563 0.001115
codec.bzip2.random.compress oxar 1.02x 0.019558 0.019876
codec.bzip2.random.decompress stdlib 1.05x 0.010774 0.010239
codec.gzip.compressible.compress stdlib 1.39x 0.000095 0.000069
codec.gzip.compressible.decompress stdlib 1.26x 0.000043 0.000034
codec.gzip.random.compress stdlib 1.22x 0.004356 0.003570
codec.gzip.random.decompress oxar 1.02x 0.000035 0.000036
codec.xz.compressible.compress stdlib 2.53x 0.011231 0.004444
codec.xz.compressible.decompress oxar 1.09x 0.000126 0.000137
codec.xz.random.compress stdlib 1.10x 0.051843 0.047256
codec.xz.random.decompress stdlib 1.10x 0.000042 0.000038
codec.zstd.compressible.compress oxar 1.00x 0.000034 0.000034
codec.zstd.compressible.decompress oxar 1.14x 0.000098 0.000112
codec.zstd.random.compress oxar 1.22x 0.000067 0.000082
codec.zstd.random.decompress oxar 1.31x 0.000019 0.000025
tar_gz.extract oxar 2.87x 0.005504 0.015795
tar_gz.list oxar 2.38x 0.002263 0.005376
tar_gz.read_middle oxar 6.62x 0.000993 0.006578
zip.extract oxar 1.14x 0.005376 0.006121
zip.list stdlib 2.85x 0.001386 0.000487
zip.read_middle oxar 1.79x 0.000304 0.000544

limits

paired winner counts use lower medians as a descriptive rule and do not indicate statistical significance. results can vary with workload, platform, compiler, and standard library build.

these are local measurements, not a cross-machine performance claim. archive fixtures are created by python's standard library. the tested subset compares zip and tar.gz listing, member reads, and extraction, plus gzip, bzip2, xz, and, where the Python version provides it, zstd one-shot codecs on compressible and random data. it does not measure 7z, archive creation in oxar, cold-cache behavior, or the larger workloads specified for release benchmarking.