Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    andrewchambers
    @andrewchambers
    yeah, i finished for now
    Benchmark #1: ./bupstash-rel put -k /tmp/x.key -r /tmp/repox/ /tmp/linux-5.9.8/
    
      Time (mean ± σ):      7.584 s ±  0.130 s    [User: 6.714 s, System: 0.991 s]
    
      Range (minmax):    7.413 s …  7.827 s
    
    Benchmark #2: ./bupstash-new put -k /tmp/x.key -r /tmp/repox/ /tmp/linux-5.9.8/
    
      Time (mean ± σ):      7.164 s ±  0.110 s    [User: 6.254 s, System: 0.983 s]
    
      Range (minmax):    7.040 s …  7.405 s
    
    Summary
    
      './bupstash-new put -k /tmp/x.key -r /tmp/repox/ /tmp/linux-5.9.8/' ran
        1.06x faster than './bupstash-rel put -k /tmp/x.key -r /tmp/repox/ /tmp/linux-5.9.8/'
    I discovered a tool called hyperfine
    sogaiu
    @sogaiu
    don't know that one
    andrewchambers
    @andrewchambers
    just prints these nice summaries telling you which command out of a list is faster
    sogaiu
    @sogaiu
    it even outputs greek
    andrewchambers
    @andrewchambers
    lol
    sogaiu
    @sogaiu
    ok, so new seems faster :)
    andrewchambers
    @andrewchambers
    haha, yeah, i think this is muuuuuch faster than borg or restic
    actually maybe i can try.
    sogaiu
    @sogaiu
    good to have real data :)
    andrewchambers
    @andrewchambers
    darn I lied
    slower by a tad
    however, once you add networking, bupstash is faster.
    actually bupstash is faster in cpu time, slower in real time lol
    though the bupstash repository is smaller too so lol
    dunno
    sogaiu
    @sogaiu
    perhaps it will become clearer after some more experiments
    andrewchambers
    @andrewchambers
    @sogaiu just spent today playing with performance
    Benchmark #1: restic init -r /tmp/resticx ; restic backup -r /tmp/resticx/ /tmp/linux-5.9.8/
    
      Time (mean ± σ):      8.544 s ±  0.045 s    [User: 15.101 s, System: 3.395 s]
    
      Range (min … max):    8.464 s …  8.621 s
    
    Benchmark #2: bupstash init -r /tmp/repox; bupstash put -q -k /tmp/x.key -r /tmp/repox/ /tmp/linux-5.9.8/
    
      Time (mean ± σ):      5.709 s ±  0.110 s    [User: 4.767 s, System: 0.971 s]
    
      Range (min … max):    5.555 s …  5.977 s
    
    Summary
    
      'bupstash init -r /tmp/repox; bupstash put -q -k /tmp/x.key -r /tmp/repox/ /tmp/linux-5.9.8/' ran
        1.50x faster than 'restic init -r /tmp/resticx ; restic backup -r /tmp/resticx/ /tmp/linux-5.9.8/
    andrewchambers
    @andrewchambers
    I improved bupstash quite a decent amount with some benchmarking.
    now definitely faster than restic
    sogaiu
    @sogaiu
    \o/ congratulations :)
    that does look quite a bit different
    andrewchambers
    @andrewchambers
    I got to do some algorithm research and then some unsafe rust code lol
    i'm still playing with it.
    andrewchambers
    @andrewchambers
    @sogaiu I got it down to 4 seconds
    so we are 2x as fast as restic for that
    sogaiu
    @sogaiu
    lol -- i guess this is make-a-prediction-and-ensure-it-is-true-driven-development?
    but seriously -- nice!
    andrewchambers
    @andrewchambers
    make a bold claim
    then desperately google search till i can make it work lol
    so the bupstash release from 3 days ago took like 12 seconds for that benchmark
    so it got 3x as fast
    sogaiu
    @sogaiu
    wow
    andrewchambers
    @andrewchambers
    @sogaiu one of the funniest things
    was the progress bar used to take 10 percent of the time
    sogaiu
    @sogaiu
    the percent of time it takes is different now?
    andrewchambers
    @andrewchambers
    yeah, I stopped updating it so frequently
    i don't update it for every file
    instead i update it at the end of each directory
    maybe this will be too infrequently, but I dunno
    sogaiu
    @sogaiu
    it seems like it might depend on details of specific files and directories
    andrewchambers
    @andrewchambers
    yeah
    I might need to revisit that
    however, if you pass -q it wastes no time lol
    andrewchambers
    @andrewchambers
    ahh @sogaiu yeah this is not frequently enough haha
    it looks awkward
    sogaiu
    @sogaiu
    can it be time-based instead?
    andrewchambers
    @andrewchambers
    hmm maybe