https://pantsbuild.org/ logo
Powered by
#development
  • w

    witty-crayon-22786

    02/03/2018, 6:12 AM
    oh... perhaps changed should be faster than 30 seconds with the daemon.
  • w

    witty-crayon-22786

    02/03/2018, 6:12 AM
    since the materialized graph should be smaller.
  • r

    rough-minister-58256

    02/03/2018, 6:13 AM
    we should probably add graph memory size to metrics
  • w

    witty-crayon-22786

    02/03/2018, 6:13 AM
    sure
  • r

    rough-minister-58256

    02/03/2018, 6:13 AM
    which would give us a closer read to the impact there
  • r

    rough-minister-58256

    02/03/2018, 6:13 AM
    vs trying to guesstimate how often users 
    ::
    from cli
  • w

    witty-crayon-22786

    02/03/2018, 6:13 AM
    can get a pretty good approximation with the node count/warmth metric
  • r

    rough-minister-58256

    02/03/2018, 6:14 AM
    whats the average size of a node?
  • r

    rough-minister-58256

    02/03/2018, 6:15 AM
    vs variance
  • w

    witty-crayon-22786

    02/03/2018, 6:16 AM
    i was thinking "in relative terms compared to what we observe for `./pants list ::`"
  • w

    witty-crayon-22786

    02/03/2018, 6:16 AM
    unknown though.
  • r

    rough-minister-58256

    02/03/2018, 6:18 AM
    yeah
  • w

    witty-crayon-22786

    02/04/2018, 4:11 AM
  • r

    rough-minister-58256

    02/05/2018, 5:01 PM
    hah
  • r

    rough-minister-58256

    02/05/2018, 7:16 PM
    the expected net speedup there is ~1s?
  • w

    witty-crayon-22786

    02/05/2018, 7:17 PM
    yea, for any daemonized run
  • w

    witty-crayon-22786

    02/05/2018, 7:17 PM
    unscientifically. it was about 800ms i think.
  • w

    witty-crayon-22786

    02/05/2018, 7:18 PM
    (not sure why initializing the engine is that expensive: maybe worth another ticket, but less critical if we enable the daemon by default)
  • r

    rough-minister-58256

    02/05/2018, 7:18 PM
    feels pretty micro in the grand scheme of things
  • w

    witty-crayon-22786

    02/05/2018, 7:19 PM
    hm, which?
  • r

    rough-minister-58256

    02/05/2018, 7:19 PM
    the 800ms-1s speedup
  • r

    rough-minister-58256

    02/05/2018, 7:20 PM
    but also prob a quick one
  • w

    witty-crayon-22786

    02/05/2018, 7:20 PM
    running a python test takes about 6 seconds currently, so that's 15%
  • r

    rough-minister-58256

    02/05/2018, 7:20 PM
    ic
  • w

    witty-crayon-22786

    02/05/2018, 7:20 PM
    but more embarassingly, 
    ./pants help
    currently takes 3-4 seconds
  • r

    rough-minister-58256

    02/05/2018, 7:20 PM
    ah, gotcha
  • w

    witty-crayon-22786

    02/05/2018, 7:22 PM
    @rough-minister-58256: if it's hard to do, then we don't need to prioritize it. it just seemed clear that if the daemon was already running we didn't need to initialize one
  • r

    rough-minister-58256

    02/05/2018, 7:22 PM
    I just think we have different definitions of ‘juicy’ 😉
  • r

    rough-minister-58256

    02/05/2018, 7:23 PM
    (have been testing in the higher end of runs, and thinking about the whole LBG construction costs and Target in the PG etc)
  • w

    witty-crayon-22786

    02/05/2018, 7:24 PM
    yep, i agree there is optimization available there.
1...444445446...906Latest