Daniel Naber <daniel.na...@languagetool.org> wrote: On 2015-10-25 05:13, Dominique Pellé wrote: > > Hi Dominique, > > > I measured LT speed using command line version of LanguageTool. > > Recorded numbers are user time reported by Linux time command. > > thanks, that's interesting. I think one issue would need to be fixed > before this can be automated: Maybe we need to run the test even more > often - currently it's not clear whether a change from e.g. 5.2 seconds > to 5.3 is a (small) regression, or whether it's just random. >
Yes, numbers vary a bit too much, that's why I gave min and max with 4 measurements. Random errors happen I think because I measured on a laptop, where using too much CPU causes the CPU to scale down frequency. I also ran it on a file too small I think (500 lines only) so startup overhead is quite significant. I did not feel like waiting with long files :-) Anyway, the numbers still see some trends despite the random errors. I'll find time later to write a script later so it can be run on desktops where it might be more accurate, more languages and more versions. Creating graphs might show interesting things, especially if run on more versions. We could also measure the number of CPU instructions (with linux perf stat) which would allow seeing minor regressions. Regards Dominique
------------------------------------------------------------------------------
_______________________________________________ Languagetool-devel mailing list Languagetool-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/languagetool-devel
