> On 03 Jan 2016, at 21:52, Parrot Raiser <1parr...@gmail.com> wrote: > > While running simple tests with essentially empty Perl6 programs, I > noticed some odd variations in run timing. Occasionally, when a test > was repeated, it would take something like 3 times as long as the > previous run. > > Using the "time" command confirmed the phenomenon, but showed no > patterns I could detect in repetition, particular test, or anything > else. I was initially looking at Linux' caching as a cause, but that > didn't seem to be a factor. > > Has anyone else noticed this, and/or found a cause? (The following > tests happened not to vary much. I repeated them with previous > command-line recall when I had the anomalous results.) > > E.g. $ cat null.p6 > #! /home/guru/bin/perl6 > # End program Last changed: 2016-01-01 16:30:32 > > $./null.p6 > vs > $perl6 null.p6 > vs > $perl6 -e "" > > 1. real 0m0.313s user 0m0.240s sys 0m0.050s > 2. real 0m0.325s user 0m0.220s sys 0m0.070s > 3. real 0m0.317s user 0m0.260s sys 0m0.030s > > No significant difference there, but occasionally, "real" would jump > to > 1 second
Could the jumps to >1 seconds be explained by automatic pre-compilation taking place after you re-compiled rakudo ? Also, those timings seem rather high to me: compare to my timings: 1. real 0m0.114s user 0m0.088s sys 0m0.022s 2. real 0m0.116s user 0m0.091s sys 0m0.022s 3. real 0m0.110s user 0m0.087s sys 0m0.020s Is this running on a real slow VM? Or an older version of rakudo? Liz