On Thu Nov 20 17:26:37 2008, bacek wrote: > chromatic wrote: > > On Thursday 20 November 2008 00:21:01 Andreas J. Koenig via RT wrote: > > > >> According to examples/benchmarks/primes2.pir there was a nearly 30x > >> slowdown in the last weeks. Two random data points: > >> > >> https://svn.perl.org/parrot/tr...@31633 was fast (~ 8 secs) > >> https://svn.perl.org/parrot/tr...@32918 was slow (> 220 secs) > >> > >> Between the two I found too many segfaults and gave up on bisecting. > >> > >> Let me know if you need further data. > > > > How about between r31633 and r31688? Between r31670 and r31749? > > > > (My guess is r31688, but if you can narrow it down further, even better.) > > Yes, it was 31668. > > ba...@haste:~/src/parrot$ git bisect good > Bisecting: 0 revisions left to test after this > [aba7ce1b95674884e5fecc138fd601e8128d232c] [pdd27mmd] Merging the > pdd27mmd branch into trunk for r30235 to r31667. > >
Tested on feather, using --optimize r31667 - 5.9s,6.0s,6.1s r31668 - 2m32.838s, panic out of memory (1 run) r41391 - .106s, .107s, .241s So it's now running in 1/40 of the time it was then. Closing ticket. -- Will "Coke" Coleda _______________________________________________ http://lists.parrot.org/mailman/listinfo/parrot-dev
