A Monday 14 June 2010 09:36:59 escriguéreu: > On 10 June 2010 02:40, Francesc Alted <fal...@pytables.org> wrote: > [snip..] > > > In order to check that I've introduced no new bugs (I have frozen any > > further development in Blosc until 1.0 would be out), it would be very > > interesting if you can participate in the second round of tests. The > > procedure is very similar as the first round, just go to: > > > > http://blosc.pytables.org/trac/wiki/SyntheticBenchmarks > > Still working on OSX, with some minor performance improvements: > > CPU: Core 2 Duo 2.8 GHz > OS: OS X 10.6.3 > Compilers - hardsuite: > > Previous tests: > GCC 4.0 Elapsed time: 6409.3 s, 444.3 MB/s > GCC 4.2 Elapsed time: 4108.4 s, 693.1 MB/s > GCC 4.5 Elapsed time: 3632.0 s, 784.0 MB/s > Clang Elapsed time: 4849.8 s, 587.1 MB/s > > Current tests: > GCC 4.0 Elapsed time: 5944.9 s, 456.2 MB/s > GCC 4.2 Elapsed time: 3838.5 s, 706.5 MB/s > GCC 4.5 Elapsed time: 3377.2 s, 803.0 MB/s > Clang Elapsed time: 4438.9 s, 610.9 MB/s
Mmh, these improvements are consequence of fine-tuning in compiler flags or due only to the recent changes in Blosc? If the later, I would not expect that, so it's a welcome surprise! > I think I'll have to revisit my practice of using GCC 4.0 for everything. Well, out of curiosity, I've tried gcc 4.5 in my openSUSE/Core2 box, but I have not noticed an important improvement with respect to the original gcc 4.4. The bump in optimization should have occurred somewhere between GCC 4.2 and 4.4. Anyway, using 4.0 creates clear quite unoptimized binaries in Mac OSX (at least when using Blosc). -- Francesc Alted ------------------------------------------------------------------------------ ThinkGeek and WIRED's GeekDad team up for the Ultimate GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the lucky parental unit. See the prize list and enter to win: http://p.sf.net/sfu/thinkgeek-promo _______________________________________________ Pytables-users mailing list Pytables-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/pytables-users
