A Saturday 05 June 2010 18:58:02 escriguéreu: > On 6 June 2010 02:09, Francesc Alted <fal...@pytables.org> wrote: > > A Saturday 05 June 2010 13:06:45 escriguéreu: > >> That seems to have fixed it, the hardsuite now completes fully > >> (attached). > > > > Wonderful news. My preliminary look into this says that performance in > > both cases is pretty good (OSX had a reputation to have a pretty bad > > implementation of pthreads, but it does not seems the case here). > > Well, you've had to do several fixes for it recently - so it's not the > best portability wise ;) Thanks for looking at it.
Well, to say the truth, standard POSIX does not enforce the implementation of barriers (despite it is in the standard). But my benchmarks says that barriers (at least the implementation on Linux and pthreads-win32) can make threaded blosc to work up to 25% faster. It is unfortunate that OSX does not provide support for barriers, but that's life! > Apple seem to have > gone their own way with Grand Central Dispatch, it would be > interesting to see if CUDA or OpenCL were applicable to blosc or > pytables. Definitely very interesting, and probably quite a lot of work too ;-) -- 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
