Re: [cctbxbb] Removal of Boost.Thread support

...

...

A secondary problem is that OpenMP is incompatible with the multiprocessing module Could you enlighten me as to the exact nature of the problem? Well, you were the one who figured out the explanation, so I'm just

On Thu, Aug 16, 2012 at 4:58 AM, Luc Bourhis wrote: paraphrasing (poorly) what you told me, but: conventional multithreading does not play nice with OpenMP threads. Since Python multiprocessing (and/or threading, which is used by multiprocessing) involves multiple threads (an implementation detail that will not be obvious to the casual Python programmer, thanks to the GIL), it crashes OpenMP. I can dig up the original emails which explained this if you want.

-Nat _______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb I thought multiprocessing was to get around the issues of the GIL and use processes instead of threads, but provide the same (or as similar as

On 8/16/12 11:17 AM, Nathaniel Echols wrote: possible) api as the threading module. At least, that is what is stated on the multiprocessing module page; I haven't tried multiprocessing + OpenMP so I cannot verify that the statements on the multiprocessing page are indeed correct. --Jeff