On 5/22/2011 4:54 AM, Daniel Carrera wrote:
Another question: Does PDL just use threads behind the scenes, or does the programmer have to do something? For example, if I have two big matrices and I multiply them "$A x $B", will that operation somehow be distributed between multiple processors?
PDL has some level of pthreads support but, by default it is not detected correctly on cygwin so I don't have recent experience with it. Check the indicated POD from the announcement or you can follow the link from Recent News on our web site to get the announcement to perldl by John. One thing this discussion has made me aware of is we should probably caveat the new feature with experimental since it is very new and untested. In addition, I'm not sure of the level of support across all PDL platforms (e.g., win32 and macosx). --Chris
Cheers, Daniel. On 21 May 2011 16:04, chm<[email protected]> wrote:This release introduces multi-processor support using pthreads as well as a number of bugs fixed and other enhancements. The git commits for this release came from John Cerney, Dima Kogan, Doug Hunt, and Craig DeForest. Many thanks to the feedback and discussions on our perldl and pdl-porters lists as well. Enjoy and Happy PDL-ing! Chris Release Notes for PDL 2.4.9_002 -------------------------- General Notes: * Another quick CPAN developers release - It is a snapshot of the current git development tree and everything may not work correctly or have complete documentation. - These release notes may not be fully complete. Please see Changes (from the git log) for full details. - All tests may not pass, especially ones corresponding to issues in Known_problems. - Manual build/install is recommended although the cpan shell may be used by specifying the distribution path rather than just the module name, e.g.: cpan> get CHM/PDL-2.4.9_002.tar.gz cpan> make CHM/PDL-2.4.9_002.tar.gz cpan> test CHM/PDL-2.4.9_002.tar.gz cpan> look CHM/PDL-2.4.9_002.tar.gz Highlights: * New multi-processor support for PDL computations using pthreads. This can improve performance up to 2X-4X for some problems. See pdldoc ParallelCPU for information. * Fix PDL::NiceSlice bugs sf.net #3295544 and #3300467 * Fix wcols() regression where file handle arguments were not being handled correctly. * Various documentation and distributions fixes. _______________________________________________ Perldl mailing list [email protected] http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
_______________________________________________ Perldl mailing list [email protected] http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
