In article <hfwdntzvr8unnwlunz2dnuvz_uown...@posted.visi>, Grant Edwards <inva...@invalid> wrote: >On 2009-02-20, Aahz <a...@pythoncraft.com> wrote: >> Steven D'Aprano <ste...@remove.this.cybersource.com.au> wrote: >>> >>> As I understand it, there's very little benefit to multi-cores in >>> Python due to the GIL. >> >> As phrased, your statement is completely wrong. Here's a more >> correct phrasing: "For threaded compute-bound applications written >> in pure Python, there's very little benefit to multiple cores." But >> threaded I/O-bound applications do receive some benefit from multiple >> cores, and using multiple processes certainly leverages multiple >> cores. If boosting the performance of a threaded compute-bound >> application is important, one can always write the critical parts in >> C/C++. > >Do the crunchy bits of scipy/numpy, scientific python, vtk and other >compute-intensive libraries tend to release the GIL while they're busy >"computing"? > >[Perhaps using them doesn't count as "pure Python", but...]
They definitely do not count as pure Python -- but I probably should have mentioned that there are pre-existing libraries that can be used. It's just that I/O is about the only area where there is a concerted effort to ensure that GIL gets released. -- Aahz (a...@pythoncraft.com) <*> http://www.pythoncraft.com/ Weinberg's Second Law: If builders built buildings the way programmers wrote programs, then the first woodpecker that came along would destroy civilization. -- http://mail.python.org/mailman/listinfo/python-list
