Leonardo Santagada wrote: > On 17/03/2008, at 13:23, Carl Friedrich Bolz wrote: >> What do you all think of these ideas? Are there other ideas around? I >> know that Jakub posted something a while ago (haven't looked in detail >> yet, though). In my opinion we should try to avoid "yet another >> incomplete interpreter implentation" for this year. > > > - Threads using the new gc seems pretty important
That's true, but I fear that this is a very hard (probably too hard) task. > - Numeric on pypy - the scientific comunity could use pypy because > they don't need many extensions, mostly scipy wich numeric is the center > - Porting ZODB to pypy - this might show a lot of bugs of pypy as it > is a really complex extension. Both of these are potentially very annoying. Either you write a PyPy-specific extension module (using rffi), then you have tons of fun because we don't have separate compilation yet. Or you do this with ctypes, which probably has way too much overhead for computation-bound applications, especially with our slow implementation. I guess rewriting big projects in ctypes should only be done for projects where it seems likely that the original authors of the project will later care for the ctypes rewrite. Otherwise you get the effect that now happens for pygame: the ctypes rewrite of it is more or less abandoned. Cheers, Carl Friedrich _______________________________________________ pypy-dev@codespeak.net http://codespeak.net/mailman/listinfo/pypy-dev