[pypy-dev] JIT SoC

Carl Friedrich Bolz cfbolz at gmx.de
Fri Mar 28 15:42:53 CET 2008

Gary Robinson wrote:
> I saw the suggestion "A good project would be to give the 386 backend
> a good refactoring, simplifying it, then add support for missing
> features like floating-point arithmetic." on the list of PyPy
> independent project suggestions.
> I'm wondering two things:
> 1) Is it realistically possible that after the SoC, if that work was
> taken on by someone, PyPy would be usable in an environment where
> high-speed floating-point calcs were needed? (The environment I have
> in mind wouldn't be putting much other stress on python -- it doesn't
> need a GUI, etc.) It sounds to me like it's far too big a project,
> but I don't know the existing code at all so I don't want to assume
> -- it would be great if it could be done.

That really depends. In the mid-term this should be possible, and one of
the things missing is indeed a better assembler backend (since right now
we have no float support at all). Note however that PyPy doesn't support
Numeric Python, so depending what you want to do this might be a problem.

> 2) I understand Google is suppling $5000 per project ($4500 for the
> student). Is there any reason an outside company can't throw a little
> extra money into the pot for a particular task to sweeten the
> motivation for that task to be done in the SoC?

I guess there is no reason, apart from the fact that it sounds a bit
unfair to other students participating in SoC. What I could imagine is
having a sprint during the SoC-period and asking that company to pay for
the travel of the student or something similar. If you are concretely
interested, you might want to take it off-list and discuss with the
caretaker group: pypy-ct at codespeak.net


Carl Friedrich

More information about the Pypy-dev mailing list