[pypy-dev] How's the JIT coming along?

Carl Friedrich Bolz cfbolz at gmx.de
Wed Aug 22 23:02:28 CEST 2007


Lenard Lindstrom wrote:
 > Gary Robinson wrote:
 >>> If i might ask - what are the use cases you are thinking about when
 >>> talking about the JIT
 >>> (and/or the JIT generator?).
 >>>
 >> Numerical computing. I have a lot of floating point arithmetic. I was
 >> also going to mention the function calling overhead that I see was
 >> already discussed (yesterday) in this thread!
 >
 > I would say that Python function calls are not the only calls to
 > consider. At some point the ctypes module will need porting to PyPy. A
 > JIT could remove the overhead of libffi.

Absolutely. This is part of the more longer-term plans for the JIT
(especially since it involves re-implementing ctypes, which will be its
own kind of pain). In theory the JIT can not only remove some of the
overhead of libffi but also be faster than statically compiled extension
modules, since in some cases it can know in advance that some runtime
type checks that a static extension does are always true and skip them.

Cheers,

Carl Friedrich



More information about the Pypy-dev mailing list