[pypy-dev] How's the JIT coming along?
david at ar.media.kyoto-u.ac.jp
Wed Aug 22 04:22:41 CEST 2007
Beatrice Düring wrote:
> Sorry for not having responded on this issue even prior to your email.
> We have had discussions some of us to go for a sprint in end week of
> probably in Germany, where we would focus on clean-up work/refactoring and
> also to discuss plans for a 1.1 release (scope, content, time). The
> sprint plan need to
> be discussed in the upcoming week with a sprint announcement going out.
> _My guess_ is that the JIT of course would also be involved in the scope
> of that sprint
> (clean up/refactoring) and that the 1.1 release would also contain
> results from such a clean up.
> Some of the reasons for the "silence" in the JIT area is partly because
> of people having vacation
> and post-eu sabbaticals, we have been going to conferences such as
> (Europython, Dyla) to do
> talks about the JIT to get input as well as focusing on getting external
> functions, RPython and
> PyPy to work more smoothly (which was feedback from users and would be
> at Europython in Vilnius). But yes - we might have been better at
> expressing this.
> If i might ask - what are the use cases you are thinking about when
> talking about the JIT
> (and/or the JIT generator?).
I cannot speak for Garry, but I myself would be interested in pypy for
numerical computing in python. Basically, there are cases where numpy is
not enough and require coding in C. The two cases I am thinking are:
- recursive algorithms: this means many functions calls, which are
too expensive in python (eg: imagine you have a two buffers of many
float x, and you want to compute f(x[i], [y[i], nu[i]) = x[i+1] = x[i] +
nu[i] * (x[i] - y[i]); even using ctypes for the trivial computation in
C kills performances because of the many calls)
- algorithms which require many temporaries to be efficient in numpy.
Both of them, if my understanding is right, would be perfect exemples of
easy to optimize using JIT.
More information about the Pypy-dev