[Python-ideas] A General Outline for Just-in-Time Acceleration of Python

[Joseph Martinot-Lagarde]

Le 14/06/2014 09:53, David Mertz a écrit :
> Is there ever a case where removing all the type annotations from Cython
> code does not produce code that can run in PyPy?  I don't know Cython
> well enough to be certain the answer is 'no', but I think so.

Cython without annotations is just python, so it can be run in pypy.

> So a
> function a little like my 'silly()' function--but that did something
> actually interesting in the loop--might run faster by removing the
> annotation and running it in PyPy.  Or is might NOT, of course; the
> answer is not obvious without looking at the exact code in question, and
> probably not without actually timing it.
>
> But the idea is that let's say I have some code with a loop and some
> numeric operations inside that loop that I'm currently running using
> CPython.  There are at least two ways I might speed up that code:
>
> A) Edit the code to contain some type annotations, and compile it with
> Cython.  However, if I do this, I *might* have to modify some other
> constructs in the overall code block to get it to compile (i.e. if
> there's any polymorphism about variable types).
>
> B) Run the unchanged code using PyPy.
>
> Well, in this description, PyPy sounds better... but it's not better if
> option (A) makes faster code in *your* specific code, of course.  And
> moreover, (B) is not true if your existing code relies on C extensions,
> such as NumPy, which mostly aren't going to run on PyPy.

Actually I don't see it like pypy vs cython. The only common point 
between these projects is that they both try to optimize python code, 
but in such different ways that the outcome completely depends on the 
code. For my problems cython is an obvious choice, for others it could 
be useless.

>
> However, I do know about https://bitbucket.org/pypy/numpy.  At least
> some substantial part of NumPy has been ported to PyPy.  This may or may
> not support the code *you* need to run.

Right now it doesn't ! ;)
I have the same problem with numba and numexpr, it seems to rarely be 
compatible with my real world use cases.

>
>
> On Sat, Jun 14, 2014 at 12:38 AM, Joseph Martinot-Lagarde
> <joseph.martinot-lagarde at m4x.org
> <mailto:joseph.martinot-lagarde at m4x.org>> wrote:
>
>     Le 14/06/2014 09:30, David Mertz a écrit :
>
>         On Fri, Jun 13, 2014 at 11:54 PM, Joseph Martinot-Lagarde
>         <joseph.martinot-lagarde at m4x.__org
>         <mailto:joseph.martinot-lagarde at m4x.org>
>         <mailto:joseph.martinot-__lagarde at m4x.org
>         <mailto:joseph.martinot-lagarde at m4x.org>>>
>         wrote:
>
>              Cython compiles all python, it is not restricted.
>
>
>         Well, kinda yes and no.  You are correct of course, that
>         anything that
>         you can execute with 'python someprog' you can compile with 'cython
>         someprog'.  However, there is an obvious sense in which adding an
>         annotation (which is, of course, a syntax error for Python itself)
>         "restricts" the code in Cython.  E.g.:
>
>              def silly():
>                   cdef int n, i
>                   for i in range(10):
>                       if i < 5:
>                           n = i + 1
>                       else:
>                           n = str(i)
>
>         This *silly* function isn't really Python code at all, of
>         course.  But
>         if you ignore the annotation, it would be--pointless code, but
>         valid. As
>         soon as you add the annotation, you *restrict* the type of code
>         you can
>         write in the scope of the annotation.
>
>
>     Yeah, the point is that *you* restrict, not cython. From your
>     previous post I understood that you meant "pypy runs all python but
>     cython doesn't, it is restricted".
>
>     I use numpy regularely, and in this case it is the other way around:
>     I can optimize my code using cython but I can't run it with pypy at all.
>
>
>     ---
>     Ce courrier électronique ne contient aucun virus ou logiciel
>     malveillant parce que la protection avast! Antivirus est active.
>     http://www.avast.com
>
>
>     _________________________________________________
>     Python-ideas mailing list
>     Python-ideas at python.org
>     <mailto:Python-ideas at python.org>
>     https://mail.python.org/__mailman/listinfo/python-ideas
>     <https://mail.python.org/mailman/listinfo/python-ideas>
>     Code of Conduct: http://python.org/psf/__codeofconduct/
>     <http://python.org/psf/codeofconduct/>
>
>
>
>
> --
> Keeping medicines from the bloodstreams of the sick; food
> from the bellies of the hungry; books from the hands of the
> uneducated; technology from the underdeveloped; and putting
> advocates of freedom in prisons.  Intellectual property is
> to the 21st century what the slave trade was to the 16th.
>
>
> _______________________________________________
> Python-ideas mailing list
> Python-ideas at python.org
> https://mail.python.org/mailman/listinfo/python-ideas
> Code of Conduct: http://python.org/psf/codeofconduct/
>


---
Ce courrier électronique ne contient aucun virus ou logiciel malveillant parce que la protection avast! Antivirus est active.
http://www.avast.com