poppopjmp / shedskin

Automatically exported from code.google.com/p/shedskin
0 stars 0 forks source link

Support compile shedskin itself to c++ #67

Closed GoogleCodeExporter closed 8 years ago

GoogleCodeExporter commented 8 years ago
It would be great to be able to compile the shedskin itself into 
c++/binary after installation. It often cost minitues of time to compile 
relatively biger code. 
I think it's intresting and a good opportunity to shine.

BTW, I'm very excied to see the efforts of developing python accelarator, 
and I'm planning to devote some efforts to refine it, something like 
enhancing the dict implementation.

Thanks a lot, :)

Original issue reported on code.google.com by jason.mi...@gmail.com on 5 May 2010 at 1:58

GoogleCodeExporter commented 8 years ago
hi jason,

thanks for the suggestion. shedskin cannot compile itself at the moment, and 
probably
never will, because of several limitations (see the tutorial for details). it's
really not meant to compile such large and dynamic programs. perhaps it could 
be used
to compile its type inference part, because that's where most time is spent, 
but I
don't think that would be worth the effort at the moment.. please see issue 40 
where
this was suggested as well.

there are better ways imo to speed up type inference, and make it scale better:
profiling, to make type inference dramatically easier (by improving its starting
point), and storing type inference results between compile sessions. I may have 
a go
at the latter at some point, but mostly because of the scaling part. I don't 
really
mind compilation taking minutes, because ideally you develop and test things 
with
cpython anyway (much more convenient), and only compile things for a final 
speedup. 

Original comment by mark.duf...@gmail.com on 7 May 2010 at 9:20

GoogleCodeExporter commented 8 years ago
oh, note that ffao has just sent in a new dict implementation, so I don't think 
that
will be necessary anymore. your help would still be welcome in other areas, of
course. the best way to start is probably to try and locate bugs or performance
problems, and then fix them yourself or report them. I'd be happy to explain 
parts of
the code to you.

Original comment by mark.duf...@gmail.com on 7 May 2010 at 9:24