Re: [Swig-devel] GSoC - Python 3.0 support progress

[Haoyu B. <div...@gm...>]

Nitro wrote:
> Am 15.07.2008, 04:54 Uhr, schrieb Haoyu Bai <div...@gm...>:
> 
>> Here's further works I plan to do in the next half of GSoC:
> 
> This is great news, you've made a very nice progress, thanks for all the 
> work you put into this!
> 
> I have one question: Do you do any performance/profiling tests? Any idea 
> how much faster/slower the python 3 backend is compared to the old backend?
> 
> -Matthias
> 

There's a hugemod test case in SWIG which create a huge module and can 
be used as performance test. The result for Python 2.5 is:

sh -c "time python hugemod" 
 


real    0m1.177s
user    0m0.670s
sys     0m0.050s
sh -c "time python hugemod"

real    0m0.287s
user    0m0.237s
sys     0m0.007s

(the first time Python execute the module by parsing source code, and 
then it compiled into byte code so the second time it get much faster)

The result for Python 3.0 is:

sh -c "time python3 hugemod"

real    0m1.495s
user    0m1.083s
sys     0m0.050s
sh -c "time python3 hugemod"

real    0m0.479s
user    0m0.393s
sys     0m0.017s

The execution time for Python 2.x is much faster than Python 3, that 
would be mostly due to Python 3 use Unicode to represent all internal 
strings.


Best regards,

Haoyu Bai
7/15/2008