[SciPy-dev] matlab io - request for testing

[Antonio]

Matthew Brett <matthew.brett <at> gmail.com> writes:

> 
> Hi,
> 
> I have been beating up the matlab io rather severely in order to
> implement some cleanups, fixes, and add new options.
> 
> I would very much appreciate it if people could pick up the current
> SVN and let me know whether they have any problems.

I tried the SVN version and found it very fast. Even putting 1M as blocksize in
scipy 0.7.0 the new version is a lot faster. Here there are benchmarks loading a
50MB matlab file:

*SCIPY 0.7.0 modified with blocksize=1M*

4771 function calls (4768 primitive calls) in 1.318 CPU seconds

   Ordered by: internal time
   List reduced from 49 to 3 due to restriction <3>

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
      400    1.011    0.003    1.011    0.003 {built-in method decompress}
       10    0.086    0.009    0.158    0.016
/usr/lib/python2.5/StringIO.py:95(seek)
        5    0.072    0.014    0.072    0.014 {method 'join' of 'str' objects}


*SCIPY '0.8.0.dev5592'*

582 function calls (579 primitive calls) in 2.957 CPU seconds

   Ordered by: internal time
   List reduced from 40 to 3 due to restriction <3>

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
       27    1.823    0.068    2.846    0.105 gzipstreams.py:77(__fill)
       52    0.963    0.019    0.963    0.019 {built-in method decompress}
        9    0.065    0.007    0.065    0.007 {method 'copy' of 'numpy.ndarray'
objects}


> Thanks a lot,

Thanks for you work :)

> Matthew
> 

  ~ Antonio

PS: put me in CC since I'm not a SciPy subscriber