[SciPy-User] Contribution to Performance Python

denis denis-bz-gg@t-online...
Tue Nov 10 09:53:56 CST 2009


Ramon,
  not an answer to your question, just a word of agreement:

FORALL is a fine construct, natural to write
    in pseudocode (I don't use fortran)
    in python, generators
    in C++: for( j k x y in parallel) is very fast on superscalar cpus
       (I use e.g. #define Forjkxy(...)  a[j][k] = f(x,y),
       don't tell the C++ police)
    in testing: for a, b, c in product( [a...], [b...], [c...] )

Would you have any more examples ?  They'd help spread FORALL.

Apropos examples, one (1) laplace is a bit meager;
has anyone (flameproof) done others, weave vs cython vs C or fortran ?

cheers
  -- denis


More information about the SciPy-User mailing list