Pauli Virtanen wrote:
On 2009-07-17, Hans Meine <me...@informatik.uni-hamburg.de> wrote:If I understood Travis' comments in the above-mentioned thread [1] correctly, this would already fix some of the performance issues along the way (since it would suddenly allow the use of special, optimized code paths).I was wondering about this too, when working on improving the cache coherency of the reduction operations. Also these would be more efficient if the striding of the output array could be chosen freely.I wonder if it would be OK to make this change... This is something that I would like to see since I also am using python wrapped fortran. My python is sprinkled with transposes, before and after many numpy operations, to avoid the performance problems. Most of the arrays are created with fortran ordering to avoid copies when passed into fortran. My fortran wrapper, Forthon, automatically handles the ordering conversion, copying if needed, but I try to avoid the copying as much as possible. It would be very nice if some of the ordering issues could be handled under the covers by numpy. Dave |
_______________________________________________ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion