Julien Fiore wrote: > Hi, > I wrote a function to compute openness (a digital elevation model > attribute). The function opens the gdal-compatible input raster and > reads the values in a square window around each cell. The results are > stored in a "1 line buffer". At the end of the line, the buffer is > written to the output raster. > > It runs very slowly and I suspect the raster access to be the main > bottleneck. Do you have any idea to speed-up the process ? >
Psyco doesn't understand numpy AFAIK. So you have you have two options: 1) Rewrite the code to use array.array from stdlib instead of numpy, then (hopefully) psyco will optimize it. Since array.array is one-dimensional, it's not going to be fun. 2) Rewrite the code to be vectorized (don't use psyco) Right now your code *doesn't* get any speed benefit from numpy -- Serge -- http://mail.python.org/mailman/listinfo/python-list
