Flavio Coelho wrote: > Thanks for the pointer, > > but randstate.pyx, which allows one to choose between differents RNGs, > offers the built-in python RNG as a python object. > > on line 561 it does a > import random > rand = random.Random() > return rand > > What I am looking for is a way to call the C function behind > random.random directly from Cython, thus saving the overhead of a > python call. check the performance difference of the two calls from > the cython snippets below: > > %cython > %timeit > from random import random > cdef int i > for i in xrange(100000000): > random() > > CPU time: 11.78 s, Wall time: 13.13 s > > %cython > %timeit > cdef extern from "stdlib.h": > long c_libc_random "random"() > void c_libc_srandom "srandom"(unsigned int seed) > cdef int i > for i in xrange(100000000): > c_libc_random() > > CPU time: 1.85 s, Wall time: 3.11 s > > Now, the above code would suffice me if the documentation of > randstate.pyx didn't mention explicitly that the RNG from libc is of > poor quality and to be avoided. Moreover, by using the RNG built into > Python, I guarantee reproducibility across platforms (I think) and > avoid external dependencies...
If you look at the sources for the random module in Python you'll see that this essentially can't be done -- it's all declared static in the C file creating the "_random" module. Of course, if it is for some reason *very* important to have the Python RNG, you could copy&paste the source code from Python. Then you have full control. (But I think you could probably just as well use another of Sage's PRNG). Then there's NumPy's PRNG (yet another one) where you can draw as many numbers at the time as you wish in an array, then access them afterwards using Cython's fast array access. Then the Python call is only done once regardless of N (but you need enough memory...) randnums = numpy.random.uniform(size=100000000) -- Dag Sverre --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-support@googlegroups.com To unsubscribe from this group, send email to sage-support-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-support URL: http://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
