[Numpy-discussion] Numpy: controlling ATLAS multithreading

Mark Dixon m.c.dixon@leeds.ac...
Wed Mar 23 04:11:54 CDT 2011


I've built Numpy 1.5.1 against a copy of ATLAS 3.8.3 and LAPACK 3.3.0, 
it's passing all tests and is looking good. Great.

I note that Numpy preferentially links against the multi-threaded ATLAS 
libraries; however, we sometimes want to constrain a process to a single 
thread, and sometimes not. I see that ATLAS doesn't have the capability of 
setting the maximum number of threads used with, say, an environment 

So, here comes the long-shot question:

Does Numpy have the capability to choose between the single-threaded and 
multi-threaded ATLAS libraries at runtime? Or do I need to make this 
decision at build time?


Mark Dixon                       Email    : m.c.dixon@leeds.ac.uk
HPC/Grid Systems Support         Tel (int): 35429
Information Systems Services     Tel (ext): +44(0)113 343 5429
University of Leeds, LS2 9JT, UK

More information about the NumPy-Discussion mailing list