Commercial Compilers and Parallel Random Numbers

[Scott Robert Ladd]

Lars Segerlund wrote:
> Perhaps it would be better to have a mutex in the random routine ?
> It's an easier implementation.

Most parallel implementations (SPRNG, for example) use separate 
generators or states for each thread. These are external, non-standard 
generators, similar to the one I use. A mutex or other blocking 
mechanism is likely to severely limit the benefit of parallelization.

We could provide a thread-safe RANDOM_NUMBER in the OpenMP library, 
though that wouldn't solve the problem for manually-threaded programs 
that use random numbers.

..Scott

-- 
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Software Invention for High-Performance Computing