random - Concurrent Communicating Kernels in CUDA? -
lets have device 2880 cuda cores.
i want run monte carlo simulation where:
- 2000 threads each running sample
- 880 threads generating random numbers
this because:
- i want 2000 samples therefore other 880 sitting idle
- i know generating random numbers can slow
therefore want make pool of random numbers replenished continuously 880 threads 2000 sample threads can take when required.
is possible? if so, please provide example.
strictly speaking propose not seem possible in cuda and, others point out, it's surely not idea. may want pick book or online course first familiarize gpu programming concepts.
more point, if want dive in straight, here's mc pi example solved cuda, openacc , thrust.
Comments
Post a Comment