Open stiber opened 11 years ago
This will be especially important as we move to support multiple GPUs, so it makes sense to get it working with a single GPU first.
Probably should be generalized as applying to any situation with multiple computing resources, including multicore.
Can we automate determining number of threads per block, etc. based on GPU capabilities? We should certainly, at least, query the GPU capabilities and gracefully exit, with informative feedback, if those capabilities don't match simulation requirements.