Open stiber opened 8 years ago
Preliminarily, we are adding a number of threads per block member variable to the ClusterInfo object. However, that really should be calculated separately for each kernel. Need to think about what to store and where to enable this during system initialization.
This should do the following:
How do we specify the specific devices to use in a way that makes sense and will work on multiple machines (not just raiju or cerberus)?