Question

How can I compile MPI/CUDA and UPC/CUDA hybrid code? Do I have to separately compile them or can I use language constructs interchangeably and compile as a single source file? Could someone with previous experience in this area help? Thanks in advance

Was it helpful?

Solution

MPI/CUDA - As JackOLantern has pointed out, can write MPI and CUDA code in separate files, compile them and link them. For UPC, if it is Berkeley UPC, same procedure can be done but have to do a small change at the initial configuration. When defining the compiler parameters, have to provide NVCC as both C and C++ compilers.

Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow
scroll top