What is the easiest way to port a whole C++ project in GPU? I’m in a big problem to port a C++ project in GPU. My assignment is to port a project (only porting not optimization). It may be in single GPU thread. Is there any tool or tricks for only porting a code???
Please help.
You would probably need to profile the application and see what takes most time ?
Try to see if u can GPU-ize that portion alone without destroying the architecture of the application.
If so, do it.
If not, break the architecture to suit the GPU and port it.
It is basically impossible to port an entire application or code base to run on a GPU, especially something written in C++ which will probably use language features that the GPU doesn’t support. What makes sense is to profile the application and work out what is slow, and analyze the data structures used in the code to see what is feasible for the GPU and go from there.