I am working on HPC software and I am trying to parallelize some C++ STL lines to be computed on Tesla C2070.
The main issue is that originally I have several C++ containers based on STL libraries and I am not able to use this containers on CUDA code (.cu) to be computed on a kernel. It seems that nvcc can not compile STL containers.
I have been reading a lot about Thrust work, but as far as I know, I have to make new Thrust containers as similar as possible, and then copy each element, one by one, from original container to Thurst containers.
Is this the solution? C++ compiler is going to work with Thrust .cpp archive?
Do you know any solution to work with STL containers under .cu archive to be compiled with nvcc?