The computing speed is increased by multi-GPU, and the platform on which the program runs is cuda8 optix6 vs2015
Could you provide more details? Is the program a Cuda program you have written? Is it a bare C/C++ program or using frameworks? You mention Optix.
Do you want to know the inner workings or want to know how to access multi-gpu as programmer?
This is the architecture of my program code, which is based on the optix case, and I want to use multiple GPUs to improve the efficiency of program running, can you talk about how to use multiple GPUs to achieve acceleration, and what is the internal working, I want to know the purpose of figuring out how threads are allocated
The optix forum may also be a resource. See here for an example.
These cases of using multiple GPUs are outdated and can’t be seen