there is two different CUDA program ， if i want to run them concurrently in one GPU,is is possible?
i.e.
for example , run two qft program concurrently,is is possible that allocate independent GPU resource for them like SMs?
You can run two different CUDA programs on the same GPU. They will share the memory space, so both will run as long as you have enough GPU memory to meet the memory requirements of both programs at once. The two programs will not share time on the GPU, however. Instead, the GPU driver will first execute a kernel from program 1, then the driver will switch and execute a kernel from program 2, and so on. Each kernel will use all of the SMs on the GPU. There is some overhead when the GPU switches between programs, so you will see a drop in performance if your programs have a lot of short kernels.