I have 1 GPU.
I have a lot of threads in the program that need access to the gpu.
Is there any disadvantage, if all the threads use(share) 1 single cuda context, as opposed to having their own independent ones?
I have 1 GPU.
I have a lot of threads in the program that need access to the gpu.
Is there any disadvantage, if all the threads use(share) 1 single cuda context, as opposed to having their own independent ones?