Concurrent execution of more than one CUDA application

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,
Does CUDA and the CUDA enabled GPUs have capability to store the context of an application running.
Suppose I have more than one (CUDA)application running on my host, can CUDA switch between the contexts
of the two applications automatically?

Thanks,
Priya

2

Of course. You can run as many CUDA processes you want on a single GPU simultaneously, as long as there is enough GPU memory to do so.

3

What I intended to ask was that suppose there are two CUDA applications running - to be more precise each of these CUDA app has a different “kernel” code. Will the driver submit each of these kernels serially ( i.e. after kernel 1 finishes, only then submit the next job to GPU) OR can the two kernel operations be interleaved, like on a convetional CPU. In the latter case, when the GPU switches between the kernels - it needs to store the context and the state of the kernel1. Is this possible? If yes, will the host code have access to this context?

Thanks.

4

The GPU will switch between different contexts only after a kernel invocation has completed.

5

I have a follow on question. Since it appears you can have two threads running concurrently (albeit one executed at a time) on the GPU via different host threads and CUDA contexts, is it possible to have two CUDA contexts running in parallel. For example one context using 1/2 of the chip and one using the other half? Obviously these would be entirely independent computations.

6

No, it is not possible. Each kernel utilizes the entire GPU