Suppose a cuda kernel/cuda thread is executing the following code:
a = a + 1;
What will happen to the cuda core that is executing this infinite thread/loop.
Will this cuda core execute this thread only ? or will the cuda core switch to other threads to execute ?
So what is the executing model of cuda and the hardware ?
Does it constantly switch to other threads, or will it try to execute a single thread for as long as possible until a memory stall happens (in the latter case if the memory stall never happens then the cuda kernel would hang indefinetly) ?