Can it occur that 2 kernels run at the same time if the 2 kernels are continuously launched?
For instance, Device Spec is as follows for brevity.
num of SM = 4
CUDA Cores / SM = 1
CUDA Cores = 4
The code is as follows.
Kernel1 <<< 6 , 1 >>> ();
Kernel2 <<< 6 , 1 >>> ();
Which of the following is actual behavior?
case A
SM1 SM2 SM3 SM4
-------------------------------------------------------
time1 Kernel1 Kernel1 Kernel1 Kernel1
time2 Kernel1 Kernel1
time3 Kernel2 Kernel2 Kernel2 Kernel2
time4 Kernel2 Kernel2
case B
SM1 SM2 SM3 SM4
-------------------------------------------------------
time1 Kernel1 Kernel1 Kernel1 Kernel1
time2 Kernel1 Kernel1 Kernel2 Kernel2 <- simultaneously
time3 Kernel2 Kernel2 Kernel2 Kernel2