unknown error, cudaGetLastError = 30

Hi,

What are the different reasons for this kind of errors?

I am getting this when i lauch kernel with large block numbers.
with less block number my kernel goes fine.

Thanks,
Nabarun.

If you are running this on a GPU which also drives a display, it is probably the display driver watchdog timer. There is a limit of a few seconds on execution time when the display driver also running on a GPU. If your kernel takes to long, it will be killed by the display driver. So your choices are either to make the code run faster, split the work up over multiple kernel launches, or use a dedicated compute GPU.

Hi,

Thanks for the reply.

But there is a different error msg for the long running kernels.

CUDA_ERROR_LAUNCH_TIMEOUT Launch exceeded timeout.

error no- 28.

is this the only reason for this error?

thanks,

Nabarun.

oh. i was trying to access data more than i allotted.

i was trying to explore the array with a wrong offset value which is much bigger than what it actually should be.

for higher block numbers it was trying to access data beyond its limit.

and so the kernel was crashing with error no 30, unknown error.

thanks,

nabarun.