Closed edejong-caltech closed 3 years ago
One example which can turn very puzzling: if the loop range parameter for launch_n
is set to zero (what would intuitively result in no operation being performed), an "Internal error happened" SystemError
is raised without any diagnostic message.
Will take a look at this soon.
I've just checked it version 0.3.11. Now the error message should look like:
cuLaunchKernel() failed with Error code: 1
Error Name: CUDA_ERROR_INVALID_VALUE
Error Description: invalid argument
Traceback (most recent call last):
File "test_trtc.py", line 30, in
Thank you!
Looks great, thank you!
I guess this can be closed than
It would be helpful if the error message
an internal error happend
reported incheck_i
inThrustRTC/Native.py
could provide more useful debug feedback.For instance, is it an error with argument length/type? Is there an illegal operation in the kernel? Is there a typecast error in the kernel?