CUDA and cuFFT counterparts of m_logger's check_mpi: turn a failed return code into a report naming the operation, the rank and the error, followed by an abort of the whole run.
A CUDA failure is local to the rank that sees it, while its partners carry on into MPI and would wait for it forever, so these always abort the job rather than stopping the one rank.
Abort the run if a CUDA runtime call did not return cudaSuccess.
| Type | Intent | Optional | Attributes | Name | ||
|---|---|---|---|---|---|---|
| integer, | intent(in) | :: | ierr | |||
| character(len=*), | intent(in) | :: | operation |
Abort the run if a cuFFT call did not return CUFFT_SUCCESS (0). cuFFT has no error-string function, so only the code is reported.
| Type | Intent | Optional | Attributes | Name | ||
|---|---|---|---|---|---|---|
| integer, | intent(in) | :: | ierr | |||
| character(len=*), | intent(in) | :: | operation |