Hi, I have questions regarding symbols with suffix _exec_mode. 1) Are symbols with suffix _exec_mode always generated? If not, in which condition are they generated? 2) I call cuModuleGetGlobal() for symbols with suffix _exec_mode. cuModuleGetGlobal() will fail if the symbol specified is not found. Although my code checks the return code from cuModuleGetGlobal() to determine whether it was successful and handles it accordingly, cuda-memcheck reports there were errors calling cuModuleGetGlobal(). Is there a way in which we can check if an _exec_mode symbol is available and then call cuModuleGetGlobal() so that there won’t be any errors from cuda-memcheck? Thanks in advance!
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| [driver api][ptx]: cuModuleGetFunction fails with CUDA_ERROR_NOT_FOUND ... but cuModuleGetGlobal wor | 8 | 7176 | September 13, 2010 | |
| Pycuda gives 'cuModuleGetFunction named symbol not found error !!! | 0 | 2249 | September 21, 2017 | |
| Program control doesn't enter into __global__ function in CUDA C | 1 | 1017 | May 5, 2015 | |
| about execution (always error) | 0 | 1441 | February 13, 2009 | |
| CUDA 3.0: cudaGetSymbolSize() reports cudaErrorInvalidSymbol | 6 | 10938 | October 24, 2010 | |
| CUDA bug in dealing with cudaMemcpyToSymbol | 0 | 6852 | December 28, 2009 | |
| Problem with device emulation | 1 | 1756 | June 8, 2007 | |
| CUDA Runtime Error Only With Memcheck | 4 | 1020 | February 27, 2019 | |
| Cannot generate cuda coredump when cuda kernel access illegal address | 8 | 1542 | December 9, 2021 | |
| How can I add symbols information when debugging a GPU core dump with cuda-gdb? | 3 | 2657 | January 26, 2022 |