You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
At least in the installation done by spack, libcuda is not present, there is just libcudart.
This needs to be changed in cmake/FindCUDACompilerNVCC.cmake.
Don't know if -lcuda is needed for older cuda versions.
The text was updated successfully, but these errors were encountered:
libcuda.so is part of the driver, not part of the CUDA SDK. The driver is not installed on the headnode of our cluster because it does not have a GPU, so there is no libcuda.so. There is .../lib64/stubs/libcuda.so, which might be sufficient to link to. Generally, libcudart provides all the cuda[A-Z][A-Za-z]+ functions and libcuda provices all the cu[A-Z][A-Za-z]+ functions, so software like Espresso should not need to link to libcuda.
Description of changes:
- convert CUDA error codes into runtime errors using the message associated to the error code
- properly handle CUDA errors by halting the flow of the program instead of ignoring them (Barnes-Hut, LB GPU)
- remove superfluous CUDA global variables (partial fix for #2628)
- restore `cuda_gather_gpus()` functionality and convert the `device_list()` getter to a regular function (API change)
- remove unused `libcuda` dependency (fixes#4085)
At least in the installation done by spack, libcuda is not present, there is just libcudart.
This needs to be changed in cmake/FindCUDACompilerNVCC.cmake.
Don't know if -lcuda is needed for older cuda versions.
The text was updated successfully, but these errors were encountered: