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
Compiling the GPU version, then logging out, then logging back in to recompile, will generate an error like:
make
make: *** No rule to make target '/tmp/pbs.7324411.pbspl1.nas.nasa.gov/tmpxft_00013835_00000000-5_HostCUDA.cudafe1.cpp', needed by 'HostCUDA.o'. Stop.
A workaround is to remove the HostCUDA.d file, since it has the dependency on the tmp file.
The text was updated successfully, but these errors were encountered:
This is something I had seen before, but didn't make a note of. Since the temp file is created by nvcc and it's the only way the compiler can track dependencies automatically, there are only two solutions, and neither is very pleasant. The first is to have the user delete the dependency file manually. The second is to not have the compiler automatically generate dependencies for .cu files. This, of course, means that all CUDA code will be re-compiled for every CUDA build.
Compiling the GPU version, then logging out, then logging back in to recompile, will generate an error like:
A workaround is to remove the HostCUDA.d file, since it has the dependency on the tmp file.
The text was updated successfully, but these errors were encountered: