-
Notifications
You must be signed in to change notification settings - Fork 8
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Include DPCT headers from system installation and remove the local copy #23
Conversation
@joeatodd Can you please confirm you are happy with the approach? |
Why did the project use its own local copy of some of the SDK files? |
Taking the zip of this version of the project and coping it over to Slurm to build in parallel with the 'main' version, I got the following error when building the nbody_dpcpp.sh: (using module cuda/11.7.1) [ 93%] Building CXX object src_sycl/CMakeFiles/nbody_dpcpp.dir/shader.cpp.o |
Hi @irudkin, the linker error has a workaround described here: |
Can I suggest, it would be good to have this linking issue with nvvm-reflect-ftz mentioned in the readme.md build section (mentioning the version of oneAPI/DPCPP) for others who may run into this problem. |
The cmake scripts use dpcpp the compiler. Should the project be changed to use the Intel icpx compiler and so remove any "dpcpp is being depreciated" warnings? |
src_sycl/include
src_sycl/CMakeLists.txt
from the cmake moduleThe DPCT is part of the oneAPI base toolkit so if
dpcpp
/icpx
is available, it is almost certaindpct
is also installed. The find module looks in the most common oneAPI installation paths and standard system header locations. It works out of the box in the oneAPI environment (after sourcing thesetvars.sh
script). It is also possible to specify the location during cmake configuration using-Ddpct_LOCATION=/my/custom/location
.