correaa
commented
2 years ago Hi Logan, welcome.
Now, I am new to using c++ code, and I am getting confused on the process to get the code to run. I know you have to compile the code, but do not know which compiler to use. I assumed since we had to install CUDA it is the one they provide, "nvcc". When I run the command "nvcc aluminum.cpp" I get the following error:
aluminum.cpp:21:10: fatal error: inq/inq.hpp: No such file or directory 21 | #include <inq/inq.hpp>
This means it is not finding the directory inq, so would that mean I need to add a path variable to find this? I tried this on "silicon_davidson.cpp", and the compiler is unable to find the system directory as well.
Yes, exactly, this is because nvcc (like any compiler) needs to know were to find the "include" files, this is passed by "-I" option.
You will need several of these options (and more) to actually compile a single .cpp or .cu file.
You can see the full "nvcc" line in this "CI output" (search for "aluminum.cpp").
For this reason we generate a script called "inq++" that has all the necessary options to compile you "inq" cpp sources. It may not work perfectly; if you find a problem open another issue with detailed error messages, etc.
I am thinking there was an issue making the code, so here are the following commands I used.
git clone --recurse-submodules git@gitlab.com:npneq/inq.git cd inq mkdir build && cd build ../configure --prefix=/usr/local make -j4 make install.
Did that work? This should have generated (and installed I believe) the
inq++script.
(BTW, we would appreciate if you run make test as well, this way we can be sure that the examples you are talking about compile and work in your machine).
Take into account that this way of "compiling" would prepare everything (e.g. inq++) to compile without GPU support. Which could be fine for experimenting a bit at first.
To compile for GPU you need the option --enable-cuda in the configure command.
See complete recipes here: https://gitlab.com/npneq/inq/-/blob/master/.gitlab-ci.yml#L560
I also installed the code on the Windows Subsystem for Linux Ubuntu, which I have had success using other DFT code such as Abinit and Quantum Espresso.
I expect the code to compile in WSL, but as far as I know WSL doesn't support running on GPU, does it?
Let us know if this answers your questions.
lllangWV-zz
Hello INQ team!
I was recently introduced to this code my graduate advisor and was told to try out the code. I have installed the code and it appears it finished without bringing up any errors.
Now, I am new to using c++ code, and I am getting confused on the process to get the code to run. I know you have to compile the code, but do not know which compiler to use. I assumed since we had to install CUDA it is the one they provide, "nvcc". When I run the command "nvcc aluminum.cpp" I get the following error:
aluminum.cpp:21:10: fatal error: inq/inq.hpp: No such file or directory 21 | #include <inq/inq.hpp>
This means it is not finding the directory inq, so would that mean I need to add a path variable to find this? I tried this on "silicon_davidson.cpp", and the compiler is unable to find the system directory as well. I am thinking there was an issue making the code, so here are the following commands I used.
git clone --recurse-submodules git@gitlab.com:npneq/inq.git cd inq mkdir build && cd build ../configure --prefix=/usr/local make -j4 make install.
I also installed the code on the Windows Subsystem for Linux Ubuntu, which I have had success using other DFT code such as Abinit and Quantum Espresso.
If you all could help, It would be greatly appreciated!
Logan