Qsim pre-release not building #457
Assignees
Comments
|
I'll assume based on the logs that you're doing this on a Linux machine. With that being the case, there are a couple of possible issues you could check for:
|
|
It's also possible that you need a newer C++ compiler version; for example, GCC versions older than 5.0 do not support C++14, and C++14 only became the default after the GCC 6.1 release. An aside: I received a previous comment via email, but it's not showing up here. If it's relevant, could you copy that comment here, along with any subsequent replies? Thanks! |
|
Yes, sorry, I accidentally deleted the comment haha. I ran the command However, I could not find the CUDA installation in the directory /usr/local (I am working on a remote server), so I am not sure exactly where it is installed. Further, I could not find the cuda directory in the $PATH variable. Perhaps that is the issue. |
|
I found my CUDA installation directory /usr/lib/nvidia-cuda-toolkit/bin and added it to the $PATH variable. However, now I am getting a different error. But, |
|
...That said, our CMake pipeline should apply the |
|
I tried inserting the statements into Makefile, but I am getting the same error when running |
|
From what I see above, the flag -std=c++14 is being applied even when I don't explicitly specify it. |
|
To clarify: the The necessary flag would need to be set in our CUDA CMakeLists file, and would need to explicitly set the CUDA flags to enable C++14, as CUDA will ignore the CXX flags. |
|
So, in that case, would the only solution for me be updating GNU? |
That's likely going to be the fastest workaround, unless you're familiar with CMake and are interested in contributing. I intend to address this issue otherwise, but likely won't be able to for another couple of weeks due to conflicting priorities. |
|
I tried injecting the CMake flags, and I also tried running with an upgraded version of GNU. There is now a new error which seems even more opaque. The issue still seems to be stemming from cmake... Any ideas? |
|
Apologies for the long silence on this. The documentation I can find on CMAKE_CUDA_ARCHITECTURES suggests that it was introduced in CMake v3.18, and prior to that version it was not required. It's possible that our build process assumes an older version of CMake, and we need to update. You can try adding this block from the above link to the top of the CUDA CMakeLists.txt to resolve the issue: if(NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
set(CMAKE_CUDA_ARCHITECTURES 75)
endif() |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Hello,
I was attempting to install the pre-release version of qsim in a fresh anaconda environment by running
git clone https://github.com/quantumlib/qsim.git cd qsim make pip install .This gave me the following error:
After that, running
python3 -c "import qsimcirq"gives the errorCould I get some assistance on what is going wrong? I am attempting to build the pre-release version of qsimcirq so that I can use GPU acceleration for circuit simulations.
The text was updated successfully, but these errors were encountered: