configure failed to locate cudart lib

[ewtang]

Describe the bug
A clear and concise description of what the bug is.
Run "configure --enable-cuda" got "configure: error: cannot find cuda runtime libraries in /usr/local/cuda-12.6/lib /usr/local/cuda-12.6/lib64:

To Reproduce
Steps to reproduce the behavior:
Run "configure --enable-cuda"

Expected behavior
A clear and concise description of what you expected to happen.
Should be able to locate cudart lib without error.

Logs
If applicable, add logfiles to help explain your problem.
config.log

Environment (please complete the following information):

• OS and version: Fedora 41
• Compiler: gcc 14.2.1
• amuse-2024.6.0

Additional context
Add any other context about the problem here.
Machine with CUDA TK installed.

[LourensVeen]

Hi, thank you for reporting this! We're aware of this issue, the CUDA detection logic was written more than a decade ago, and things have changed since then. I'm working on a new build system with an updated CUDA detection macro that should fix this, which will be released soon.

Unfortunately, most of the CUDA-enabled codes use an older CUDA API that no longer exists, so they will still be broken. ph4 should work in the new release.

I have two questions that will help me help you:

• Which of the CUDA-enabled codes would you like to use?
• How did you install CUDA on this machine?

[ewtang]

Hi,

• Which of the CUDA-enabled codes would you like to use?
  I am interested in star cluster but have the faintest idea.

• How did you install CUDA on this machine?
  My laptop has a legacy GPU so I installed the following rpmfusion packages:
  xorg-x11-drv-nvidia-390xx
  akmod-nvidia-390xx
  xorg-x11-drv-nvidia-390xx-cuda

Then I installed the Nvidia CUDA tool kit for Fedora 39.

Thanks.

[LourensVeen]

Okay, I've had a look at the nVidia CUDA packages for Fedora 39 and how they install it, and I think that'll work fine with the new release when it comes out.

I'm a software engineer, not an astrophysicist, so I'm going to refer you to the AMUSE book (you can download it for free from there) for choosing an appropriate n-body code for star clusters, it has a whole section on them that also explains all the different codes and how they work.

As I mentioned ph4 should work in the new release, and Huayno's OpenCL mode should work with CUDA as well actually, and I'll take this as a request to get Bonsai running again, which we need to do anyway, in case you want to make some really big star clusters.

[ewtang]

I look forward to the forthcoming release. In the meantime, let me check out the AMUSE book.

Thanks again.