Install proprietary Amber16 with GPU support

[Weizheng Lu]

Hi there,

I am here again...I am trying to install Amber on my Slurm cluster. Currently I only have Amber16 purchased years before. Amber is not open-source and I just put it in my current working directory.

Here are some of my basic info:

Intel Xeon 5218

CentOS 7.8 built-in GCC 4.8.5

GCC 9.3.0 installed by Spack

Spack 0.15.3

After install GCC 9.3.0, my default gcc is 9.3.0.

I tried to install both non-gpu version and gpu version.

When I executed 'spack install amber@16' which will install non-gpu amber. I got the following:

==> Installing amber

==> No binary for amber found: installing from source

==> Warning: There is no checksum on file to fetch amber@16 safely.

==>   Fetch anyway? [y/N]

But I have already put Amber16.tar.bz2 in the current working directory.

And when I executed 'spack install amber@16 +cuda', I got the following:

1. "%gcc@7:" conflicts with "amber+cuda ^cuda@:9.1 arch=linux-None-x86_64:"

2. "%gcc@8:" conflicts with "amber+cuda ^cuda@:10.0.130 arch=linux-None-x86_64:"

3. "%gcc@6:" conflicts with "amber+cuda ^cuda@:8 arch=linux-None-x86_64:"

4. "%gcc@9:" conflicts with "amber+cuda ^cuda@:10.2.89 arch=linux-None-x86_64:"

5. "%gcc@5:" conflicts with "amber+cuda ^cuda@:7.5 arch=linux-None-x86_64:"

Is it because Spack do not support Amber16? Or my gcc version is too high?

[Pariksheet Nanda]

Hi Weizheng,

The output is telling you that CUDA's nvcc compiler is pinned against specific *older* version ranges of gcc.  For example, the output is telling you can't install CUDA 10.2 because thats only compatible with gcc@8 and earlier because gcc@9: onwards is incompatible.  The cuda package.py file has a comment detailing nvidia's gcc support matrix http://docs.nvidia.com/cuda/cuda-installation-guide-linux/#system-requirements

Though cuda@11 on spack's develop branch I think supports your gcc@9 (or at least the nvidia documentation says it should).

Pariksheet

On 8/12/20 1:11 AM, Weizheng Lu wrote:

*Message sent from a system outside of UConn.*

--
You received this message because you are subscribed to the Google Groups "Spack" group.
To unsubscribe from this group and stop receiving emails from it, send an email to spack+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/spack/48bfbaf0-a2b1-446f-882f-77104bd614a4o%40googlegroups.com.