Linux-Bench: suggestion

JAW

New Member
Mar 1, 2017
3
0
1
Hamburg
First of all, the link from Linux-Bench pointing to the forum

https://forums.servethehome.com/linux-bench/

is dead. So I hope I'm at the right place here.

I am wondering if there is a particular reason you have NAMD as a molecular dynamics benchmark over other programs?
I would like to suggest to add a Gromacs Molecular dynamics benchmark and I'm happy to provide a system.
(http://manual.gromacs.org/documentation ,
GitHub - gromacs/gromacs: Public/backup repository of the gromacs molecular simulation toolkit )

The reasons for the suggestion are the following:

- Gromacs is LGPLv2.1 vs. NAMDs own licence (NAMD License)
- Gromacs is much easier to install using cmake, can run threaded without MPI
- Gromacs GPU code path has almost feature parity of CUDA and OpenCL, NAMD is CUDA only (GROMACS 2016 Release Notes — GROMACS 2016 Release Notes 2016 documentation)
- Gromacs is FAST and widely adopted
- Gromacs makes heavy usage of the newest vector extensions (AVX2, AVX512ER/F, Neon/asimd in Aarch64, VSX in Power8, New low-level SIMD support — GROMACS 5.1 Release Notes 5.1 documentation)
- Gromacs needs no other external dependencies, only fetches and compiles fftw (FFTW Home Page) via cmake installation

If you are interested, I'm happy to provide a simple install/compile script and a production test system!

cheers, Johannes
 

Patrick

Administrator
Staff member
Dec 21, 2010
11,805
4,760
113
@JAW Thanks for the feedback. I am heading to India to meet with development firms next week for the next rev.

If you do have an install/ compile script I am happy to profile it for the next Linux-Bench revision. We have 20+ systems with different CPUs online so it is fairly easy to run profiling runs.
 
  • Like
Reactions: JAW

JAW

New Member
Mar 1, 2017
3
0
1
Hamburg
That sounds great. How do I get this to you? The download from github or gromacs.org and install script is like 10 lines, I can post this right in here. But the molecluar dynamics system to run is ~5MB, could make a tarball with a little readme.txt and also upload here to the forum?
 
Last edited:

Patrick

Administrator
Staff member
Dec 21, 2010
11,805
4,760
113
That sounds great. How do I get this to you? The download from github or gromacs.org and install script is like 10 lines, I can post this right in here. But the molecluar dynamics system to run is ~5MB, could make a tarball with a little readme.txt and also upload here to the forum?
Let me shoot you a note.
 

pbuscemi

New Member
Dec 27, 2018
4
0
1
First of all, the link from Linux-Bench pointing to the forum

https://forums.servethehome.com/linux-bench/

is dead. So I hope I'm at the right place here.

I am wondering if there is a particular reason you have NAMD as a molecular dynamics benchmark over other programs?
I would like to suggest to add a Gromacs Molecular dynamics benchmark and I'm happy to provide a system.
(http://manual.gromacs.org/documentation ,
GitHub - gromacs/gromacs: Public/backup repository of the gromacs molecular simulation toolkit )

The reasons for the suggestion are the following:

- Gromacs is LGPLv2.1 vs. NAMDs own licence (NAMD License)
- Gromacs is much easier to install using cmake, can run threaded without MPI
- Gromacs GPU code path has almost feature parity of CUDA and OpenCL, NAMD is CUDA only (GROMACS 2016 Release Notes — GROMACS 2016 Release Notes 2016 documentation)
- Gromacs is FAST and widely adopted
- Gromacs makes heavy usage of the newest vector extensions (AVX2, AVX512ER/F, Neon/asimd in Aarch64, VSX in Power8, New low-level SIMD support — GROMACS 5.1 Release Notes 5.1 documentation)
- Gromacs needs no other external dependencies, only fetches and compiles fftw (FFTW Home Page) via cmake installation

If you are interested, I'm happy to provide a simple install/compile script and a production test system!

cheers, Johannes
 

pbuscemi

New Member
Dec 27, 2018
4
0
1
@JAW Thanks for the feedback. I am heading to India to meet with development firms next week for the next rev.

If you do have an install/ compile script I am happy to profile it for the next Linux-Bench revision. We have 20+ systems with different CPUs online so it is fairly easy to run profiling runs.