Skip to content
A GPU implementation of the sampling algorithm ACHR.
Cuda MATLAB C Shell Makefile TeX
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
ACHRcu
VFWarmup
docs
paper
travis
.gitignore
.travis.yml
CONDUCT.md
LICENSE.txt
Makefile
README.md

README.md

DOI DOI TRAVIS codecov License: MIT

Description

ACHR.cu is a General Purpose Graphical Processing Unit (GP-GPU) implementation of the popular sampling algorithm of metabolic models ACHR.

The code comes with the peer-reviewed software note ACHR.cu: GPU accelerated sampling of metabolic networks..

Sampling is the tool of choice in metabolic modeling in unbiased analysis as it allows to explore the solution space constrained by the linear bounds without necessarily assuming an objective function. As metabolic models grow in size to represent communities of bacteria and complex human tissues, sampling became less used because of the large analysis time. With ACHR.cu we can achieve at least a speed up of 10x in the sampling process.

Technically sampling is a two step process:

  1. The generation of warmup points.

  2. The actual sampling starting from the previously generated warmup points.

Installation

The software can be installed via cloning this repository to your local machine and compiling VFWarmup (for step 1) and ACHR.cu (for step 2) at their root folders. More details can be found in the documentation.

Dependencies

  • IBM CPLEX v12.6 (free for academics)

  • GSL linear algebra library

  • CUDA v8.0

  • MPI and OpenMP

Harwarde requirements

  • Nvidia GPU with sm_35 architecture and above. Check the documentation for more details on the requirements.

Reproducibility

Have a look at the code ocean capusle to run ACHR.cu in an interactive container with a sample example. The capsule has an access to an NVIDIA GPU with all software dependencies cached.

Quick guide

Sampling is a two-step process:

  1. The generation of warmup points.

Quick installation

cd VFWarmup
source ./install.sh
make

Make sure to run source on the install script because it exports environment variables. Then test your installation:

make test

Then you can run the generation of warmup points

mpirun -np nCores --bind-to none -x OMP_NUM_THREADS=nThreads createWarmupPts model.mps

This command allows to generate warmup points given by the user in runtime of the model in model.mps file using dynamically load balanced nCores and nThreads through a hybrid MPI/OpenMP.

  1. The actual sampling starting from the previously generated warmup points.

Quick install first.

cd ACHRcu
source ./install.sh
make

Also here, make sure to run source on the install script because it exports environment variables. Then, test your installation:

make test

Then you can perform the sampling.

./ACHRCuda model.mps warmuppoints.csv nFiles nPoints nSteps

This command allows to generate the actual sampling points starting from the previously generated sampling points stored in warmuppoints.csv to generate a total of nFiles*nPoints with each point requiring nSteps to converge.

Acknowledgments

The experiments were carried out using the HPC facilities of the University of Luxembourg

License

The software is free ańd is licensed under the MIT license, see the file LICENSE for details.

Feedback/Issues

Please report any issues to the issues page.

Code of Conduct

Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.

You can’t perform that action at this time.