Singular/GPI-Space Framework

TRR195_logo_edit4.png

 

Download


Shortcuts

 


 

modular

 

https://github.com/singular-gpispace/modular

pfd-parallel

 

https://github.com/singular-gpispace/pfd-parallel

framework

 

https://github.com/singular-gpispace/framework

spack packages

 

https://github.com/singular-gpispace/spack-packages

 


Roadmap

 


 

Our aim is to persue the development of the Singular/GPI-Space Framework along the following lines:

1)

Provide access from Singular to GPI-Space to call again libSingular for massively parallel computations. 

2)

Provide a sample application in form of the smoothness test algorithm.

3)

Provide the most fundamental programming constructs in form of a parallelWaitAll and parallelWaitFirst.

4)

Provide more programming constructs, e.g., a massively parallel graph expansion.

5)

Provide access to more of our applications, including the GIT-Fan, Tropicalization and Feynman Integral Applications.

6)

Provide integration between different programming constructs to allow for easy design of new applications.

7)

Provide integration with the next generation open source computer algebra system OSCAR.

8)

Provide more fundamental algorithms in a massively parallel way.

 


Github Repository

 


 

The current version of the publicly available code can be downloaded from our Github site 

https://github.com/singular-gpispace

The code is released under the GPL3 license.

The README.md of the repository

https://github.com/singular-gpispace/framework

gives detailed and explicit instructions how to install and examples how to use the Singular/GPI-Space framework.

The Singular/GPI-Space framework code base is currently migrating to the supercomputing package manager Spack, which provides an easy one-line installer experience in any environment from a personal computer to a supercomputer setting. Our repository with Spack packages

https://github.com/singular-gpispace/spack-packages

currently provides some of our applications, e.g., pfd-parallel, and all dependencies including Singular, GPI-Space, GPI-2 and Flint.

In case of problems, please open an issue on Github, or feel free to contact us directly via email.

 


Release 0.1.0 of modular

 


 

This is the first release if the modular framework for massively parallel modular computations in commutative algebra and algebraic geometry.
The release can be obtained via

git clone https://github.com/singular-gpispace/modular.git

Please refer to the Readme.md for instructions how to install and use the framework. After obtaining the package mangager Spack and registering our repository (as described in the Readme.md), you can install the package and all its dependencies by just typing

spack install modular

When doing this for the first time, plan in some time, since all dependencies are installed in the Spack environment.

 


Release 0.2.0 of pfd-parallel

 


 

This is the first release if the pfd-parallel framework for massively parallel partial fraction decomposition.
The release can be obtained via

git clone --branch v0.2.0 https://github.com/singular-gpispace/pfd-parallel.git

Please refer to the Readme.md for instructions how to install and use the framework. After obtaining the package mangager Spack and registering our repository (as described in the Readme.md), you can install the package and all its dependencies by just typing

spack install pfd-parallel

When doing this for the first time, plan in some time, since all dependencies are installed in the Spack environment.

 


Release 0.1.0 of pfd-parallel

 


 

This is the first release if the pfd-parallel framework for massively parallel partial fraction decomposition.
The release can be obtained via

git clone --branch v0.1.0 https://github.com/singular-gpispace/pfd-parallel.git

Please refer to the Readme.md for instructions how to install and use the framework. After obtaining the package mangager Spack and registering our repository (as described in the Readme.md), you can install the package and all its dependencies by just typing

spack install pfd-parallel

When doing this for the first time, plan in some time, since all dependencies are installed in the Spack environment.

 


Release 0.2.0

 


 

The release includes the first couple of the predefined workflows. These are the wait all and the wait first workflows, which apply a Singular computation to a tuple of input tokens and wait until all computations finish resp. until one computation finishes. The computations can be administered in a convinient way completely from Singular.
The release can be obtained via

git clone --branch v0.2.0 https://github.com/singular-gpispace/framework.git

Please refer to the README.md for instructions how to build Singular, GPI-Space, necessary non-standard dependencies, and the Singular/GPI-Space framework, and how to use the framework (illustrated at the example of the smoothness test, the wait all and the wait first).

 

 


Release 0.1.2

 


 

The release includes minor changes and is the last version before significant additions.
The release can be obtained via

git clone --branch v0.1.2 https://github.com/singular-gpispace/framework.git

Please refer to the README.md for instructions how to build Singular, GPI-Space, necessary non-standard dependencies, and the Singular/GPI-Space framework, and how to use the framework (illustrated at the example of the smoothness test).

 

 


Release 0.1.1

 


 

The release includes minor changes, bugfixes and improvements of the user interface.
The release can be obtained via

git clone --branch v0.1.1 https://github.com/singular-gpispace/framework.git

Please refer to the README.md for instructions how to build Singular, GPI-Space, necessary non-standard dependencies, and the Singular/GPI-Space framework, and how to use the framework (illustrated at the example of the smoothness test).

 

 


Release 0.1.0

 


 

We address items 1) and 2) of the above roadmap by providing a Singular library and accompanying dynamical module for Singular, which provides access from Singular to GPI-Space to do massively parallel computations. We use GPI-Space for the coordination of the computation and Singular as the frontend and the computational backend.

Configuration of the system can easily be done by using a so-called configuration token in Singular and specifying a nodefile file (with one hostname per line), which contains the names of the computation nodes to be used.
The release can be obtained via

git clone --branch v0.1.0 https://github.com/singular-gpispace/framework.git

Please refer to the README.md for instructions how to build Singular, GPI-Space, necessary non-standard dependencies, and the Singular/GPI-Space framework, and how to use the framework (illustrated at the example of the smoothness test).