Installation¶

Requirements¶

To run VSDP, the following requirements have to be fulfilled:

A recent version of GNU Octave or MATLAB has to be installed.
The interval toolbox INTLAB is required.
At least one of the following approximate solvers has to be installed:
- CSDP,
- GLPK,
- LINPROG,
- lp_solve,
- MOSEK,
- SDPA,
- SDPT3, or
- SeDuMi.

Obtaining VSDP¶

ZIP-File¶

The most recent version of VSDP and this manual are available at https://vsdp.github.io. There you can download a ZIP-file vsdp-2020-master.zip and extract it to an arbitrary location.

Legacy versions of VSDP are available from https://www.tuhh.de/ti3/software/.

Using git¶

If you have git installed and about 700 MB of disk space available, you can easily obtain a full bundle of VSDP 2006, 2012, 2020, including some aforementioned approximate solvers, and some benchmark libraries by the command

git clone --recurse-submodules https://github.com/vsdp/vsdp-bundle

In the cloned directory vsdp-bundle/vsdp/2020 you find the latest version of VSDP.

Installing VSDP¶

If all requirements are fulfilled, just call from the MATLAB or GNU Octave command prompt inside the VSDP directory

install_vsdp;

and all necessary paths are set and VSDP is fully functional. To test the latter, you can run the small builtin test suite from MATLAB via

runtests ('testVSDP')

  Totals:
    5 Passed, 0 Failed, 0 Incomplete.
    8.2712 seconds testing time.

or from GNU Octave via

testVSDP;

Test summary
------------
 
testSINDEX   	PASSED	0.322136
testSVEC_SMAT	PASSED	0.549652
testLP       	PASSED	3.108967
testSOCP     	PASSED	2.231160
testSDP      	PASSED	16.238318

VSDP 2020 manual

Installation

Contents