PyPLASMA

PyPLASMA is a python wrapper for some functions of PLASMA. This is specially made for some linear regression work.

Requirements

• swig
• python 3.6.7
• Intel's math kernel library (MKL)
• plasma-17.1 (ICL, already pre-included)
• prodtools (custom library)
• CMAKE 3.14.x
• numpy

Math Kernel Library

Please install intel's math kernel library (https://software.seek.intel.com/performance-libraries) run install.sh (if you run it with root permission, it installs onto /opt, else it will install on the user's home directory). The installation path is essential as we need to link the library later

Compilation

Install the pre-requisites

apt-get install swig
apt-get install python3-dev
apt-get install libopenblas-dev liblapacke-dev liblapack-dev
apt-get install python3-pip
pip3 install numpy

Compile the custom library prodtools

git clone https://github.com/ToraNova/library
cd library/prodtools && ./cmake_build.sh

Clone this repository

git clone https://github.com/ToraNova/pyplasma
cd pyplasma

Edit the preset_env file to point to the roots for MKL (use your favourite editor), afterwhich, source the preset_env.sh file to your environment.

. preset_env.sh

Compile

make

Run a test

To ensure compilation and setup is successful, please run a test.

make runtest

Expect the weight array to have almost 1 on the first element. The rest of the test should return OK or 0

Troubleshooting

see here

Disclaimer !

I take no credit for ICL's work. This is just a wrapper for their awesome library

PLASMA's super fast linear algebra software library

https://bitbucket.org/icl/plasma/src/default/

Version 1.0 - Git initialized and minimal version