FemtoLab Software Docs

This is the first in a series of doxygen docs detailing the software developed at the FemtoLab at IIT Kanpur under Prof. Debabarata Goswami.

yodaChill

A collaborative effort on teaching high-performance-computing techniques in a domain specific manner. This documentation site is meant to introduce and supplement the teaching methodology for HPC computing at EduHiPC 2019.

Paper Abstract

We describe a general work-flow which scales intuitively to high-performance computing (HPC) clusters for different domains of scientific computation. We demonstrate our methodology with a radial distribution function calculation in C++, with mental models for FORTRAN and Python as well. We present a pedagogical framework for the development of guided concrete incremental techniques to incorporate domain specific knowledge and transfer existing expertise for developing high-performance, platform-independent, reproducible scientific software. This is effected by presenting the acceleration of a radial distribution function, a well known algorithm in computational chemistry. Thus we assert that for domain specific algorithms, there is a language-independent pedagogical methodology which may be leveraged to ensure best practices for the scientific HPC community with minimal cognitive dissonance for practitioners and students.

Authors: Rohit Goswami, Sonaly Goswami, Pranay Baldev, Shaivya Anand, Debabrata Goswami