INQ: reinventing the electronic-structure code
ORAL
Abstract
In the last few years we have seen a radical change in the design of HPC systems.
Graphical processing units (GPUs) have become the standard in supercomputers and clusters.
At the same time, new ideas have appeared in software engineering that have revolutionized code development.
We apply these new principles to the electronic structure problem in a new program called INQ.
INQ is a compact code that implements density functional theory (DFT) and time-dependent functional theory (TDDFT).
It uses modern code design features and techniques that make the development fast and simple, and ensure the quality of the program.
By designing the code around algorithms, rather than against specific implementations and numerical libraries, we provide a concise and modular code that is simple to understand, flexible, and extensible.
INQ contains by a large set of tests that check the individual components and the program as a whole.
INQ is based on the plane wave approach and can simulate both finite and periodic systems.
It implements the standard DFT semi-local functionals as well as hybrid functionals.
Hybrids are applied using the adaptively compressed exchange (ACE) approach to make the calculations computationally affordable.
Its current focus is the simulations of the excited state of large systems using real-time TDDFT.
However, it is a general code that can calculate ground state properties as well.
INQ was designed from the scratch to run in parallel on multiple GPUs.
In TDDFT simulations on GPU-based supercomputers INQ achieves excellent performance.
It can handle hundreds and thousands of atoms, with simulation times of a second or less per time-step, and scale to thousands of GPUs.
The code is open source and freely accessible at http://gitlab.com/npneq/inq
Graphical processing units (GPUs) have become the standard in supercomputers and clusters.
At the same time, new ideas have appeared in software engineering that have revolutionized code development.
We apply these new principles to the electronic structure problem in a new program called INQ.
INQ is a compact code that implements density functional theory (DFT) and time-dependent functional theory (TDDFT).
It uses modern code design features and techniques that make the development fast and simple, and ensure the quality of the program.
By designing the code around algorithms, rather than against specific implementations and numerical libraries, we provide a concise and modular code that is simple to understand, flexible, and extensible.
INQ contains by a large set of tests that check the individual components and the program as a whole.
INQ is based on the plane wave approach and can simulate both finite and periodic systems.
It implements the standard DFT semi-local functionals as well as hybrid functionals.
Hybrids are applied using the adaptively compressed exchange (ACE) approach to make the calculations computationally affordable.
Its current focus is the simulations of the excited state of large systems using real-time TDDFT.
However, it is a general code that can calculate ground state properties as well.
INQ was designed from the scratch to run in parallel on multiple GPUs.
In TDDFT simulations on GPU-based supercomputers INQ achieves excellent performance.
It can handle hundreds and thousands of atoms, with simulation times of a second or less per time-step, and scale to thousands of GPUs.
The code is open source and freely accessible at http://gitlab.com/npneq/inq
*The work was supported by the Center for Non-Perturbative Studies of Functional Materials Under Non-Equilibrium Conditions (NPNEQ) funded by the Computational Materials Sciences Program of the US Department of Energy (DOE), Office of Science, Basic Energy Sciences, Materials Sciences and Engineering Division. The work was performed under the auspices of the U.S. DOE by Lawrence Livermore National Laboratory (LLNL) under Contract DE-AC52-07NA27344. Computing support for this work came from the LLNL Institutional Computing Grand Challenge program.
–
Presenters
-
Xavier Andrade
- Lawrence Livermore Natl Lab