Implementation of relativistic coupled cluster theory for massively parallel GPU-accelerated computing architectures

TL;DR

This paper presents a GPU-accelerated, parallel implementation of relativistic coupled cluster theory for heavy-element molecules, improving computational efficiency and accuracy on modern high-performance architectures.

Contribution

It introduces a new ExaTENSOR-based implementation of relativistic coupled cluster algorithms optimized for GPU and heterogeneous systems, enabling advanced molecular calculations.

Findings

Demonstrates high accuracy of the new implementation.

Shows significant performance improvements on GPU architectures.

Provides a flexible software module for relativistic electronic structure calculations.

Abstract

In this paper, we report a reimplementation of the core algorithms of relativistic coupled cluster theory aimed at modern heterogeneous high-performance computational infrastructures. The code is designed for efficient parallel execution on many compute nodes with optional GPU coprocessing, accomplished via the new ExaTENSOR back end. The resulting ExaCorr module is primarily intended for calculations of molecules with one or more heavy elements, as relativistic effects on electronic structure are included from the outset. In the current work, we thereby focus on exact 2-component methods and demonstrate the accuracy and performance of the software. The module can be used as a stand-alone program requiring a set of molecular orbital coefficients as starting point, but is also interfaced to the DIRAC program that can be used to generate these. We therefore also briefly discuss an improvement of the parallel computing aspects of the relativistic self-consistent field algorithm of the DIRAC program.