Massively Parallel Computational Chemistry with the Super Instruction Architecture and ACES4

Area of Science:

• Computational chemistry
• High-performance computing

Background:

• Scientific advancement relies on efficient parallel software for large-scale computing.
• Existing software limitations hinder the full exploitation of modern computer systems.

Purpose of the Study:

• To introduce the Super Instruction Architecture (SIA) as a parallel programming platform.
• To present the ACES4 quantum chemistry software package, built upon a reimplemented SIA.
• To demonstrate the capabilities of SIA and ACES4 through benchmark calculations.

Main Methods:

• Description of the Super Instruction Architecture (SIA) design.
• Reimplementation of SIA to address limitations of previous versions.
• Application of SIA in the development of the ACES4 quantum chemistry package.
• Performance evaluation using difficult quantum chemistry open-shell coupled-cluster benchmark calculations.

Main Results:

• The SIA platform facilitates the development of high-performing parallel software.
• The reimplemented SIA overcomes limitations of the original ACESIII program.
• ACES4, utilizing the enhanced SIA, demonstrates successful application in quantum chemistry.
• Benchmark calculations showcase the capabilities of SIA and ACES4 for complex problems.

Conclusions:

• The SIA platform offers a more cost-effective solution for developing parallel software.
• The SIA and ACES4 package represent a significant advancement for large-scale quantum chemistry computations.
• This work enables better utilization of advanced computing resources for scientific discovery.