TAMM: Tensor algebra for many-body methods.

Area of Science:

• Computational Chemistry
• High-Performance Computing
• Electronic Structure Theory

Background:

• Tensor algebra operations, like contractions, are computationally intensive in electronic structure theory.
• Existing tensor algebra frameworks face challenges in targeting diverse, heterogeneous computing platforms.

Purpose of the Study:

• Introduce Tensor Algebra for Many-body Methods (TAMM), a novel framework for developing scalable computational chemistry methods.
• Enhance productivity and performance portability across various high-performance computing systems.

Main Methods:

• TAMM decouples the specification of tensor computations from their execution on hardware.
• The framework provides a tensor algebra interface for domain scientists and optimization hooks for HPC developers.
• Its modular design supports diverse hardware architectures and facilitates integration of new algorithms.

Main Results:

• TAMM enables domain scientists to focus on algorithms while HPC experts optimize for specific hardware.
• Demonstrated ease of use, performance gains, and productivity improvements compared to existing frameworks.
• Facilitates sustainable development of scalable ground- and excited-state electronic structure methods.

Conclusions:

• TAMM provides a flexible and efficient solution for tensor algebra in computational chemistry.
• The framework's design promotes collaboration between domain scientists and HPC developers.
• TAMM is poised to advance the development of complex electronic structure calculations on modern computing platforms.