Resolutions of the Coulomb operator: VIII. Parallel implementation using the modern programming language X10

Area of Science:

• Computational Chemistry
• High-Performance Computing
• Parallel Programming

Background:

• Computing long-range Coulomb and exchange interactions is computationally intensive.
• Efficient parallelization strategies are crucial for large-scale molecular simulations.

Purpose of the Study:

• To present the use of the X10 parallel programming language for computing long-range interactions.
• To demonstrate the parallelization of the Ewald operator using X10.

Main Methods:

• Utilized X10, a partitioned global address space language, for task parallelism and data locality.
• Implemented and evaluated four dynamic load balancing schemes with X10's work-stealing runtime.
• Performed long-range Hartree-Fock (HF) energy calculations for large molecules.

Main Results:

• Achieved straightforward parallelization of the Ewald operator using both intranode and internode parallelism.
• Demonstrated efficient performance for large molecule/high-quality basis HF energy calculations.
• Successfully scaled computations up to 1024 cores on a high-performance cluster.

Conclusions:

• X10 provides an effective framework for parallelizing complex computational chemistry tasks.
• The presented approach enables significant acceleration of electronic structure calculations.
• X10's features facilitate efficient utilization of modern high-performance computing resources.