GronOR

GronOR performs non-orthogonal configuration interaction calculations to construct electronic wave functions from anti-symmetrized products of multi-configuration molecular fragment wave functions and to process two-electron integrals for scalable high-performance electronic-structure computations.

Key Features:

• Non-orthogonal CI: Implements non-orthogonal configuration interaction using anti-symmetrized products of multi-configuration molecular fragment wave functions.
• Two-electron integral handling: Processes two-electron integrals expressed in atomic orbitals or in an orbital basis derived from fragment molecular orbitals.
• Distributed-memory parallelism: Uses MPI+OpenACC/OpenMP to target distributed memory architectures and enable massively parallel computing.
• GPU acceleration: Supports GPU acceleration for large-scale molecular system calculations.
• Task-based execution model: Employs a task-based execution model that yields linear scaling with node count on pre-exascale architectures and provides hardware fault resiliency.
• Heterogeneous resource support: Operates on systems with distinct CPU-only and GPU-accelerated partitions.
• Interoperability: Interfaces with multi-configuration electronic structure codes to accept optimized molecular fragment orbitals, configuration interaction coefficients, and necessary integrals.
• Performance and accuracy analysis: Provides algorithmic performance benchmarks and analysis of how accuracy and computational efficiency depend on calculation thresholds.

Scientific Applications:

• Electronic-structure calculations: Enables non-orthogonal CI studies of electronic structure for molecular systems.
• Fragment-based wavefunction construction: Constructs global electronic wave functions from multi-configuration molecular fragment wave functions.
• High-performance computations: Facilitates large-molecule electronic-structure computations on distributed and GPU-accelerated architectures.
• Benchmarking and threshold studies: Supports benchmarking of parallel performance and systematic analysis of accuracy versus computational thresholds.

Methodology:

Constructs electronic wave functions from anti-symmetrized products of multi-configuration molecular fragment wave functions; processes two-electron integrals in atomic-orbital or fragment-MO orbital bases; implements distributed-memory parallelism with MPI+OpenACC/OpenMP and GPU acceleration using a task-based execution model for linear scaling and fault resiliency; interfaces with multi-configuration electronic-structure codes to ingest fragment orbitals, CI coefficients, and integrals.