Second-order Epstein-Nesbet perturbation theory (EN-PT2) is an important tool for improving selected configuration interaction (SCI) energies, but the large external determinant space creates substantial memory and communication costs on GPU clusters. We develop a fully GPU-accelerated implementation of distributed semistochastic EN-PT2, fusing excitation generation, screening, and PT2 weighting in a single kernel traversal. Fixed-size random-phase CountSketch arrays limit memory use and replace large external determinant exchange with bounded collective summation. The implementation on one A100 completes EN-PT2 calculations 8.21-37.53 times faster than Arrow on 64 CPU cores at the same requested statistical standard error, and fixed-workload strong-scaling tests achieve 93.5% parallel efficiency on 16 GPUs across two nodes. Applications to an iron-oxo complex in a (60e, 187o) active space, iron-porphyrin references containing 200 million determinants per state, and neural-network quantum states establish the method's efficiency for perturbative corrections to large reference wave functions in the GPU era.
Second-order Epstein-Nesbet perturbation theory (EN-PT2) is an important tool for improving selected configuration interaction (SCI) energies, but the large external determinant space creates substantial memory and communication costs on GPU clusters. We develop a fully GPU-accelerated implementation of distributed semistochastic EN-PT2, fusing excitation generation, screening, and PT2 weighting in a single kernel traversal. Fixed-size random-phase CountSketch arrays limit memory use and replace large external determinant exchange with bounded collective summation. The implementation on one A100 completes EN-PT2 calculations 8.21-37.53 times faster than Arrow on 64 CPU cores at the same requested statistical standard error, and fixed-workload strong-scaling tests achieve 93.5% parallel efficiency on 16 GPUs across two nodes. Applications to an iron-oxo complex in a (60e, 187o) active space, iron-porphyrin references containing 200 million determinants per state, and neural-network quantum states establish the method's efficiency for perturbative corrections to large reference wave functions in the GPU era.
Second-order Møller–Plesset perturbation theory (MP2) using the Resolution of the Identity approximation (RI-MP2) is a widely used method for computing molecular energies beyond the Hartree–Fock mean-field approximation. However, its high computational cost and lack of efficient algorithms for modern supercomputing architectures limit its applicability to large molecules. In this paper, we present the first distributed-memory many-GPU RI-MP2 algorithm explicitly designed to utilize hundreds of GPU accelerators for every step of the computation. Our novel algorithm achieves near-peak performance on GPU-based supercomputers through the development of a distributed memory algorithm for forming RI-MP2 intermediate tensors with zero internode communication, except for a single O ( N 2 ) asynchronous broadcast, and a distributed memory algorithm for the O ( N 5 ) energy reduction step, capable of sustaining near-peak performance on clusters with several hundred GPUs. Comparative analysis shows our implementation outperforms state-of-the-art quantum chemistry software by over 3.5 times in speed while achieving an 8-fold reduction in computational power consumption. Benchmarking on the Perlmutter supercomputer, our algorithm achieves 11.8 PFLOP/s (83% of peak performance) performing and the RI-MP2 energy calculation on a 314-water cluster with 7850 primary and 30,144 auxiliary basis functions in 4 min on 180 nodes and 720 A100 GPUs. This performance represents a substantial improvement over traditional CPU-based methods, demonstrating significant time-to-solution and power consumption benefits of leveraging modern GPU-accelerated computing environments for quantum chemistry calculations.
We extend the recently proposed heat-bath configuration interaction (HCI) method [Holmes, Tubman, Umrigar, J. Chem. Theory Comput. 2016, 12, 3674], by introducing a semistochastic algorithm for performing multireference Epstein–Nesbet perturbation theory, in order to completely eliminate the severe memory bottleneck of the original method. The proposed algorithm has several attractive features. First, there is no sign problem that plagues several quantum Monte Carlo methods. Second, instead of using Metropolis–Hastings sampling, we use the Alias method to directly sample determinants from the reference wave function, thus avoiding correlations between consecutive samples. Third, in addition to removing the memory bottleneck, semistochastic HCI (SHCI) is faster than the deterministic variant for many systems if a stochastic error of 0.1 mHa is acceptable. Fourth, within the SHCI algorithm one can trade memory for a modest increase in computer time. Fifth, the perturbative calculation is embarrassingly parallel. The SHCI algorithm extends the range of applicability of the original algorithm, allowing us to calculate the correlation energy of very large active spaces. We demonstrate this by performing calculations on several first row dimers including F 2 with an active space of (14e, 108o), Mn-Salen cluster with an active space of (28e, 22o), and Cr 2 dimer with up to a quadruple-ζ basis set with an active space of (12e, 190o). For these systems we were able to obtain better than 1 mHa accuracy with a wall time of merely 55 s, 37 s, and 56 min on 1, 1, and 4 nodes, respectively.
An algorithm is presented for the coupled-cluster singles, doubles, and perturbative triples correction [CCSD(T)] method based on the density fitting or the resolution-of-the-identity (RI) approximation for performing calculations on heterogeneous computing platforms composed of multicore CPUs and graphics processing units (GPUs). The directive-based approach to GPU offloading offered by the OpenMP application programming interface has been employed to adapt the most compute-intensive terms in the RI-CCSD amplitude equations with computational costs scaling as O ( N O 2 N V 4 ), O ( N O 3 N V 3 ), and O ( N O 4 N V 2 ) (where N O and N V denote the numbers of correlated occupied and virtual orbitals, respectively) and the perturbative triples correction to execute on GPU architectures. The pertinent tensor contractions are performed using an accelerated math library such as cuBLAS or hipBLAS. Optimal strategies are discussed for splitting large data arrays into tiles to fit them into the relatively small memory space of the GPUs, while also minimizing the low-bandwidth CPU–GPU data transfers. The performance of the hybrid CPU–GPU RI-CCSD(T) code is demonstrated on pre-exascale supercomputers composed of heterogeneous nodes equipped with NVIDIA Tesla V100 and A100 GPUs and on the world’s first exascale supercomputer named “Frontier”, the nodes of which consist of AMD MI250X GPUs. Speedups within the range 4–8× relative to the recently reported CPU-only algorithm are obtained for the GPU-offloaded terms in the RI-CCSD amplitude equations. Applications to polycyclic aromatic hydrocarbons containing 16–66 carbon atoms demonstrate that the acceleration of the hybrid CPU–GPU code for the perturbative triples correction relative to the CPU-only code increases with the molecule size, attaining a speedup of 5.7× for the largest circumovalene molecule (C 66 H 20 ). The GPU-offloaded code enables the computation of the perturbative triples correction for the C 60 molecule using the cc-pVDZ/aug-cc-pVTZ-RI basis sets in 7 min on Frontier when using 12,288 AMD GPUs with a parallel efficiency of 83.1%.
Computational acceleration of orbital-invariant local correlation methods on graphics processing units (GPUs) has remained largely unexplored due to substantial algorithmic complexities. The runtime efficiency of GPU-implemented local correlation theories can be significantly constrained by the degree of parallelism in the orbital localization procedure, the iterative solution of the local wave function, and the adaptation of CUDA kernels to inherently local or sparse operations. Using the second-order Møller-Plesset perturbation (MP2) theory, we present a multi-GPU implementation for large-scale third-order many-body expansion orbital-specific virtual MP2 (MBE(3)-OSV-MP2) energy calculations. Accordingly, our algorithms and implementation address the GPU parallelization to maximize peak utilization and the parallelism of local MP2 computation in several aspects, including Jacobi-Pipek-Mezey localization, randomized OSV (rOSV) generation, direct MP2 integral regeneration, and CUDA kernel adaptation for local operations. The GPU-based MBE(3)-OSV-MP2 energy computation achieves an effective empirical N1.9 scaling up to (Gly)40/def2-TZVP and 84% parallel efficiency without localization up to 24 GPUs distributed on multiple nodes. The present implementation delivers 40-fold wall time speedup of the canonical RI-MP2 and 10-fold speedup of the central processing units (CPU)-based MBE(3)-OSV-MP2 for (H2O)128/cc-pVDZ and (H2O)190/cc-pVDZ, respectively. A large-scale computation of a 784-atom insulin peptide yields the full MBE(3)-OSV-MP2 energies in 24 min with cc-pVDZ (7571 basis functions) and in 6.4 h with cc-pVTZ (17,448 basis functions) on 8 NVIDIA A800 GPUs. Our work opens new possibilities for fast GPU-based local correlation calculations on real-world macromolecules.
The interplay between advances in stochastic and deterministic algorithms has recently led to development of interesting new selected configuration interaction (SCI) methods for solving the many body Schrödinger equation. The performance of these SCI methods can be greatly improved with a second order perturbation theory (PT2) correction, which is often evaluated in a stochastic or hybrid-stochastic manner. In this work, we present a highly efficient, fully deterministic PT2 algorithm for SCI methods and demonstrate that our approach is orders of magnitude faster than recent proposals for stochastic SCI+PT2. We also show that it is important to have a compact reference SCI wave function, in order to obtain optimal SCI+PT2 energies. This indicates that it advantageous to use accurate search algorithms such as 'ASCI search' rather than more approximate approaches. Our deterministic PT2 algorithm is based on sorting techniques that have been developed for modern computing architectures and is inherently straightforward to use on parallel computing architectures. Related architectures such as GPU implementations can be also used to further increase the efficiency. Overall, we demonstrate that the algorithms presented in this work allow for efficient evaluation of trillions of PT2 contributions with modest computing resources.
AI-driven methods have demonstrated considerable success in tackling the central challenge of accurately solving the Schrödinger equation for complex many-body systems. Among neural network quantum state (NNQS) approaches, the NNQS-SCI (Selected Configuration Interaction) method stands out as a state-of-the-art technique, recognized for its high accuracy and scalability. However, its application to larger systems is severely constrained by a hybrid CPU-GPU architecture. Specifically, centralized CPU-based global de-duplication creates a severe scalability barrier due to communication bottlenecks, while host-resident coupled-configuration generation induces prohibitive computational overheads. We introduce QiankunNet-cuSCI, a fully GPU-accelerated SCI framework designed to overcome these bottlenecks. It first integrates a distributed, load-balanced global de-duplication algorithm to minimize redundancy and communication overhead at scale. To address compute limitations, it employs specialized, fine-grained CUDA kernels for exact coupled configuration generation. Finally, to break the single-GPU memory barrier exposed by this full acceleration, it incorporates a GPU memory-centric runtime featuring GPU-side pooling, streaming mini-batches, and overlapped offloading. This design enables much larger configuration spaces and shifts the bottleneck from host-side limitations back to on-device inference. Our evaluation demonstrates that our work fundamentally expands the scale of solvable problems. On an NVIDIA A100 cluster with 64 GPUs, our work achieves up to 2.32 × end-to-end speedup over the highly-optimized NNQS-SCI baseline while preserving the same chemical accuracy. Furthermore, it demonstrates excellent distributed performance, maintaining over 90% parallel efficiency in strong scaling tests.