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 inter-node 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 eightfold 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 7,850 primary and 30,144 auxiliary basis functions in 4 minutes 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.
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 inter-node 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 eightfold 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 7,850 primary and 30,144 auxiliary basis functions in 4 minutes 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.
This article presents a novel algorithm for the calculation of analytic energy gradients from second-order Møller-Plesset perturbation theory within the Resolution-of-the-Identity approximation (RI-MP2), which is designed to achieve high performance on clusters with multiple graphical processing units (GPUs). The algorithm uses GPUs for all major steps of the calculation, including integral generation, formation of all required intermediate tensors, solution of the Z-vector equation and gradient accumulation. The implementation in the EXtreme Scale Electronic Structure System (EXESS) software package includes a tailored, highly efficient, multistream scheduling system to hide CPU-GPU data transfer latencies and allows nodes with 8 A100 GPUs to operate at over 80% of theoretical peak floating-point performance. Comparative performance analysis shows a significant reduction in computational time relative to traditional multicore CPU-based methods, with our approach achieving up to a 95-fold speedup over the single-node performance of established software such as Q-Chem and ORCA. Additionally, we demonstrate that pairing our implementation with the molecular fragmentation framework in EXESS can drastically lower the computational scaling of RI-MP2 gradient calculations from quintic to subquadratic, enabling further substantial savings in runtime while retaining high numerical accuracy in the resulting gradients.
Second-order Møller-Plesset perturbation theory using the Resolution-of-the-Identity approximation (RI-MP2) is a state-of-the-art approach to accurately estimate many-body electronic correlation effects. This is critical for predicting the physicochemical properties of complex molecular systems; however, the scale of these calculations is limited by their extremely high computational cost. In this paper, a novel many-GPU algorithm and implementation of a molecular-fragmentation-based RI-MP2 method are presented that enable correlated calculations on over 180,000 electrons and 45,000 atoms using up to the entire Summit supercomputer in 12 minutes. The implementation demonstrates remarkable speedups with respect to other current GPU and CPU codes, excellent strong scalability on Summit achieving 89.1% parallel efficiency on 4600 nodes, and shows nearly-ideal weak scaling up to 612 nodes. This work makes feasible ab initio correlated quantum chemistry calculations on significantly larger molecular scales than before on both large supercomputing systems and on commodity clusters, with a potential for major impact on progress in chemical, physical, biological and engineering sciences.
Post-Hartree-Fock methods are wavefunction-based quantum chemistry approaches that explicitly account for the electron correlation beyond the mean-field approximation, allowing accurate determination of electron orbitals and energies in a given system. However, their computational cost scales as the fourth power or higher with respect to the number of atoms, severely limiting their practical applications. This paper proposes an efficient GPU implementation of the Møller-Plesset method, a post-Hartree-Fock approach based on perturbation theory. We focus on the computation of second- and thirdorder perturbation energies (MP2 and MP3) for closed-shell systems, which incorporate electron correlation on top of the Hartree-Fock energy. To accelerate the extensive contraction operations arising in the Møller-Plesset method, our implementation employs a formulation that maximizes the use of highly optimized matrix multiplication routines. Furthermore, we effectively utilize computational capabilities of GPU architecture through a hierarchical reduction using warp-level primitives and shared memory, as well as contiguous coalesced access to matrix elements. Performance evaluation on an NVIDIA A100 GPU shows up to 255.5 and 71.1 times speedups for the MP2 correlation energy computation, compared with the widely used quantum chemistry software package PySCF and its GPU plugin GPU4PySCF, respectively. In the MP3 calculations, our DGEMM-based implementation achieves a maximum speedup of 96.2 times over PySCF.
In Computational Quantum Chemistry, the restricted second-order M0ller-Plesset theory (RMP2) has garnered significant attention for performing high-accuracy computations on closed-shell molecular systems. In particular, the RMP2 with Resolution-of-the-Identity approximation (RI-RMP2) reduces memory usage by approximating the O(M4) tensor with an O(M3) intermediate tensor. However, this approach still requires a high computational cost, making further acceleration necessary. In this paper, we propose a GPU-based implementation to efficiently compute the RI-RMP2 correlation energy. The key idea of our implementation lies in reconstructing the O(M4) tensor via a vendor-optimized matrix multiplication routine, while limiting the required memory usage to O(M3). Furthermore, we formulate an equation to compute the RI-RMP2 correlation energy via a four-term decomposition (summation) with mutually exclusive index ranges, create separate CUDA kernels for each decomposed term, and compute them by different CUDA streams. Our method reduces the required number of threads and maximizes GPU resource utilization. Experimental evaluations conducted on an NVIDIA A100 GPU using relevant molecules with different basis sets demonstrate that our implementation achieves up to a 23.18× speedup over a naive implementation, and up to 165.16× and 4.04× speedups over the state-of-the-art versions of existing quantum chemistry software packages, PySCF and GPU4PySCF, respectively.
The development of novel double-hybrid density functionals offers new levels of accuracy and is leading to fresh insights into the fundamental properties of matter. Hartree-Fock exact exchange and correlated wave function methods, such as second-order Møller-Plesset (MP2) and direct random phase approximation (dRPA), are usually required to build such functionals. Their high computational cost is a concern, and their application to large and periodic systems is, therefore, limited. In this work, low-scaling methods for Hartree-Fock exchange (HFX), SOS-MP2, and direct RPA energy gradients are developed and implemented in the CP2K software package. The use of the resolution-of-the-identity approximation with a short range metric and atom-centered basis functions leads to sparsity, allowing for sparse tensor contractions to take place. These operations are efficiently performed with the newly developed Distributed Block-sparse Tensors (DBT) and Distributed Block-sparse Matrices (DBM) libraries, which scale to hundreds of graphics processing unit (GPU) nodes. The resulting methods, resolution-of-the-identity (RI)-HFX, SOS-MP2, and dRPA, were benchmarked on large supercomputers. They exhibit favorable sub-cubic scaling with system size, good strong scaling performance, and GPU acceleration up to a factor of 3. These developments will allow for double-hybrid level calculations of large and periodic condensed phase systems to take place on a more regular basis.