Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

One programming model to rule them all: High-performance computational chemistry and applications via OpenMP offloading mode

Melisa Alkan

Published Jan 1, 2023• 0 citations
Publisher

Abstract

This dissertation is focused around enabling large-scale quantum chemistry calculations by developing novel algorithms on new heterogeneous hardware architectures. Section 2 contains a study on assessing the importance of many-body dispersion phenomena in molecular clusters and understanding the need to incorporate the phenomena into existing quantum chemistry methods to accurately capture long-range interactions. It is known that highly accurate quantum chemistry methods are computationally expensive; therefore, there is a need for new computer algorithms that will be able to vastly improve the time to solution and enable correlated quantum chemistry calculations on large molecular systems. Sections 3 and 4 focus on the development and implementation of various algorithms that enable quantum chemistry calculations on molecules of unprecedent sizes. Specifically, Section 3 describes a novel CUDA-based algorithm that enables GPU-acceleration of the main parts of the Self-Consistent Field (SCF) cycle, resulting in the fastest SCF code to date. Section 4 introduces a novel OpenMP-based implementation of the Fock build formation, which enables GPU-accelerated calculations on all existing GPU and CPU hardware. Section 5 expands upon the OpenMP work and describes a new GPU-accelerated library named LibERI for evaluating molecular integrals – often the main computational bottleneck in many quantum chemistry methods. LibERI is intended to serve as a library for fast quantum chemistry calculations that can be interfaced with any existing quantum chemistry software packages through the MolSSI Driver Interface (MDI). MDI is a standard interface between scientific computational codes. One of LibERI’s objectives is to seamlessly integrate with a Molecular Mechanics (MM) package using MDI to enable fast QM/MM calculations. This integration would allow for efficient GPU-accelerated treatment of large QM regions via LibERI. Lastly, Section 6 concludes this dissertation with a study that applies existing quantum chemistry methods to unravel the mechanism of a Pd-catalyzed carboacylation reaction involving the activation of ester C-O bonds.

DOI
10.31274/td-20240329-562

Related papers

The design and development of GPU accelerated algorithms for ab initio integrals and integral derivatives illustrated on ab initio quantum and hybrid QM/MM dynamics

C. Alicia Renison

Graphical Processing Units (GPUs) are highly parallel, programmable accelerators boasting high peak floating point performance. Over the last couple of years the use of GPUs for general purpose computing have revolutionized quantum chemistry. The computational bottleneck in an ab-initio quantum method is the calculation of a large number of twoelectron integrals. To date, a number of GPU accelerated two-electron integral implementations have been developed significantly improving the performance of a static quantum mechanical (QM) calculation. However, when performing an ab-initio QM gradient calculation, optimization, QM or Hybrid Quantum Mechanical/Molecular Mechanical (QM/MM) dynamics simulation the twoelectron integral derivatives arise as an additional bottleneck. Hybrid QM/MM methods particularly dynamics methods are commonly used to study large chemical/biological systems. These methods are a popular choice used for studying reaction mechanisms, conformational and configurational structures important in glycobiology. Usually a semiempirical QM method is used however, these have shown variable accuracy for the study of carbohydrate conformation and prevents an analytical investigation of electronic structure. The use of a higher level of theory, such as an ab initio method, is desirable however this comes at a much greater computational cost. Together with the bottlenecks above this cost results from the polarization of the QM region within an electrostatic embedding scheme, which requires the calculation of a large number of one-electron integral derivatives. Thus for QM/MM calculations the one-electron integral derivatives becomes a third bottleneck together with the two mentioned above. Recently, the above bottlenecks have become popular GPU acceleration targets. This thesis describes an extension of the GPU based Quantum Supercharger Library (QSL) to perform the above calculations/simulations. In contrast to GPU packages developed from the ground up, the QSL is a library of routines aimed at accelerating legacy codes, such as GAMESS-UK, GAMESS-US and NWChem, used in electronic structure calculations. Algorithms are presented for accelerating the one- and two-electron integral derivatives on a GPU. In addition to the derivatives, the one-electron integral calculation was ported to the GPU in order to remove a small additional cost arising for large QM/MM systems. Furthermore, the use of automatic code generation for generating GPU kernels was explored and compared to the original approach. Using the QSL library implemented in GAMESS-UK several benchmark calculations and simulations were performed. These were performed in double precision on a single GPU (Kepler K20) and compared to a single CPU using the 6-31G basis set. A speedup of up to 9.3X is achieved for an ab initio gradient calculation compared to the CPU running an optimized serial version of GAMESS-UK using the Schlegel method. For a single point QM/MM calculation of cellobiose in different sized water spheres (3267-24843 point charges) speedups of between 13X and 34X is achieved. QSL/GAMESS-UK coupled to the CHARMM molecular dynamics package was then used in order to perform accelerated molecular dynamics simulations. Benchmark QM and QM/MM molecular dynamics simulations were performed on cellobiose in vacuo and in a water sphere (45 QM atoms and 24843 point charges respectively). The QSL is able to perform 9.7 ps/day of ab initio QM dynamics and 6.4 ps/day of QM/MM dynamics. Testing of the auto-generated version of the QSL showed better performance for lower angular momentum classes but reduced performance for higher angular momentum classes. The efficiency of the integral and derivative routines within the QSL library was tested on a computationally intense realistic glycobiological condensed phase free energy computation. Ab-initio pucker free energy surfaces/volumes of !-Ribofuranose and !-Glucopyranose in vacuum and in water were computed. These are the first converged Hartree-Fock/MM free energy simulations for these carbohydrates performed in solution. The value of the ab initio Free energy surfaces/volumes was demonstrated through an analysis of solvent polarization effects that are evident through a comparison of the vacuum and solution minimum free energy pathways. In particular the water structure around the pucker conformers, analysis of the primary alcohol distribution as well as analysis of the electronic structure of these conformers reveals that water significantly affects the free energy pathways, primary alcohol distribution and the barriers for inter-conversion between pucker conformers.

•Jan 1, 2016•0 citations

Many-body quantum chemistry on graphics processing units

A. Eugene DePrince, Jeff R. Hammond, Stephen K. Gray

Heterogeneous nodes composed of a multicore CPU and at least one graphics processing unit (GPU) are increasingly common in high-performance scientific computing, and significant programming effort is currently being undertaken to port existing scientific algorithms to these unique architectures. We present implementations for two many-body quantum chemistry methods on heterogeneous nodes: the coupled-cluster with single and double excitations (CCSD) and time-dependent configuration interaction with single and double excitations (TD-CISD) methods. Both methods can be implemented on a computer as a series of dense matrix-matrix multiplications, operations that GPUs are particularly adept at performing. The GPU-accelerated CCSD algorithm is as much as 4.3 times faster than the corresponding CPU algorithm and 9.7 times faster than the algorithm in the Molpro package. The TD-CISD algorithm is accelerated by as much as a factor of 3.9 by GPUs. Enhanced performance is achieved by overlapping CPU and GPU computations.

•Jan 1, 2011•1 citation

High-Performance, Accurate Large-Scale Quantum Chemistry Calculations on GPU Supercomputers using Coulomb-Perturbed Fragmentation

Fazeleh Kazemian, Jorge Luis Galvez Vallejo, Giuseppe M. J. Barca

Predicting the chemico-physical properties of large molecular systems is a formidable challenge in chemistry and materials science. Traditional quantum mechanical methods, while accurate, have impractical scaling for large molecules with thousands of atoms, which are crucial in the development of novel therapeutics, catalysts, and nanomaterials. To address this, molecular fragmentation algorithms have been proposed to improve scalability and enable extensive parallelism. In this article, we introduce a significant enhancement to the Fragment Molecular Orbital (FMO) method, termed the Coulomb-Perturbed Fragmentation (CPF) method. CPF incorporates algorithmic improvements and implementation enhancements to optimize performance on heterogeneous computing systems equipped with a large number of GPUs. Key developments include a significant simplification of iteratitve self-consistent field (SCF) algorithm, advanced data management through a one-sided communication model, topology-aware optimizations, and a hybrid communication strategy for intra-group exchanges. Moreover, CPF integrates a distributed dynamic multi-layer load balancing scheme to optimise fragment distribution and workload management across nodes and GPUs. Performance evaluations on a 420-atom benzene molecule system comprising 35 fragments reveal that CPF outperforms existing GPU/CPU-based FMO algorithms in both efficiency and accuracy. When deployed on the Gadi supercomputer, CPF achieves over 97% parallel efficiency on 20 nodes, with scalability maintaining above 98% and 90% efficiency in weak-scaling tests for smaller and larger systems, respectively. Notably, CPF matches or exceeds the computational accuracy of conventional FMO methods, marking a substantial progress in the field of computational chemistry for fragmentation-based large-scale molecular modelling.

Proceedings of the International Conference on Parallel Processing•Aug 8, 2024•6 citations

Accelerating fundamental components of a quantum chemical multilevel workflow by means of GPU offloading

Pit Steinbach

The dynamic rearrangement of a molecule at finite temperature is not captured by a single static input geometry to a quantum chemical calculation. Multilevel workflows have established themselves as standard procedures to capture the dependence of the molecule’s configuration on computed properties. Each step within this workflow uses an electronic structure theory method that features the optimal computational time to accuracy ratio. For exploratory steps the computational expense of a method needs to be minimal as several thousand energy and nuclear gradient evaluations are needed. Semiempirical tight binding methods, such as GFN2-xTB are often used for such purposes. However, these methods usually lack the accuracy to rank conformers relative to each other within chemical accuracy (2 kcal/mol). Therefore, so-called “low-cost” DFT methods are commonly used for geometry optimizations and conformer ranking. The combination of (semi)empirical correction models with small basis sets leads to good accuracy at moderate computational cost. However, even these highly efficient methods reach the limits of feasibility for very large systems (> 200 atoms). This work proposes solutions to partially alleviate those restrictions and allow routine calculations of these system sizes outside of large high-performance computing facilities. The key to unlock those calculations is to combine tailored algorithmic developments with consumer-grade GPUs. A GPU-enabled linear algebra library is developed to accelerate further development and make the entry to GPU-accelerated software development easier. To demonstrate the ease of implementation, the addition of Mulliken-approximated exchange to GFN2-xTB will be shown. In most semiempirical methods, the general eigenvalue solver algorithm becomes the computational bottleneck, dominating the walltime with increasing system size. A density matrix purification algorithm is tailored to give up to six-fold acceleration of GFN2-xTB calculations using a mixed-precision approach. Lastly, 3c-methods are added to the GPU-accelerated TeraChem software package. In combination with the “mixed-precision” scheme for computing the electron repulsion integrals, this addition decreases the runtime of DFT calculations by up to an order of magnitude on mid-level consumer hardware. The overall goal of these projects is to make GPU-accelerated calculations in quantum chemistry more available. This achieve by exposing algorithmic developments that are performant on reasonably priced hardware and making GPU computing more accessible to non-experts.

•Jan 1, 2026•0 citations

Development of Efficient Algorithms for Quantum Chemistry Calculations of Large Molecules

和也 石村, カズヤ イシムラ, Kazuya Ishimura

Quantum chemistry plays an important role in elucidating molecular geometries, electronic states, and reaction mechanisms, because of the developments of a variety of theoretical methods, such as Hartree-Fock (HF), Møler-Plesset (MP) perturbation, configuration interaction (CI), coupled-cluster (CC), and density functional theory (DFT) methods. Electronic structure calculations have been carried out by not only theoretical chemists but also experimental chemists. DFT is currently most widely used to investigate large molecules in the ground state as well as small molecules because of the low computational cost. However, the generally used functionals fail to describe correctly non-covalent interactions that are important for host-guest molecules, self-assembly, and molecular recognition, and they tend to underestimate reaction barriers. Many attempts have been made to develop new functionals and add semiempirical or empirical correction terms to standard functionals, but no generally accepted DFT method has emerged yet. Second-order Møler-Plesset perturbation theory (MP2) is the simplest method that includes electron correlation important for non-covalent interactions and reaction barriers nonempirically. However, the computational cost of MP2 is considerably higher than that of DFT. In addition, much larger sizes of fast memory and hard disk are required in MP2 calculations. These make MP2 calculations increasingly difficult for larger molecules. Since workstation or personal computer (PC) clusters have become popular for quantum chemistry calculations, an efficient parallel calculation is a solution of the problem. Therefore, new parallel algorithms for MP2 energy and gradient calculations are presented in this thesis. Furthermore, an efficient algorithm for the generation of two-electron repulsion integrals (ERIs) which is important in quantum chemistry calculations is also presented. For the calculations of excited states, different approaches are required: for example, CI, multi-configuration self-consistent field (MCSCF), time-dependent DFT (TDDFT), and symmetry adapted cluster (SAC)/SAC-CI methods. One of the most accurate methods is SAC/SAC-CI, as demonstrated for many molecules. In this thesis, SAC/SAC-CI calculations of ground, ionized, and excited states are presented. This thesis consists of five chapters: a new algorithm of two-electron repulsion integral calculations (Chapter I), a new parallel algorithm of MP2 energy calculations (Chapter II), a new parallel algorithm of MP2 energy gradient calculations (Chapter III), applications of MP2 calculations (Chapter IV), and SAC/SAC-CI calculations of ionized and excited states (Chapter V). In quantum chemistry calculations, the generation of ERIs is one of the most basic subjects and is the most time-consuming step especially in direct SCF calculations. Many algorithms have been developed to reduce the computational cost. In Pople-Hehre algorithm, Cartesian axes are rotated to make several coordinate components zero or constant, so that these components are skipped in the generation of ERIs. In McMurchie-Davidson algorithm, ERIs are generated from (ss|ss) type integrals using a recurrence relation derived from Hermite polynomials. By combining these two algorithms, a new algorithm is developed in Chapter I. The results show that the new algorithm reduces the computational cost by 10 - 40%, as compared with the original algorithms. It is notable that the generation of ERIs including d functions is considerably fast. The program implemented officially in GAMESS in 2004 has been used all over the world. In quantum mechanics, perturbation methods can be used for adding corrections to reference solutions. In the MP perturbation method, a sum over Fock operators is used as the reference term, and the exact two-electron repulsion operator minus twice the average two-electron repulsion operator is used as the perturbation term. It is the advantage that the MP perturbation method is size consistent and size extensive, unlike truncated CI methods. The zero-order wave function is the HF Slater determinant, and the zero-order energy is expressed as a sum of occupied molecular orbital (MO) energies. The first-order perturbation is the correction for the overcounting of two-electron repulsions at zero-order, and the first-order energy corresponds to the HF energy. The MP correlation starts at second-order. In general, second-order (MP2) accounts for 80 - 90% of electron correlation. Therefore, MP2 is focused in this thesis since it is applicable to large molecules with considerable reliability and low computational cost. The formal computational scaling of MP2 energy calculations with respect to molecular size is fifth order, much higher than that of DFT energy calculations. Therefore, less expensive methods, such as Local MP2, density fitting (resolution of identity, RI) MP2, and Laplace Transform MP2, have been developed. However, all of these methods include approximations or cut-offs that need to be checked against full MP2 energies. An alternative approach to reduce the computational cost is to parallelize MP2 energy calculations. A number of papers on parallel MP2 energy calculations have been published. Almost all of them are based on simple parallelization methods that distribute only atomic orbital (AO) or MO indices to each processor. These methods have a disadvantage since intermediate integrals are broadcasted to all CPUs or the same AO ERIs are generated in all processors. Baker and Pulay developed a new parallel algorithm using SaebøAlmlöf integral transformation method. This algorithm parallelizes the first half transformation by AO indices and the second half transformation by MO indices. The advantages are that the total amount of network communication is independent of the number of processors and the AO integrals are generated only once. The disadvantage is the I/O overhead for the sorting of half-transformed integrals. A new parallel algorithm for MP2 energy calculations based on the two-step parallelization idea is presented in Chapter II. In this algorithm, AO indices are distributed in the AO integral generation and the first three quarter transformation, and MO indices are distributed in the last quarter transformation and MP2 energy calculation. Because the algorithm makes the sorting of intermediate integrals very simple, the parallel efficiency is highly improved and the I/O overhead is removed. Furthermore, the algorithm reduces highly the floating point operation (FLOP) count as well as the required memory and hard disk space, in comparison with other algorithms. Test calculations of taxol (C47H51NO14) and luciferin (C11H8N2O3S2) were performed on a cluster of Pentium 4 computers connected by gigabit Ethernet. The parallel scaling of the developed code is excellent up to the largest number of processors we have tested. For instance, the elapsed time for the MP2 energy calculations on 16 processors is on average 15.4 times faster than that on the single-processor. Determination of molecular geometries and reaction paths is a fundamental task in quantum chemistry and requires energy gradients with respect to nuclear coordinates. In Chapter III, a new parallel algorithm for MP2 energy gradient calculations is presented. The algorithm consists of 5 steps, the integral transformation, the MP2 amplitude calculation, the MP2 Lagrangian calculation, the coupled-perturbed HF calculation, and the integral derivative calculation. All steps are parallelized by distributing AO or MO indices. The algorithm also reduces the FLOP count, the required memory, and hard disk space. Test calculations of MP2 energy gradients were performed for taxol and luciferin on a cluster of Pentium 4 computers. The speedups are very good up to 80 CPU cores we have tested. For instance, the speedup ratios are 28.2 - 33.0 on 32 processors, corresponding to 88% - 103% of linear speedup. This indicates the high parallel efficiency of the present algorithm. The calculation of taxol with 6-31G(d) (1032 contracted basis functions) finishes within 2 hours on 32 processors, which requires only 1.8GB memory and 13.4GB hard disk per processor. Therefore, geometry optimization of molecules with 1000 basis functions can be easily performed using standard PC clusters. In Chapter IV, several applications of MP2 are performed using the program developed in Chapters II and III. Some molecules that DFT cannot treat well are optimized at the MP2 level. Geometry optimization is also carried out using the spin-component scaled (SCS) MP2 method. In this method, a different scaling is employed for the same and opposite spin components of the MP2 energy, so that SCS-MP2 performs as well as the much more costly CCSD(T) method at a high level of theory. SAC theory is developed for ground states and based on CC theory that describes higher-order electron correlation. The main factor of electron correlation is collisions of two electrons. In CC theory, most collisions of four electrons can be taken in as the product of collisions of tw

•Sep 28, 2007•0 citations

Accelerating Coupled-Cluster Calculations with GPUs: An Implementation of the Density-Fitted CCSD(T) Approach for Heterogeneous Computing Architectures Using OpenMP Directives

Dipayan Datta, Mark S. Gordon

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%.

Journal of Chemical Theory and Computation•Oct 25, 2023•21 citations
This dissertation is focused around enabling large-scale quantum chemistry calculations by developing novel algorithms on new heterogeneous hardware architectures. Section 2 contains a study on assessing the importance of many-body dispersion phenomena in molecular clusters and understanding the need to incorporate the phenomena into existing quantum chemistry methods to accurately capture long-range interactions. It is known that highly accurate quantum chemistry methods are computationally expensive; therefore, there is a need for new computer algorithms that will be able to vastly improve the time to solution and enable correlated quantum chemistry calculations on large molecular systems. Sections 3 and 4 focus on the development and implementation of various algorithms that enable quantum chemistry calculations on molecules of unprecedent sizes. Specifically, Section 3 describes a novel CUDA-based algorithm that enables GPU-acceleration of the main parts of the Self-Consistent Field (SCF) cycle, resulting in the fastest SCF code to date. Section 4 introduces a novel OpenMP-based implementation of the Fock build formation, which enables GPU-accelerated calculations on all existing GPU and CPU hardware. Section 5 expands upon the OpenMP work and describes a new GPU-accelerated library named LibERI for evaluating molecular integrals – often the main computational bottleneck in many quantum chemistry methods. LibERI is intended to serve as a library for fast quantum chemistry calculations that can be interfaced with any existing quantum chemistry software packages through the MolSSI Driver Interface (MDI). MDI is a standard interface between scientific computational codes. One of LibERI’s objectives is to seamlessly integrate with a Molecular Mechanics (MM) package using MDI to enable fast QM/MM calculations. This integration would allow for efficient GPU-accelerated treatment of large QM regions via LibERI. Lastly, Section 6 concludes this dissertation with a study that applies existing quantum chemistry methods to unravel the mechanism of a Pd-catalyzed carboacylation reaction involving the activation of ester C-O bonds.
Graphical Processing Units (GPUs) are highly parallel, programmable accelerators boasting high peak floating point performance. Over the last couple of years the use of GPUs for general purpose computing have revolutionized quantum chemistry. The computational bottleneck in an ab-initio quantum method is the calculation of a large number of twoelectron integrals. To date, a number of GPU accelerated two-electron integral implementations have been developed significantly improving the performance of a static quantum mechanical (QM) calculation. However, when performing an ab-initio QM gradient calculation, optimization, QM or Hybrid Quantum Mechanical/Molecular Mechanical (QM/MM) dynamics simulation the twoelectron integral derivatives arise as an additional bottleneck. Hybrid QM/MM methods particularly dynamics methods are commonly used to study large chemical/biological systems. These methods are a popular choice used for studying reaction mechanisms, conformational and configurational structures important in glycobiology. Usually a semiempirical QM method is used however, these have shown variable accuracy for the study of carbohydrate conformation and prevents an analytical investigation of electronic structure. The use of a higher level of theory, such as an ab initio method, is desirable however this comes at a much greater computational cost. Together with the bottlenecks above this cost results from the polarization of the QM region within an electrostatic embedding scheme, which requires the calculation of a large number of one-electron integral derivatives. Thus for QM/MM calculations the one-electron integral derivatives becomes a third bottleneck together with the two mentioned above. Recently, the above bottlenecks have become popular GPU acceleration targets. This thesis describes an extension of the GPU based Quantum Supercharger Library (QSL) to perform the above calculations/simulations. In contrast to GPU packages developed from the ground up, the QSL is a library of routines aimed at accelerating legacy codes, such as GAMESS-UK, GAMESS-US and NWChem, used in electronic structure calculations. Algorithms are presented for accelerating the one- and two-electron integral derivatives on a GPU. In addition to the derivatives, the one-electron integral calculation was ported to the GPU in order to remove a small additional cost arising for large QM/MM systems. Furthermore, the use of automatic code generation for generating GPU kernels was explored and compared to the original approach. Using the QSL library implemented in GAMESS-UK several benchmark calculations and simulations were performed. These were performed in double precision on a single GPU (Kepler K20) and compared to a single CPU using the 6-31G basis set. A speedup of up to 9.3X is achieved for an ab initio gradient calculation compared to the CPU running an optimized serial version of GAMESS-UK using the Schlegel method. For a single point QM/MM calculation of cellobiose in different sized water spheres (3267-24843 point charges) speedups of between 13X and 34X is achieved. QSL/GAMESS-UK coupled to the CHARMM molecular dynamics package was then used in order to perform accelerated molecular dynamics simulations. Benchmark QM and QM/MM molecular dynamics simulations were performed on cellobiose in vacuo and in a water sphere (45 QM atoms and 24843 point charges respectively). The QSL is able to perform 9.7 ps/day of ab initio QM dynamics and 6.4 ps/day of QM/MM dynamics. Testing of the auto-generated version of the QSL showed better performance for lower angular momentum classes but reduced performance for higher angular momentum classes. The efficiency of the integral and derivative routines within the QSL library was tested on a computationally intense realistic glycobiological condensed phase free energy computation. Ab-initio pucker free energy surfaces/volumes of !-Ribofuranose and !-Glucopyranose in vacuum and in water were computed. These are the first converged Hartree-Fock/MM free energy simulations for these carbohydrates performed in solution. The value of the ab initio Free energy surfaces/volumes was demonstrated through an analysis of solvent polarization effects that are evident through a comparison of the vacuum and solution minimum free energy pathways. In particular the water structure around the pucker conformers, analysis of the primary alcohol distribution as well as analysis of the electronic structure of these conformers reveals that water significantly affects the free energy pathways, primary alcohol distribution and the barriers for inter-conversion between pucker conformers.
Heterogeneous nodes composed of a multicore CPU and at least one graphics processing unit (GPU) are increasingly common in high-performance scientific computing, and significant programming effort is currently being undertaken to port existing scientific algorithms to these unique architectures. We present implementations for two many-body quantum chemistry methods on heterogeneous nodes: the coupled-cluster with single and double excitations (CCSD) and time-dependent configuration interaction with single and double excitations (TD-CISD) methods. Both methods can be implemented on a computer as a series of dense matrix-matrix multiplications, operations that GPUs are particularly adept at performing. The GPU-accelerated CCSD algorithm is as much as 4.3 times faster than the corresponding CPU algorithm and 9.7 times faster than the algorithm in the Molpro package. The TD-CISD algorithm is accelerated by as much as a factor of 3.9 by GPUs. Enhanced performance is achieved by overlapping CPU and GPU computations.
Predicting the chemico-physical properties of large molecular systems is a formidable challenge in chemistry and materials science. Traditional quantum mechanical methods, while accurate, have impractical scaling for large molecules with thousands of atoms, which are crucial in the development of novel therapeutics, catalysts, and nanomaterials. To address this, molecular fragmentation algorithms have been proposed to improve scalability and enable extensive parallelism. In this article, we introduce a significant enhancement to the Fragment Molecular Orbital (FMO) method, termed the Coulomb-Perturbed Fragmentation (CPF) method. CPF incorporates algorithmic improvements and implementation enhancements to optimize performance on heterogeneous computing systems equipped with a large number of GPUs. Key developments include a significant simplification of iteratitve self-consistent field (SCF) algorithm, advanced data management through a one-sided communication model, topology-aware optimizations, and a hybrid communication strategy for intra-group exchanges. Moreover, CPF integrates a distributed dynamic multi-layer load balancing scheme to optimise fragment distribution and workload management across nodes and GPUs. Performance evaluations on a 420-atom benzene molecule system comprising 35 fragments reveal that CPF outperforms existing GPU/CPU-based FMO algorithms in both efficiency and accuracy. When deployed on the Gadi supercomputer, CPF achieves over 97% parallel efficiency on 20 nodes, with scalability maintaining above 98% and 90% efficiency in weak-scaling tests for smaller and larger systems, respectively. Notably, CPF matches or exceeds the computational accuracy of conventional FMO methods, marking a substantial progress in the field of computational chemistry for fragmentation-based large-scale molecular modelling.
The dynamic rearrangement of a molecule at finite temperature is not captured by a single static input geometry to a quantum chemical calculation. Multilevel workflows have established themselves as standard procedures to capture the dependence of the molecule’s configuration on computed properties. Each step within this workflow uses an electronic structure theory method that features the optimal computational time to accuracy ratio. For exploratory steps the computational expense of a method needs to be minimal as several thousand energy and nuclear gradient evaluations are needed. Semiempirical tight binding methods, such as GFN2-xTB are often used for such purposes. However, these methods usually lack the accuracy to rank conformers relative to each other within chemical accuracy (2 kcal/mol). Therefore, so-called “low-cost” DFT methods are commonly used for geometry optimizations and conformer ranking. The combination of (semi)empirical correction models with small basis sets leads to good accuracy at moderate computational cost. However, even these highly efficient methods reach the limits of feasibility for very large systems (> 200 atoms). This work proposes solutions to partially alleviate those restrictions and allow routine calculations of these system sizes outside of large high-performance computing facilities. The key to unlock those calculations is to combine tailored algorithmic developments with consumer-grade GPUs. A GPU-enabled linear algebra library is developed to accelerate further development and make the entry to GPU-accelerated software development easier. To demonstrate the ease of implementation, the addition of Mulliken-approximated exchange to GFN2-xTB will be shown. In most semiempirical methods, the general eigenvalue solver algorithm becomes the computational bottleneck, dominating the walltime with increasing system size. A density matrix purification algorithm is tailored to give up to six-fold acceleration of GFN2-xTB calculations using a mixed-precision approach. Lastly, 3c-methods are added to the GPU-accelerated TeraChem software package. In combination with the “mixed-precision” scheme for computing the electron repulsion integrals, this addition decreases the runtime of DFT calculations by up to an order of magnitude on mid-level consumer hardware. The overall goal of these projects is to make GPU-accelerated calculations in quantum chemistry more available. This achieve by exposing algorithmic developments that are performant on reasonably priced hardware and making GPU computing more accessible to non-experts.
Quantum chemistry plays an important role in elucidating molecular geometries, electronic states, and reaction mechanisms, because of the developments of a variety of theoretical methods, such as Hartree-Fock (HF), Møler-Plesset (MP) perturbation, configuration interaction (CI), coupled-cluster (CC), and density functional theory (DFT) methods. Electronic structure calculations have been carried out by not only theoretical chemists but also experimental chemists. DFT is currently most widely used to investigate large molecules in the ground state as well as small molecules because of the low computational cost. However, the generally used functionals fail to describe correctly non-covalent interactions that are important for host-guest molecules, self-assembly, and molecular recognition, and they tend to underestimate reaction barriers. Many attempts have been made to develop new functionals and add semiempirical or empirical correction terms to standard functionals, but no generally accepted DFT method has emerged yet. Second-order Møler-Plesset perturbation theory (MP2) is the simplest method that includes electron correlation important for non-covalent interactions and reaction barriers nonempirically. However, the computational cost of MP2 is considerably higher than that of DFT. In addition, much larger sizes of fast memory and hard disk are required in MP2 calculations. These make MP2 calculations increasingly difficult for larger molecules. Since workstation or personal computer (PC) clusters have become popular for quantum chemistry calculations, an efficient parallel calculation is a solution of the problem. Therefore, new parallel algorithms for MP2 energy and gradient calculations are presented in this thesis. Furthermore, an efficient algorithm for the generation of two-electron repulsion integrals (ERIs) which is important in quantum chemistry calculations is also presented. For the calculations of excited states, different approaches are required: for example, CI, multi-configuration self-consistent field (MCSCF), time-dependent DFT (TDDFT), and symmetry adapted cluster (SAC)/SAC-CI methods. One of the most accurate methods is SAC/SAC-CI, as demonstrated for many molecules. In this thesis, SAC/SAC-CI calculations of ground, ionized, and excited states are presented. This thesis consists of five chapters: a new algorithm of two-electron repulsion integral calculations (Chapter I), a new parallel algorithm of MP2 energy calculations (Chapter II), a new parallel algorithm of MP2 energy gradient calculations (Chapter III), applications of MP2 calculations (Chapter IV), and SAC/SAC-CI calculations of ionized and excited states (Chapter V). In quantum chemistry calculations, the generation of ERIs is one of the most basic subjects and is the most time-consuming step especially in direct SCF calculations. Many algorithms have been developed to reduce the computational cost. In Pople-Hehre algorithm, Cartesian axes are rotated to make several coordinate components zero or constant, so that these components are skipped in the generation of ERIs. In McMurchie-Davidson algorithm, ERIs are generated from (ss|ss) type integrals using a recurrence relation derived from Hermite polynomials. By combining these two algorithms, a new algorithm is developed in Chapter I. The results show that the new algorithm reduces the computational cost by 10 - 40%, as compared with the original algorithms. It is notable that the generation of ERIs including d functions is considerably fast. The program implemented officially in GAMESS in 2004 has been used all over the world. In quantum mechanics, perturbation methods can be used for adding corrections to reference solutions. In the MP perturbation method, a sum over Fock operators is used as the reference term, and the exact two-electron repulsion operator minus twice the average two-electron repulsion operator is used as the perturbation term. It is the advantage that the MP perturbation method is size consistent and size extensive, unlike truncated CI methods. The zero-order wave function is the HF Slater determinant, and the zero-order energy is expressed as a sum of occupied molecular orbital (MO) energies. The first-order perturbation is the correction for the overcounting of two-electron repulsions at zero-order, and the first-order energy corresponds to the HF energy. The MP correlation starts at second-order. In general, second-order (MP2) accounts for 80 - 90% of electron correlation. Therefore, MP2 is focused in this thesis since it is applicable to large molecules with considerable reliability and low computational cost. The formal computational scaling of MP2 energy calculations with respect to molecular size is fifth order, much higher than that of DFT energy calculations. Therefore, less expensive methods, such as Local MP2, density fitting (resolution of identity, RI) MP2, and Laplace Transform MP2, have been developed. However, all of these methods include approximations or cut-offs that need to be checked against full MP2 energies. An alternative approach to reduce the computational cost is to parallelize MP2 energy calculations. A number of papers on parallel MP2 energy calculations have been published. Almost all of them are based on simple parallelization methods that distribute only atomic orbital (AO) or MO indices to each processor. These methods have a disadvantage since intermediate integrals are broadcasted to all CPUs or the same AO ERIs are generated in all processors. Baker and Pulay developed a new parallel algorithm using SaebøAlmlöf integral transformation method. This algorithm parallelizes the first half transformation by AO indices and the second half transformation by MO indices. The advantages are that the total amount of network communication is independent of the number of processors and the AO integrals are generated only once. The disadvantage is the I/O overhead for the sorting of half-transformed integrals. A new parallel algorithm for MP2 energy calculations based on the two-step parallelization idea is presented in Chapter II. In this algorithm, AO indices are distributed in the AO integral generation and the first three quarter transformation, and MO indices are distributed in the last quarter transformation and MP2 energy calculation. Because the algorithm makes the sorting of intermediate integrals very simple, the parallel efficiency is highly improved and the I/O overhead is removed. Furthermore, the algorithm reduces highly the floating point operation (FLOP) count as well as the required memory and hard disk space, in comparison with other algorithms. Test calculations of taxol (C47H51NO14) and luciferin (C11H8N2O3S2) were performed on a cluster of Pentium 4 computers connected by gigabit Ethernet. The parallel scaling of the developed code is excellent up to the largest number of processors we have tested. For instance, the elapsed time for the MP2 energy calculations on 16 processors is on average 15.4 times faster than that on the single-processor. Determination of molecular geometries and reaction paths is a fundamental task in quantum chemistry and requires energy gradients with respect to nuclear coordinates. In Chapter III, a new parallel algorithm for MP2 energy gradient calculations is presented. The algorithm consists of 5 steps, the integral transformation, the MP2 amplitude calculation, the MP2 Lagrangian calculation, the coupled-perturbed HF calculation, and the integral derivative calculation. All steps are parallelized by distributing AO or MO indices. The algorithm also reduces the FLOP count, the required memory, and hard disk space. Test calculations of MP2 energy gradients were performed for taxol and luciferin on a cluster of Pentium 4 computers. The speedups are very good up to 80 CPU cores we have tested. For instance, the speedup ratios are 28.2 - 33.0 on 32 processors, corresponding to 88% - 103% of linear speedup. This indicates the high parallel efficiency of the present algorithm. The calculation of taxol with 6-31G(d) (1032 contracted basis functions) finishes within 2 hours on 32 processors, which requires only 1.8GB memory and 13.4GB hard disk per processor. Therefore, geometry optimization of molecules with 1000 basis functions can be easily performed using standard PC clusters. In Chapter IV, several applications of MP2 are performed using the program developed in Chapters II and III. Some molecules that DFT cannot treat well are optimized at the MP2 level. Geometry optimization is also carried out using the spin-component scaled (SCS) MP2 method. In this method, a different scaling is employed for the same and opposite spin components of the MP2 energy, so that SCS-MP2 performs as well as the much more costly CCSD(T) method at a high level of theory. SAC theory is developed for ground states and based on CC theory that describes higher-order electron correlation. The main factor of electron correlation is collisions of two electrons. In CC theory, most collisions of four electrons can be taken in as the product of collisions of tw
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%.