First Experiences with ab initio Molecular Dynamics on OpenPOWER: The Case of CPMD

Weber, Valéry; Malossi, A. Cristiano I.; Tavernelli, Ivano; Laino, Teodoro; Bekas, Costas; Modani, Manish; Wilner, Nina; Heller, Tom; Curioni, A.

doi:10.1007/978-3-319-46079-6_16

Cited by 2 publications

(3 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The largest speed-up of almost three orders of magnitude was achieved for the routine hnlmat (square symbols), which computes the nonlocal part of the Hamiltonian matrix H (13) stemming from the sum expression in (6). In the original algorithm the respective matrix elements have been updated in the innermost part of a nested loop construct.…”

Section: Further Uspp Specific Optimizationsmentioning

confidence: 99%

“…One of the main computational bottlenecks in CPMD are the 3-d FFTs needed to transform the electronic states (1024 for our 256 water benchmark) from momentum space to real space. The states in real space are needed twice per MD step, once for the calculation of the electronic density (4) and once for the interaction with the effective potential V eff (6). Subsequently, the product of the wave function with V eff is transformed from real to momentum space to calculate the gradients on the electronic states.…”

Section: Batched Multi 3-d Fftmentioning

confidence: 99%

“…A key factor to its success always was the porting and optimization for new compute architectures, like vector machines or the multi-core architecture of the IBM Blue Gene series. More recent porting efforts focus on the support of accelerators, namely GPUs [6], which are one aspect of a general trend in high performance computing. In the last decade, the compute power but also the complexity of the nodes has grown much stronger than the speed of the interconnects.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Integrating State of the Art Compute, Communication, and Autotuning Strategies to Multiply the Performance of the Application Programm CPMD for Ab Initio Molecular Dynamics Simulations

Klöffel¹,

Mathias²,

Meyer³

2020

Preprint

View full text Add to dashboard Cite

We present our recent code modernizations of the of the ab initio molecular dynamics program CPMD (www.cpmd.org) with a special focus on the ultra-soft pseudopotential (USPP) code path. Following the internal instrumentation of CPMD, all time critical routines have been revised to maximize the computational throughput and to minimize the communication overhead for optimal performance. Throughout the program missing hybrid MPI+OpenMP parallelization has been added to optimize scaling. For communication intensive routines, as the multiple distributed 3-d FFTs of the electronic states and distributed matrix-matrix multiplications related to the β-projectors of the pseudopotentials, this MPI+OpenMP parallelization now overlaps computation and communication. The necessary partitioning of the workload is optimized by an auto-tuning algorithm. In addition, the largest global MPI Allreduce operation has been replaced by highly tuned node-local parallelized operations using MPI shared-memory windows to avoid inter-node communication. A batched algorithm for the multiple 3-d FFTs improves the throughput of the MPI Alltoall communication and, thus, the scalability of the implementation, both for USPP and for the frequently used normconserving pseudopotential code path. The enhanced performance and scalability is demonstrated on a mid-sized benchmark system of 256 water molecules and further water systems of from 32 up to 2048 molecules.

show abstract

Section: Further Uspp Specific Optimizationsmentioning

confidence: 99%

Section: Batched Multi 3-d Fftmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation