Accelerating Large-Scale Excited-State GW Calculations on Leadership HPC Systems

Ben, Mauro Del; Yang, Charlene; Li, Zhenglu; Jornada, Felipe H. da; Louie, Steven G.; Deslippe, Jack

doi:10.1109/sc41405.2020.00008

Cited by 28 publications

(29 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…This operation appears very often as a building block of linear algebra calculations like computation of the wavefunction overlap matrix, S = ψ † ψ, in quantum mechanics. [26,27] In those applications, ψ may consist of hundreds of thousands of tiles like A and B, giving rise to millions of tasks. In these workflows, however, neglect the hypothetical communication steps needed to sum the results in computing S. Each task management system was benchmarked using a weak-scaling methodology -where number of tasks scaled with processors.…”

Section: Evaluation Methodsmentioning

confidence: 99%

Three Practical Workflow Schedulers for Easy Maximum Parallelism

Rogers¹

2021

Preprint

View full text Add to dashboard Cite

Runtime scheduling and workflow systems are an increasingly popular algorithmic component in HPC because they allow full system utilization with relaxed synchronization requirements. There are so many special-purpose tools for task scheduling, one might wonder why more are needed. Use cases seen on the Summit supercomputer needed better integration with MPI and greater flexibility in job launch configurations. Preparation, execution, and analysis of computational chemistry simulations at the scale of tens of thousands of processors revealed three distinct workflow patterns. A separate job scheduler was implemented for each one using extremely simple and robust designs: file-based, task-list based, and bulk-synchronous. Comparing to existing methods shows unique benefits of this work, including simplicity of design, suitability for HPC centers, short startup time, and well-understood per-task overhead. All three new tools have been shown to scale to full utilization of Summit, and have been made publicly available with tests and documentation. This work presents a complete characterization of the minimum effective task granularity for efficient scheduler usage scenarios. These schedulers have the same bottlenecks, and hence similar task granularities as those reported for existing tools following comparable paradigms.

show abstract

Section: Evaluation Methodsmentioning

confidence: 99%

Three Practical Workflow Schedulers for Easy Maximum Parallelism

Rogers¹

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…This work involves three major computation steps: 1) DFPT calculation using the ABINIT code [29]; 2) GWPT calculation using the BerkeleyGW code [19,20,30,31]; and 3) Wannier interpolation using the EPW code [32,59]. We have developed wrappers to link the three codes.…”

Section: Computational Detailsmentioning

confidence: 99%

Unmasking the Origin of Kinks in the Photoemission Spectra of Cuprate Superconductors

Chan

et al. 2021

Phys. Rev. Lett.

Self Cite

View full text Add to dashboard Cite

The origin of a ubiquitous bosonic coupling feature in the photoemission spectra of high-Tc cuprates, an energy-momentum dispersion 'kink' observed at ~70 meV binding energy, remains a two-decade-old mystery. Understanding this phenomenon requires an accurate description of the coupling between the electron and some collective modes. We report here ab initio calculations based on GW perturbation theory and show that correlation-enhanced electron-phonon interaction in cuprates gives rise to the strong kinks, which not only explains quantitatively the observations but provides new understanding of experiments. Our results reveal it is the electron density-of-states being the predominant factor in determining the dopingdependence of the kink size, manifesting the multi-band nature of the cuprates, as opposed to the prevalent belief of it being a measure of the mode-coupling strength.

show abstract

“…For these reasons, MBPT calculations typically require computing architectures with a much larger number of nodes and larger memory. Efficient numerical schemes implemented in several codes [114][115][116] have partly alleviated these problems, enabling a significant speed-up on highly parallelized hardware. All in all, at present, MBPT calculations (GW and BSE) on systems with about 50 atoms are performed on the timescale of the order of a few tens of hours.…”

Section: Computational Costsmentioning

confidence: 99%

Ab Initio Quantum-Mechanical Predictions of Semiconducting Photocathode Materials

Cocchi

Saßnick

2021

Micromachines

View full text Add to dashboard Cite

Ab initio quantum–mechanical methods are well-established tools for material characterization and discovery in many technological areas. Recently, state-of-the-art approaches based on density-functional theory and many-body perturbation theory were successfully applied to semiconducting alkali antimonides and tellurides, which are currently employed as photocathodes in particle accelerator facilities. The results of these studies have unveiled the potential of ab initio methods to complement experimental and technical efforts for the development of new, more efficient materials for vacuum electron sources. Concomitantly, these findings have revealed the need for theory to go beyond the status quo in order to face the challenges of modeling such complex systems and their properties in operando conditions. In this review, we summarize recent progress in the application of ab initio many-body methods to investigate photocathode materials, analyzing the merits and the limitations of the standard approaches with respect to the confronted scientific questions. In particular, we emphasize the necessary trade-off between computational accuracy and feasibility that is intrinsic to these studies, and propose possible routes to optimize it. We finally discuss novel schemes for computationally-aided material discovery that are suitable for the development of ultra-bright electron sources toward the incoming era of artificial intelligence.

show abstract

Accelerating Large-Scale Excited-State GW Calculations on Leadership HPC Systems

Cited by 28 publications

References 34 publications

Three Practical Workflow Schedulers for Easy Maximum Parallelism

Three Practical Workflow Schedulers for Easy Maximum Parallelism

Unmasking the Origin of Kinks in the Photoemission Spectra of Cuprate Superconductors

Ab Initio Quantum-Mechanical Predictions of Semiconducting Photocathode Materials

Contact Info

Product

Resources

About