Efficient Calculation of Pairwise Nonbonded Forces

Chiu, Matt; Khan, Md. Ashfaquzzaman; Herbordt, Martin C.

doi:10.1109/fccm.2011.34

Cited by 22 publications

(15 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Finally, these restructurings sometimes suggest improvements to the hardware (phase 2). Overall we find that the concern of remaining in a local minimum is unfounded: per chip performance is competitive [6] while the intrinsic communication support is unparalleled for COTS parts.…”

Section: Discussionmentioning

confidence: 92%

“…3. MD software system = NAMD [12], FPGA = Altera Stratix IV [6]. The major change with moving to NAMD is that the range-limited force calculation is substantially more complex.…”

Section: Polynomial Evaluationmentioning

confidence: 99%

“…These are computed using the expressions given in the previous section plus correction terms to smooth the cut-off and simplify the long-range force. These expressions are sufficiently complex that a large number of choices emerge in their evaluation (see, e.g., [2], [6], [7], [8] and references therein). For FPGA-based systems the object is to minimize pipeline size to enable more computation per area and thus more parallelism and faster execution.…”

Section: Polynomial Evaluationmentioning

confidence: 99%

See 2 more Smart Citations

Architecture/algorithm codesign of molecular dynamics processors

Herbordt

2013

2013 Asilomar Conference on Signals, Systems and Computers

Self Cite

View full text Add to dashboard Cite

Molecular Dynamics is notable in High PerformanceComputing in that it is both sufficiently critical, and not wellenough served by off-the-shelf processors, that it has been continually targeted with domain-specific architectures. These range from the common, i.e., use of accelerators, to FPGA-centric servers describe here, to full-scale ASIC-based systems (e.g., the Anton processor). The strict constraints on achieving strong scaling leads to novel designs, which in turn suggest restructurings of the underlying algorithms. These in turn suggest some changes to the hardware. We discuss the technological issues involved, the implications of various approaches on architecture/algorithm codesign, and how these are likely to affect future high-end processors, both domain-specific and off-the-shelf.

show abstract

Section: Discussionmentioning

confidence: 92%

“…3. MD software system = NAMD [12], FPGA = Altera Stratix IV [6]. The major change with moving to NAMD is that the range-limited force calculation is substantially more complex.…”

Section: Polynomial Evaluationmentioning

confidence: 99%

Section: Polynomial Evaluationmentioning

confidence: 99%

See 1 more Smart Citation

Architecture/algorithm codesign of molecular dynamics processors

Herbordt

2013

2013 Asilomar Conference on Signals, Systems and Computers

Self Cite

View full text Add to dashboard Cite

show abstract

“…implementation possibilities. If logic cells alone are used, then ASICtype designs would be preferred for fixed [28,36,37] or floatingpoint [29,38]. Modern FPGAs, however, also have many thousands of embedded ASIC blocks, viz.…”

Section: Particle-pairmentioning

confidence: 99%

“…In CPU implementations this can be handled by periodically creating neighbor lists. In FPGAs, the preferred method is to do this on-the-fly [27][28][29] through filtering. Boundary Conditions.…”

Section: Introductionmentioning

confidence: 99%

Fully integrated FPGA molecular dynamics simulations

Yang

Geng

Wang

et al. 2019

Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis

View full text Add to dashboard Cite

The implementation of Molecular Dynamics (MD) on FPGAs has received substantial attention. Previous work, however, has consisted of either proof-of-concept implementations of components, usually the range-limited force; full systems, but with much of the work shared by the host CPU; or prototype demonstrations, e.g., using OpenCL, that neither implement a whole system nor have competitive performance. In this paper, we present what we believe to be the first full-scale FPGA-based simulation engine, and show that its performance is competitive with a GPU (running Amber in an industrial production environment). The system features on-chip particle data storage and management, short-and long-range force evaluation, as well as bonded forces, motion update, and particle migration. Other contributions of this work include exploring numerous architectural trade-offs and analysis on various mappings schemes among particles/cells and the various on-chip compute units. The potential impact is that this system promises to be the basis for long timescale Molecular Dynamics with a commodity cluster.

show abstract