D. Brown, H. Minoux, and B. Maigret. A domain decomposition parallel processing algorithm for molecular dynamics simulations of systems of arbitrary connectivity. Computer Physics Communications, 103:170-186, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....decomposition by reducing communication costs through the use of a block decomposition, as in LAMMPS [13] and CHARMM [6] and as discussed in reference [12] d) Spatial decomposition, which uses linked lists and scales very well, but it is more difficult to write and extend. It is used in ddgmq [3], SIGMA, NAMD2 [8] PMD [19] EulerGromos [4] and it is evaluated in references [1, 9] 3.1 Work Distribution and Data Transfer Work distribution is performed by the master, through a short message (workcommand) Slaves retrieve the required data independently, carry out the work, and notify ....

D. Brown, H. Minoux, and B. Maigret. A domain decomposition parallel processing algorithm for molecular dynamics simulations of systems of arbitrary connectivity. Computer Physics Communications, 103:170--186, 1997.

....by reducing communication costs through the use of a block decomposition, as in LAMMPS [92] CHARMM [55] and PROTOMOL [D] and as discussed in [87] Spatial decomposition uses linked lists and scales very well [80] but it is more difficult to implement and extend. It is used in ddgmq [16], SIGMA [53] NAMD2 [66] PMD [125] EulerGromos [22] DMMD [4] and evaluated in [9] SPRINGS (see paper [C] 4.1 Load balancing A good load balancing of the computational work in a parallel environment becomes more important for a large number of processors [66, 86] see Table 1 in paper ....

D. Brown, H. Minoux, and B. Maigret. A domain decomposition parallel processing algorithm for molecular dynamics simulations of systems of arbitrary connectivity. Computer Physics Communications, 103:170--186, 1997.

....Each time step itself is relatively small. Since each time step must complete before the next one can begin, there is little opportunity to parallelize across time steps, requiring extensive parallelization of computation within each timestep. For example, simulating a biomolecular system with 3762 atoms sequentially required only about 1 second for a single time step in a particular simulation. To run this computation e#ectively on a large parallel machine implies that each time step must complete within tens of milliseconds. Given the communication requirements of molecular dynamics, ....

....the communication to computation ratio of # P . While this is much better than RD, it is still not scalable according to our strict criteria: as we increase the number of processors, the proportion of communication cost increases even if we simulate a larger biomolecule. Research by Plimpton [38, 37] and Hwang et al. [19] shows that this method provides a better speedup than RD, and can be used with good speedups up to hundreds of processors. Results obtained by Hwang show that the proportion of communication cost goes from 6 percent of the computation cost on 32 processors to 36 percent on ....

[Article contains additional citation context not shown here]

David Brown, Julian H. R. Clarke nad Motoi Okuda, and Takao Yamazaki. A domain decomposition parallel processing algorithm for molecular dynamics simulations of polymers. Computer Physics Communications, 83:1--13, 1994.

.... Skeel, Milind Bhandarkar, Robert Brunner, Attila Gursoy, Neal Krawetz, James Phillips, Aritomo Shinozaki, Krishnan Varadarajan, and Klaus Schulten Theoretical Biophysics Group, Beckman Institute for Advanced Science and Technology, University of Illinois at Urbana Champaign, Urbana, IL 61801. kale,skeel,milind,brunner,gursoy, nealk,jim,ari,krishnan,kschulte ks.uiuc.edu March 22, 1999 1 Abstract Molecular dynamics programs simulate the behavior of biomolecular systems, leading to insights and understanding of their functions. However, the computational complexity of such ....

....Each time step itself is relatively small. Since each time step must complete before the next one can begin, there is little opportunity to parallelize across time steps, requiring extensive parallelization of computation within each timestep. For example, simulating a biomolecular system with 3762 atoms sequentially required only about 1 second for a single time step in a particular simulation. To run this computation e#ectively on a large parallel machine implies that each time step must complete within tens of milliseconds. Given the communication requirements of molecular dynamics, and ....

[Article contains additional citation context not shown here]

David Brown, Herve Minoux, and Bernard Maigret. A domain decomposition parallel processing algorithm for molecular dynamics simulations of systems of arbitrary connectivity. Computer Physics Communications, 103:170--186, 1997.

....atom decomposition by reducing communication costs through the use of a block decomposition, as in LAMMPS [13] and CHARMM [6] and as discussed in reference [12] d) Spatial decomposition, which uses linked lists and scales very well, but it is more dicult to write and extend. It is used in ddgmq [3], SIGMA, NAMD2 [8] PMD [19] EulerGromos [4] and it is evaluated in references [1, 9] 3.1 Work Distribution and Data Transfer Work distribution is performed by the master, through a short message (workcommand) Slaves retrieve the required data independently, carry out the work, and notify ....

D. Brown, H. Minoux, and B. Maigret. A domain decomposition parallel processing algorithm for molecular dynamics simulations of systems of arbitrary connectivity. Computer Physics Communications, 103:170-186, 1997.

....computes forces on only the atoms in its sub domain. In the large N limit, the technique scales optimally as N=P , 2 since each processor need only acquire information from processors who own neighboring sub domains. However, SD algorithms have not been as widely used for molecular simulation [8, 9, 10, 14, 33] due to their complexity (particularly for bonded force computation) and the difficulty of achieving good load balance across processors for irregularly shaped or dynamically changing domains. We have developed a new parallel algorithm, called force decomposition (FD) which is particularly ....

....connectivity information must be exchanged and updated between processors. The extra coding to manipulate the appropriate data structures and optimize the communication performance of the data exchange subtracts from the parallel efficiency of the algorithm. We discuss several SD implementations [8, 10, 14] further in Section 7. 5 Force Decomposition Algorithm We now present the new force decomposition (FD) algorithm. Its communication cost lies in between that of the RD and SD approaches. The FD method partitions the force matrix F by blocks rather than by rows as in Section 3. ....

[Article contains additional citation context not shown here]

D. Brown, J. H. R. Clarke, M. Okuda, and T. Yamazaki. A domain decomposition parallel processing algorithm for molecular dynamics simulations of polymers. Comp. Phys. Comm., 83:1--13, 1994.

No context found.

D. Brown, H. Minoux, and B. Maigret. A domain decomposition parallel processing algorithm for molecular dynamics simulations of systems of arbitrary connectivity. Computer Physics Communications, 103:170-186, 1997.

No context found.

D. Brown, H. Minoux, and B. Maigret. A domain decomposition parallel processing algorithm for molecular dynamics simulations of systems of arbitrary connectivity. Computer Physics Communications, 103:170--186, 1997.

No context found.

D. Brown, H. Minoux, and B. Maigret. A domain decomposition parallel processing algorithm for molecular dynamics simulations of systems of arbitrary connectivity. Computer Physics Communications, 103:170-186, 1997.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC