Parallelization strategies

In the distributed FMM scheme, it is also beneficial to use the same FMM trees to compute the electrostatic potential on the surface grid. This involves a procedure similar to that described above, in which all solute point charges traverse the distributed surface grid. In our implementation, we let each MPI rank store all atomic coordinates and related information, so that each MPI rank can independently compute its portion of v in Eq. (11.22), without communication. Allowing all MPI ranks to store the global atomic coordinates and charges is usually feasible because there are far fewer atoms than surface grid points. 

What we have described above applies only to solute point charges, whereas an electron density must be treated differently. A simple procedure for the latter is to let the QM charge density interact explicitly (rather than through the distributed FMM trees) with all surface grid points within a certain pre-determined cutoff distance, which maintains a nearly constant amount of CPU time for a fixed QM region. The electrostatic potential thus computed can then be communicated as needed to the appropriate MPI ranks. 

For QM/MM/PCM jobs, this means that the FMM trees are used to compute interactions between the surface charges themselves, and between the surface charges and the MM charges, but not for interactions that involve the QM region. While this approach clearly could be improved, we often find that the purely classical steps are the bottlenecks when the MM region is large. 

---

Esselink K, Smit B and Hilbers P A J 1993 Efficient parallel implementation of molecular dynamics on a toroidal network. I. Parallelizing strategy J. Comput. Phys. 106 101-7... 

An alternative route to allenyl stannanes involves organocuprate displacements on propargylic chlorides bearing an alkynyl PhsSn substituent (equation 39)79. Interestingly, transmetallation by attack of the cuprate on the tin substituent is not observed in these systems. A parallel strategy can be employed for allenylgermanes (equation 39)79. The... 

Vibrational Eigenstates of Four-Atom Molecules A Parallel Strategy Employing the Implicitly Restarted Lanczos Method. 

Brown, Clarke, Okuda, and Yamazaki also adopted a domain decomposition parallelization strategy on the Fujitsu API 000 machine. " They describe an algorithm that is a development of a spatial decomposition technique due to Liem and co-workers but incorporated decomposition of three-dimensional space in all three dimensions with linked-cell and neighbor table techniques for enhanced efficiency. Communications between processors were minimized without incurring the penalties associated with redundant force calculation. 

D. Brown, J. H. R. Clarke, M. Okuda, and T. Yamazaik, Comput. Phys. Commun., 74, 67 (1993). A Domain Decomposition Parallelization Strategy for Molecular Dynamics Simulations on Distributed Memory Machines. 

The shift of the post-processing to an HPC system, parallelization strategies, and the innovative data management mainly aimed at the reduction of total runtime. On the other hand, these approaches are not sufficient to fulfill the demand of short system response times needed for interactive exploration tasks. Because of the size of today s datasets, it is not possible to meet this criterion fully since the speed-up of an algorithm cannot be increased significantly. However, a fast representation of first temporary or approximate results leads to a considerable reduction of noticeable delays by simply decreasing the latency time of an algorithm. 

Three parallel strategies have been developed to access oligodeoxynucleotides... 

In the following two sections, these two parallel strategies will be more closely described and discussed. 

Unfortunately, automated parallel analytical separation systems were commercially not available at this time. However, in seeking to create the parallel strategy, we initiated studies on a manual-based injection system to validate the approach. It was thus possible to develop a manual parallel analytical separation system that allowed two columns to be used for the analysis of two samples concurrently. If this process could be automated and a greater number of analytes could be examined simultaneously by the same methodology, a quantal leap in productivity could be achieved. As previously discussed, the sample presentation typically occurs in an MTP format and so the simultaneous parallel analysis of a column or row from such a plate would seem a logical approach. 

In the example of the 500-g tissue extract above, multiple SPE runs are required to process the entire 500 g-Eq. Separate stations can be set up to run multiple, parallel SPE columns or the 40 g-Eq aliquots can be processed in series on the same SPE column (after the 80% MeCN/0.1 % TFA elution, 50 mL of 0.1 % TFA is passed through the column followed by another 40 g-Eq of the extract). A compromise between the series/parallel strategy is usually the most efficient. In the example above, the authors would set up four SPE columns and use each for three consecutive 41.6 g-Eq runs (a total of 12 runs). The lipid-pigment compo-... 

Second-order Moller-Plesset (MP2) perturbation theory is a widely used quantum chemical method for incorporating electron correlation. This chapter considers parallel computation of MP2 energies, comparing the performance achievable wifh simple and more sophisticated parallelization strategies. 

I conclude that universities and colleges that ate vastly different in orientation, markets served, and available resources are using roughly parallel strategies in positioning for prestige, having framed their aspirations in a similar manner (Toma 2008, p. 1). 

Two forms of parallelism are possible. Each processors can generate the n trial replicas of a new atom in parallel. However, a communication step is then required to choose one trial position and update the coordinates on each processor. This synchronisation must occur before a processor can proceed to the next atom. A second parallel strategy involves allowing each processor to grow one or several copies of a new chain. 

A DPD simulation may involve as few as 1000 particles for siirple equilibrium simulations to several million particles for simulations of complex fluids. Hence, it is important to run DPD codes in a parallel environment. There are several parallelization strategies, such as domain decomposition and force decomposition methods. One method that is appealing because of the particle-based nature of the method is the atom decomposition algorithm in which the computations for the N particles are split among P processors. Further advantage can be taken from the fact that DPD is a short-range method, i. e. each particle does... 

---