Parallelization of a molecular dynamics non-bonded force algorithm for MIMD architecture

Revisiting and parallelizing SHAKE

“…Indeed, significant advances were made in algorithms that compute Molecular Dynamic simulations in parallel. Implementations of parallel algorithms for Molecular Dynamic simulations focus on the calculation of the forces (Nguyen, Khanmohammadbaigi et al 1985;Clark and J.A. 1990;Mertz, Tobias D.J.…”

Application of Pfortran and Co‐Array Fortran in the Parallelization of the GROMOS96 Molecular Dynamics Module

“…The parallelization of molecular dynamics has been explored widely in the literature [4,6,8,9,11,15,19,23,27,[29][30][31]. Fortunately, molecular dynamics simulations of biomolecular systems are well suited for parallel computation since the forces acting on each atom can be calculated independently with a small amount of boundary information consisting of a neighborhood of atomic coordinates and in some cases, velocities.…”

“…Parallel algorithms are critical to the application and progress of MD in order to 1) improve the accuracy of simulation models, 2) extend the length of simulations, and 3) simulate large, complex systems. Numerous MD parallelizations have been described in the literature, ranging from the easy to implement replicated algorithm [6,20] to the more difficult to implement spatial decomposition [9,30], which is generally more scalable. The force decomposition algorithm is an intermediate approach in that it is generally more efficient than the replicated algorithm and easier to implement than the spatial decomposition [27].…”

Parallel atomistic simulations

“…Muller-Plathe et al [16] tested four benchmark systems, obtaining speedups of up to 38 on 65 processors of an Intel Paragon with a code named PARALLACS. The GROMOS (GROningen MOlecular Simulation Software) code was parallelized with the replicated data algorithm [17,18] yielding a speedup of about 67 on 128 processors. Follow-on work with GROMOS consisted of implementing the neighbor list/linked cell force calculation algorithm [1] in parallel and a dkcussion of parallel strategies [19] for the SHAKE algorithm [20].…”