## A Portable Parallel Particle Program (1995)

Venue: | Computer Physics Communications |

Citations: | 53 - 7 self |

### BibTeX

@ARTICLE{Warren95aportable,

author = {Michael S. Warren and John K. Salmon},

title = {A Portable Parallel Particle Program},

journal = {Computer Physics Communications},

year = {1995},

volume = {87},

pages = {266--290}

}

### Years of Citing Articles

### OpenURL

### Abstract

We describe our implementation of the parallel hashed oct-tree (HOT) code, and in particular its application to neighbor finding in a smoothed particle hydrodynamics (SPH) code. We also review the error bounds on the multipole approximations involved in treecodes, and extend them to include general cell-cell interactions. Performance of the program on a variety of problems (including gravity, SPH, vortex method and panel method) is measured on several parallel and sequential machines. 1 Introduction There are two strategies that can be applied in the quest for more knowledge from bigger and better particle simulations. One can use the brute force approach; simple algorithms on bigger and faster machines (and bigger and faster now means massively parallel). To compute the gravitational force and potential for a single interaction takes 28 floating point operations (here we count a division as 4 floating point operations and a square root as 4 floating point operations). A typical grav...

### Citations

1174 |
The Design and Analysis of Spatial Data Structures
- Samet
- 1990
(Show Context)
Citation Context ...-bit key. In practice, we use 21 bit values and a 64-bit key Apart from the choice of origin and coordinate system, this is identical to Morton ordering (also called Z or N ordering, see Chapter 1 of =-=[17]-=- and references therein, and also [12]). By a suitable modification of this scheme, we can represent internal nodes of an oct-tree using the same type of key. We prepend an additional 1-bit to the mos... |

349 |
Computer Simulation Using Particles
- Hockney, Eastwood
(Show Context)
Citation Context ...f the sphere, and the second is the maximum allowed error in the gradient of the potential. 10 Conclusion Particle methods are useful for studying an enormous variety of systems. Hockney and Eastwood =-=[28]-=- discuss applications in plasma physics, device physics, astrophysics and material science. Our code has allowed us to make significant progress in the study of galaxy dynamics [22] and cosmology [23]... |

147 | A parallel hashed oct-tree n-body algorithm
- Warren, Salmon
- 1993
(Show Context)
Citation Context ...ection through a hash table can also be used to catch accesses to non-local data, and allows us to request and receive data from other processors. This is the basis of the hashed oct-tree method (see =-=[16]-=- for further details). It is very important for many parallel computational algorithms to efficiently represent irregular data structures. While it is possible to represent cleanly and efficiently reg... |

138 |
Smoothed particle hydrodynamics
- Monaghan
- 1992
(Show Context)
Citation Context ...cle Hydrodynamics Smoothed particle hydrodynamics was introduced by [2, 3] and was combined with a hierarchical tree method by [4, 5]. The derivation of the SPH formalism has been presented elsewhere =-=[6, 7, 8]-=-, and is beyond the scope of this paper. Our implementation of SPH follows that described in [7] for the most part. We jump immediately to the bottom line, which is a formula for the acceleration on p... |

117 |
Smoothed particle hydrodynamics : Theory and application to non-spherical stars
- Gingold, Monaghan
- 1977
(Show Context)
Citation Context ...ly describe two other applications to which our code has been applied, a vortex particle method and a panel method. 2 Smoothed Particle Hydrodynamics Smoothed particle hydrodynamics was introduced by =-=[2, 3]-=- and was combined with a hierarchical tree method by [4, 5]. The derivation of the SPH formalism has been presented elsewhere [6, 7, 8], and is beyond the scope of this paper. Our implementation of SP... |

93 |
A hierarchical o(nlogn) force-calculation algorithm
- Barnes, Hut
- 1986
(Show Context)
Citation Context ...ultipole acceptance criterion (MAC) then becomes R 2 0 ? r 2 c (32) for each displacement R 0 and critical radius r c (Fig. 3). This is computationally very similar to the BarnesHut opening criterion =-=[13]-=-, where instead of using a multiple of the box size, s=`, we use the distance r c , derived from the contents of the cell and the error tolerance. We have also implemented a gravitational code that co... |

91 |
Parallel hierarchical N-body methods
- Salmon
- 1991
(Show Context)
Citation Context ...ome of the mathematics behind multipole methods, and present error bounds suitable for computing local expansions arising from distributed sources for arbitrary Green's functions. In our earlier work =-=[10, 11]-=-, we restricted our attention to methods which computed only Body-Cell or Body-Body interactions (where the sources are single particles, i.e. not distributed). Here we extend the analysis to include ... |

86 | Astrophysical n-body simulations using hierarchical tree data structures
- Warren, Salmon
- 1992
(Show Context)
Citation Context ...zed pieces, while minimizing the surface area of each processor domain. Several varieties of domain decompositions have been used. In our first generation code, we used orthogonal recursive bisection =-=[14]-=-. This splits the work in half along the first spatial dimension of the entire domain, and then independently splits the two halves along the second dimension and so on until each processor has a piec... |

44 |
Load balancing and data locality in hierarchical N-body methods
- Singh, Holt, et al.
- 1992
(Show Context)
Citation Context ...sociated with each item. We have also begun to investigate using the tree structure itself to provide pertinent information for the next domain decomposition stage (an idea that was first proposed in =-=[15]-=-), since in an O(N) method the work is not so much associated with particles as it is with cells. 5 Tree Construction In a conventional oct-tree data structure, the topology of the tree is indicated b... |

40 | Skeletons from the treecode closet
- Salmon, Warren
- 1994
(Show Context)
Citation Context ...ome of the mathematics behind multipole methods, and present error bounds suitable for computing local expansions arising from distributed sources for arbitrary Green's functions. In our earlier work =-=[10, 11]-=-, we restricted our attention to methods which computed only Body-Cell or Body-Body interactions (where the sources are single particles, i.e. not distributed). Here we extend the analysis to include ... |

32 |
Computing three-dimensional incompressible flows with vortex elements
- Leonard
- 1985
(Show Context)
Citation Context ...ur code has allowed us to make significant progress in the study of galaxy dynamics [22] and cosmology [23]. In addition, we note applications in molecular dynamics [29], computational fluid dynamics =-=[30, 31]-=- and partial differential equations relevant to biology [32]. Fast multipole methods have been used to address two-dimensional problems in potential flows [33], and electromagnetic scattering [34]. Mu... |

25 |
TREESPH: a unification of SPH with the hierarchical tree method
- Hernquist, Katz
- 1989
(Show Context)
Citation Context ...en applied, a vortex particle method and a panel method. 2 Smoothed Particle Hydrodynamics Smoothed particle hydrodynamics was introduced by [2, 3] and was combined with a hierarchical tree method by =-=[4, 5]-=-. The derivation of the SPH formalism has been presented elsewhere [6, 7, 8], and is beyond the scope of this paper. Our implementation of SPH follows that described in [7] for the most part. We jump ... |

25 |
A modified tree code: Don’t laugh; it runs
- Barnes
- 1990
(Show Context)
Citation Context ...ch neighbors must be found in SPH. A cell in the sink is a collection of bodies whose interactions are computed together. In the language of [18], a sink is the center of a "local expansion"=-=. Barnes' [19]-=- simultaneous application of the multipole acceptability criterion to collections of bodies is similar in spirit to our use of sinks. Usually, when called from an application code, the source and sink... |

24 |
The fast multipole method (FMM) for electromagnetic cattering problems
- Engheta, Murphy, et al.
- 1992
(Show Context)
Citation Context ...[30, 31] and partial differential equations relevant to biology [32]. Fast multipole methods have been used to address two-dimensional problems in potential flows [33], and electromagnetic scattering =-=[34]-=-. Much chemistry is done with small N , and these may not benefit much from treecodes. Nevertheless, there are some important problems with larger N [35], which should run well with a parallel treecod... |

22 |
Atomic level simulations on a million particles: The cell multipole method for Coulomb and London nonbond interactions
- Ding, Karasawa, et al.
- 1992
(Show Context)
Citation Context ...ial flows [33], and electromagnetic scattering [34]. Much chemistry is done with small N , and these may not benefit much from treecodes. Nevertheless, there are some important problems with larger N =-=[35]-=-, which should run well with a parallel treecode. Although we have described our methods in the context of particle data, treecodes are also useful in contexts completely outside dynamics. Oct-trees a... |

20 |
Accelerated Molecular Dynamics Simulation with the Fast Multipole Algorithm
- Board, Causey, et al.
- 1992
(Show Context)
Citation Context ...strophysics and material science. Our code has allowed us to make significant progress in the study of galaxy dynamics [22] and cosmology [23]. In addition, we note applications in molecular dynamics =-=[29]-=-, computational fluid dynamics [30, 31] and partial differential equations relevant to biology [32]. Fast multipole methods have been used to address two-dimensional problems in potential flows [33], ... |

17 |
A Special-Purpose Computer for Gravitational Many-Body Problems,” Nature
- Sugimoto, Chikada, et al.
- 1990
(Show Context)
Citation Context ... a year to complete the simulation. Thus, it should be clear that the simple bruteforce approach is not a viable option for large N-body simulations (although using specialized hardware such as GRAPE =-=[1]-=- makes it a possibility). The second approach is to try to develop better algorithms that can solve problems to the desired accuracy using much less computational power. In this paper we show that a m... |

16 |
Dark halos formed via dissipationless collapse. I - Shapes and alignment of angular momentum
- Warren, Quinn, et al.
- 1992
(Show Context)
Citation Context ...700 timesteps was completed in about 15 hours on 128 processors of an Intel Paragon. Additional applications of earlier versions of our treecode to large astrophysical N-body problems may be found in =-=[22, 23]-=- and [24]. 9.1 The Vortex Particle Method The vorticity equation (! = r \Theta u, and hence r \Delta ! = 0) for an incompressible fluid (r \Delta u = 0) is obtained from taking the curl of the momentu... |

16 |
Panel Methods in Computational Fluid Dynamics
- Hess
- 1990
(Show Context)
Citation Context ...solutions of Eq. 35 in such a way that the boundary conditions are properly met. These so-called panel methods are used extensively for aerodynamics modeling in the aircraft and automobile industries =-=[26]-=-. Here we give a brief overview of the application of our treecode to the solution of this problem. A more in-depth analysis may be found in [27]. A distribution of panels in a uniform stream parallel... |

15 |
Particle Methods for Hydrodynamics
- Monaghan
- 1985
(Show Context)
Citation Context ...cle Hydrodynamics Smoothed particle hydrodynamics was introduced by [2, 3] and was combined with a hierarchical tree method by [4, 5]. The derivation of the SPH formalism has been presented elsewhere =-=[6, 7, 8]-=-, and is beyond the scope of this paper. Our implementation of SPH follows that described in [7] for the most part. We jump immediately to the bottom line, which is a formula for the acceleration on p... |

13 |
Fast Parallel Treecodes for Gravitational and Fluid Dynamical N-body Problems” Intl
- Salmon, Winckelmanns, et al.
- 1992
(Show Context)
Citation Context ...roximate the summation with an expression involving the multipole moments of the vorticity distribution. A more detailed description of the application of our treecode to this problem may be found in =-=[25]-=-. We carried out a series of timings for a problem representing the evolution of an initially spherical vorticity distribution. Figure 10 shows the initial positions of vortex particles representing a... |

10 |
Dynamic mass exchange in doubly degenerate binaries I. 0.9 and 1.2 M⊙ stars, Astrophys
- Benz, Bowers, et al.
- 1990
(Show Context)
Citation Context ...en applied, a vortex particle method and a panel method. 2 Smoothed Particle Hydrodynamics Smoothed particle hydrodynamics was introduced by [2, 3] and was combined with a hierarchical tree method by =-=[4, 5]-=-. The derivation of the SPH formalism has been presented elsewhere [6, 7, 8], and is beyond the scope of this paper. Our implementation of SPH follows that described in [7] for the most part. We jump ... |

9 |
Topics in vortex methods for the computation of three- and two-dimensional incompressible unsteady flows
- Winckelmans
- 1989
(Show Context)
Citation Context ...ur code has allowed us to make significant progress in the study of galaxy dynamics [22] and cosmology [23]. In addition, we note applications in molecular dynamics [29], computational fluid dynamics =-=[30, 31]-=- and partial differential equations relevant to biology [32]. Fast multipole methods have been used to address two-dimensional problems in potential flows [33], and electromagnetic scattering [34]. Mu... |

8 |
An efficient N-body algorithm for a finegrain parallel computer
- Barnes
- 1986
(Show Context)
Citation Context ... i.e., the B (p) . Computing this information takes place in the tree construction stage, and takes very little time compared with the later phases of the algorithm. Our implementation of the N log N =-=[12, 11]-=- method associates a critical radius (r c ) with each cell, and demands that any particle that interacts with the cell lie outside the sphere defined by r c . In this method, all sinks are individual ... |

8 |
Large scale structure after COBE: Peculiar velocities and correlations of cold dark matter halos
- Zurek, Quinn, et al.
- 1994
(Show Context)
Citation Context ...700 timesteps was completed in about 15 hours on 128 processors of an Intel Paragon. Additional applications of earlier versions of our treecode to large astrophysical N-body problems may be found in =-=[22, 23]-=- and [24]. 9.1 The Vortex Particle Method The vorticity equation (! = r \Theta u, and hence r \Delta ! = 0) for an incompressible fluid (r \Delta u = 0) is obtained from taking the curl of the momentu... |

5 | Potential flow in channels
- Greengard
- 1990
(Show Context)
Citation Context ...s [29], computational fluid dynamics [30, 31] and partial differential equations relevant to biology [32]. Fast multipole methods have been used to address two-dimensional problems in potential flows =-=[33]-=-, and electromagnetic scattering [34]. Much chemistry is done with small N , and these may not benefit much from treecodes. Nevertheless, there are some important problems with larger N [35], which sh... |

3 |
Speeding up N-body calculations on machines without hardware square root
- Karp
- 1992
(Show Context)
Citation Context ...peration was coded in assembly language using a Newton-Raphson iteration. For the SP-1 code, the same operation was coded in C using a Chebychev polynomial approximation and one Newton iteration (see =-=[20]-=- for 18 details). On the CM-5, no attempt was made to use the vector units. The YMP performance could possibly be improved a great deal by tuning of a few critical functions where vectorization was in... |

3 |
The fast solution of three-dimensional boundary integral equations in potential flow aerodynamics using parallel and sequential tree codes
- Winckelmans, Salmon, et al.
- 1994
(Show Context)
Citation Context ...ics modeling in the aircraft and automobile industries [26]. Here we give a brief overview of the application of our treecode to the solution of this problem. A more in-depth analysis may be found in =-=[27]-=-. A distribution of panels in a uniform stream parallel to the x-axis produces a potential \Phi(~r k ) = U1x k + 1 4 N X j=1 oe j Z j 1 r kj ds j (36) 21 Figure 11: On the left, the time per timestep ... |

2 |
Numerical Modeling of Stellar Pulsations
- Benz
- 1989
(Show Context)
Citation Context ...cle Hydrodynamics Smoothed particle hydrodynamics was introduced by [2, 3] and was combined with a hierarchical tree method by [4, 5]. The derivation of the SPH formalism has been presented elsewhere =-=[6, 7, 8]-=-, and is beyond the scope of this paper. Our implementation of SPH follows that described in [7] for the most part. We jump immediately to the bottom line, which is a formula for the acceleration on p... |

2 | A gradient random walk method for two-dimensional reaction-diffusion equations
- Sherman, Mascagni
- 1994
(Show Context)
Citation Context ... of galaxy dynamics [22] and cosmology [23]. In addition, we note applications in molecular dynamics [29], computational fluid dynamics [30, 31] and partial differential equations relevant to biology =-=[32]-=-. Fast multipole methods have been used to address two-dimensional problems in potential flows [33], and electromagnetic scattering [34]. Much chemistry is done with small N , and these may not benefi... |

1 |
The Rapid Evalution of Potential Fields in Particle Systems
- Greengard
- 1987
(Show Context)
Citation Context ...in multipole codes. It is the collection of bodies from which neighbors must be found in SPH. A cell in the sink is a collection of bodies whose interactions are computed together. In the language of =-=[18], a sink i-=-s the center of a "local expansion". Barnes' [19] simultaneous application of the multipole acceptability criterion to collections of bodies is similar in spirit to our use of sinks. Usually... |

1 |
uller, "A comparison between SPH and PPM: Simulation of stellar collisions
- Davies, Ruffert, et al.
- 1991
(Show Context)
Citation Context ...ectness of our implementation of the SPH method. It took about 40 seconds per timestep on a 32 processor machine. This problem was previously used as the basis for a comparison between SPH and PPM in =-=[21]-=-. Figure 9 also shows a galaxy cluster simulation with 1.1 million particles computed using local expansions. This simulation of 2700 timesteps was completed in about 15 hours on 128 processors of an ... |

1 |
Experimental Cosmology Using Fast Parallel N-body Methods
- Warren
- 1994
(Show Context)
Citation Context ... was completed in about 15 hours on 128 processors of an Intel Paragon. Additional applications of earlier versions of our treecode to large astrophysical N-body problems may be found in [22, 23] and =-=[24]-=-. 9.1 The Vortex Particle Method The vorticity equation (! = r \Theta u, and hence r \Delta ! = 0) for an incompressible fluid (r \Delta u = 0) is obtained from taking the curl of the momentum equatio... |