List of Figures

1. RMI Architecture.
2. A parallel matrix addition.
3. Red-black iteration.
4. A pipelined matrix multiplication program.
5. HPJava data parallel version of the N-body.
6. Data parallel version of the ``N-body'' example with MPI communications.
7. Principal classes of mpiJava
8. Minimal mpiJava program (run in two processes)
9. Software Layers
10. PingPong Results in Shared Memory (SM) mode
11. PingPong Results in Distributed Memory (DM) mode
12. Send and receive operations for various array shapes.
13. Communication times from ping-pong benchmark in non-shared-memory case. The lines represent the model defined by Equations 5.1 to 5.3 in the text, with parameters from Table 5.1.
14. Communication times from ping-pong benchmark in shared-memory case. The lines represent the model defined by Equations 5.1 to 5.3 in the text, with parameters from Table 5.1.
15. Improved protocol for handling arrays of primitive elements.
16. Pseudocode for ArrayOutputStream and ArrayInputStream
17. Ping-pong timings with primitive array data sent separately (solid points) in distributed memory mode, compared with the unoptimized results from Figure 5.2 (open points). Recall that the goal is to bring times for ``object-oriented'' sends of arrays down to the ``native'' send times, most closely approximated by the triangular points.
18. Ping-pong timings with primitive array data sent separately (solid points) in shared memory mode, compared with the unoptimized results from Figure 5.3 (open points). Recall that the goal is to bring times for ``object-oriented'' sends of arrays down to the ``native'' send times, most closely approximated by the triangular points.
19. Timings allowing two-dimensional array proxies in the object stream (solid points) in distributed memory mode, compared with the unoptimized results from Figure 5.2 (open points). Sends of two-dimensional Java arrays (solid circles) are now much closer to the native bandwidth (of which the triangular points are representative).
20. Timings allowing two-dimensional array proxies in the object stream (solid points) in shared memory mode, compared with the unoptimized results from Figure 5.3 (open points). Sends of two-dimensional Java arrays (solid circles) are now much closer to the native bandwidth (of which the triangular points are representative).
21. Parallel Multiplicative Linear Congruential Generators (MLCG).
22. The Main Procedure of Potts Model Simulation using Metropolis Algorithm: One starts with an initial configuration of spins and repeats these procedures.
23. Checkerboard Partition and Blocked Communication.
24. The Main Procedure of Sequential Swendsen-Wang Algorithm.
25. MIMD Component Labeling. The bonds are shown as the thick lines.
26. The Potts Model Monte Carlo Simulation.
27. Metropolis Performance with Lattice Size $32^2$.
28. Metropolis Performance with Lattice Size $64^2$.
29. Metropolis Performance with Lattice Size $128^2$.
30. Metropolis Performance with Lattice Size $256^2$.
31. Metropolis Performance with Lattice Size $512^2$.
32. Metropolis Performance with Lattice Size $1024^2$.
33. Swendsen-Wang Performance with Lattice Size $32^2$.
34. Swendsen-Wang Performance with Lattice Size $64^2$.
35. Swendsen-Wang Performance with Lattice Size $128^2$.
36. Swendsen-Wang Performance with Lattice Size $256^2$.
37. Swendsen-Wang Performance with Lattice Size $512^2$.
38. Swendsen-Wang Performance with Lattice Size $1024^2$.
39. Speedup of Metropolis by using mpiJava as compared with serial Java.
40. Speedup of Swendsen-Wang by using mpiJava as compared with serial Java.