next up previous
Next: Summary Up: Performance Previous: Protocol Performance

Simulator SpeedUps

is the execution time of the simulator using N processors divided by the execution time of the simulator using 1 processor. The speedups of the simulators are shown in Figure 5.8. For the LU and MG benchmarks, speedups are shown for the fast simulators, since the NMP+CEP+Det mode could not detect the determinism in the applications. The speedups for the LU benchmarks are relative to the smallest host processor configuration that could be used to run the simulator. For example, the 8 target processor simulator could be executed on 2, 4 or 8 host processors. Hence, the reference execution time is of the 2 processor simulation.

Notice that the speedups achieved by the simulation are characteristic of the application itself, since the simulation overhead is very little.

  
Figure 5.8: Fast Simulator Speedups



next up previous
Next: Summary Up: Performance Previous: Protocol Performance



Andy Kahn
Wed Jun 25 20:28:02 PDT 1997