论文信息 - Task Level Parallelization for Seismic Modeling and Inversion

Task Level Parallelization for Seismic Modeling and Inversion

This paper present experience with using PVM to parallelize \Dseismic inversion code developed in the Rice Inversion Project. We use coarse grain parallelism to dynamically distribute simulations over several workstations. When doing modeling, this strategy works efficiently, allowing us to reach an speedup of almost 4.5 on a cluster of 6 IBM RS6000 workstations. When doing inversion, however, we are currently limited to speedups of 2.4 on 3 workstations.

[1] John E. Dennis,et al. Direct Search Methods on Parallel Machines , 1991, SIAM J. Optim..

[2] Jack Dongarra,et al. Integrated Pvm Framework Supports Heterogeneous Network Computing , 1993 .

[3] W. W. Symes. Task Level Parallelization of Dso , 1993 .

[4] Jack Dongarra,et al. Pvm 3 user's guide and reference manual , 1993 .

[5] C. A. R. Hoare,et al. Communicating sequential processes , 1978, CACM.

[6] G. Amdhal,et al. Validity of the single processor approach to achieving large scale computing capabilities , 1967, AFIPS '67 (Spring).

[7] Jack J. Dongarra,et al. The PVM Concurrent Computing System: Evolution, Experiences, and Trends , 1994, Parallel Comput..

[8] Oscar Naim D'Paola. Performance visualization of parallel programs , 1995 .

[9] Michel Kern,et al. Shot level parallelization of a seismic inversion code using PVM , 1994 .

[10] William W. Symes,et al. Loop Level Parallelization of a Seismic Inversion Code , 1993 .

[11] J. Bell,et al. Parallel distributed seismic migration , 1992, Future Gener. Comput. Syst..