| Co-Investigator(Kenkyū-buntansha) |
NISHIMURA Toshikazu Ritsumeikan Univ., Fac. Science and Engineering, Associate Professor, 理工学部, 助教授 (00273483)
NIIMI Haruo Kyoto Sangyo Univ., Fac. Engineering, Professor, 工学部, 教授 (40144331)
OHNISHI Atsushi Ritsumeikan Univ., Fac. Science and Engineering, Professor, 理工学部, 教授 (50160560)
|
|---|
| Budget Amount *help |
¥3,600,000 (Direct Cost: ¥3,600,000)
Fiscal Year 2003: ¥1,400,000 (Direct Cost: ¥1,400,000)
Fiscal Year 2002: ¥2,200,000 (Direct Cost: ¥2,200,000)
|
|---|
| Research Abstract |
This research developed teaching materials for parallel computing using synchronized multi-media, and parallelized several subjects on a PC cluster.
1.We built animations for basic concepts of parallel computing, such as load balancing, synchronization and pipelining based on the models of bowling, athletics, and distribution of handouts. Then we developed teaching materials for primers by synchronizing figures, voices and texts with the animations. Moreover, we confirmed that streaming is more useful than down loading based on small-scale experiments of streaming. These experiments showed that teaching materials of parallel computing using synchronized multi-media are very useful for educating primers of parallel programming.
2.We developed a parallel programming environment on a PC cluster in which 16 PCs are connected via fast network cards, called Myrinet, and cluster system software, called SCore, is installed. Several subjects such as JPEG/JPEG2000 encoders, the Kerman vortex using the finite element method, the systematic dither method, three dimensional drawing of images, n-queen problem, spline interpolation, and the Gauss-Jordan iteration method were parallelized using OpenMP on the PC cluster. We could get reasonable speedups for these subjects. Moreover, we confirmed that pipelining is essential for three dimensional spline curves and the Gauss-Jordan iteration method. Thus, MPI is more suitable for these programs.
|
