Abstract
A simulation program and an assembler were prepared for a parallel computer with 128{times} 256 operators operating under a numerical simulator at National Aerospace Laboratory. Examples of evaluation on the parallel computer, estimation of its maximum performance, and the case comparing the parallel computer with one processing element one described. The developed assembler operates on FACOM-M780, and converts the starting program into a machine language for the parallel computer. The verifying simulation used a program to derive a maximum value in data in the array data memory elements and a program to analyze flow fields around a two-dimensional cylinder. The former simulation presented that consistency in the command system for mask operation is more important than network improvement. This should be given due considerations when handling a three-dimensional problem with a large number of data in the memory elements. In the latter simulation, such results were obtained that vorticity computation can be expected of performance corresponding to the number of operators, that computing the flow functions achieved the expectation at 60% (performance improvement is possible), and computing surface pressure, resistance and lift resulted in low efficiency (because of data transfer processing time). 17 refs., 28 figs., 8 tabs.
Citation Formats
Harada, K, Yoshida, M, and Nakamura, K.
Simulation of a parallel computer; Heiretsu keisanki no kensho mogi.
Japan: N. p.,
1992.
Web.
Harada, K, Yoshida, M, & Nakamura, K.
Simulation of a parallel computer; Heiretsu keisanki no kensho mogi.
Japan.
Harada, K, Yoshida, M, and Nakamura, K.
1992.
"Simulation of a parallel computer; Heiretsu keisanki no kensho mogi."
Japan.
@misc{etde_10156068,
title = {Simulation of a parallel computer; Heiretsu keisanki no kensho mogi}
author = {Harada, K, Yoshida, M, and Nakamura, K}
abstractNote = {A simulation program and an assembler were prepared for a parallel computer with 128{times} 256 operators operating under a numerical simulator at National Aerospace Laboratory. Examples of evaluation on the parallel computer, estimation of its maximum performance, and the case comparing the parallel computer with one processing element one described. The developed assembler operates on FACOM-M780, and converts the starting program into a machine language for the parallel computer. The verifying simulation used a program to derive a maximum value in data in the array data memory elements and a program to analyze flow fields around a two-dimensional cylinder. The former simulation presented that consistency in the command system for mask operation is more important than network improvement. This should be given due considerations when handling a three-dimensional problem with a large number of data in the memory elements. In the latter simulation, such results were obtained that vorticity computation can be expected of performance corresponding to the number of operators, that computing the flow functions achieved the expectation at 60% (performance improvement is possible), and computing surface pressure, resistance and lift resulted in low efficiency (because of data transfer processing time). 17 refs., 28 figs., 8 tabs.}
place = {Japan}
year = {1992}
month = {May}
}
title = {Simulation of a parallel computer; Heiretsu keisanki no kensho mogi}
author = {Harada, K, Yoshida, M, and Nakamura, K}
abstractNote = {A simulation program and an assembler were prepared for a parallel computer with 128{times} 256 operators operating under a numerical simulator at National Aerospace Laboratory. Examples of evaluation on the parallel computer, estimation of its maximum performance, and the case comparing the parallel computer with one processing element one described. The developed assembler operates on FACOM-M780, and converts the starting program into a machine language for the parallel computer. The verifying simulation used a program to derive a maximum value in data in the array data memory elements and a program to analyze flow fields around a two-dimensional cylinder. The former simulation presented that consistency in the command system for mask operation is more important than network improvement. This should be given due considerations when handling a three-dimensional problem with a large number of data in the memory elements. In the latter simulation, such results were obtained that vorticity computation can be expected of performance corresponding to the number of operators, that computing the flow functions achieved the expectation at 60% (performance improvement is possible), and computing surface pressure, resistance and lift resulted in low efficiency (because of data transfer processing time). 17 refs., 28 figs., 8 tabs.}
place = {Japan}
year = {1992}
month = {May}
}