نحو التنفيذ الأمثل لحزمة Wien2k باستخدام نماذج البرمجة المتوازية ( دراسة مقارنة)
اريج مصطفى محمود جابر
Areej Mustafa Mahmoud Jabir
In the present work, a comparison between two parallel methods have been done using message passing with OpenMPI API. Both Distribute k-pointmethod and Data distribution method have been used to run physical package which is used to study the physical and chemical properties of the materials which is called Wien2k package. Two data set size are used to be as a benchmark of this study, execution time of running with respect to RAM size either in shared or distributed case has been studied, two different size of RAM per CPU has been tested in Distribute k-point for the two benchmark, respectively. Network speed and its effect on calculation time has been studied, the network speed used in the study was 100 Mbps and 1 Gbps, the network effect has been tested on Distribute k-point method for two classes of CPUs and different RAM size per CPUs . The effect of CPUs speed on execution time has been studied by distribute k-point and Data distribution methods; different speed of CPUs has been used to study homogenous and heterogeneous effects on execution time. In all tests, RAM size considerably affect the time of calculation effectively, it’s found that increasing size of RAM available per CPU will cause considerable decrease in the calculation time, the study showed a small effect for the network speed on the calculation time, the effect of network can be neglected with respect to RAM size effect, The execution time showed that Data distribution gives better reduction in the time of calculation and higher speed up factor with increasing number of CPU’s, two scalable quantity has been used to compare and analyze the results, speed up factor and the power factor of decaying formula for time of execution.
علم الحاسوب , Computer Science