Estimation of optimum size of hypercube for a given parallel algorithm /
Abstract: "In an ideal situation, with increase in the number of processors in a machine, the execution time should decrease proportionally. But in reality with increase in the number of processors, the time to communicate intermediate computed results also increases. For every algorithm, bey...
| Main Author: | |
|---|---|
| Format: | Book |
| Language: | English |
| Published: |
College Station, Tex. :
Texas A & M University, Computer Science Dept.,
[1992]
|
| Series: | Technical report (Texas A & M University. Computer Science Department) ;
92-023. |
| Subjects: |
| Summary: | Abstract: "In an ideal situation, with increase in the number of processors in a machine, the execution time should decrease proportionally. But in reality with increase in the number of processors, the time to communicate intermediate computed results also increases. For every algorithm, beyond a certain number of processors, the gain due to decrease in computation time is overshadowed by the increase in communication time. This is especially true in message passing multiprocessors. Thus in order to minimize total execution time for a given parallel algorithm and to reduce over utilization of processing resources, it is essential to be able to estimate the optimum number of processors required. For this reason we develop a communication and a computation model for a hypercube based nCUBE machine. The communication model includes expressions for calculating time for one to one communication, broadcast, multinode broadcast and concentration time. Where as the computation model allows us to approximate the execution time of a block of instructions written in higher level language. Backpropagation, a neural network learning algorithm is used to illustrate the procedure for calculating the execution time and then finding the optimum size of the cube for a given size of the problem." |
|---|---|
| Item Description: | "October 1992." |
| Physical Description: | 16 leaves : illustrations ; 28 cm. |
| Bibliography: | Includes bibliographical references. |