Studi Komparasi Kinerja Algoritma Reduksi Siklis Dengan Algoritma Pemisahan Rekursif Pada Sistem Multiprosesor Berbasis Pvm

Prabawa, Tri - 941059 (2016) Studi Komparasi Kinerja Algoritma Reduksi Siklis Dengan Algoritma Pemisahan Rekursif Pada Sistem Multiprosesor Berbasis Pvm. STMIK AKAKOM Yogyakarta, Yogyakarta.

[img]
Preview
Text
1_941059_HALAMAN_DEPAN.pdf - Published Version

Download (774kB) | Preview
[img]
Preview
Text
2_941059_BAB_I.pdf - Published Version

Download (296kB) | Preview
[img]
Preview
Text
3_941059_BAB_II.pdf - Published Version

Download (556kB) | Preview
[img] Text
4_941059_BAB_III.pdf - Published Version
Restricted to Repository staff only

Download (93kB)
[img] Text
5_941059_BAB_IV.pdf - Published Version
Restricted to Repository staff only

Download (662kB)
[img]
Preview
Text
6_941059_BAB_V.pdf - Published Version

Download (90kB) | Preview
[img]
Preview
Text
DAFTAR PUSTAKA.pdf - Published Version

Download (253kB) | Preview

Abstract

This study discusses the comparison of the performance of the settlement system of linear equations Au = d, where the coefficient matrix A tridiagonal, with a cyclic reduction method and recursive decoupling on multiprocessor systems. The basic idea reduction cyclic reduction method is lowered rows independently by reduction row indexed odd or even, while for the recursive decoupling is lowered rows independent based on a strategy of rank-one updating and the partitioning process repeated on the matrix system in order to obtain the form of a matrix diagonal blocks of 2x2. Solving problems in parallel system is to perform the decomposition problems in algorithmic or geometric, in order to identify the characteristics of parallelism. The performance characteristics of parallel algorithms can be seen from the measurement execution time, and communication, speed-up, and efficiency. To determine these characteristics will be carried out a comparative study of the results of two studies have been carried out. The first study discusses the results of a cyclic reduction algorithm performance and performance results of a second study on the recursive decoupling algorithms are implemented on a system of parallel virtual machine (PVM), which is a model of a single distributed parallel processor. From the test results it appears that there is an increase of acceleration as the number of processors used. Acceleration for algorithms cyclical reduction ranged from 1.61 (2 processors) to 4.22 (8 processors), Whereas for the recursive decoupling algorithm separation between 1.61 (2 processors) to 5.90 (8 processors). But on the contrary, by increasing the number of processors used a drop in efficiency. Level of efficiency for cyclic reduction algorithm 88.38% (2 processors) and the lowest was 35.58% (8 processors), Whereas for the recursive decoupling algorithm 80.43% (2 processors) and the lowest was 36.25% (8 processors) , It is heavily influenced the higher the communication time caused the synchronous process occurs repeatedly.

Item Type: Other
Uncontrolled Keywords: tridiagonal system, cyclic reduction, recursive decoupling, speed-up, efficiency, and PVM.
Subjects: D Ilmu Sosial (Social Sciences) > D1 Statistik (Statistics)
F Ilmu Alam Dan Matematika (Natural Science And Mathematics) > F1 Matematika (Mathematics)
A Karya Umum (General) > Ilmu Komputer (Computer Science) > Pemrograman (Programming)
Divisions: Laporan Penelitian Dan Pengabdian
Depositing User: Wahyu Firmanto
Date Deposited: 12 Feb 2018 07:46
Last Modified: 15 Mar 2018 02:54
URI: http://eprints.akakom.ac.id/id/eprint/6980

Actions (login required)

View Item View Item