Graduation Semester and Year

2012

Language

English

Document Type

Thesis

Degree Name

Master of Science in Electrical Engineering

Department

Electrical Engineering

First Advisor

Kamisetty R Rao

Abstract

This thesis is aimed at reducing the encoding time for a group of pictures (GOP) for h.264 baseline profile on a general purpose CPU. This thesis also aims at drawing a comparison between task based parallelism and data based parallelism. The reduction is achieved by encoding the frames in parallel rather than serially. Task based parallelism is achieved by equally dividing the GOP's in two different threads and running them on the underlying hardware at the same time using Open MP software. Data based parallelism is achieved by finding the hot spots in the software and then parallelizing it using task based approach as stated above. JM 18.0 [1] is the reference software used to obtain the results. By adopting task based parallelism methodology the encoding time for a group of pictures is reduced approximately by 50%. But this reduction comes at a cost of increased CPU power consumption by approximately 50% and a marginal 2% increase in the physical memory usage. Data based parallelism does not show any improvement in the time complexity reduction as thread creation for every hot spot and every frame adds up to the time complexity overhead.

Disciplines

Electrical and Computer Engineering | Engineering

Comments

Degree granted by The University of Texas at Arlington

Download

Share

COinS