Date

2015

Document Type

Thesis

Degree

Master of Science

Department

Computer Science

First Adviser

Spear, Michael F.

Abstract

This paper implemented four transactional memory algorithms, one hardware transaction algorithm and three software transaction algorithms. The goal of this research was to investigate the cost of determinism in a parallel world. The current standard for top per- forming parallelism is working on jobs that are independent such as a Map Reduce, it isolates each job. This paper attempts to investigate the cost of obtaining deterministic output on independent and non-independent tasks while being parallel.It is obvious that the cost of determinism will be steep, and the results presented will not exceed the independent parallel cases. But how damaging is determinism, can it exceed serial executions, if so when is it appropriate to run a deterministic parallel execution over a serial execution.

Share

COinS