موازی سازی حلقه های تودرتوی چندسطحی با وابستگی غیریکنواخت
Parallel Title Proper
Parallelization of Multi-Level Nested Loops with Non-Uniform Dependences
First Statement of Responsibility
/زهرا عبدی ریحان
.PUBLICATION, DISTRIBUTION, ETC
Name of Publisher, Distributor, etc.
: علوم ریاضی
Date of Publication, Distribution, etc.
، ۱۳۹۹
Name of Manufacturer
، کبیری
PHYSICAL DESCRIPTION
Specific Material Designation and Extent of Item
۱۵۲ص
NOTES PERTAINING TO PUBLICATION, DISTRIBUTION, ETC.
Text of Note
چاپی - الکترونیکی
CONTENTS NOTE
Text of Note
چکیده فارسی پی دی اف و فایل ورد قابل کپی نبود.
DISSERTATION (THESIS) NOTE
Dissertation or thesis details and type of degree
دکتری
Discipline of degree
علوم کامپیوتر
Date of degree
۱۳۹۹/۰۶/۲۶
Body granting the degree
تبریز
SUMMARY OR ABSTRACT
Text of Note
Most important scientific and engineering applications such as numerical analysis, simulation, image processing, etc have complex computations or extensive data. In most of these applications, a considerable amount of time is consumed by nested loops. Therefore, loops are the primary source of the parallelization of scientific and engineering programs. Loops are divided into two groups of uniform and non-uniform. Many parallelization compilers focus on the parallelization of nested loops with uniform dependences, and parallelization of nested loops with non-uniform dependences has not been extensively investigated. The purpose of this thesis is to parallelization of multi-level nested loops with non-uniform dependences on parallel systems with the aim of speeding up execution time. In this thesis, using the bin packing problem, two different approaches for this purpose are presented. In the first approach called intra-tiling, parallelization is done on iterations of a tile, and independent iterations are in a tile. In the second approach called inter-tiling, parallelization is done on tiles. In these approaches, by increasing load-balancing and decreasing inter-processor communication, total execution time is minimized. In the second approach increasing data-locality is also considered. In this approach, by increasing load-balancing, reducing inter-processor, and increasing data-locality total execution time is minimized. In this thesis, for each nested loops with non-uniform dependences, a tradeoff is considered between two proposed approaches, and these approaches are compared in terms of speed up, and one of them is selected as the parallelization method. Simulation and experimental results show that proposed approaches effectively reduce the total execution time of several benchmarks compared to the other tiling methods
PARALLEL TITLE PROPER
Parallel Title
Parallelization of Multi-Level Nested Loops with Non-Uniform Dependences