Časovno neomejena optimizacija strojne kode

magistrsko delo

Žan Magerl (Author), Boštjan Slivnik (Mentor)

Abstract

Eden izmed najpomembnejših korakov v postopku prevajanja je optimizacija, s pomočjo katere prevajalnik proizvede hitrejšo in učinkovitejšo strojno kodo. Prevajalniki morajo optimizacijo izvajati hitro, saj programer ne želi dolgo čakati na prevod svojega programa. V magistrskem delu želimo ugotoviti, če lahko s tem, da se ne oziramo na čas, izboljšamo že prevedene programe. Za odgovor na to vprašanje razvijemo optimizator za optimizacijo MMIX-ove strojne kode. Optimizator preizkusimo na množici testnih programov in analiziramo dobljene rezultate. Ugotovimo, da je optimizacija strojne kode veliko bolj odvisna od vrste in kvalitete optimizacij, ki jih izvajamo, kot pa od časa, ki ji ga namenimo. Kljub temu uspemo razviti optimizator, ki zmanjša tako dolžino programa kot tudi število izvedenih ukazov programa med njegovim izvajanjem.

Keywords

optimizacija;preiskovanje;strojna koda;MMIX;magisteriji;

Data

Language:	Slovenian
Year of publishing:	2022
Typology:	2.09 - Master's Thesis
Organization:	UL FRI - Faculty of Computer and Information Science
Publisher:	[Ž. Magerl]
UDC:	004.4(043.2)
COBISS:	128933891
Views:	54
Downloads:	19
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Machine code optimization without time constraints
Secondary abstract:	One of the most important steps in the process of compilation is optimization, with which compiler outputs quicker and more efficient machine code. Compilers have to optimize quickly, since programmers do not want to wait long to get compiled programs. In master thesis we want to find out if we can improve already compiled programs when we are not constrained with time. In order to find the answer to this question we have developed an optimizer that optimizes MMIX machine code. We test the optimizer on the set of test programs and analyse the results. We find out that machine code optimization is not constrained with time, but with the quality of designed optimizations. Despite that, we develop an optimizer, that successfully reduces the length of the program and the number of instructions during execution of the program.
Secondary keywords:	compilers;optimization;search;machine code;MMIX;computer science;master's degree;Prevajalniki (programska oprema);Računalništvo;Univerzitetna in visokošolska dela;
Type (COBISS):	Master's thesis/paper
Study programme:	1000471
Embargo end date (OpenAIRE):	1970-01-01
Thesis comment:	Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages:	55 str.
ID:	16853633