國立虎尾科技大學 |

Generating Thread-Level Parallelism in Nondeterministic Programs.

Record Type:	Language materials, manuscript : Monograph/item
Title/Author:	Generating Thread-Level Parallelism in Nondeterministic Programs./
Author:	Deiana, Enrico Armenio.
Description:	1 online resource (139 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 85-06, Section: A.
Contained By:	Dissertations Abstracts International85-06A.
Subject:	Design. -
Online resource:	click for full text (PQDT)
ISBN:	9798381175103

Generating Thread-Level Parallelism in Nondeterministic Programs.
Deiana, Enrico Armenio.

Generating Thread-Level Parallelism in Nondeterministic Programs. - 1 online resource (139 pages)

Source: Dissertations Abstracts International, Volume: 85-06, Section: A.

Thesis (Ph.D.)--Northwestern University, 2023.

Includes bibliographical references

Chip Multiprocessors (CMP) are everywhere, from mobile systems to servers. Thread-Level Parallelism (TLP) is the characteristic of a program that makes use of the parallel cores of a CMP to improve performance. Programming language abstractions are a way to generate TLP, which allows CMP to reach their full potential.This dissertation focuses on two main contributions:• STATS, a parallelizing compiler designed to extract TLP from nondeterministic programs by leveraging a novel programming language abstraction.• CARMOT, a tool intended to aid programmers in effectively utilizing this innovative programming language abstraction.TLP in today's programs is limited by data dependences that must be satisfied as the program executes. Nondeterministic programs suffer from this same limitation, but the nondeterminism gives them an additional degree of freedom that deterministic programs do not have: the ability to satisfy some dependences with many different data, which results in different outputs even when they run with the same input. Some of these outputs can be generated more quickly in parallel than others can. STATS is the first compiler to generate a new source of parallelism that has never been explored before, and it does so by taking advantage of this extra degree of freedom in nondeterministic programs. This resulted in STATS being able to achieve significant performance improvements in nondeterministic programs. To use STATS, developers have to express this additional degree of freedom in the code explicitly. STATS enables developers to encode this knowledge by extending the C++ language with a new abstraction.While STATS allows developers to obtain significantly more performance, using the C++ abstraction we introduced can be challenging in large code-bases. To assist developers in using this abstraction, we created a new tool called CARMOT. We found that CARMOT can also support developers in using many other programming language abstractions beyond the STATS abstraction. Hence, we developed an approach that generalizes the STATS-specific needs to reach many other modern programming language abstractions such as those offered by OpenMP pragmas and C++ features such as smart pointers.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2024

Mode of access: World Wide Web

ISBN: 9798381175103Subjects--Topical Terms:

595500
Design.
Subjects--Index Terms:

CompilerIndex Terms--Genre/Form:

554714
Electronic books.

Generating Thread-Level Parallelism in Nondeterministic Programs.
LDR:03614ntm a22003977 4500 001 1143895
005 20240517105021.5
006 m o d
007 cr mn ---uuuuu
008 250605s2023 xx obm 000 0 eng d
020 $a 9798381175103
035 $a (MiAaPQ)AAI30810799
035 $a AAI30810799
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Deiana, Enrico Armenio. $3 1468704
245 1 0 $a Generating Thread-Level Parallelism in Nondeterministic Programs.
264 0 $c 2023
300 $a 1 online resource (139 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 85-06, Section: A.
500 $a Advisor: Campanoni, Simone.
502 $a Thesis (Ph.D.)--Northwestern University, 2023.
504 $a Includes bibliographical references
520 $a Chip Multiprocessors (CMP) are everywhere, from mobile systems to servers. Thread-Level Parallelism (TLP) is the characteristic of a program that makes use of the parallel cores of a CMP to improve performance. Programming language abstractions are a way to generate TLP, which allows CMP to reach their full potential.This dissertation focuses on two main contributions:• STATS, a parallelizing compiler designed to extract TLP from nondeterministic programs by leveraging a novel programming language abstraction.• CARMOT, a tool intended to aid programmers in effectively utilizing this innovative programming language abstraction.TLP in today's programs is limited by data dependences that must be satisfied as the program executes. Nondeterministic programs suffer from this same limitation, but the nondeterminism gives them an additional degree of freedom that deterministic programs do not have: the ability to satisfy some dependences with many different data, which results in different outputs even when they run with the same input. Some of these outputs can be generated more quickly in parallel than others can. STATS is the first compiler to generate a new source of parallelism that has never been explored before, and it does so by taking advantage of this extra degree of freedom in nondeterministic programs. This resulted in STATS being able to achieve significant performance improvements in nondeterministic programs. To use STATS, developers have to express this additional degree of freedom in the code explicitly. STATS enables developers to encode this knowledge by extending the C++ language with a new abstraction.While STATS allows developers to obtain significantly more performance, using the C++ abstraction we introduced can be challenging in large code-bases. To assist developers in using this abstraction, we created a new tool called CARMOT. We found that CARMOT can also support developers in using many other programming language abstractions beyond the STATS abstraction. Hence, we developed an approach that generalizes the STATS-specific needs to reach many other modern programming language abstractions such as those offered by OpenMP pragmas and C++ features such as smart pointers.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Design. $3 595500
650 4 $a Computer engineering. $3 569006
650 4 $a Computer science. $3 573171
653 $a Compiler
653 $a Nondeterminism
653 $a Parallel programming languages
653 $a Thread-Level Parallelism
653 $a Chip Multiprocessors
655 7 $a Electronic books. $2 local $3 554714
690 $a 0984
690 $a 0464
690 $a 0389
710 2 $a Northwestern University. $b Computer Science. $3 1188771
710 2 $a ProQuest Information and Learning Co. $3 1178819
773 0 $t Dissertations Abstracts International $g 85-06A.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30810799 $z click for full text (PQDT)