Grant: Discovery Grants Program - Individual
University of Alberta
$31,000.00
Automatically exploring parameter space for code transformations in compilers
EF 2013-14
Résumé
Compilers are responsible for a significant portion of the performance improvement observed in computer systems each year. The quality of the code generated by a commercial compiler has a very significant impact in the amount of resources --- time, storage and energy consumption --- required to perform computing tasks. Traditionally compiler developers use limited manual experimentation to determine a reasonable set of parameters for the code transformations that are built into contemporary compilers. This research aims to demonstrate that there are advantages to replace this ad-hoc selection of parameters with an automated selection based on machine learning. Future compilers that implement such an automated approach will potentially identify sets of parameters that result in generated code that consumes less resources. These compilers will more readily adapt to new sets of program workloads. Moreover, the compiler development process itself will be more efficient, thus increasing the productivity of compiler developers.The proposed research will also address the issue of combining information collected from multiple profiling runs of the same computer program in a way that provide the most information about the variations of execution due to changes in the data set processed by the program. Better profiling tools will enable compiler developers to more aggressively use speculative code transformations that are based on feedback collected from previous executions of the program.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Alberta, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
239013-2010
Name nom
Amaral, Jose
Department département
Computing Science
Organization id
9
Institution établissement
University of Alberta
Province en
Alberta
Province fr
Alberta
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2013
Competition year année de concours
2010
Award amount
31000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
Discovery grants programs
Group fr
Programmes de subventions à la découverte
Committee code
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
801
Area of application group en
Information and communication services
Area of application group fr
Services d'information et de télécommunications
Area of application en
Computer software
Area of application fr
Logiciel
Research subject code
2720
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Computer systems software
Research subject fr
Logiciel de systèmes informatiques
Partie
4
Nb partie
5
Application title
Automatically exploring parameter space for code transformations in compilers
Keyword
NSERC
Application summary
Compilers are responsible for a significant portion of the performance improvement observed in computer systems each year. The quality of the code generated by a commercial compiler has a very significant impact in the amount of resources --- time, storage and energy consumption --- required to perform computing tasks. Traditionally compiler developers use limited manual experimentation to determine a reasonable set of parameters for the code transformations that are built into contemporary compilers. This research aims to demonstrate that there are advantages to replace this ad-hoc selection of parameters with an automated selection based on machine learning. Future compilers that implement such an automated approach will potentially identify sets of parameters that result in generated code that consumes less resources. These compilers will more readily adapt to new sets of program workloads. Moreover, the compiler development process itself will be more efficient, thus increasing the productivity of compiler developers.The proposed research will also address the issue of combining information collected from multiple profiling runs of the same computer program in a way that provide the most information about the variations of execution due to changes in the data set processed by the program. Better profiling tools will enable compiler developers to more aggressively use speculative code transformations that are based on feedback collected from previous executions of the program.