Grant: Discovery Grants Program - Individual
McMaster University
$36,000.00
Code Specialization and Data Compression for Heterogeneous Computing
FY 2023-24
Summary
The growth of computing systems, arguably the most important means for human discoveries, has been constrained by fundamental physical boundaries and limited energy resources. The ability for transistor density, the mainstream methodology to increase computing power for decades, is saturating. Furthermore, the energy consumed by computing systems outpaces energy production and contributes to carbon emissions, which are increasingly alarming due to their impact on climate change. The solution is the intelligent use of transistors by building application-specific accelerators alongside existing generic processors, leading to heterogeneous processors. However, currently, applications and algorithms must be manually restructured to use these diverse resources, which is tedious. Automated systems, such as compilers, facilitate the re-developing applications for heterogeneous processors, but they do not keep up with the fast pace of emerging new accelerators and communication restrictions of these processors. My research program builds software and hardware systems to accelerate scientific and machine learning applications on future heterogeneous architectures to deliver high-performance and resilient systems with sustainable energy consumption. This research program focuses on building software systems and advances the field of computing systems on three fronts: 1) code specialization techniques to close the gap between application-specific accelerators with the application's computation pattern, 2) compression algorithms to reduce the memory footprints and the number of operations by trading accuracy for performance, and 3) algorithm comprehension for compilers and frameworks to automate algorithm level decision making for numerical solvers. This research program builds the next generation of software systems meeting ever-growing computational needs for artificial intelligence and scientific simulations. The novel automated software systems in this proposal provide a fast and portable code using accelerators in heterogeneous processors. The proposed systems also reduce communication between computation resources to reduce power consumption. This research program has the potential to make a significant impact on the socioeconomic of Canadians and people across the globe. The efficient code from this research will be used in computing servers and devices, accelerating scientific discoveries and enhancing human equity. The COVID vaccine discovery is a recent successful example of acceleration using fast supercomputers and simulation software. The reduced power consumption as a result of the systems of this proposal will save on energy resources and reduce carbon emissions. This research program also trains software developers to address the shortage of software experts anticipated by the job bank of Canada. This program also improves inclusion in the technology sector of Canada by preparing HQP from women and under-represented groups.
Program
Discovery Grants Program - Individual
Recipient
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
RGPIN-2023-04897
Name nom
Cheshmi, Kazem
Department département
Electrical and Computer Engineering
Organization id
27
Institution établissement
McMaster University
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2023
Competition year année de concours
2023
Award amount
36000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR 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
2950
Research subject group en
Applied mathematics
Research subject group fr
Applied mathematics
Research subject en
Mathématiques appliquées
Research subject fr
Mathématiques appliquées
Application title
Code Specialization and Data Compression for Heterogeneous Computing
Keyword
code generation; heterogeneous computing; machine learning; parallel programming model; power consumption; runtime systems; scalable algorithms; scientific computing; sparse computation; verification
Application summary
The growth of computing systems, arguably the most important means for human discoveries, has been constrained by fundamental physical boundaries and limited energy resources. The ability for transistor density, the mainstream methodology to increase computing power for decades, is saturating. Furthermore, the energy consumed by computing systems outpaces energy production and contributes to carbon emissions, which are increasingly alarming due to their impact on climate change. The solution is the intelligent use of transistors by building application-specific accelerators alongside existing generic processors, leading to heterogeneous processors. However, currently, applications and algorithms must be manually restructured to use these diverse resources, which is tedious. Automated systems, such as compilers, facilitate the re-developing applications for heterogeneous processors, but they do not keep up with the fast pace of emerging new accelerators and communication restrictions of these processors. My research program builds software and hardware systems to accelerate scientific and machine learning applications on future heterogeneous architectures to deliver high-performance and resilient systems with sustainable energy consumption. This research program focuses on building software systems and advances the field of computing systems on three fronts: 1) code specialization techniques to close the gap between application-specific accelerators with the application's computation pattern, 2) compression algorithms to reduce the memory footprints and the number of operations by trading accuracy for performance, and 3) algorithm comprehension for compilers and frameworks to automate algorithm level decision making for numerical solvers. This research program builds the next generation of software systems meeting ever-growing computational needs for artificial intelligence and scientific simulations. The novel automated software systems in this proposal provide a fast and portable code using accelerators in heterogeneous processors. The proposed systems also reduce communication between computation resources to reduce power consumption. This research program has the potential to make a significant impact on the socioeconomic of Canadians and people across the globe. The efficient code from this research will be used in computing servers and devices, accelerating scientific discoveries and enhancing human equity. The COVID vaccine discovery is a recent successful example of acceleration using fast supercomputers and simulation software. The reduced power consumption as a result of the systems of this proposal will save on energy resources and reduce carbon emissions. This research program also trains software developers to address the shortage of software experts anticipated by the job bank of Canada. This program also improves inclusion in the technology sector of Canada by preparing HQP from women and under-represented groups.