Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Alliance Grants

University of Toronto

$119,378.00

Next Generation Reconfigurable Acceleration Devices: Architectures, Tools and Applications

FY 2024-25

Summary
Computing has transformed society, enabling ubiquitous communications, on-demand entertainment, speech recognition, and more. While computation demand is growing for deep learning (DL) and other applications, the traditional efficiency gains afforded by transistor scaling are slowing down. To fill this gap, we need computational devices that are more efficient yet reprogrammable to allow new applications. Field-Programmable Gate Arrays (FPGAs) can be reconfigured at the hardware level, enabling energy-efficiency gains of 10x or more vs. processors for many deep learning and datacenter applications. We seek to advance on three related fronts: architecting better reconfigurable devices, enhancing computer-aided design (CAD) tools to explore and enable these new architectures, and implementing key applications like deep learning inference and datacenter infrastructure acceleration on these new devices. We will explore multiple opportunities to make more efficient reconfigurable architectures, which we term RADs. First, while FPGAs are highly programmable, not every aspect of the device needs to be reconfigurable: by judiciously adding "hardened" accelerator blocks, we seek to find the sweet spot between programmability and performance. Second, while current FPGAs connect everything with a very flexible, but low level, programmable interconnect, we believe the new RADs we explore will benefit from pre-built networks-on-chip (NoCs) and will explore this new style communication. Finally, while transistor scaling is slowing down, we can now stack multiple silicon dice to create a 3D system; we will explore how RADs can leverage this new capability. To determine which new chip ideas are the most efficient, we must create new Computer-Aided Design (CAD) tools that can map applications into these new architectures and quantitatively evaluate the resulting efficiency. Similarly, to ensure new architectures indeed perform well on key applications like deep learning inference, we will create multiple types of deep learning accelerators that leverage our new RADs.
Program
Alliance Grants
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
590565-2023
Name nom
Betz, Vaughn VT
Department département
Electrical & Computer Engineering, Edward S. Rogers Sr. Department of
Organization id
31
Institution établissement
University of Toronto
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2023
Award amount
119378
Program id
ALLRP
Program name en
Alliance Grants
Program name fr
Subventions Alliance
Group en
RESEARCH PARTNERSHIPS
Group fr
PARTENARIATS DE RECHERCHE
Committee code
96
Committee name en
RPP Internal Decision Cttee
Committee name fr
Cté de décision interne - PPR
Area of application code
1207
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Information, computer and communication technologies
Area of application fr
Technologie de l'information, des ordinateurs et des réseaux de communications
Research subject code
2719
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Computer architecture and design
Research subject fr
Architecture d'ordinateur et conception
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Next Generation Reconfigurable Acceleration Devices: Architectures, Tools and Applications
Keywords
cad; compute acceleration; computer-aided design; deep learning acceleration; field programmable gate array; fpga; fpga architecture; integrated circuits; multi-die integration; network on chip
Application summary
Computing has transformed society, enabling ubiquitous communications, on-demand entertainment, speech recognition, and more. While computation demand is growing for deep learning (DL) and other applications, the traditional efficiency gains afforded by transistor scaling are slowing down. To fill this gap, we need computational devices that are more efficient yet reprogrammable to allow new applications. Field-Programmable Gate Arrays (FPGAs) can be reconfigured at the hardware level, enabling energy-efficiency gains of 10x or more vs. processors for many deep learning and datacenter applications. We seek to advance on three related fronts: architecting better reconfigurable devices, enhancing computer-aided design (CAD) tools to explore and enable these new architectures, and implementing key applications like deep learning inference and datacenter infrastructure acceleration on these new devices. We will explore multiple opportunities to make more efficient reconfigurable architectures, which we term RADs. First, while FPGAs are highly programmable, not every aspect of the device needs to be reconfigurable: by judiciously adding "hardened" accelerator blocks, we seek to find the sweet spot between programmability and performance. Second, while current FPGAs connect everything with a very flexible, but low level, programmable interconnect, we believe the new RADs we explore will benefit from pre-built networks-on-chip (NoCs) and will explore this new style communication. Finally, while transistor scaling is slowing down, we can now stack multiple silicon dice to create a 3D system; we will explore how RADs can leverage this new capability. To determine which new chip ideas are the most efficient, we must create new Computer-Aided Design (CAD) tools that can map applications into these new architectures and quantitatively evaluate the resulting efficiency. Similarly, to ensure new architectures indeed perform well on key applications like deep learning inference, we will create multiple types of deep learning accelerators that leverage our new RADs.