Evolution in Architectures and Programming Methodologies of Coarsegrained Reconfigurable Computing

1
Evolution in Architectures and Programming
Methodologies of Coarse-grained Reconfigurable
Computing

• By Zain-ul-Abdin and Bertil Svensson

2
Outline

• Motivation
• Scope
• Architectural Characteristics
• Computation Models
• Coarse-grained Reconfigurable Archietctures
• Discussion
• Future Trends

3
Motivation

• Emergence of architectures different from Von
  Neumans paradigm of computing
• Growing focus towards Reconfigurable computing
  based on data-streaming
• Increased Performance
• Run-time Reconfiguration
• Power Efficiency
• Need for Models of Computations
• Enhance the understanding of the developer to
  organize computations
• Utilizes the expilicit concurrency in the
  underlying HW

4
Scope
5
Architectural Characteristics

• Granularity
• Fine-grain/Coarse-grain
• Reconfigurability
• Reconfiguration Overheads Latency, Speed
• Interconnection Networks
• Characteristics Network topology, Flow control,
  Routing, QoS
• Energy Efficiency
• Scalability

6
Computation Models

• Stream Processing Model
• Languages StreamIt, StreamC/KernelC, TDF
• CSP based Model
• Languages Handel-C, Streams-C, Mobius
• Spatial Computation Model
• Pegasus
• Kahn Process Networks
• Compaan, Ambric structural object programming
  language

7
Coarse-grained Reconfigurable Architectures

• Classification
• Hybrid architectures
• Array of functional units
• Array of processors
• Array of soft processors

8
Coarse-grained Reconfigurable Architectures

• Array of Soft Processors
• Mitrionics Virtual Processor

9
Discussion

• Summary of features of MoC

10
Discussion

• Relationship between MoC and CG Arch.

11
Discussion

• Trends in Relative complexity of CG Arch.

12
Future trends

• Architectural trends (GALS)
• Computing Model trends (Streaming/CSP/Process
  Networks)
• Technology trends (i.e. Nanoscale devices)