RASSP Token-based Performance Modeling Appnote

Abstract

Efficient design of complex systems requires modeling at multiple abstraction levels. This RASSP application note describes token-based performance modeling at an abstract level and how to use it within a rapid-prototyping environment. It describes rapid-prototyping methods developed and demonstrated on the RASSP project that improve simulation speed by several orders of magnitude.

Purpose

This application note should be read by system architects, software designers who are responsible for deciding how to partition software among multiple computer nodes, and hardware designers who are responsible for selecting appropriate network configurations and components. The basic concepts can be applied to other systems and other levels of abstraction.

Roadmap

1.0 Introduction to Token-based Performance Modeling

• 1.1 What is it
• 1.2 Why and when is it needed
• 1.3 How it fits into overall design process

• 2.1 Starting Information
• 2.2 Result Information

3.0 Metrics

• 3.1 Throughput
• 3.2 Latency
• 3.3 Utilization
• 3.4 Time-Lines

4.0 Performance Modeling Environments

• 4.1 Language and Tool Requirements
• 4.2 Proprietary Languages/Tools
• 4.3 Open-Standard Languages
• 4.4 Open-Standard Language-based Environments

5.0 Method

• 5.1 Descriptive Paradigm
• 5.2 HW/SW-Codesign process
• 5.3 Steps for Token-Based Performance Modeling
  • 5.3.1 Hardware Description
  • 5.3.2 Software Description
  • 5.3.3 Simulation
  • 5.3.4 Postprocessing/Analysis
  • 5.3.5 Recursion
  • 5.3.6 Model Validation/Maintenance

6.0 Example - SAR System

7.0 References