MDI_Logo
      About Us   |   Contact Us   |   Evaluation
MDI_Topstrip
Mirabilis Design Navigation
MDI - About Us MDI - Technology MDI - Presentations MDI - Products MDI - Demonstrations

      Evaluation of Software flow on a Hardware Platform

Click here to view the interactive VisualSim Block Diagram and model

Click here to execute the VisualSim model and view analysis results

The system is a hardware platform that can be utilized to execute a variety of software applications. The model constructed here studies the bus utilization for a varying amount MPEG traffic load. The rest of the system from the MailBox and the Application Processor Bus is abstracted as a traffic model.

The purpose of this experiment is demonstrate the use of the Schedulers for exploring hardware architectures and to describe the data flow model.

System Overview

The block diagram in Figure 1 shows the platform architecture. The architecture consists of the separate bus and processor system interfaced using a MailBox with associated memory. The incoming frames are stored in the Memory (MDP). Depending on the type of data (Audio or Video), the data is processed by DP1 or DP2. The processors can make request to their individual caches (CL1 and CL2) to retrieve the application software. The two section work identical.


Figure 1: Block Diagram of the Hardware Platform

Model Description:


Figure 2: Translation of the Dual Processor portion of the MPEG system architecture into VisualSim

The data flow in ( Model 1 )is described above.


Model Notes:

  • Requests made to the Cache and processors are made via the Bus. Request can arrive at the Bus in parallel from multiple sources or operations. The Schedulers handle the contention and arbitrate between them.
  • The model describes the flow of data through the system. If a specific application and data frame is considered, the Schedulers must be refined to make a detailed definition of the processor and bus architectures.
  • The current model utilizes a statistical model to generate the traffic. A more realistic traffic could be generated from a software routine scripted in a text file or using trace data.
  • The Application Processor Bus and the MailBox are modeled as overhead and contend only for the Bus resource. An expanded model may consider adding more details for that section of the design for increased accuracy.

Analysis

Two results are captured in ( Model 2 ).

  • Timeline plot showing the period of time that various devices were turned on and off. This gives an indication on where the power consumption is and the devices that have the greatest potential for bootleneck.
  • A summary of the utilization of the different devices including the Dual Processors, Caches and the Bus.
The results indicate the bus is heavily utilized. This may indicate that the processors must have a point-to-point connection with the cache. Moreover the processor is not being utilized at all. Depending on the application, it is possible to utilize these elements for other resources or go with a lower priced processor.
  Copyright 2008© Mirabilis Design Inc. All Rights Reserved. Best Viewed in 800x600 resolution. | Site Map | Technical Support