10.01.2013 Views

ComputerAided_Design_Engineering_amp_Manufactur.pdf

ComputerAided_Design_Engineering_amp_Manufactur.pdf

ComputerAided_Design_Engineering_amp_Manufactur.pdf

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

• Programming logic and VLSI arrays,<br />

• Communication networks and protocols,<br />

• Neural networks,<br />

• Digital filters, and<br />

• Decision models.<br />

However, in most of these applications, a different graphical model is used instead, such as data<br />

flow for digital signal processing and state diagrams for communication protocols. As a result, it is<br />

desired that the tool can also draw and analyze these graphical models. The tool has been enhanced<br />

in this respect: for rapid development, the tool converts these graphs into Petri nets internally for<br />

analysis. Therefore, there is no need to develop new algorithms and codes for each type of graph.<br />

Features of the tool include:<br />

A. Data Communication: We have developed a very powerful CAD tool for designing protocols and<br />

Petri nets. 11 Currently, the tool is able to draw Petri nets, state diagrams, data flow graphs, finite state<br />

machines, and general graphical objects. Once the graph is drawn, the tool can analyze, simulate,<br />

reduce, and synthesize it. Few existing tools are capable of such an integration. The tool has been<br />

enhanced to synthesize error-recovery protocols with great time-complexity reduction, simulate<br />

extended finite state machines, hierarchically model and simulate Petri nets, automatically generate<br />

“C” codes, and generate unique I/O (UIO) conformance test patterns. We are extending the tool to<br />

• Simulate queuing networks,<br />

• Software engineering for protocol design implementation and documentation,<br />

• Advanced synthesis and reduction of communication protocols using PN,<br />

• Implementation of synthesis of local entities,<br />

• Complexity reduction in synthesizing multiparty protocols, and<br />

• Animation of protocols with real objects.<br />

B. Network Simulation: We66 have enhanced our protocol design tool for object-oriented simulations.<br />

After drawing network objects and interconnecting them, we can perform simulation.<br />

The “Step” mode allows easy debugging and detailed trace of network behavior, which<br />

is useful for computer-aided education. To illustrate, we have applied the tool to the bitonic<br />

sorting part of the Batcher-bayan network. Automatic code generation helps produce efficient<br />

“C” code for non-X-Window environments after verification via simulation.<br />

C. Parallel Processing: The above tool has been extended65 to analyze and simulate performance<br />

of DFGs. It incorporates a unique, efficient algorithm that we developed to do rate-optimal<br />

steady state scheduling without initial transients. We have extended the tool to draw multiple<br />

rate DFGs and simulate. It can find critical and subcritical loops and perform rate-optimal<br />

scheduling. The user can randomly pick a node to check its input and output values against<br />

its functional type after the simulation. The critical loops, scheduling ranges, and processor<br />

assignments can be displayed in a graphical fashion. It can also display critical paths and Gann<br />

charts for DFGs without any loops. Few tools are capable of such integration and ease of use.<br />

The final matrix theory developed is the first of its kind and is used throughout the design. It<br />

calculates the iteration bound, finds critical loops, derives formulas for scheduling ranges, and<br />

performs a fast processor assignment with rate-optimal scheduling. This eliminates the need for<br />

inequality charts for finding the scheduling ranges by Heemstra et al. 67 Most approaches repetitively<br />

update scheduling ranges (which is very time consuming), and our theory can eliminate many of<br />

these updates. Benchmark testing indicates remarkable (thousandfolds) speedup of scheduling, especially<br />

for large DFGs that take less than 1 second compared with other approaches that take several hundred<br />

seconds. In addition, the resulting scheduling needs fewer numbers of processors than other approaches.<br />

The theory has been updated to include the case of resource constraints and iteration periods<br />

greater than iteration bounds.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!