Design Time

Table of Contents

• Data Pipeline
  • Nodes
    • .node.json to C++ Stubs
  • Graphlets
  • Application
• Schedule

When designing an application you don't have to implement any code (until you want to run the application). The following steps describe how to create an application bottom-up. As a result you can visualize the full data pipeline as well as generate and analyze the schedule.

Data Pipeline

Create JSON Descriptors Bottom-Up

Nodes

Each node can be designed using the UI. The UI allows to edit all properties of a node:

• the name
• input and output ports
• passes
• parameters

For each node this information is saved in a .node.json file.

.node.json to C++ Stubs

When starting to implement the logic for a node which has been designed before, the nodestub command line tool can be used to generate a set of C++ headers and source files from a .node.json file.

‍Note: this is not necessary to proceed with the next steps of designing graphlets and the application.

Graphlets

Similar to nodes, each graphlet can be designed using the UI. The UI allows to edit all properties of a graphlet:

• the name
• input and output ports
• parameters
• adding subcomponents (which can be either nodes and other graphlets)
• connection between ports, either from the graphlet or the subcomponents

For each graphlet this information is saved in a .graphlet.json file. There is no C++ code related to a graphlet.

Application

At the moment the application has to created by editing a .app.json file directly. In the future the UI will support visualizing and editing this file type too.

Schedule

When the worst-case execution time (WCET) is provided for each nodes pass, STM can generate a schedule for the application. See the build workflow on how to invoke the STM compiler.