Configuration

In the top-right corner, you can select which algorithm type you wish to run. For each of these algorithm types, it is possible to have differents kinds of implementations, but they all accept the same kind of input.

Each of these algorithm types can be instantiated to create a configuration for that specific type. New configurations can be added, and configurations can be saved to/loaded from disk. This is useful for restoring an experimental setup, as well as exploring the output/statistics gathered from the experiments themselves. To prevent losing your configuration, the tool will save the state of your configuration when you close the tool and will restore this state the next time you start it up.

Each configuration contains the following:

• Algorithm library: a set of configurations for specific implementations of the algorithm type.

• Input library: a set of inputs which can be used to run these algorithms. Inputs can be saved to/loaded from disk or imported using a custom importer. Furthermore, differents kinds of custom input editors may be defined for visualizing and editing the input.

• Workload: a set of outputs and statistics produced by running one or more algorithm run configurations. A run configuration is composed of:

  • A specific input chosen from the input library.
  • A specific implementation configuration chosen from the algorithm library
  • An amount of iterations for which to repeat running the given algorithm on the given input. This is useful for gathering multiple samples.

  Run configurations can be executed by selecting one or multiple of them (indicated in blue) and clicking Compute. Note that you can select multiple configurations uing CTRL+CLICK or SHIFT+CLICK on the run configuration name, akin to selecting files/folders in the Windows file explorer. Running multiple configurations at the same time will execute each run configuration in a separate thread. Using the same way of selecting multiple run configurations, it is also possible to clear the output of multiple run configurations at the same time by clicking the Reset button. Based on the color of the Amount column, you can tell what the status of a run configuration is. It is either idle (red), in progress (yellow), or green (completed all iterations).

  As alluded to earlier, when the algorithm type configuration is stored to disk, the algorithm workload configuration as well as any produced workload output/statistics are saved.

Output Visualization

In the main panel on the left, visualizations can be shown for the output of the run configurations. This can either be the output of the algorithm or (live) statistics that were gathered. The type of visualization can be chosen by hovering your mouse over the top edge and selecting the type of visualization in the dropdown menu on the left.

By default, the run configurations that are visualized inside of a visualization are chosen automatically, but you can disable this by clicking the Choose button and selecting one or more configurations. You can go back to automatically selecting the run configurations shown by a visualization by clicking the Auto button.

Finally, by clicking the Split button, it is possible to show multiple visualizations at the same time by splitting the visualization panel either vertically or horizontally. This can be done as often as you want to show many different visualizations simultaneously. Alternatively, by selecting the Unsplit option, a visualization can be removed and a neighboring visualization panel will be expanded to fill the space.

The following pre-defined visualizations are available for any algorithm type:

• Statistics: table containing running time and iterations completed, algorithm configuration, and input/output statistics. Can show statistics of multiple runs simultaneously by expanding to multiple columns.
• Log: raw logging generated by the algorithm for one specific run.