-
Notifications
You must be signed in to change notification settings - Fork 0
Design Goals
Right now this is just sort of a list of goals that we are working toward.
-
Enhance the users ability to quickly visualize the bottlenecks in their application at high and extreme processor count scale.
-
Create a modular, component-based graphical user interface framework which supports the runtime loading and management of plugins that will display an intuitive visual representation of large-scale collected performance data to the user. This will include not only data collected by the Open|SpeedShop performance tool, but can be extended to support other debuggers and performance gathering applications.
-
Provide the capability to invoke and run Open|SpeedShop GUI from a remote machine, using a Windows, Linux or Macintosh operating system, other than the machine where the performance analysis data is being gathered. This capability includes the ability to run the GUI on any desktop while the whole Open|SpeedShop framework, including the source and runtime data, is running on the HPC computer system resource.
-
Facilitate methods, like the use of performance analysis wizards, for guiding the user through the process of identifying performance bottlenecks and suggest ways to solve common problems.
-
Provide an intuitive GUI feature that compares application performance analysis data gathered by Open|SpeedShop at multiple levels of granularity
-
Provide the capability to gather multiple experiments as explicit sets which will allow noise reduction and/or scaling experiments as well as parameter studies.