In the following video, INESC-ID introduces the Cache-Aware Roofline Model (CARM) and demonstrates the integration between the CARM Tool and the Paraver suite from BSC.This video highlights key features of the integration, including coherent data representation and time synchronization between the tools.

We start with setting up the CARM Tool GUI, from cloning the repository to configuring the required PATH variables so that Paraver can interact with it. We then walk through a practical example of trace analysis: opening a trace in Paraver with CARM counters (see documentation for details), launching the CARM GUI directly from a Paraver timeline, and exploring it in the browser with the CARM GUI.

We hope that this synergy between CARM and Paraver will open new opportunities for application analysis, code optimization, and bottleneck detection for HPC centers.

Try the CARM Tool GUI and its Paraver integration: https://github.com/champ-hub/carm-paraver

Get the full CARM Tool to generate CARM results for your machine: https://github.com/champ-hub/carm-roofline

-- José Morgado, Aleksandar Ilic and Leonel Sousa (INESC-ID)