P.B. Jayaraj, R. Gopalakrishnan, and V. Jain (India)
Software Performance Engineering, Software Optimization, Software Performance Analysis, Linux, Embedded Systems
This paper1,2 presents a case study of dynamic performance analysis and bottleneck identification of a large application software stack on a Linux-based embedded platform. We use some standard techniques for performance analysis, but we also try out some techniques that are not so common for performance characterization and visualization. The methodology used for measurement collection is generic and hardware independent (software-centric), and it is relevant for any effort that develops embedded applications on a Linux platform. Using this methodology, we have analyzed the performance of the JUICE (Joint User Interface and Control Engine) software stack developed within Philips, running on a Linux-based embedded platform. We have identified relevant metrics, detailed the tools and methods by which they can be collected (given the existing open source tool chain) and discussed the salient aspects of performance analysis and visualization of the collected measurements. We conclude that the open-source tool chain provides good support for such measurement collection and can aid further performance analysis efforts. We also discuss the various gaps and issues in the existing tooling that we have come across in this effort.
Important Links:
Go Back
