14/32  

Community-wide VM Performance tracking

Mentor: Stefan Marr
Second mentor:
Level: Beginner
Invited students:
Students interested: stepa alimov, Saurav Bhattacharya(lightly)

Description

The various Smalltalk VMs are under constant development and maintenance. To avoid performance regressions, they need to be monitored and tracked with every change. The goal is to collect a set of standard benchmarks and set up the necessary infrastructure based on Jenkins/Buildbot and ReBench, SMark, and CodeSpeed.

Technical Details

The main parts of the infrastructure is already available as Smalltalk-independent projects. With ReBench and SMark, the Smalltalk community even got the bits and pieces for running benchmarks automatically. However, it still misses a setup that can benefit from the continuous integration servers to benchmark the evolving VM codebases.

Benefits to the Student

Get an insight in setting up and managing typical cross-technology projects. Get a better understanding of the pitfalls of performance evaluation and benchmarking on modern platforms.

Benefits to the Community

An additional quality insurance for core elements of the ecosystem, i.e. the VMs. Furthermore, an objective measure to compare different Smalltalk implementations.




Updated: 18.3.2012