[Project 3: Benchmarking] Full proposal

[aygarwal]

Hello!
I have been working on writing my full proposal, and I had two questions:

• What are the current parameters by which a regression is detected? Is it simply visually, by looking at the graph of the performance of that module over time, or is there a threshold value / other metric that is used to classify regressions?
• Also with regards to identifying the exact lines of code / calls that cause a slowdown, I understand that cProfile / perf are used for profiling. Similar to the first question, is there a specific agreed upon value by which one can say that this is a slowdown?

[aygarwal]

@orbeckst I'd really appreciate getting your help on my full proposal if possible.

Along with the above questions, I also wanted to know if there was a preference(if any) towards increasing more benchmark coverage, or identifying regressions?

[orbeckst]

At the moment there is not specific process for detecting regressions. You can develop one as part of your project.

[orbeckst]

Ideally your work has impact beyond the GSOC period. Think about so how you can make benchmarking and performance improvements part of the development process. For instance, if you make it easier to add new benchmarks then other contributors can work on them. If you make it easy to detect regressions and provide docs on how to fix these, other will be able to do it in the future.

[Dreamstick9]

Hey @orbeckst could you also take a look at this as well :- #5354
as this is also under Project 3, any feedback would be helpful

[aygarwal]

@orbeckst

I've submitted my full proposal. Thank for your help!