Blog

I’ve read Miguel’s post about how a few ‘details’ can impact Mono’s performance, and of course, following some mailing list discussions about small portions of code beeing changed (System.IO and StringBuilder come to mind) which should impact the perfomance as well.

So, I noticed that we have a TDD (Test Driven Development) project but we don’t usually run some benchmarks on code and Mono in general. Of course one usually tests a bit when a bottleneck is found and times like that, but there aren’t really any global benchmarking beeing performed.

In the same way such that when we fix  a regression we do a test case, why don’t we implement some benchmarks-test-cases also? We could even show the benchmark timmings in the Status Page  so users can get a picture of where imprements are from version to version.

There are a lot of people that keep running their application with older versions of Mono, the main reason beeing the previous record of regression (ASP.Net and Mono 1.1.9.x comes to mind). So if this applications aren’t beeing actively developed, or they just don’t need nothing from newer versions, why would they upgrade? Performance would be an outstanding argument.

How could we implement this?

This is a question I left to the masters.

My idea would be a server with each version of Mono, and Head too, where the benchmark tests would run for each version. After that is done we only need to keep running the benchmarks on the Head version, and when new benchmarks arrive we run them on older versions (if possible). Something like the Builds Status page.

A nice thing would be to have such benchmarking tests running on a Windows machine with both Mono and MS.Net, but that would need some requirements to be meet due to Microsoft’s benchmarking policies.

This entry was posted on Tuesday, January 24th, 2006 at 17:16 pm and is filed under Mono. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.