On 10/04/2011, at 4:38 PM, Russel Winder wrote: > As an anecdotal data point, and not an experimental result, Gradle > 1.0-milestone-2 seems faster than earlier Gradles. Moreover, it seems > just as fast as Maven for equivalent tasks/goals on the same project -- > based on timing equivalent tasks/goals on the same project.
There were some performance improvements in milestone-2, in test execution and up-to-date checking. So the build for a typical jvm-based project should be faster now (these fixes shaved about 9% off the clean build time for a large enterprise build, and about the same for my benchmarking projects). There's a couple more improvements I want to make which didn't make it into milestone-2. btw, are you using the daemon for these comparisons? I think, to be fair, we should use the maven shell to compare against the gradle daemon. At least, until the daemon becomes the default 'out-of-the-box' behaviour. > > It is worth noting that the time of execution that Maven reports is not > real time in the way that Gradle reports time. People can therefore be > lulled into the mis-perception that Maven is faster because they believe > the times reported by the two tools. Even though this is clearly > inaccurate, it will drive perception of the masses. people may will > believe Maven is faster than Gradle even though it isn't. Not sure we can do much about this. I think if we're going to show a build time, it should be the most accurate we can manage. One option, perhaps, would be to fix maven so that it does a better job of reporting the elapsed build time. Another option might be to remove the build time from the default gradle output, so that you have to explicitly 'time' both commands. -- Adam Murdoch Gradle Co-founder http://www.gradle.org VP of Engineering, Gradleware Inc. - Gradle Training, Support, Consulting http://www.gradleware.com
