28. January 2014 21:24
This is embarrassing. For the n-th time during the past couple of years I've felt an unease waiting for our projects (read: solutions) to compile. I kept seeing this:
This is MSBuild using 1 (!), yes, one!, of the 8 CPU cores I've sitting in my machine to get my work done. What about the other 7? Why don't you use them, MSBuild? With that single core, currently my simple local build of our project discoverize takes around 36 seconds:
Tell MSBuild to use all cpu cores
Well, it's as easy as adding /m or /maxcpucount to your msbuild command line build to boost your build times:
Down to 8 seconds with 3 additional characters: [space]/m. That's easily a 4.5 times improvement!
Your mileage may vary
Of course, every project is different, so your speed increase might be higher or a lot lower than what I've seen. But it's an easy measure to get at least some improvement in build times with very little effort. Don't trust Visual Studio on that one, though – the solution builds slowly there, still.
For reference, let me tell you, that the /maxcpucount switch can actually take a parameter value like so: /maxcpucount:4. So if you lots of other stuff going on in the background or I don't know for what reason, really, you can limit the number of cpus used by MSBuild.
Props to the Orchard team for a highly parallelizable build
One of the specifics of the Orchard source code that's the base for discoverize is the very loose coupling between the 70+ projects in the solution. This allows MSBuild to distribute the compilation work to a high number of threads because there are almost no dependencies between the projects that MSBuild would have to respect. Great work!