Ninja is faster than Visual Studio, especially when generating the WebCore bindings.
Created attachment 257358 [details] Patch
(In reply to comment #0) > Ninja is faster than Visual Studio, especially when generating the WebCore > bindings. Do you have clean / incremental numbers?
Created attachment 257391 [details] Patch
(In reply to comment #2) > (In reply to comment #0) > > Ninja is faster than Visual Studio, especially when generating the WebCore > > bindings. > > Do you have clean / incremental numbers? I don't yet, but ninja is probably about twice as fast as sequentially generating all the bindings for WebCore, which is what the CMake build does without this. I would estimate this brings the CMake build time on Windows from about an hour to about a half hour.
Comment on attachment 257391 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=257391&action=review > Source/cmake/OptionsWindows.cmake:30 > + endif endif ()
Created attachment 257618 [details] Patch
Created attachment 257619 [details] Patch
Comment on attachment 257619 [details] Patch Seems like a safe bet.
http://trac.webkit.org/changeset/187470
(In reply to comment #2) > (In reply to comment #0) > > Ninja is faster than Visual Studio, especially when generating the WebCore > > bindings. > > Do you have clean / incremental numbers? Today, a clean build with CMake and MSBuild took 66.7 minutes. A clean build with CMake and Ninja took 27.7 minutes. That's just over 2x as fast, mostly because it parallelizes the WebCore generating phase. MSBuild: https://build.webkit.org/builders/WinCairo%2064-Bit%20Release/builds/47680/steps/compile-webkit/logs/stdio Ninja: https://build.webkit.org/builders/WinCairo%2064-Bit%20Release/builds/47698/steps/compile-webkit/logs/stdio