Performance Analysis

Eclipse vs. bake

Test system: Intel Xeon W3520 (4x 2.67 GHz), 4GB Ram, Windows XP, HDD Raid
Real test application with ~ 200 projects
ruby 1.9.2

Eclipse 3.6.1, CDT 6bake
Build whole workspace13:01 min7:19 min
Build again without clean3:05 min0:50 min
Clean whole workspace0:41 min0:08 min
Clean the cleaned workspace again0:35 min0:02 min
Memory consumption of build environment
(exclusive compiler, etc.)		Eclipse needs additional
300MB, stays in memory		35MB peak,freed after build

ruby vs. ruby

Test system: Intel Xeon W3520 (4x 2.67 GHz), 4GB Ram, Windows XP, HDD Raid
Real test application with ~ 200 projects
Results averaged out of 30 runs

ruby 1.8.6p398ruby 1.8.7p352ruby 1.9.2p180ruby 1.9.3p0
Build whole workspace13:58 min8:07 min7:19 min7:28 min
Build single file1,20 sec1,87 sec2,38 sec1,29 sec

ruby 1.8.6 uses only ONE native thread for multiple ruby threads.
ruby 1.9.2 for Windows needs very long to startup for complex applications, which is fixed in 1.9.3

CMake with Unix makefiles vs. bake

Test system: Intel Xeon W3520 (4x 2.67 GHz), 4GB Ram, Windows XP, HDD Raid
Real test application with ~ 100 projects

  • Build whole workspace: bake is ~10% faster
  • Build again without clean: bake is ~50% faster
  • Clean whole workspace: bake is ~80% faster

    More details will follow.