Hi, I did some more test with different recent versions of trunk. In all tests I used -J-server and -J-Djruby.objectspace.enabled=false. I've included the "full action" line only for the sake of readability.
The first line represents JIT turned on, the second line is with -J-Djruby.jit.enabled=false Here we go: rev 4383 full action: 43.340000 0.000000 43.340000 ( 43.340000) full action: 36.770000 0.000000 36.770000 ( 36.770000) rev 4409 full action: 45.034000 0.000000 45.034000 ( 45.034000) full action: 45.921000 0.000000 45.921000 ( 45.922000) rev 4414 full action: 42.952000 0.000000 42.952000 ( 42.952000) full action: 38.571000 0.000000 38.571000 ( 38.571000) rev 4415 full action: 35.767000 0.000000 35.767000 ( 35.767000) full action: 31.382000 0.000000 31.382000 ( 31.382000) rev 4419 full action: 36.063000 0.000000 36.063000 ( 36.063000) full action: 30.328000 0.000000 30.328000 ( 30.328000) rev 4423 full action: 36.255000 0.000000 36.255000 ( 36.255000) full action: 30.551000 0.000000 30.551000 ( 30.551000) rev 4447 full action: 35.245000 0.000000 35.245000 ( 35.245000) full action: 30.266000 0.000000 30.266000 ( 30.266000) rev 4463 full action: 33.966000 0.000000 33.966000 ( 33.966000) full action: 30.188000 0.000000 30.188000 ( 30.188000) rev 4470 full action: 32.928000 0.000000 32.928000 ( 32.928000) full action: 29.895000 0.000000 29.895000 ( 29.895000) and also fixes-1_0 rev 4385: full action: 38.209000 0.000000 38.209000 ( 38.209000) full action: 32.002000 0.000000 32.002000 ( 32.002000) and MRI full action: 17.120000 1.010000 18.130000 ( 19.108413) The module inclusion improvements of rev 4415 seem to pay off a lot. For the first time a trunk build is faster than the fixes-1_0 build (with my setup). Still it runs faster without JIT. And I'm curious how fixes-1_0 would perform if 4415 changeset was backported. Christian
