On Sun, Dec 15, 2013 at 03:55:39PM +0000, Mel Gorman wrote: > <SNIP> > tlbflush micro benchmark > 3.13.0-rc3 3.13.0-rc3 3.4.69 > vanilla nowalk-v2r7 vanilla > Min 1 7.00 ( 0.00%) 6.00 ( 14.29%) 5.00 ( 28.57%) > Min 2 8.00 ( 0.00%) 6.00 ( 25.00%) 4.00 ( 50.00%) > Min 3 13.00 ( 0.00%) 11.00 ( 15.38%) 9.00 ( 30.77%) > Min 4 17.00 ( 0.00%) 19.00 (-11.76%) 15.00 ( 11.76%) > Mean 1 11.28 ( 0.00%) 10.66 ( 5.48%) 5.17 ( 54.13%) > Mean 2 11.42 ( 0.00%) 11.52 ( -0.85%) 9.04 ( 20.82%) > Mean 3 23.43 ( 0.00%) 21.64 ( 7.64%) 10.92 ( 53.39%) > Mean 4 35.33 ( 0.00%) 34.17 ( 3.28%) 19.55 ( 44.67%) > Range 1 6.00 ( 0.00%) 7.00 (-16.67%) 4.00 ( 33.33%) > Range 2 23.00 ( 0.00%) 36.00 (-56.52%) 19.00 ( 17.39%) > Range 3 15.00 ( 0.00%) 17.00 (-13.33%) 10.00 ( 33.33%) > Range 4 29.00 ( 0.00%) 26.00 ( 10.34%) 9.00 ( 68.97%) > Stddev 1 1.01 ( 0.00%) 1.12 ( 10.53%) 0.57 (-43.70%) > Stddev 2 1.83 ( 0.00%) 3.03 ( 66.06%) 6.83 (274.00%) > Stddev 3 2.82 ( 0.00%) 3.28 ( 16.44%) 1.21 (-57.14%) > Stddev 4 6.65 ( 0.00%) 6.32 ( -5.00%) 1.58 (-76.24%) > Max 1 13.00 ( 0.00%) 13.00 ( 0.00%) 9.00 ( 30.77%) > Max 2 31.00 ( 0.00%) 42.00 (-35.48%) 23.00 ( 25.81%) > Max 3 28.00 ( 0.00%) 28.00 ( 0.00%) 19.00 ( 32.14%) > Max 4 46.00 ( 0.00%) 45.00 ( 2.17%) 24.00 ( 47.83%) > > <SNIP> > > 3.13.0-rc3 3.13.0-rc3 3.4.69 > vanilla nowalk-v2r7 vanilla > User 179.36 165.25 97.29 > System 153.59 155.07 128.32 > Elapsed 1439.52 1437.69 2802.01 > After I ran the test, I looked closer at the elapsed times and it was due to a bug in the test setup itself. The tlbflush tests will need to be rerun but ebizzy still has the problem where threads see very different performance. -- Mel Gorman SUSE Labs -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>