From: Con Kolivas on 8 Aug 2010 19:20 On Sat, 7 Aug 2010 10:07:09 am Con Kolivas wrote: > This is to announce the availability of the updated BFS CPU scheduler for > linux kernel v2.6.35. > > > http://ck.kolivas.org/patches/bfs/2.6.35-sched-bfs-323.patch In response to requests I get for some kind of basic benchmarks, attached are the kernbench results on a quad core Q9650 of 2.6.35 and 2.6.35-ck1, same config. (Summary, BFS slightly slower -j2, and slightly faster on -j1 , -j4 and -j). This benchmark is not remotely meant to be a comprehensive summary of performance, but a simple regression test I do regularly. While the throughput of BFS has decreased slightly since dropping the rr interval, it is still performing well. But then latency is my prime concern. On the -ck1 release thread I will post some interbench results as well, addressing this. Thu Aug 5 20:59:36 EST 2010 2.6.35 Average Single threaded Run (std deviation): Elapsed Time 89.302 (0.0396232) User Time 78.272 (0.0593296) System Time 10.676 (0.060663) Percent CPU 99 (0) Context Switches 5537 (17.6352) Sleeps 12383.8 (55.1788) Average Half load -j 2 Run (std deviation): Elapsed Time 45.56 (0.100995) User Time 78.336 (0.0756968) System Time 10.96 (0.0935414) Percent CPU 195.6 (0.547723) Context Switches 5462.2 (16.5136) Sleeps 12578 (13.0958) Average Optimal load -j 4 Run (std deviation): Elapsed Time 25.632 (0.131985) User Time 79.706 (1.44625) System Time 11.46 (0.535703) Percent CPU 279 (87.9192) Context Switches 7780.7 (2444.47) Sleeps 12750.8 (192.797) Average Maximal load -j Run (std deviation): Elapsed Time 26.65 (0.0244949) User Time 81.372 (2.70062) System Time 11.92 (0.800206) Percent CPU 307.867 (82.1878) Context Switches 12080.7 (6592.71) Sleeps 12632.1 (233.545) Thu Aug 5 21:13:03 EST 2010 2.6.35-ck1 Average Single threaded Run (std deviation): Elapsed Time 89.05 (0.0324037) User Time 79.514 (0.0890505) System Time 9.3 (0.103682) Percent CPU 99 (0) Context Switches 210.4 (12.5419) Sleeps 14390.6 (11.1041) Average Half load -j 2 Run (std deviation): Elapsed Time 45.826 (0.0931665) User Time 79.95 (0.0620484) System Time 9.772 (0.0576194) Percent CPU 195.2 (0.447214) Context Switches 217 (18.775) Sleeps 14467.6 (47.8153) Average Optimal load -j 4 Run (std deviation): Elapsed Time 25.598 (0.109179) User Time 81.265 (1.38736) System Time 10.247 (0.50339) Percent CPU 279.5 (88.8673) Context Switches 4438.1 (4449.75) Sleeps 14033.3 (466.044) Average Maximal load -j Run (std deviation): Elapsed Time 26.408 (0.0614003) User Time 82.7227 (2.40661) System Time 10.5467 (0.59639) Percent CPU 308.333 (82.8162) Context Switches 7810.2 (6091.3) Sleeps 13889.3 (445.423) -- -ck -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
|
Pages: 1 Prev: [PATCH] hrtimer: make epoll_wait() use the hrtimer range feature Next: 2.6.35-ck1 |