Rick broke the non HT parts of D7 out into patches 2,3,4,5 for us.
This has been useful in all sorts of ways ...
http://marc.theaimsgroup.com/?l=lse-tech&m=104436251927067&w=2
(I'll attatch his patches here for convenience, they're small, and I don't
trust the archives not to munge them).
Below I've done two runs on each interesting combo (2 is a no-op).
Overall results are a slight improvement in kernbench and a degredation
in SDET. My impressions from the data below (16x NUMA-Q)
3 - gives a good boost to kernbench, marginal degreadation on SDET.
4 - doesn't do much.
5 - gives some degredation to kernebench, most of the degredation of SDET.
Conclusion: 3 = good (mostly). 5 = bad.
key:
2.5.59 - virgin
2.5.59-sched3 - patch 2 & 3
2.5.59-sched3-2 - patch 2 & 3
2.5.59-sched4 - patch 2 & 3 & 4
2.5.59-sched4-2 - patch 2 & 3 & 4
2.5.59-ricksched - patch 2 & 3 & 4 & 5
2.5.59-ricksched2 - patch 2 & 3 & 4 & 5
Oh, and the runs for 4-2 don't have any schedbench numbers because the
machine crashed, and I'm too lazy to rerun, as it wasn't showing much
anyway.
Kernbench-2: (make -j N vmlinux, where N = 2 x num_cpus)
Elapsed User System CPU
2.5.59 46.08 563.88 118.38 1480.00
2.5.59-sched3 46.12 567.52 115.82 1480.83
2.5.59-sched3-2 46.37 567.81 116.48 1475.00
2.5.59-sched4 46.00 566.28 116.30 1483.17
2.5.59-sched4-2 46.30 567.29 117.43 1478.33
2.5.59-ricksched 46.11 568.54 117.21 1486.67
2.5.59-ricksched2 46.54 568.72 117.43 1473.50
Kernbench-16: (make -j N vmlinux, where N = 16 x num_cpus)
Elapsed User System CPU
2.5.59 47.45 568.02 143.17 1498.17
2.5.59-sched3 47.59 573.48 139.78 1498.33
2.5.59-sched3-2 47.61 573.63 139.97 1498.17
2.5.59-sched4 47.65 573.47 140.47 1497.67
2.5.59-sched4-2 47.73 573.62 141.07 1496.83
2.5.59-ricksched 47.16 571.72 140.02 1508.50
2.5.59-ricksched2 47.19 571.60 138.00 1503.00
DISCLAIMER: SPEC(tm) and the benchmark name SDET(tm) are registered
trademarks of the Standard Performance Evaluation Corporation. This
benchmarking was performed for research purposes only, and the run results
are non-compliant and not-comparable with any published results.
Results are shown as percentages of the first set displayed
SDET 1 (see disclaimer)
Throughput Std. Dev
2.5.59 100.0% 0.8%
2.5.59-sched3 98.2% 0.8%
2.5.59-sched3-2 97.2% 2.1%
2.5.59-sched4 95.3% 3.5%
2.5.59-sched4-2 97.1% 3.6%
2.5.59-ricksched 89.2% 4.3%
2.5.59-ricksched2 89.4% 2.9%
SDET 2 (see disclaimer)
Throughput Std. Dev
2.5.59 100.0% 0.6%
2.5.59-sched3 93.3% 7.7%
2.5.59-sched3-2 92.8% 8.5%
2.5.59-sched4 91.0% 10.1%
2.5.59-sched4-2 96.4% 7.4%
2.5.59-ricksched 89.7% 5.6%
2.5.59-ricksched2 84.5% 2.1%
SDET 4 (see disclaimer)
Throughput Std. Dev
2.5.59 100.0% 5.7%
2.5.59-sched3 96.9% 4.4%
2.5.59-sched3-2 98.6% 5.2%
2.5.59-sched4 102.5% 7.2%
2.5.59-sched4-2 97.0% 1.6%
2.5.59-ricksched 91.4% 3.5%
2.5.59-ricksched2 92.6% 1.5%
SDET 8 (see disclaimer)
Throughput Std. Dev
2.5.59 100.0% 1.4%
2.5.59-sched3 103.7% 4.8%
2.5.59-sched3-2 102.8% 4.2%
2.5.59-sched4 104.5% 4.3%
2.5.59-sched4-2 102.0% 2.9%
2.5.59-ricksched 102.2% 4.6%
2.5.59-ricksched2 102.5% 1.1%
SDET 16 (see disclaimer)
Throughput Std. Dev
2.5.59 100.0% 1.7%
2.5.59-sched3 98.9% 1.9%
2.5.59-sched3-2 101.9% 1.8%
2.5.59-sched4 99.1% 1.7%
2.5.59-sched4-2 99.5% 0.9%
2.5.59-ricksched 98.9% 1.1%
2.5.59-ricksched2 78.7% 50.1%
SDET 32 (see disclaimer)
Throughput Std. Dev
2.5.59 100.0% 1.5%
2.5.59-sched3 98.9% 2.0%
2.5.59-sched3-2 99.7% 0.8%
2.5.59-sched4 97.7% 1.1%
2.5.59-sched4-2 98.7% 1.4%
2.5.59-ricksched 96.7% 1.6%
2.5.59-ricksched2 89.7% 14.9%
SDET 64 (see disclaimer)
Throughput Std. Dev
2.5.59 100.0% 0.7%
2.5.59-sched3 98.8% 0.6%
2.5.59-sched3-2 98.9% 0.4%
2.5.59-sched4 97.8% 0.7%
2.5.59-sched4-2 96.7% 0.9%
2.5.59-ricksched 98.8% 0.7%
2.5.59-ricksched2 97.7% 1.1%
SDET 128 (see disclaimer)
Throughput Std. Dev
NUMA schedbench 4:
AvgUser Elapsed TotalUser TotalSys
2.5.59 0.00 38.88 82.78 0.65
2.5.59-sched3 0.00 30.13 93.10 1.02
2.5.59-sched3-2 0.00 37.47 111.83 0.69
2.5.59-sched4 0.00 30.92 90.07 0.90
2.5.59-sched4-2 0.00 34.20 84.41 0.70
2.5.59-ricksched 0.00 26.40 69.10 0.68
2.5.59-ricksched2 0.00 60.15 170.80 0.98
NUMA schedbench 8:
AvgUser Elapsed TotalUser TotalSys
2.5.59 0.00 49.30 247.80 1.93
2.5.59-sched3 0.00 54.05 306.14 1.50
2.5.59-sched3-2 0.00 39.71 241.97 2.19
2.5.59-sched4 0.00 54.05 305.21 1.35
2.5.59-sched4-2 0 0 0 0
2.5.59-ricksched 0.00 47.32 261.69 1.59
2.5.59-ricksched2 0.00 45.88 260.44 1.70
NUMA schedbench 16:
AvgUser Elapsed TotalUser TotalSys
2.5.59 0.00 57.37 843.12 3.77
2.5.59-sched3 0.00 57.96 849.88 2.81
2.5.59-sched3-2 0.00 57.85 853.58 3.28
2.5.59-sched4 0.00 57.63 847.41 2.78
2.5.59-sched4-2 0 0 0 0
2.5.59-ricksched 0.00 57.48 850.36 3.19
2.5.59-ricksched2 0.00 57.29 848.65 4.21
NUMA schedbench 32:
AvgUser Elapsed TotalUser TotalSys
2.5.59 0.00 116.99 1805.79 6.05
2.5.59-sched3 0.00 120.15 1839.49 6.70
2.5.59-sched3-2 0.00 126.20 1843.77 6.42
2.5.59-sched4 0.00 119.90 1825.33 7.05
2.5.59-sched4-2 0 0 0 0
2.5.59-ricksched 0.00 122.43 1821.62 6.80
2.5.59-ricksched2 0.00 125.44 1821.75 6.96
NUMA schedbench 64:
AvgUser Elapsed TotalUser TotalSys
2.5.59 0.00 235.18 3632.73 15.45
2.5.59-sched3 0.00 238.90 3714.23 17.74
2.5.59-sched3-2 0.00 235.20 3699.45 18.75
2.5.59-sched4 0.00 241.24 3689.19 15.65
2.5.59-sched4-2 0 0 0 0
2.5.59-ricksched 0.00 242.80 3708.15 16.97
2.5.59-ricksched2 0.00 235.71 3693.12 16.50
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Feb 15 2003 - 22:00:50 EST