The O(1) J9 scheduler with changes to generic context-switch definitely
provides the best context-switch latency time based on the lat_ctx test
from LMBench benchmark. Test hardware was a 2-way SMP system with 512MB
memory. The results obtained from a 2.4.17 SMP kernel built with O(1)J2,
O(1)J4 and O(1)J9 versions, compared against 2.2.19 base and 2.4.17 base.
The data are time in microsecs. In almost all test cases, context switches
under O(1)J9 seem much faster than 2.4.17 base, and as fast as 2.2.19 (or
better at heavy load). The 2.2.19 measurements were used as a reference
point for 2.2.x.
Ratio of
Base Base O(1)J2 O(1)J4 O(1)J9 O(1)J9 /
Kernel 2.2.19 2.4.17 2.4.17 2.4.17 2.4.17
2417Base
lat_ctx -s 0 2 4 8 16 32 64
2 1.46 3.03 7.38 7.38 0.97 32.0%
4 2.04 3.55 4.84 4.89 4.97 140.2%
8 2.74 4.41 6.04 4.96 4.43 100.6%
16 3.01 4.78 4.93 5.46 3.82 80.0%
32 5.48 7.29 4.80 5.75 4.56 62.6%
64 5.74 8.37 5.86 6.12 5.62 67.1%
lat_ctx -s 16 2 4 8 16 32 64
2 14.33 16.15 15.62 15.11 13.93 86.3%
4 14.30 16.13 17.83 16.27 16.51 102.4%
8 14.39 16.38 17.58 17.35 15.20 92.8%
16 16.43 19.34 17.75 17.70 16.54 85.6%
32 39.92 39.93 24.63 24.95 32.65 81.8%
64 53.67 49.87 45.34 42.76 49.91 100.1%
lat_ctx -s 32 2 4 8 16 32 64
2 22.86 24.73 27.56 27.03 24.29 98.2%
4 22.85 24.85 25.98 25.78 25.22 101.5%
8 25.18 30.94 26.51 26.65 25.04 81.0%
16 58.70 74.32 38.69 35.07 48.32 65.0%
32 99.16 94.41 74.32 74.85 89.62 94.9%
64 99.28 96.91 98.45 97.75 94.29 97.3%
lat_ctx -s 64 2 4 8 16 32 64
2 40.24 42.06 44.57 44.11 40.72 96.8%
4 49.10 43.28 43.23 43.99 42.85 99.0%
8 111.28 105.56 49.43 45.81 58.41 55.3%
16 185.12 182.84 127.11 124.74 169.94 92.9%
32 185.20 182.81 184.97 182.92 175.70 96.1%
64 185.26 184.46 186.54 186.22 178.63 96.8%
Duc.