ULE/SCHED_SMP diff for 7.0

Marinos Ilias marinosi at ceid.upatras.gr
Tue Jul 17 21:03:32 UTC 2007


Hello list,
although I run -CURRENT on a pentium4 , as I have HyperThreading enabled (ok I know about the issues :-P ) I tried Jeff's patch.It works for me :-P
  It works fine , although I cannot figure out if I can have any improve with my cpu.
I attach the results for 4 sysbench tests (for both unpatched and patched ULE) .A little bit weird results for 800 and 1000 threads test.Congratulations Jeff.Nice work!


Cheers,
Ilias
-------------- next part --------------
sysbench v0.4.8:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 800

Doing thread subsystem performance test
Thread yields per test: 1000 Locks used: 8
Threads started!
Done.


Test execution summary:
    total time:                          115.8402s
    total number of events:              10000
    total time taken by event execution: 89360.0746
    per-request statistics:
         min:                            0.0128s
         avg:                            8.9360s
         max:                            92.6393s
         approx.  95 percentile:         32.3754s

Threads fairness:
    events (avg/stddev):           12.5000/4.36
    execution time (avg/stddev):   111.7001/2.50

sysbench v0.4.8:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 1000

Doing thread subsystem performance test
Thread yields per test: 1000 Locks used: 8
Threads started!
Done.


Test execution summary:
    total time:                          119.3290s
    total number of events:              10000
    total time taken by event execution: 114023.2052
    per-request statistics:
         min:                            0.0230s
         avg:                            11.4023s
         max:                            119.0953s
         approx.  95 percentile:         40.3187s

Threads fairness:
    events (avg/stddev):           10.0000/3.88
    execution time (avg/stddev):   114.0232/3.17

sysbench v0.4.8:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 1000

Doing CPU performance benchmark

Threads started!
Done.

Maximum prime number checked in CPU test: 10000


Test execution summary:
    total time:                          38.5863s
    total number of events:              10000
    total time taken by event execution: 6625.7610
    per-request statistics:
         min:                            0.0071s
         avg:                            0.6626s
         max:                            38.3246s
         approx.  95 percentile:         0.9653s

Threads fairness:
    events (avg/stddev):           10.0000/17.95
    execution time (avg/stddev):   6.6258/12.28

sysbench v0.4.8:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 100

Doing CPU performance benchmark

Threads started!
Done.

Maximum prime number checked in CPU test: 10000


Test execution summary:
    total time:                          38.3250s
    total number of events:              10000
    total time taken by event execution: 3321.8330
    per-request statistics:
         min:                            0.0072s
         avg:                            0.3322s
         max:                            25.5190s
         approx.  95 percentile:         0.9860s

Threads fairness:
    events (avg/stddev):           100.0000/56.36
    execution time (avg/stddev):   33.2183/4.89

-------------- next part --------------
sysbench v0.4.8:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 800

Doing thread subsystem performance test
Thread yields per test: 1000 Locks used: 8
Threads started!
Done.


Test execution summary:
    total time:                          118.3846s
    total number of events:              10000
    total time taken by event execution: 91204.1873
    per-request statistics:
         min:                            0.0033s
         avg:                            9.1204s
         max:                            103.2080s
         approx.  95 percentile:         34.6621s

Threads fairness:
    events (avg/stddev):           12.5000/4.79
    execution time (avg/stddev):   114.0052/2.56

sysbench v0.4.8:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 1000

Doing thread subsystem performance test
Thread yields per test: 1000 Locks used: 8
Threads started!
Done.


Test execution summary:
    total time:                          119.2878s
    total number of events:              10000
    total time taken by event execution: 113845.7329
    per-request statistics:
         min:                            0.0232s
         avg:                            11.3846s
         max:                            118.5737s
         approx.  95 percentile:         40.6458s

Threads fairness:
    events (avg/stddev):           10.0000/3.87
    execution time (avg/stddev):   113.8457/3.21

sysbench v0.4.8:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 1000

Doing CPU performance benchmark

Threads started!
Done.

Maximum prime number checked in CPU test: 10000


Test execution summary:
    total time:                          39.7346s
    total number of events:              10594
    total time taken by event execution: 8634.6490
    per-request statistics:
         min:                            0.0067s
         avg:                            0.8151s
         max:                            39.3339s
         approx.  95 percentile:         3.4416s

Threads fairness:
    events (avg/stddev):           10.5940/11.65
    execution time (avg/stddev):   8.6346/7.60

sysbench v0.4.8:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 100

Doing CPU performance benchmark

Threads started!
Done.

Maximum prime number checked in CPU test: 10000


Test execution summary:
    total time:                          37.8735s
    total number of events:              10000
    total time taken by event execution: 3497.9287
    per-request statistics:
         min:                            0.0069s
         avg:                            0.3498s
         max:                            15.2557s
         approx.  95 percentile:         1.0344s

Threads fairness:
    events (avg/stddev):           100.0000/58.32
    execution time (avg/stddev):   34.9793/2.34



More information about the freebsd-current mailing list