GNU/kFreeBSD threading performance benchmark
For potential future investigation, I noticed that the sysbench utility
records notably lower performance in its threading benchmark on
GNU/kFreeBSD systems, than on somewhat somewhat older GNU/Linux systems.
GNU/Linux, 4-way Opteron 285 2.6 GHz => 2.4383s total:
> # sysbench --num-threads=64 --test=threads --thread-yields=100 --thread-locks=2 run
> sysbench 0.4.12: multi-threaded system evaluation benchmark
>
> Running the test with following options:
> Number of threads: 64
>
> Doing thread subsystem performance test
> Thread yields per test: 100 Locks used: 2
> Threads started!
> Done.
>
>
> Test execution summary:
> total time: 2.4383s
> total number of events: 10000
> total time taken by event execution: 155.2520
> per-request statistics:
> min: 0.05ms
> avg: 15.53ms
> max: 131.32ms
> approx. 95 percentile: 56.51ms
>
> Threads fairness:
> events (avg/stddev): 156.2500/16.61
> execution time (avg/stddev): 2.4258/0.00
>
GNU/kFreeBSD, 4-way Opteron 8380 2.5 GHz => 13.4990s total:
> # sysbench --num-threads=64 --test=threads --thread-yields=100 --thread-locks=2 run
> sysbench 0.4.12: multi-threaded system evaluation benchmark
>
> Running the test with following options:
> Number of threads: 64
>
> Doing thread subsystem performance test
> Thread yields per test: 100 Locks used: 2
> Threads started!
> Done.
>
>
> Test execution summary:
> total time: 13.4990s
> total number of events: 10000
> total time taken by event execution: 861.9675
> per-request statistics:
> min: 0.04ms
> avg: 86.20ms
> max: 207.55ms
> approx. 95 percentile: 174.67ms
>
> Threads fairness:
> events (avg/stddev): 156.2500/1.40
> execution time (avg/stddev): 13.4682/0.02
--
Steven Chamberlain
steven@pyro.eu.org
Reply to: