Re: [PATCH 4/4] sched: bias to target cpu load to reduce task moving

[Alex Shi <alex.shi@linaro.org>]

> We obsevered 150% performance gain with vm-scalability/300s-mmap-pread-seq
> testcase with this patch applied. Here is a list of changes we got so far:
> 
> testbox : brickland

got some explain of brickland on wiki:
High-end server platform based on the Ivy Bridge-EX processor
> testcase: vm-scalability/300s-mmap-pread-seq

https://github.com/aristeu/vm-scalability

Thanks a lot for testing! :)
Do you have data of base upstream commit?

> 
> 
>     f1b6442c7dd12802e622      d70495ef86f397816d73  
>        (parent commit)            (this commit)
> ------------------------  ------------------------  
>              26393249.80      +150.9%  66223933.60  vm-scalability.throughput
> 
>                   225.12       -49.9%       112.75  time.elapsed_time
>                 36333.40       -90.7%      3392.20  vmstat.system.cs
>                     2.40      +375.0%        11.40  vmstat.cpu.id
>               3770081.60       -97.7%     87673.40  time.major_page_faults
>               3975276.20       -97.0%    117409.60  time.voluntary_context_switches
>                     3.05      +301.7%        12.24  iostat.cpu.idle
>                 21118.41       -70.3%      6277.19  time.system_time
>                    18.40      +130.4%        42.40  vmstat.cpu.us
>                    77.00       -41.3%        45.20  vmstat.cpu.sy
>                 47459.60       -31.3%     32592.20  vmstat.system.in
>                 82435.40       -12.1%     72443.60  time.involuntary_context_switches
>                  5128.13       +14.0%      5848.30  time.user_time
>                 11656.20        -7.8%     10745.60  time.percent_of_cpu_this_job_got
>            1069997484.80        +0.3% 1073679919.00 time.minor_page_faults
> 

-- 
Thanks
    Alex