[PATCH v2] Improve the performance of --num-threads -d 31

[Minfei Huang]

On 02/23/16 at 12:58am, Minfei Huang wrote:
> On 02/17/16 at 03:05pm, Zhou Wenjian wrote:
> > The new implementation won't do the extra work for filtered pages any
> > more. So the performance of -d 31 is close to that of serial processing.
> 
> Hi, Wenjian.
> 
> kdump:/# time bash -x a.sh makedumpfile vmcore02 1 --num-threads 128
> + makedumpfile --num-threads -l --message-level 1 -d 31 /proc/vmcore /kdumproot/home//var/crash/127.0.0.1-2016-02-22-23:22:36/vmcore02

Please ignore this benchmark, since there is no parameter passed to
option num-threads. Following is my new test result which is generated
in 1st kernel.

127.0.0.1-2016-02-22-19\:08\:17/vmcore01 is generated by makedumpfile
with -d 31 in 2nd kernel.
> 
> Command makedumpfile is complied with this patch on version 1.5.9, and
> makedumpfile.backup is complied on version 1.5.7.
> 
> The compressed file vmcore is 67G filtered out from 4T memory.

[crash]# time makedumpfile -l --message-level 1 -d 31 127.0.0.1-2016-02-22-19\:08\:17/vmcore01 vmcore10
Copying data                       : [100.0 %] /

real    7m25.492s
user    6m37.386s
sys 0m47.801s
[crash]# time makedumpfile.backup -l --message-level 1 -d 31 127.0.0.1-2016-02-22-19\:08\:17/vmcore01 vmcore11
Copying data                       : [100.0 %] -

real    7m15.784s
user    6m28.829s
sys 0m46.656s

[crash]# time makedumpfile --num-threads 128 -l --message-level 1 -d 31 127.0.0.1-2016-02-22-19\:08\:17/vmcore01
vmcore14
Copying data                       : [ 99.7 %] |

Never return from makedumpfile.

There are more than 128 cpus plugged in this machine.

Thanks
Minfei