Re: [PATCH v2] locking/rwsem: Add reader-owned state to the owner field

[Waiman Long <waiman.long@hpe.com>]

[-- Attachment #1: Type: text/plain, Size: 882 bytes --]

On 05/07/2016 12:56 AM, Ingo Molnar wrote:
> * Waiman Long<Waiman.Long@hpe.com>  wrote:
>
>> On a 4-socket Haswell machine running on a 4.6-rc1 based kernel, the
>> fio test with multithreaded randrw and randwrite tests on the same
>> file on a XFS partition on top of a NVDIMM were run, the aggregated
>> bandwidths before and after the patch were as follows:
>>
>>    Test      BW before patch     BW after patch  % change
>>    ----      ---------------     --------------  --------
>>    randrw         988 MB/s          1192 MB/s      +21%
>>    randwrite     1513 MB/s          1623 MB/s      +7.3%
> What testcase/suite is this? I'd like to run this on other machines as well.
>
> Thanks,
>
> 	Ingo

I just used fio on a nvdimm based xfs filesystem. It is essentially like 
a ramfs filesystem in term of performance. Attached were config files 
that I used.

Cheers,
Longman

[-- Attachment #2: rw.fio --]
[-- Type: text/plain, Size: 420 bytes --]

[global]
direct=1
ioengine=libaio
norandommap
randrepeat=0
bs=4K
size=1G
iodepth=1	# pmem has no queue depth
runtime=30
time_based=1
group_reporting
thread=1
gtod_reduce=1	# reduce=1 except for latency test
gtod_cpu=1


## cross-CPU combinations
numjobs=18	
cpus_allowed=0-39

cpus_allowed_policy=split

[drive_0]
filename=/mnt/fio
cpus_allowed=0-17
rw=randrw

[drive_1]
filename=/mnt/fio2
cpus_allowed=18-35
rw=randrw


[-- Attachment #3: wo.fio --]
[-- Type: text/plain, Size: 426 bytes --]

[global]
direct=1
ioengine=libaio
norandommap
randrepeat=0
bs=4K
size=1G
iodepth=1	# pmem has no queue depth
runtime=30
time_based=1
group_reporting
thread=1
gtod_reduce=1	# reduce=1 except for latency test
gtod_cpu=1


## cross-CPU combinations
numjobs=18	
cpus_allowed=0-39

cpus_allowed_policy=split

[drive_0]
filename=/mnt/fio
cpus_allowed=0-17
rw=randwrite

[drive_1]
filename=/mnt/fio2
cpus_allowed=18-35
rw=randwrite