2014-10-27 02:18:13

by Fengguang Wu

[permalink] [raw]
Subject: [LKP] [futex] 76835b0ebf8: -12.1% will-it-scale.per_process_ops

_______________________________________________
LKP mailing list
[email protected]

Attachments:
(No filename) (28.29 kB)
 job.yaml (1.66 kB)
 reproduce (615.00 B)
 (No filename) (85.00 B)
 Download all attachments

2014-10-27 02:26:15

by Davidlohr Bueso

[permalink] [raw]
Subject: Re: [LKP] [futex] 76835b0ebf8: -12.1% will-it-scale.per_process_ops

On Mon, 2014-10-27 at 10:18 +0800, kernel test robot wrote:
> FYI, we noticed the below changes on
>
> commit 76835b0ebf8a7fe85beb03c75121419a7dec52f0 ("futex: Ensure get_futex_key_refs() always implies a barrier")

fwiw I was also able to reproduce similar results, with the hashing
costing (unsurprisingly) around an extra ~10% with the new barrier, not
much we can do about that. In any case these are pretty low level
measurements so shouldn't impact too much. Additionally, any user that
cares about performance will try to avoid incurring futex calls into
kernel space.

2014-10-30 04:39:22

by Darren Hart

[permalink] [raw]
Subject: Re: [LKP] [futex] 76835b0ebf8: -12.1% will-it-scale.per_process_ops

On Mon, Oct 27, 2014 at 10:18:22AM +0800, kernel test robot wrote:
> FYI, we noticed the below changes on
>
> commit 76835b0ebf8a7fe85beb03c75121419a7dec52f0 ("futex: Ensure get_futex_key_refs() always implies a barrier")

Wow, this is great information to get for free. Thanks Fengguang!

--
Darren Hart
Intel Open Source Technology Center