[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openbsd-tech
Subject:    Re: timekeep: fixing large skews on amd64 with RDTSCP
From:       Hrvoje Popovski <hrvoje () srce ! hr>
Date:       2020-08-23 22:55:12
Message-ID: 778e779f-70bc-041c-b9be-63ad2d5fb2cd () srce ! hr
[Download RAW message or body]

On 23.8.2020. 16:50, Claudio Jeker wrote:
> On Sun, Aug 23, 2020 at 04:06:01PM +0200, Christian Weisgerber wrote:
>> Scott Cheloha:
>>
>>> This "it might slow down the network stack" thing keeps coming up, and
>>> yet nobody can point to (a) who expressed this concern or (b) what the
>>> penalty is in practice.
>>
>> It was kettenis@ who simply raised the question and asked for
>> comments from the network people.
>>
>> I think we should just go ahead and use rdtsc_lfence() in
>> tsc_get_timecount().  It is *correct*.
> 
> I agree. For the network stack there are bigger fishes to fry befor such a
> micro optimisation would even matter.
> 

Hi all,

as you said, with this diff forwarding performance is little slower.

6 x Intel(R) Xeon(R) CPU E5-2643 v2 @ 3.50GHz, 3600.54 MHz, 06-3e-04
without diff 1.11 Mpps
with diff 1.04 Mpps

12 x Intel(R) Xeon(R) CPU E5-2620 v2 @ 2.10GHz, 2400.01 MHz, 06-3e-04
without diff 650 Kpps
with diff 600 Kpps


[prev in list] [next in list] [prev in thread] [next in thread]