* 3.0.0-rc6 ~9X system time degradation on kernbench
@ 2011-07-08 19:36 Prarit Bhargava
2011-07-09 0:45 ` Rik van Riel
0 siblings, 1 reply; 4+ messages in thread
From: Prarit Bhargava @ 2011-07-08 19:36 UTC (permalink / raw)
To: Linux Kernel
Just some numbers out of kernbench. System Time & Context Switches
climb dramatically between 2.6.38 and 3.0.0-rc6. Currently bisecting to
find out why. .config is based on Fedora's x86_64 .config.
Thu Jul 7 18:23:48 EDT 2011
2.6.38.8-32.fc15.x86_64
Average Half load -j 64 Run (std deviation):
Elapsed Time 155.376 (3.77835)
User Time 4536.18 (11.5886)
System Time 1810.42 (27.0879)
Percent CPU 4086 (100.252)
Context Switches 794941 (3757.3)
Sleeps 754029 (10516)
Thu Jul 7 19:50:38 EDT 2011
3.0.0-rc6
Average Half load -j 64 Run (std deviation):
Elapsed Time 466.362 (2.78605)
User Time 8616.43 (22.5528)
System Time 14583.6 (63.2273)
Percent CPU 4974.2 (19.6265)
Context Switches 2.69157e+06 (7679.26)
Sleeps 827950 (3050.16)
Thu Jul 7 21:30:22 EDT 2011
2.6.39
Average Half load -j 64 Run (std deviation):
Elapsed Time 250.044 (2.55373)
User Time 5361.27 (14.6996)
System Time 5173.44 (14.9366)
Percent CPU 4212.8 (40.4314)
Context Switches 1.28385e+06 (1610.42)
Sleeps 741129 (3339.19)
Fri Jul 8 12:21:29 EDT 2011
3.0.0-rc1
Average Optimal load -j 64 Run (std deviation):
Elapsed Time 435.338 (5.45314)
User Time 6125.78 (15.2816)
System Time 8346.45 (62.5065)
Percent CPU 3324.2 (29.0637)
Context Switches 6.42863e+06 (23420)
Sleeps 996764 (3744.98)
P.
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: 3.0.0-rc6 ~9X system time degradation on kernbench
2011-07-08 19:36 3.0.0-rc6 ~9X system time degradation on kernbench Prarit Bhargava
@ 2011-07-09 0:45 ` Rik van Riel
2011-07-09 22:48 ` Kyle McMartin
0 siblings, 1 reply; 4+ messages in thread
From: Rik van Riel @ 2011-07-09 0:45 UTC (permalink / raw)
To: Prarit Bhargava; +Cc: Linux Kernel
On 07/08/2011 03:36 PM, Prarit Bhargava wrote:
> Just some numbers out of kernbench. System Time& Context Switches
> climb dramatically between 2.6.38 and 3.0.0-rc6. Currently bisecting to
> find out why. .config is based on Fedora's x86_64 .config.
For the record, I am also seeing dramatically increased
CPU use with the 3.0-rc kernel, vs 2.6.38.
I first attributed it to me upgrading my video card, but
Prarit is seeing slowdowns too, so it's not just me...
--
All rights reversed
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: 3.0.0-rc6 ~9X system time degradation on kernbench
2011-07-09 0:45 ` Rik van Riel
@ 2011-07-09 22:48 ` Kyle McMartin
2011-07-14 22:14 ` Prarit Bhargava
0 siblings, 1 reply; 4+ messages in thread
From: Kyle McMartin @ 2011-07-09 22:48 UTC (permalink / raw)
To: Rik van Riel; +Cc: Prarit Bhargava, Linux Kernel
On Fri, Jul 08, 2011 at 08:45:51PM -0400, Rik van Riel wrote:
> On 07/08/2011 03:36 PM, Prarit Bhargava wrote:
> >Just some numbers out of kernbench. System Time& Context Switches
> >climb dramatically between 2.6.38 and 3.0.0-rc6. Currently bisecting to
> >find out why. .config is based on Fedora's x86_64 .config.
>
Er, if you're using Fedora's config, did you 'make release' first?
grep DEBUG in the config and see why... ;-)
> For the record, I am also seeing dramatically increased
> CPU use with the 3.0-rc kernel, vs 2.6.38.
>
> I first attributed it to me upgrading my video card, but
> Prarit is seeing slowdowns too, so it's not just me...
>
> --
> All rights reversed
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: 3.0.0-rc6 ~9X system time degradation on kernbench
2011-07-09 22:48 ` Kyle McMartin
@ 2011-07-14 22:14 ` Prarit Bhargava
0 siblings, 0 replies; 4+ messages in thread
From: Prarit Bhargava @ 2011-07-14 22:14 UTC (permalink / raw)
To: Kyle McMartin, Linux Kernel; +Cc: Rik van Riel
On 07/09/2011 06:48 PM, Kyle McMartin wrote:
> On Fri, Jul 08, 2011 at 08:45:51PM -0400, Rik van Riel wrote:
>
>> On 07/08/2011 03:36 PM, Prarit Bhargava wrote:
>>
>>> Just some numbers out of kernbench. System Time& Context Switches
>>> climb dramatically between 2.6.38 and 3.0.0-rc6. Currently bisecting to
>>> find out why. .config is based on Fedora's x86_64 .config.
>>>
>>
> Er, if you're using Fedora's config, did you 'make release' first?
> grep DEBUG in the config and see why... ;-)
>
That did it :)
Thu Jul 14 15:33:27 EDT 2011
3.0.0-rc7+
Average Optimal load -j 48 Run (std deviation):
Elapsed Time 178.384 (0.618692)
User Time 3970.09 (1.93208)
System Time 587.498 (0.511488)
Percent CPU 2554.2 (8.58487)
Context Switches 1.08771e+06 (9697.64)
Sleeps 871265 (13137.6)
Thu Jul 14 16:06:39 EDT 2011
2.6.38.8-35.fc15.x86_64
Average Optimal load -j 48 Run (std deviation):
Elapsed Time 184.6 (8.64522)
User Time 3961.07 (5.05)
System Time 637.382 (11.7187)
Percent CPU 2494.4 (102.827)
Context Switches 1.09046e+06 (3986.03)
Sleeps 873607 (4756.41)
Still a bit higher ... but not 9x ... thanks Kyle.
P.
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2011-07-14 22:14 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2011-07-08 19:36 3.0.0-rc6 ~9X system time degradation on kernbench Prarit Bhargava
2011-07-09 0:45 ` Rik van Riel
2011-07-09 22:48 ` Kyle McMartin
2011-07-14 22:14 ` Prarit Bhargava
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox