* [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness
@ 2013-01-19 20:17 John Morris
2013-01-19 20:25 ` Gilles Chanteperdrix
2013-01-19 22:12 ` John Morris
0 siblings, 2 replies; 12+ messages in thread
From: John Morris @ 2013-01-19 20:17 UTC (permalink / raw)
To: Xenomai
Hi list,
These are the final tests on 32-bit before the initial RedHat packages
can be released for wider testing.
On this host, a Dell Celeron with ICH5 chipset, the CLOCK_REALTIME
numbers look funny. All other tests run correctly.
++ /usr/lib/xenomai/clocktest -T 30
== Tested clock: 0 (CLOCK_REALTIME)
CPU ToD offset [us] ToD drift [us/s] warps max delta [us]
--- -------------------- ---------------- ---------- --------------
0 0.0 0.000 0 0.0
^[[1A 0 -1074721.3 39.245 0 0.0
^[[1A 0 -1074711.5 39.172 0 0.0
^[[1A 0 -1074701.8 39.096 0 0.0
^[[1A 0 -1074691.9 39.184 0 0.0
This is Gilles's i-pipe dev tree and xenomai master pulled yesterday,
kernel 3.5.7.
John
http://www.zultron.com/static/2013/01/xenomai/3.5.7-test-32-bit/config-3.5.7.txt
http://www.zultron.com/static/2013/01/xenomai/3.5.7-test-32-bit/dmesg.log
http://www.zultron.com/static/2013/01/xenomai/3.5.7-test-32-bit/xeno-regression-test.log
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness
2013-01-19 20:17 [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness John Morris
@ 2013-01-19 20:25 ` Gilles Chanteperdrix
2013-01-19 20:32 ` John Morris
2013-01-19 22:12 ` John Morris
1 sibling, 1 reply; 12+ messages in thread
From: Gilles Chanteperdrix @ 2013-01-19 20:25 UTC (permalink / raw)
To: John Morris; +Cc: Xenomai
On 01/19/2013 09:17 PM, John Morris wrote:
> Hi list,
>
> These are the final tests on 32-bit before the initial RedHat packages
> can be released for wider testing.
>
> On this host, a Dell Celeron with ICH5 chipset, the CLOCK_REALTIME
> numbers look funny. All other tests run correctly.
>
> ++ /usr/lib/xenomai/clocktest -T 30
> == Tested clock: 0 (CLOCK_REALTIME)
> CPU ToD offset [us] ToD drift [us/s] warps max delta [us]
> --- -------------------- ---------------- ---------- --------------
> 0 0.0 0.000 0 0.0
> ^[[1A 0 -1074721.3 39.245 0 0.0
> ^[[1A 0 -1074711.5 39.172 0 0.0
> ^[[1A 0 -1074701.8 39.096 0 0.0
> ^[[1A 0 -1074691.9 39.184 0 0.0
>
> This is Gilles's i-pipe dev tree and xenomai master pulled yesterday,
> kernel 3.5.7.
Nothing really wrong here, the ToD offset is due to the fact that the
realtime clock was changed between the time Xenomai read it to set its
own realt-ime clock, and the time clocktest is run, and the drift is due
to the tsc frequency adjustments made by Linux after Xenomai is started.
--
Gilles.
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness
2013-01-19 20:25 ` Gilles Chanteperdrix
@ 2013-01-19 20:32 ` John Morris
2013-01-19 20:34 ` Gilles Chanteperdrix
0 siblings, 1 reply; 12+ messages in thread
From: John Morris @ 2013-01-19 20:32 UTC (permalink / raw)
To: Gilles Chanteperdrix; +Cc: Xenomai
On 01/19/2013 02:25 PM, Gilles Chanteperdrix wrote:
> On 01/19/2013 09:17 PM, John Morris wrote:
>
>> Hi list,
>>
>> These are the final tests on 32-bit before the initial RedHat packages
>> can be released for wider testing.
>>
>> On this host, a Dell Celeron with ICH5 chipset, the CLOCK_REALTIME
>> numbers look funny. All other tests run correctly.
> Nothing really wrong here, the ToD offset is due to the fact that the
> realtime clock was changed between the time Xenomai read it to set its
> own realt-ime clock, and the time clocktest is run, and the drift is due
> to the tsc frequency adjustments made by Linux after Xenomai is started.
>
Woo hoo! That means it's time to start working on the package repos.
I'll report back soon. Thanks for all the help.
John
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness
2013-01-19 20:32 ` John Morris
@ 2013-01-19 20:34 ` Gilles Chanteperdrix
0 siblings, 0 replies; 12+ messages in thread
From: Gilles Chanteperdrix @ 2013-01-19 20:34 UTC (permalink / raw)
To: John Morris; +Cc: Xenomai
On 01/19/2013 09:32 PM, John Morris wrote:
>
>
> On 01/19/2013 02:25 PM, Gilles Chanteperdrix wrote:
>> On 01/19/2013 09:17 PM, John Morris wrote:
>>
>>> Hi list,
>>>
>>> These are the final tests on 32-bit before the initial RedHat packages
>>> can be released for wider testing.
>>>
>>> On this host, a Dell Celeron with ICH5 chipset, the CLOCK_REALTIME
>>> numbers look funny. All other tests run correctly.
>
>> Nothing really wrong here, the ToD offset is due to the fact that the
>> realtime clock was changed between the time Xenomai read it to set its
>> own realt-ime clock, and the time clocktest is run, and the drift is due
>> to the tsc frequency adjustments made by Linux after Xenomai is started.
>>
>
> Woo hoo! That means it's time to start working on the package repos.
> I'll report back soon. Thanks for all the help.
It would make sense to wait for Jan's results on monday.
--
Gilles.
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness
2013-01-19 20:17 [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness John Morris
2013-01-19 20:25 ` Gilles Chanteperdrix
@ 2013-01-19 22:12 ` John Morris
2013-01-19 22:14 ` Gilles Chanteperdrix
1 sibling, 1 reply; 12+ messages in thread
From: John Morris @ 2013-01-19 22:12 UTC (permalink / raw)
To: Xenomai
On 01/19/2013 02:17 PM, John Morris wrote:
> Hi list,
>
> These are the final tests on 32-bit before the initial RedHat packages
> can be released for wider testing.
>
> On this host, a Dell Celeron with ICH5 chipset, the CLOCK_REALTIME
> numbers look funny. All other tests run correctly.
Whoops, how embarrassing, I reported the wrong thing: it's the
native/tsc numbers that look funny.
++ /usr/lib/xenomai/regression/native/tsc
Checking tsc for 1 minute(s)
min: 4294967295, max: 0, avg: -nan
min: 4294967295, max: 0, avg: -nan
min: 4294967295, max: 0, avg: -nan
min: 4294967295, max: 0, avg: -nan
min: 4294967295, max: 0, avg: -nan
min: 4294967295, max: 0, avg: -nan
[...]
min: 4294967295, max: 0, avg: -nan -> -nan us
John
> http://www.zultron.com/static/2013/01/xenomai/3.5.7-test-32-bit/config-3.5.7.txt
>
> http://www.zultron.com/static/2013/01/xenomai/3.5.7-test-32-bit/dmesg.log
>
> http://www.zultron.com/static/2013/01/xenomai/3.5.7-test-32-bit/xeno-regression-test.log
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness
2013-01-19 22:12 ` John Morris
@ 2013-01-19 22:14 ` Gilles Chanteperdrix
2013-01-19 22:16 ` Gilles Chanteperdrix
0 siblings, 1 reply; 12+ messages in thread
From: Gilles Chanteperdrix @ 2013-01-19 22:14 UTC (permalink / raw)
To: John Morris; +Cc: Xenomai
On 01/19/2013 11:12 PM, John Morris wrote:
> On 01/19/2013 02:17 PM, John Morris wrote:
>> Hi list,
>>
>> These are the final tests on 32-bit before the initial RedHat packages
>> can be released for wider testing.
>>
>> On this host, a Dell Celeron with ICH5 chipset, the CLOCK_REALTIME
>> numbers look funny. All other tests run correctly.
>
> Whoops, how embarrassing, I reported the wrong thing: it's the
> native/tsc numbers that look funny.
>
> ++ /usr/lib/xenomai/regression/native/tsc
> Checking tsc for 1 minute(s)
> min: 4294967295, max: 0, avg: -nan
> min: 4294967295, max: 0, avg: -nan
> min: 4294967295, max: 0, avg: -nan
> min: 4294967295, max: 0, avg: -nan
> min: 4294967295, max: 0, avg: -nan
> min: 4294967295, max: 0, avg: -nan
> [...]
> min: 4294967295, max: 0, avg: -nan -> -nan us
You are probably missing --enable-x86-tsc on configure command line.
--
Gilles.
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness
2013-01-19 22:14 ` Gilles Chanteperdrix
@ 2013-01-19 22:16 ` Gilles Chanteperdrix
2013-01-19 22:50 ` John Morris
0 siblings, 1 reply; 12+ messages in thread
From: Gilles Chanteperdrix @ 2013-01-19 22:16 UTC (permalink / raw)
To: John Morris; +Cc: Xenomai
On 01/19/2013 11:14 PM, Gilles Chanteperdrix wrote:
> On 01/19/2013 11:12 PM, John Morris wrote:
>
>> On 01/19/2013 02:17 PM, John Morris wrote:
>>> Hi list,
>>>
>>> These are the final tests on 32-bit before the initial RedHat packages
>>> can be released for wider testing.
>>>
>>> On this host, a Dell Celeron with ICH5 chipset, the CLOCK_REALTIME
>>> numbers look funny. All other tests run correctly.
>>
>> Whoops, how embarrassing, I reported the wrong thing: it's the
>> native/tsc numbers that look funny.
>>
>> ++ /usr/lib/xenomai/regression/native/tsc
>> Checking tsc for 1 minute(s)
>> min: 4294967295, max: 0, avg: -nan
>> min: 4294967295, max: 0, avg: -nan
>> min: 4294967295, max: 0, avg: -nan
>> min: 4294967295, max: 0, avg: -nan
>> min: 4294967295, max: 0, avg: -nan
>> min: 4294967295, max: 0, avg: -nan
>> [...]
>> min: 4294967295, max: 0, avg: -nan -> -nan us
>
>
> You are probably missing --enable-x86-tsc on configure command line.
>
(or rather have passed --disable-x86-tsc as --enable-x86-tsc should be
the default).
--
Gilles.
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness
2013-01-19 22:16 ` Gilles Chanteperdrix
@ 2013-01-19 22:50 ` John Morris
2013-01-19 22:52 ` Gilles Chanteperdrix
0 siblings, 1 reply; 12+ messages in thread
From: John Morris @ 2013-01-19 22:50 UTC (permalink / raw)
To: Gilles Chanteperdrix; +Cc: Xenomai
On 01/19/2013 04:16 PM, Gilles Chanteperdrix wrote:
> On 01/19/2013 11:14 PM, Gilles Chanteperdrix wrote:
>
>> On 01/19/2013 11:12 PM, John Morris wrote:
>>
>>> On 01/19/2013 02:17 PM, John Morris wrote:
>>>> Hi list,
>>>>
>>>> These are the final tests on 32-bit before the initial RedHat packages
>>>> can be released for wider testing.
>>>>
>>>> On this host, a Dell Celeron with ICH5 chipset, the CLOCK_REALTIME
>>>> numbers look funny. All other tests run correctly.
>>>
>>> Whoops, how embarrassing, I reported the wrong thing: it's the
>>> native/tsc numbers that look funny.
>>>
>>> ++ /usr/lib/xenomai/regression/native/tsc
>>> Checking tsc for 1 minute(s)
>>> min: 4294967295, max: 0, avg: -nan
>>> min: 4294967295, max: 0, avg: -nan
>>> min: 4294967295, max: 0, avg: -nan
>>> min: 4294967295, max: 0, avg: -nan
>>> min: 4294967295, max: 0, avg: -nan
>>> min: 4294967295, max: 0, avg: -nan
>>> [...]
>>> min: 4294967295, max: 0, avg: -nan -> -nan us
>>
>>
>> You are probably missing --enable-x86-tsc on configure command line.
>>
>
> (or rather have passed --disable-x86-tsc as --enable-x86-tsc should be
> the default).
>
--enable-x86-tsc is in there, and configure finds TSC support:
+ ./configure --build=i686-koji-linux-gnu --host=i686-koji-linux-gnu
--target=i686-redhat-linux-gnu --program-prefix= --prefix=/usr
--exec-prefix=/usr --bindir=/usr/bin --sbindir=/usr/sbin
--sysconfdir=/etc --datadir=/usr/share
--includedir=/usr/include/xenomai/ --libdir=/usr/lib
--libexecdir=/usr/libexec --localstatedir=/var --sharedstatedir=/var/lib
--mandir=/usr/share/man --infodir=/usr/share/info --enable-x86-tsc
--enable-dlopen-skins --with-testdir=/usr/lib/xenomai
[...]
checking for x86 TSC support... y
Build log follows. What else can I check?
John
Build log:
http://www.zultron.com/static/2013/01/xenomai/3.5.7-test-32-bit/xenomai-build.log
>
http://www.zultron.com/static/2013/01/xenomai/3.5.7-test-32-bit/config-3.5.7.txt
>
> http://www.zultron.com/static/2013/01/xenomai/3.5.7-test-32-bit/dmesg.log
>
>
http://www.zultron.com/static/2013/01/xenomai/3.5.7-test-32-bit/xeno-regression-test.log
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness
2013-01-19 22:50 ` John Morris
@ 2013-01-19 22:52 ` Gilles Chanteperdrix
2013-01-19 22:57 ` John Morris
0 siblings, 1 reply; 12+ messages in thread
From: Gilles Chanteperdrix @ 2013-01-19 22:52 UTC (permalink / raw)
To: John Morris; +Cc: Xenomai
On 01/19/2013 11:50 PM, John Morris wrote:
>
>
> On 01/19/2013 04:16 PM, Gilles Chanteperdrix wrote:
>> On 01/19/2013 11:14 PM, Gilles Chanteperdrix wrote:
>>
>>> On 01/19/2013 11:12 PM, John Morris wrote:
>>>
>>>> On 01/19/2013 02:17 PM, John Morris wrote:
>>>>> Hi list,
>>>>>
>>>>> These are the final tests on 32-bit before the initial RedHat packages
>>>>> can be released for wider testing.
>>>>>
>>>>> On this host, a Dell Celeron with ICH5 chipset, the CLOCK_REALTIME
>>>>> numbers look funny. All other tests run correctly.
>>>>
>>>> Whoops, how embarrassing, I reported the wrong thing: it's the
>>>> native/tsc numbers that look funny.
>>>>
>>>> ++ /usr/lib/xenomai/regression/native/tsc
>>>> Checking tsc for 1 minute(s)
>>>> min: 4294967295, max: 0, avg: -nan
>>>> min: 4294967295, max: 0, avg: -nan
>>>> min: 4294967295, max: 0, avg: -nan
>>>> min: 4294967295, max: 0, avg: -nan
>>>> min: 4294967295, max: 0, avg: -nan
>>>> min: 4294967295, max: 0, avg: -nan
>>>> [...]
>>>> min: 4294967295, max: 0, avg: -nan -> -nan us
>>>
>>>
>>> You are probably missing --enable-x86-tsc on configure command line.
>>>
>>
>> (or rather have passed --disable-x86-tsc as --enable-x86-tsc should be
>> the default).
>>
>
> --enable-x86-tsc is in there, and configure finds TSC support:
>
> + ./configure --build=i686-koji-linux-gnu --host=i686-koji-linux-gnu
> --target=i686-redhat-linux-gnu --program-prefix= --prefix=/usr
> --exec-prefix=/usr --bindir=/usr/bin --sbindir=/usr/sbin
> --sysconfdir=/etc --datadir=/usr/share
> --includedir=/usr/include/xenomai/ --libdir=/usr/lib
> --libexecdir=/usr/libexec --localstatedir=/var --sharedstatedir=/var/lib
> --mandir=/usr/share/man --infodir=/usr/share/info --enable-x86-tsc
> --enable-dlopen-skins --with-testdir=/usr/lib/xenomai
> [...]
> checking for x86 TSC support... y
>
> Build log follows. What else can I check?
Could you send me the "tsc" program?
--
Gilles.
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness
2013-01-19 22:52 ` Gilles Chanteperdrix
@ 2013-01-19 22:57 ` John Morris
2013-01-19 23:03 ` Gilles Chanteperdrix
0 siblings, 1 reply; 12+ messages in thread
From: John Morris @ 2013-01-19 22:57 UTC (permalink / raw)
To: Gilles Chanteperdrix; +Cc: Xenomai
On 01/19/2013 04:52 PM, Gilles Chanteperdrix wrote:
> On 01/19/2013 11:50 PM, John Morris wrote:
>
>>
>>
>> On 01/19/2013 04:16 PM, Gilles Chanteperdrix wrote:
>>> On 01/19/2013 11:14 PM, Gilles Chanteperdrix wrote:
>>>
>>>> On 01/19/2013 11:12 PM, John Morris wrote:
>>>>
>>>>> On 01/19/2013 02:17 PM, John Morris wrote:
>>>>>> Hi list,
>>>>>>
>>>>>> These are the final tests on 32-bit before the initial RedHat packages
>>>>>> can be released for wider testing.
>>>>>>
>>>>>> On this host, a Dell Celeron with ICH5 chipset, the CLOCK_REALTIME
>>>>>> numbers look funny. All other tests run correctly.
>>>>>
>>>>> Whoops, how embarrassing, I reported the wrong thing: it's the
>>>>> native/tsc numbers that look funny.
>>>>>
>>>>> ++ /usr/lib/xenomai/regression/native/tsc
>>>>> Checking tsc for 1 minute(s)
>>>>> min: 4294967295, max: 0, avg: -nan
>>>>> min: 4294967295, max: 0, avg: -nan
>>>>> min: 4294967295, max: 0, avg: -nan
>>>>> min: 4294967295, max: 0, avg: -nan
>>>>> min: 4294967295, max: 0, avg: -nan
>>>>> min: 4294967295, max: 0, avg: -nan
>>>>> [...]
>>>>> min: 4294967295, max: 0, avg: -nan -> -nan us
>>>>
>>>>
>>>> You are probably missing --enable-x86-tsc on configure command line.
>>>>
>>>
>>> (or rather have passed --disable-x86-tsc as --enable-x86-tsc should be
>>> the default).
>>>
>>
>> --enable-x86-tsc is in there, and configure finds TSC support:
>>
>> + ./configure --build=i686-koji-linux-gnu --host=i686-koji-linux-gnu
>> --target=i686-redhat-linux-gnu --program-prefix= --prefix=/usr
>> --exec-prefix=/usr --bindir=/usr/bin --sbindir=/usr/sbin
>> --sysconfdir=/etc --datadir=/usr/share
>> --includedir=/usr/include/xenomai/ --libdir=/usr/lib
>> --libexecdir=/usr/libexec --localstatedir=/var --sharedstatedir=/var/lib
>> --mandir=/usr/share/man --infodir=/usr/share/info --enable-x86-tsc
>> --enable-dlopen-skins --with-testdir=/usr/lib/xenomai
>> [...]
>> checking for x86 TSC support... y
>>
>> Build log follows. What else can I check?
>
>
> Could you send me the "tsc" program?
>
http://www.zultron.com/static/2013/01/xenomai/3.5.7-test-32-bit/tsc.gz
John
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness
2013-01-19 22:57 ` John Morris
@ 2013-01-19 23:03 ` Gilles Chanteperdrix
2013-01-20 19:22 ` John Morris
0 siblings, 1 reply; 12+ messages in thread
From: Gilles Chanteperdrix @ 2013-01-19 23:03 UTC (permalink / raw)
To: John Morris; +Cc: Xenomai
On 01/19/2013 11:57 PM, John Morris wrote:
> On 01/19/2013 04:52 PM, Gilles Chanteperdrix wrote:
>> On 01/19/2013 11:50 PM, John Morris wrote:
>>
>>>
>>>
>>> On 01/19/2013 04:16 PM, Gilles Chanteperdrix wrote:
>>>> On 01/19/2013 11:14 PM, Gilles Chanteperdrix wrote:
>>>>
>>>>> On 01/19/2013 11:12 PM, John Morris wrote:
>>>>>
>>>>>> On 01/19/2013 02:17 PM, John Morris wrote:
>>>>>>> Hi list,
>>>>>>>
>>>>>>> These are the final tests on 32-bit before the initial RedHat packages
>>>>>>> can be released for wider testing.
>>>>>>>
>>>>>>> On this host, a Dell Celeron with ICH5 chipset, the CLOCK_REALTIME
>>>>>>> numbers look funny. All other tests run correctly.
>>>>>>
>>>>>> Whoops, how embarrassing, I reported the wrong thing: it's the
>>>>>> native/tsc numbers that look funny.
>>>>>>
>>>>>> ++ /usr/lib/xenomai/regression/native/tsc
>>>>>> Checking tsc for 1 minute(s)
>>>>>> min: 4294967295, max: 0, avg: -nan
>>>>>> min: 4294967295, max: 0, avg: -nan
>>>>>> min: 4294967295, max: 0, avg: -nan
>>>>>> min: 4294967295, max: 0, avg: -nan
>>>>>> min: 4294967295, max: 0, avg: -nan
>>>>>> min: 4294967295, max: 0, avg: -nan
>>>>>> [...]
>>>>>> min: 4294967295, max: 0, avg: -nan -> -nan us
>>>>>
>>>>>
>>>>> You are probably missing --enable-x86-tsc on configure command line.
>>>>>
>>>>
>>>> (or rather have passed --disable-x86-tsc as --enable-x86-tsc should be
>>>> the default).
>>>>
>>>
>>> --enable-x86-tsc is in there, and configure finds TSC support:
>>>
>>> + ./configure --build=i686-koji-linux-gnu --host=i686-koji-linux-gnu
>>> --target=i686-redhat-linux-gnu --program-prefix= --prefix=/usr
>>> --exec-prefix=/usr --bindir=/usr/bin --sbindir=/usr/sbin
>>> --sysconfdir=/etc --datadir=/usr/share
>>> --includedir=/usr/include/xenomai/ --libdir=/usr/lib
>>> --libexecdir=/usr/libexec --localstatedir=/var --sharedstatedir=/var/lib
>>> --mandir=/usr/share/man --infodir=/usr/share/info --enable-x86-tsc
>>> --enable-dlopen-skins --with-testdir=/usr/lib/xenomai
>>> [...]
>>> checking for x86 TSC support... y
>>>
>>> Build log follows. What else can I check?
>>
>>
>> Could you send me the "tsc" program?
>>
>
> http://www.zultron.com/static/2013/01/xenomai/3.5.7-test-32-bit/tsc.gz
Ok, this program is compiled correctly, the error you have suggests that
on your machine reading the tsc takes systematically more than 80
cycles, a hardcoded value that is well, a bit too arbitrary.
--
Gilles.
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness
2013-01-19 23:03 ` Gilles Chanteperdrix
@ 2013-01-20 19:22 ` John Morris
0 siblings, 0 replies; 12+ messages in thread
From: John Morris @ 2013-01-20 19:22 UTC (permalink / raw)
To: Gilles Chanteperdrix; +Cc: Xenomai
On 01/19/2013 05:03 PM, Gilles Chanteperdrix wrote:
> On 01/19/2013 11:57 PM, John Morris wrote:
>
>> On 01/19/2013 04:52 PM, Gilles Chanteperdrix wrote:
>>> On 01/19/2013 11:50 PM, John Morris wrote:
>>>
>>>>
>>>>
>>>> On 01/19/2013 04:16 PM, Gilles Chanteperdrix wrote:
>>>>> On 01/19/2013 11:14 PM, Gilles Chanteperdrix wrote:
>>>>>
>>>>>> On 01/19/2013 11:12 PM, John Morris wrote:
>>>>>>
>>>>>>> On 01/19/2013 02:17 PM, John Morris wrote:
>>>>>>>> Hi list,
>>>>>>>>
>>>>>>>> These are the final tests on 32-bit before the initial RedHat packages
>>>>>>>> can be released for wider testing.
>>>>>>>>
>>>>>>>> On this host, a Dell Celeron with ICH5 chipset, the CLOCK_REALTIME
>>>>>>>> numbers look funny. All other tests run correctly.
>>>>>>>
>>>>>>> Whoops, how embarrassing, I reported the wrong thing: it's the
>>>>>>> native/tsc numbers that look funny.
>>>>>>>
>>>>>>> ++ /usr/lib/xenomai/regression/native/tsc
>>>>>>> Checking tsc for 1 minute(s)
>>>>>>> min: 4294967295, max: 0, avg: -nan
>>>>>>> min: 4294967295, max: 0, avg: -nan
>>>>>>> min: 4294967295, max: 0, avg: -nan
>>>>>>> min: 4294967295, max: 0, avg: -nan
>>>>>>> min: 4294967295, max: 0, avg: -nan
>>>>>>> min: 4294967295, max: 0, avg: -nan
>>>>>>> [...]
>>>>>>> min: 4294967295, max: 0, avg: -nan -> -nan us
>>>>>>
>>>>>>
>>>>>> You are probably missing --enable-x86-tsc on configure command line.
>>>>>>
>>>>>
>>>>> (or rather have passed --disable-x86-tsc as --enable-x86-tsc should be
>>>>> the default).
>>>>>
>>>>
>>>> --enable-x86-tsc is in there, and configure finds TSC support:
>>>>
>>>> + ./configure --build=i686-koji-linux-gnu --host=i686-koji-linux-gnu
>>>> --target=i686-redhat-linux-gnu --program-prefix= --prefix=/usr
>>>> --exec-prefix=/usr --bindir=/usr/bin --sbindir=/usr/sbin
>>>> --sysconfdir=/etc --datadir=/usr/share
>>>> --includedir=/usr/include/xenomai/ --libdir=/usr/lib
>>>> --libexecdir=/usr/libexec --localstatedir=/var --sharedstatedir=/var/lib
>>>> --mandir=/usr/share/man --infodir=/usr/share/info --enable-x86-tsc
>>>> --enable-dlopen-skins --with-testdir=/usr/lib/xenomai
>>>> [...]
>>>> checking for x86 TSC support... y
>>>>
>>>> Build log follows. What else can I check?
>>>
>>>
>>> Could you send me the "tsc" program?
>>>
>>
>> http://www.zultron.com/static/2013/01/xenomai/3.5.7-test-32-bit/tsc.gz
>
>
> Ok, this program is compiled correctly, the error you have suggests that
> on your machine reading the tsc takes systematically more than 80
> cycles, a hardcoded value that is well, a bit too arbitrary.
An arbitrary extra zero, raising max dt to 800, confirms your explanation.
John
$ ./tsc
Checking tsc for 1 minute(s)
min: 99, max: 666, avg: 100.002
min: 99, max: 711, avg: 100.002
min: 99, max: 630, avg: 100.002
min: 99, max: 720, avg: 100.003
min: 99, max: 657, avg: 100.003
min: 99, max: 729, avg: 100.002
min: 99, max: 603, avg: 100.002
min: 99, max: 693, avg: 100.002
min: 99, max: 657, avg: 100.001
min: 99, max: 693, avg: 100.002
min: 99, max: 675, avg: 100.003
min: 99, max: 675, avg: 100.001
min: 99, max: 756, avg: 100.001
min: 99, max: 720, avg: 100
min: 99, max: 342, avg: 100
min: 99, max: 414, avg: 100
min: 99, max: 666, avg: 100
min: 99, max: 756, avg: 100
min: 99, max: 234, avg: 100
min: 99, max: 387, avg: 100
min: 99, max: 270, avg: 100
min: 99, max: 576, avg: 100
min: 99, max: 612, avg: 100
min: 99, max: 369, avg: 100.001
min: 99, max: 369, avg: 100
min: 99, max: 576, avg: 100
min: 99, max: 729, avg: 100
min: 99, max: 513, avg: 100
min: 99, max: 450, avg: 100
min: 99, max: 495, avg: 100
min: 99, max: 441, avg: 100
min: 99, max: 351, avg: 100
min: 99, max: 738, avg: 100
min: 99, max: 567, avg: 100
min: 99, max: 468, avg: 100
min: 99, max: 567, avg: 100
min: 99, max: 513, avg: 100
min: 99, max: 585, avg: 100
min: 99, max: 441, avg: 100
min: 99, max: 261, avg: 100
min: 99, max: 405, avg: 100
min: 99, max: 225, avg: 100
min: 99, max: 459, avg: 100
min: 99, max: 279, avg: 100
min: 99, max: 378, avg: 100
min: 99, max: 558, avg: 100
min: 99, max: 504, avg: 100
min: 99, max: 720, avg: 100
min: 99, max: 378, avg: 100
min: 99, max: 387, avg: 100
min: 99, max: 225, avg: 100
min: 99, max: 333, avg: 100
min: 99, max: 333, avg: 100
min: 99, max: 342, avg: 100
min: 99, max: 648, avg: 100
min: 99, max: 432, avg: 100
min: 99, max: 288, avg: 100
min: 99, max: 414, avg: 100
min: 99, max: 603, avg: 100
min: 99, max: 513, avg: 100
min: 99, max: 756, avg: 100 -> 0.0417693 us
^ permalink raw reply [flat|nested] 12+ messages in thread
end of thread, other threads:[~2013-01-20 19:22 UTC | newest]
Thread overview: 12+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-01-19 20:17 [Xenomai] 32-bit regression tests: CLOCK_REALTIME wonkiness John Morris
2013-01-19 20:25 ` Gilles Chanteperdrix
2013-01-19 20:32 ` John Morris
2013-01-19 20:34 ` Gilles Chanteperdrix
2013-01-19 22:12 ` John Morris
2013-01-19 22:14 ` Gilles Chanteperdrix
2013-01-19 22:16 ` Gilles Chanteperdrix
2013-01-19 22:50 ` John Morris
2013-01-19 22:52 ` Gilles Chanteperdrix
2013-01-19 22:57 ` John Morris
2013-01-19 23:03 ` Gilles Chanteperdrix
2013-01-20 19:22 ` John Morris
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.