Re: [oe-commits] Andy Voltz : tune-cortexa.inc: create a common include for cortex-a armv7a tuning

Re: [oe-commits] Andy Voltz : tune-cortexa.inc: create a common include for cortex-a armv7a tuning

[Martin Jansa]

[-- Attachment #1: Type: text/plain, Size: 2438 bytes --]

On Fri, Jun 07, 2013 at 03:43:57PM +0000, git@git.openembedded.org wrote:
> Module: openembedded-core.git
> Branch: master
> Commit: 3e4f4a1cf07ff7cf4c71566492385f8fbf581789
> URL:    http://git.openembedded.org/?p=openembedded-core.git&a=commit;h=3e4f4a1cf07ff7cf4c71566492385f8fbf581789
> 
> Author: Andy Voltz <andy.voltz@timesys.com>
> Date:   Wed Jun  5 11:52:29 2013 -0400
> 
> tune-cortexa.inc: create a common include for cortex-a armv7a tuning
> 
> The tuning files for the cortex-a* processors are mostly identical for
> the A7,A8,A9,A15 processors. Rework these files to use a CORTEX_ID
> variable to setup the tuning for each specific processor.
> 
> Signed-off-by: Andy Voltz <andy.voltz@timesys.com>
> Signed-off-by: Saul Wold <sgw@linux.intel.com>

Something is wrong with this one:

ERROR:  OE-core's config sanity checker detected a potential misconfiguration.
    Either fix the cause of this error or at your own risk disable the checker (see sanity.conf).
    Following is the list of potential problems / advisories:

    Error, the PACKAGE_ARCHS variable does not contain TUNE_PKGARCH (${ARMPKGARCH_tune-cortexa8t-neon}).Toolchain tunings invalid:
Tuning 'cortexa8t-neon' has no defined features, and cannot be used.

ERROR: Execution of event handler 'check_sanity_eventhandler' failed

Please use tests like
http://git.openembedded.org/openembedded-core-contrib/log/?h=jansa/tune2-test
to check that all tunes are still valid after such change.

> 
> ---
> 
>  meta/conf/machine/include/arm/README         |    3 ++
>  meta/conf/machine/include/tune-cortexa.inc   |   35 ++++++++++++++++++++++++++
>  meta/conf/machine/include/tune-cortexa15.inc |   35 +------------------------
>  meta/conf/machine/include/tune-cortexa7.inc  |   35 +------------------------
>  meta/conf/machine/include/tune-cortexa8.inc  |   35 +------------------------
>  meta/conf/machine/include/tune-cortexa9.inc  |   35 +------------------------
>  6 files changed, 46 insertions(+), 132 deletions(-)
> 
> Diff:   http://git.openembedded.org/?p=openembedded-core.git/?a=commitdiff;h=3e4f4a1cf07ff7cf4c71566492385f8fbf581789
> _______________________________________________
> Openembedded-commits mailing list
> Openembedded-commits@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-commits

-- 
Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 205 bytes --]

Re: [oe-commits] Andy Voltz : tune-cortexa.inc: create a common include for cortex-a armv7a tuning

[Martin Jansa]

[-- Attachment #1: Type: text/plain, Size: 2865 bytes --]

On Sat, Jun 08, 2013 at 11:10:20AM +0200, Martin Jansa wrote:
> On Fri, Jun 07, 2013 at 03:43:57PM +0000, git@git.openembedded.org wrote:
> > Module: openembedded-core.git
> > Branch: master
> > Commit: 3e4f4a1cf07ff7cf4c71566492385f8fbf581789
> > URL:    http://git.openembedded.org/?p=openembedded-core.git&a=commit;h=3e4f4a1cf07ff7cf4c71566492385f8fbf581789
> > 
> > Author: Andy Voltz <andy.voltz@timesys.com>
> > Date:   Wed Jun  5 11:52:29 2013 -0400
> > 
> > tune-cortexa.inc: create a common include for cortex-a armv7a tuning
> > 
> > The tuning files for the cortex-a* processors are mostly identical for
> > the A7,A8,A9,A15 processors. Rework these files to use a CORTEX_ID
> > variable to setup the tuning for each specific processor.
> > 
> > Signed-off-by: Andy Voltz <andy.voltz@timesys.com>
> > Signed-off-by: Saul Wold <sgw@linux.intel.com>
> 
> Something is wrong with this one:
> 
> ERROR:  OE-core's config sanity checker detected a potential misconfiguration.
>     Either fix the cause of this error or at your own risk disable the checker (see sanity.conf).
>     Following is the list of potential problems / advisories:
> 
>     Error, the PACKAGE_ARCHS variable does not contain TUNE_PKGARCH (${ARMPKGARCH_tune-cortexa8t-neon}).Toolchain tunings invalid:
> Tuning 'cortexa8t-neon' has no defined features, and cannot be used.
> 
> ERROR: Execution of event handler 'check_sanity_eventhandler' failed
> 
> Please use tests like
> http://git.openembedded.org/openembedded-core-contrib/log/?h=jansa/tune2-test
> to check that all tunes are still valid after such change.

Also it was using -mtune and documentation talks about -mtune, but
actual tune-cortexa.inc is using -mcpu, this change wasn't mentioned in
commit message and I don't think it's improvement.

> > 
> > ---
> > 
> >  meta/conf/machine/include/arm/README         |    3 ++
> >  meta/conf/machine/include/tune-cortexa.inc   |   35 ++++++++++++++++++++++++++
> >  meta/conf/machine/include/tune-cortexa15.inc |   35 +------------------------
> >  meta/conf/machine/include/tune-cortexa7.inc  |   35 +------------------------
> >  meta/conf/machine/include/tune-cortexa8.inc  |   35 +------------------------
> >  meta/conf/machine/include/tune-cortexa9.inc  |   35 +------------------------
> >  6 files changed, 46 insertions(+), 132 deletions(-)
> > 
> > Diff:   http://git.openembedded.org/?p=openembedded-core.git/?a=commitdiff;h=3e4f4a1cf07ff7cf4c71566492385f8fbf581789
> > _______________________________________________
> > Openembedded-commits mailing list
> > Openembedded-commits@lists.openembedded.org
> > http://lists.openembedded.org/mailman/listinfo/openembedded-commits
> 
> -- 
> Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com



-- 
Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 205 bytes --]

Re: [oe-commits] Andy Voltz : tune-cortexa.inc: create a common include for cortex-a armv7a tuning

[Andy Voltz]

Hi Martin,

> On Sat, Jun 08, 2013 at 11:10:20AM +0200, Martin Jansa wrote:

> > Something is wrong with this one:
> > 
> > ERROR:  OE-core's config sanity checker detected a potential misconfiguration.
> >     Either fix the cause of this error or at your own risk disable the checker (see sanity.conf).
> >     Following is the list of potential problems / advisories:
> > 
> >     Error, the PACKAGE_ARCHS variable does not contain TUNE_PKGARCH (${ARMPKGARCH_tune-cortexa8t-neon}).Toolchain tunings invalid:
> > Tuning 'cortexa8t-neon' has no defined features, and cannot be used.
> > 
> > ERROR: Execution of event handler 'check_sanity_eventhandler' failed
> > 
> > Please use tests like
> > http://git.openembedded.org/openembedded-core-contrib/log/?h=jansa/tune2-test
> > to check that all tunes are still valid after such change.
> 

I was not aware of this test. I'll look into this.

> Also it was using -mtune and documentation talks about -mtune, but
> actual tune-cortexa.inc is using -mcpu, this change wasn't mentioned in
> commit message and I don't think it's improvement.
> 

*facepalm*

This is actually my mistake when I squashed my cortex-a5 file into a general
cortexa file. I meant to keep the -mtune in place. I started out using mcpu
before realizing mtune was also valid for cortex-a5.

I have a patch to revert that change / go back to -mtune. I will send it after
I rebuild for cortex-a5.

Sorry about that. It didn't come out because the build was happy with -mcpu=cortex-a5
-- 
Andy Voltz
Timesys Corporation

Re: [oe-commits] Andy Voltz : tune-cortexa.inc: create a common include for cortex-a armv7a tuning

[Khem Raj]

On Mon, Jun 10, 2013 at 8:50 AM, Andy Voltz <andy.voltz@timesys.com> wrote:
>> > ERROR:  OE-core's config sanity checker detected a potential misconfiguration.
>> >     Either fix the cause of this error or at your own risk disable the checker (see sanity.conf).
>> >     Following is the list of potential problems / advisories:
>> >
>> >     Error, the PACKAGE_ARCHS variable does not contain TUNE_PKGARCH (${ARMPKGARCH_tune-cortexa8t-neon}).Toolchain tunings invalid:
>> > Tuning 'cortexa8t-neon' has no defined features, and cannot be used.
>> >
>> > ERROR: Execution of event handler 'check_sanity_eventhandler' failed
>> >
>> > Please use tests like
>> > http://git.openembedded.org/openembedded-core-contrib/log/?h=jansa/tune2-test
>> > to check that all tunes are still valid after such change.
>>
>
> I was not aware of this test. I'll look into this.

This patch is not going to work. How did you test it ?
Its trying to substitute into overrides which then are not expanded
before being treated as overrides.
as a result overrides variable themselves are wrong.
We have to revert this patch please and I conur with Martin it did not
seem like a significant improvement either.

Re: [oe-commits] Andy Voltz : tune-cortexa.inc: create a common include for cortex-a armv7a tuning

[Andy Voltz]

Hi Khem,

On Tue, Jun 11, 2013 at 12:12:47AM -0400, Khem Raj wrote:
> On Mon, Jun 10, 2013 at 8:50 AM, Andy Voltz <andy.voltz@timesys.com> wrote:
> >> > ERROR:  OE-core's config sanity checker detected a potential misconfiguration.
> >> >     Either fix the cause of this error or at your own risk disable the checker (see sanity.conf).
> >> >     Following is the list of potential problems / advisories:
> >> >
> >> >     Error, the PACKAGE_ARCHS variable does not contain TUNE_PKGARCH (${ARMPKGARCH_tune-cortexa8t-neon}).Toolchain tunings invalid:
> >> > Tuning 'cortexa8t-neon' has no defined features, and cannot be used.
> >> >
> >> > ERROR: Execution of event handler 'check_sanity_eventhandler' failed
> >> >
> >> > Please use tests like
> >> > http://git.openembedded.org/openembedded-core-contrib/log/?h=jansa/tune2-test
> >> > to check that all tunes are still valid after such change.
> >>
> >
> > I was not aware of this test. I'll look into this.
> 
> This patch is not going to work. How did you test it ?
> Its trying to substitute into overrides which then are not expanded
> before being treated as overrides.
> as a result overrides variable themselves are wrong.
> We have to revert this patch please and I conur with Martin it did not
> seem like a significant improvement either.

I understand that. I sent an updated patch which reverts to mtune, but I'm getting
acquainted with this tune test and it's still showing issues. I'm need to dig in more.

I tested this originally by building for the cortex-a5. I'm going to revert
locally and keep using this test to keep things consistent.

Regards
-- 
Andy Voltz
Timesys Corporation

Re: [oe-commits] Andy Voltz : tune-cortexa.inc: create a common include for cortex-a armv7a tuning

[Khem Raj]

On Tue, Jun 11, 2013 at 6:45 AM, Andy Voltz <andy.voltz@timesys.com> wrote:
> I understand that. I sent an updated patch which reverts to mtune, but I'm getting

mtune/march is small issue with this patch. I would recommend to send
a revert of the patches
since all arm architectures are broken as of now in OE-Core

Re: [oe-commits] Andy Voltz : tune-cortexa.inc: create a common include for cortex-a armv7a tuning

[Andy Voltz]

Hi Khem,

> On Tue, Jun 11, 2013 at 6:45 AM, Andy Voltz <andy.voltz@timesys.com> wrote:
> > I understand that. I sent an updated patch which reverts to mtune, but I'm getting
> 
> mtune/march is small issue with this patch. I would recommend to send
> a revert of the patches
> since all arm architectures are broken as of now in OE-Core

I see the issue now. It seems that overriding the tunes with a variable
reference does not work for all the tunings. With the tune test script, I'm
seeing that these tunes are broken:
log.fake-cortexa9.cortexa9*
log.fake-cortexa8.cortexa8*

I see that variables can be assigned with a reference based on stuff like 
FILES_${QT_BASE_NAME}-fonts = ...

(From recipes-qt/qt4/qt4-embedded.inc)

I'm not sure why it works there and not in the tunings, or how to fix.

For now I have a revert patch which removes tune-cortexa.inc and restores the
previous contents of the a7,a8,a9,15 tuning files. 

It also adds the tune-cortexa5.inc file I had sent on the list originally
which led to this code-reduction experiment. That thread is here for reference:

http://lists.openembedded.org/pipermail/openembedded-core/2013-June/079452.html

Thanks for pointing out this issue. Patch is on the way.

Regards
-- 
Andy Voltz
Timesys Corporation

Re: [oe-commits] Andy Voltz : tune-cortexa.inc: create a common include for cortex-a armv7a tuning

[Khem Raj]

On Jun 12, 2013, at 9:13 AM, Andy Voltz <andy.voltz@timesys.com> wrote:

> Hi Khem,
> 
>> On Tue, Jun 11, 2013 at 6:45 AM, Andy Voltz <andy.voltz@timesys.com> wrote:
>>> I understand that. I sent an updated patch which reverts to mtune, but I'm getting
>> 
>> mtune/march is small issue with this patch. I would recommend to send
>> a revert of the patches
>> since all arm architectures are broken as of now in OE-Core
> 
> I see the issue now. It seems that overriding the tunes with a variable
> reference does not work for all the tunings. With the tune test script, I'm
> seeing that these tunes are broken:
> log.fake-cortexa9.cortexa9*
> log.fake-cortexa8.cortexa8*
> 
> I see that variables can be assigned with a reference based on stuff like 
> FILES_${QT_BASE_NAME}-fonts = …
> 
> (From recipes-qt/qt4/qt4-embedded.inc)
> 
> I'm not sure why it works there and not in the tunings, or how to fix.
> 


its a different thing. Its constructing a variable name and not override. You were trying to create a override on the fly
that won't work.

> For now I have a revert patch which removes tune-cortexa.inc and restores the
> previous contents of the a7,a8,a9,15 tuning files. 
> 
> It also adds the tune-cortexa5.inc file I had sent on the list originally
> which led to this code-reduction experiment. That thread is here for reference:
> 
> http://lists.openembedded.org/pipermail/openembedded-core/2013-June/079452.html
> 
> Thanks for pointing out this issue. Patch is on the way.
> 

np


> Regards
> -- 
> Andy Voltz
> Timesys Corporation