Patchwork [10/22] tune/arch-powerpc: Use fpu-soft to set PPCPKGSFX_FPU

login
register
mail settings
Submitter Saul Wold
Date Aug. 1, 2011, 7:36 a.m.
Message ID <3e502d7441be9c7c25598237f79c527bdf3ce38e.1312182844.git.sgw@linux.intel.com>
Download mbox | patch
Permalink /patch/8941/
State New, archived
Headers show

Comments

Saul Wold - Aug. 1, 2011, 7:36 a.m.
From: Kumar Gala <galak@kernel.crashing.org>

Its more likely that we explicitly set soft-floating point support for a
given target than hard.  So use 'fpu-soft' in TUNE_FEATURES rather than
'fpu-hard' to determine setting 'nf' (no-float) in PPCPKGSFX_FPU.

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
---
 meta/conf/machine/include/powerpc/arch-powerpc.inc |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
Richard Purdie - Aug. 1, 2011, 1:03 p.m.
On Mon, 2011-08-01 at 00:36 -0700, Saul Wold wrote:
> From: Kumar Gala <galak@kernel.crashing.org>
> 
> Its more likely that we explicitly set soft-floating point support for a
> given target than hard.  So use 'fpu-soft' in TUNE_FEATURES rather than
> 'fpu-hard' to determine setting 'nf' (no-float) in PPCPKGSFX_FPU.
> 
> Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
> ---
>  meta/conf/machine/include/powerpc/arch-powerpc.inc |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/meta/conf/machine/include/powerpc/arch-powerpc.inc b/meta/conf/machine/include/powerpc/arch-powerpc.inc
> index 8cc76ab..013755d 100644
> --- a/meta/conf/machine/include/powerpc/arch-powerpc.inc
> +++ b/meta/conf/machine/include/powerpc/arch-powerpc.inc
> @@ -18,7 +18,7 @@ TARGET_FPU .= "${@bb.utils.contains("TUNE_FEATURES", "fpu-soft", "soft", "", d)}
>  
>  ABIEXTENSION = "${@['','spe'][d.getVar('TARGET_FPU', True) in ['ppc-efd', 'ppc-efs']]}"
>  
> -PPCPKGSFX_FPU = "${@bb.utils.contains("TUNE_FEATURES", "fpu-hard" , "", "-nf", d)}"
> +PPCPKGSFX_FPU = "${@bb.utils.contains("TUNE_FEATURES", "fpu-soft", "-nf", "", d)}"
>  
>  TUNE_PKGARCH ?= "${TUNE_ARCH}${PPCPKGSFX_FPU}"

Whilst the change itself may or may not be ok, it would need to change
the various boards that are already setting or not setting fpu-hard as a
feature so as it stands this change is incomplete.

Cheers,

Richard
Kumar Gala - Aug. 1, 2011, 2:13 p.m.
On Aug 1, 2011, at 8:03 AM, Richard Purdie wrote:

> On Mon, 2011-08-01 at 00:36 -0700, Saul Wold wrote:
>> From: Kumar Gala <galak@kernel.crashing.org>
>> 
>> Its more likely that we explicitly set soft-floating point support for a
>> given target than hard.  So use 'fpu-soft' in TUNE_FEATURES rather than
>> 'fpu-hard' to determine setting 'nf' (no-float) in PPCPKGSFX_FPU.
>> 
>> Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
>> ---
>> meta/conf/machine/include/powerpc/arch-powerpc.inc |    2 +-
>> 1 files changed, 1 insertions(+), 1 deletions(-)
>> 
>> diff --git a/meta/conf/machine/include/powerpc/arch-powerpc.inc b/meta/conf/machine/include/powerpc/arch-powerpc.inc
>> index 8cc76ab..013755d 100644
>> --- a/meta/conf/machine/include/powerpc/arch-powerpc.inc
>> +++ b/meta/conf/machine/include/powerpc/arch-powerpc.inc
>> @@ -18,7 +18,7 @@ TARGET_FPU .= "${@bb.utils.contains("TUNE_FEATURES", "fpu-soft", "soft", "", d)}
>> 
>> ABIEXTENSION = "${@['','spe'][d.getVar('TARGET_FPU', True) in ['ppc-efd', 'ppc-efs']]}"
>> 
>> -PPCPKGSFX_FPU = "${@bb.utils.contains("TUNE_FEATURES", "fpu-hard" , "", "-nf", d)}"
>> +PPCPKGSFX_FPU = "${@bb.utils.contains("TUNE_FEATURES", "fpu-soft", "-nf", "", d)}"
>> 
>> TUNE_PKGARCH ?= "${TUNE_ARCH}${PPCPKGSFX_FPU}"
> 
> Whilst the change itself may or may not be ok, it would need to change
> the various boards that are already setting or not setting fpu-hard as a
> feature so as it stands this change is incomplete.

Which boards are you speaking of?  I'm not aware of any that set soft float at this point for PPC.

- k
Mark Hatle - Aug. 1, 2011, 2:50 p.m.
On 8/1/11 9:13 AM, Kumar Gala wrote:
> 
> On Aug 1, 2011, at 8:03 AM, Richard Purdie wrote:
> 
>> On Mon, 2011-08-01 at 00:36 -0700, Saul Wold wrote:
>>> From: Kumar Gala <galak@kernel.crashing.org>
>>>
>>> Its more likely that we explicitly set soft-floating point support for a
>>> given target than hard.  So use 'fpu-soft' in TUNE_FEATURES rather than
>>> 'fpu-hard' to determine setting 'nf' (no-float) in PPCPKGSFX_FPU.
>>>
>>> Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
>>> ---
>>> meta/conf/machine/include/powerpc/arch-powerpc.inc |    2 +-
>>> 1 files changed, 1 insertions(+), 1 deletions(-)
>>>
>>> diff --git a/meta/conf/machine/include/powerpc/arch-powerpc.inc b/meta/conf/machine/include/powerpc/arch-powerpc.inc
>>> index 8cc76ab..013755d 100644
>>> --- a/meta/conf/machine/include/powerpc/arch-powerpc.inc
>>> +++ b/meta/conf/machine/include/powerpc/arch-powerpc.inc
>>> @@ -18,7 +18,7 @@ TARGET_FPU .= "${@bb.utils.contains("TUNE_FEATURES", "fpu-soft", "soft", "", d)}
>>>
>>> ABIEXTENSION = "${@['','spe'][d.getVar('TARGET_FPU', True) in ['ppc-efd', 'ppc-efs']]}"
>>>
>>> -PPCPKGSFX_FPU = "${@bb.utils.contains("TUNE_FEATURES", "fpu-hard" , "", "-nf", d)}"
>>> +PPCPKGSFX_FPU = "${@bb.utils.contains("TUNE_FEATURES", "fpu-soft", "-nf", "", d)}"
>>>
>>> TUNE_PKGARCH ?= "${TUNE_ARCH}${PPCPKGSFX_FPU}"
>>
>> Whilst the change itself may or may not be ok, it would need to change
>> the various boards that are already setting or not setting fpu-hard as a
>> feature so as it stands this change is incomplete.
> 
> Which boards are you speaking of?  I'm not aware of any that set soft float at this point for PPC.

It shouldn't be the board setting it to soft-fp, but instead the tune.
Something like qemuppc can use the nf 603e tune.

--Mark

> - k
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core
Kumar Gala - Aug. 1, 2011, 3:57 p.m.
On Aug 1, 2011, at 9:50 AM, Mark Hatle wrote:

> On 8/1/11 9:13 AM, Kumar Gala wrote:
>> 
>> On Aug 1, 2011, at 8:03 AM, Richard Purdie wrote:
>> 
>>> On Mon, 2011-08-01 at 00:36 -0700, Saul Wold wrote:
>>>> From: Kumar Gala <galak@kernel.crashing.org>
>>>> 
>>>> Its more likely that we explicitly set soft-floating point support for a
>>>> given target than hard.  So use 'fpu-soft' in TUNE_FEATURES rather than
>>>> 'fpu-hard' to determine setting 'nf' (no-float) in PPCPKGSFX_FPU.
>>>> 
>>>> Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
>>>> ---
>>>> meta/conf/machine/include/powerpc/arch-powerpc.inc |    2 +-
>>>> 1 files changed, 1 insertions(+), 1 deletions(-)
>>>> 
>>>> diff --git a/meta/conf/machine/include/powerpc/arch-powerpc.inc b/meta/conf/machine/include/powerpc/arch-powerpc.inc
>>>> index 8cc76ab..013755d 100644
>>>> --- a/meta/conf/machine/include/powerpc/arch-powerpc.inc
>>>> +++ b/meta/conf/machine/include/powerpc/arch-powerpc.inc
>>>> @@ -18,7 +18,7 @@ TARGET_FPU .= "${@bb.utils.contains("TUNE_FEATURES", "fpu-soft", "soft", "", d)}
>>>> 
>>>> ABIEXTENSION = "${@['','spe'][d.getVar('TARGET_FPU', True) in ['ppc-efd', 'ppc-efs']]}"
>>>> 
>>>> -PPCPKGSFX_FPU = "${@bb.utils.contains("TUNE_FEATURES", "fpu-hard" , "", "-nf", d)}"
>>>> +PPCPKGSFX_FPU = "${@bb.utils.contains("TUNE_FEATURES", "fpu-soft", "-nf", "", d)}"
>>>> 
>>>> TUNE_PKGARCH ?= "${TUNE_ARCH}${PPCPKGSFX_FPU}"
>>> 
>>> Whilst the change itself may or may not be ok, it would need to change
>>> the various boards that are already setting or not setting fpu-hard as a
>>> feature so as it stands this change is incomplete.
>> 
>> Which boards are you speaking of?  I'm not aware of any that set soft float at this point for PPC.
> 
> It shouldn't be the board setting it to soft-fp, but instead the tune.
> Something like qemuppc can use the nf 603e tune.
> 
> --Mark

Right, and the tune files we have do this already, tune-ppce300c2.inc.

TUNE_FEATURES_tune-ppce300c2 = "m32 fpu-soft ppce300c2"

So, see no reason not to apply this.

- k
Kumar Gala - Aug. 2, 2011, 6:31 p.m.
On Aug 1, 2011, at 8:03 AM, Richard Purdie wrote:

> On Mon, 2011-08-01 at 00:36 -0700, Saul Wold wrote:
>> From: Kumar Gala <galak@kernel.crashing.org>
>> 
>> Its more likely that we explicitly set soft-floating point support for a
>> given target than hard.  So use 'fpu-soft' in TUNE_FEATURES rather than
>> 'fpu-hard' to determine setting 'nf' (no-float) in PPCPKGSFX_FPU.
>> 
>> Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
>> ---
>> meta/conf/machine/include/powerpc/arch-powerpc.inc |    2 +-
>> 1 files changed, 1 insertions(+), 1 deletions(-)
>> 
>> diff --git a/meta/conf/machine/include/powerpc/arch-powerpc.inc b/meta/conf/machine/include/powerpc/arch-powerpc.inc
>> index 8cc76ab..013755d 100644
>> --- a/meta/conf/machine/include/powerpc/arch-powerpc.inc
>> +++ b/meta/conf/machine/include/powerpc/arch-powerpc.inc
>> @@ -18,7 +18,7 @@ TARGET_FPU .= "${@bb.utils.contains("TUNE_FEATURES", "fpu-soft", "soft", "", d)}
>> 
>> ABIEXTENSION = "${@['','spe'][d.getVar('TARGET_FPU', True) in ['ppc-efd', 'ppc-efs']]}"
>> 
>> -PPCPKGSFX_FPU = "${@bb.utils.contains("TUNE_FEATURES", "fpu-hard" , "", "-nf", d)}"
>> +PPCPKGSFX_FPU = "${@bb.utils.contains("TUNE_FEATURES", "fpu-soft", "-nf", "", d)}"
>> 
>> TUNE_PKGARCH ?= "${TUNE_ARCH}${PPCPKGSFX_FPU}"
> 
> Whilst the change itself may or may not be ok, it would need to change
> the various boards that are already setting or not setting fpu-hard as a
> feature so as it stands this change is incomplete.
> 
> Cheers,
> 
> Richard

Maybe I misunderstood when I first read this, did you mean we should remove explicitly setting 'fpu-hard' on the boards / tune files that are doing this?

- k
Richard Purdie - Aug. 2, 2011, 10:36 p.m.
On Tue, 2011-08-02 at 13:31 -0500, Kumar Gala wrote:
> On Aug 1, 2011, at 8:03 AM, Richard Purdie wrote:
> 
> > On Mon, 2011-08-01 at 00:36 -0700, Saul Wold wrote:
> >> From: Kumar Gala <galak@kernel.crashing.org>
> >> 
> >> Its more likely that we explicitly set soft-floating point support for a
> >> given target than hard.  So use 'fpu-soft' in TUNE_FEATURES rather than
> >> 'fpu-hard' to determine setting 'nf' (no-float) in PPCPKGSFX_FPU.
> >> 
> >> Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
> >> ---
> >> meta/conf/machine/include/powerpc/arch-powerpc.inc |    2 +-
> >> 1 files changed, 1 insertions(+), 1 deletions(-)
> >> 
> >> diff --git a/meta/conf/machine/include/powerpc/arch-powerpc.inc b/meta/conf/machine/include/powerpc/arch-powerpc.inc
> >> index 8cc76ab..013755d 100644
> >> --- a/meta/conf/machine/include/powerpc/arch-powerpc.inc
> >> +++ b/meta/conf/machine/include/powerpc/arch-powerpc.inc
> >> @@ -18,7 +18,7 @@ TARGET_FPU .= "${@bb.utils.contains("TUNE_FEATURES", "fpu-soft", "soft", "", d)}
> >> 
> >> ABIEXTENSION = "${@['','spe'][d.getVar('TARGET_FPU', True) in ['ppc-efd', 'ppc-efs']]}"
> >> 
> >> -PPCPKGSFX_FPU = "${@bb.utils.contains("TUNE_FEATURES", "fpu-hard" , "", "-nf", d)}"
> >> +PPCPKGSFX_FPU = "${@bb.utils.contains("TUNE_FEATURES", "fpu-soft", "-nf", "", d)}"
> >> 
> >> TUNE_PKGARCH ?= "${TUNE_ARCH}${PPCPKGSFX_FPU}"
> > 
> > Whilst the change itself may or may not be ok, it would need to change
> > the various boards that are already setting or not setting fpu-hard as a
> > feature so as it stands this change is incomplete.
>
> Maybe I misunderstood when I first read this, did you mean we should
> remove explicitly setting 'fpu-hard' on the boards / tune files that
> are doing this?

Yes, that is what I meant.

Cheers,

Richard

Patch

diff --git a/meta/conf/machine/include/powerpc/arch-powerpc.inc b/meta/conf/machine/include/powerpc/arch-powerpc.inc
index 8cc76ab..013755d 100644
--- a/meta/conf/machine/include/powerpc/arch-powerpc.inc
+++ b/meta/conf/machine/include/powerpc/arch-powerpc.inc
@@ -18,7 +18,7 @@  TARGET_FPU .= "${@bb.utils.contains("TUNE_FEATURES", "fpu-soft", "soft", "", d)}
 
 ABIEXTENSION = "${@['','spe'][d.getVar('TARGET_FPU', True) in ['ppc-efd', 'ppc-efs']]}"
 
-PPCPKGSFX_FPU = "${@bb.utils.contains("TUNE_FEATURES", "fpu-hard" , "", "-nf", d)}"
+PPCPKGSFX_FPU = "${@bb.utils.contains("TUNE_FEATURES", "fpu-soft", "-nf", "", d)}"
 
 TUNE_PKGARCH ?= "${TUNE_ARCH}${PPCPKGSFX_FPU}"