Patchwork x264: SPE is not compatible with Altivec

login
register
mail settings
Submitter David Nyström
Date Feb. 21, 2014, 1:56 p.m.
Message ID <1392991017-3524-1-git-send-email-david.c.nystrom@gmail.com>
Download mbox | patch
Permalink /patch/67115/
State New
Headers show

Comments

David Nyström - Feb. 21, 2014, 1:56 p.m.
Signed-off-by: David Nyström <david.c.nystrom@gmail.com>
---
 meta/recipes-multimedia/x264/x264_git.bb | 1 +
 1 file changed, 1 insertion(+)
Saul Wold - Feb. 21, 2014, 6:28 p.m.
On 02/21/2014 05:56 AM, David Nyström wrote:
> Signed-off-by: David Nyström <david.c.nystrom@gmail.com>
> ---
>   meta/recipes-multimedia/x264/x264_git.bb | 1 +
>   1 file changed, 1 insertion(+)
>
> diff --git a/meta/recipes-multimedia/x264/x264_git.bb b/meta/recipes-multimedia/x264/x264_git.bb
> index 5ebf885..782c26e 100644
> --- a/meta/recipes-multimedia/x264/x264_git.bb
> +++ b/meta/recipes-multimedia/x264/x264_git.bb
> @@ -23,6 +23,7 @@ inherit lib_package pkgconfig
>   X264_DISABLE_ASM = ""
>   X264_DISABLE_ASM_armv4 = "--disable-asm"
>   X264_DISABLE_ASM_armv5 = "--disable-asm"
> +X264_DISABLE_ASM_e500v2 = "--disable-asm"
>
I think that this change really belongs in the e500v2 BSP, it's not 
generic enough to include in core.

Thanks
	Sau!


>   EXTRA_OECONF = '--prefix=${prefix} \
>                   --host=${HOST_SYS} \
>
David Nyström - Feb. 21, 2014, 7:40 p.m.
On Feb 21, 2014 7:28 PM, "Saul Wold" <sgw@linux.intel.com> wrote:
>
> On 02/21/2014 05:56 AM, David Nyström wrote:
>>
>> Signed-off-by: David Nyström <david.c.nystrom@gmail.com>
>> ---
>>   meta/recipes-multimedia/x264/x264_git.bb | 1 +
>>   1 file changed, 1 insertion(+)
>>
>> diff --git a/meta/recipes-multimedia/x264/x264_git.bbb/meta/recipes-multimedia/x264/
x264_git.bb
>> index 5ebf885..782c26e 100644
>> --- a/meta/recipes-multimedia/x264/x264_git.bb
>> +++ b/meta/recipes-multimedia/x264/x264_git.bb
>> @@ -23,6 +23,7 @@ inherit lib_package pkgconfig
>>   X264_DISABLE_ASM = ""
>>   X264_DISABLE_ASM_armv4 = "--disable-asm"
>>   X264_DISABLE_ASM_armv5 = "--disable-asm"
>> +X264_DISABLE_ASM_e500v2 = "--disable-asm"
>>
> I think that this change really belongs in the e500v2 BSP, it's not
generic enough to include in core.

Perhaps you know this already, but e500v2 is a cpu arch, same as armv5.

For future reference, should all ppc related CPU-arch specific stuff be
directed to the FSL bsp ?

Br,
David

>
> Thanks
>         Sau!
>
>
>
>>   EXTRA_OECONF = '--prefix=${prefix} \
>>                   --host=${HOST_SYS} \
>>
Phil Blundell - Feb. 21, 2014, 9:20 p.m.
On Fri, 2014-02-21 at 20:40 +0100, David Nystrom wrote:
> Perhaps you know this already, but e500v2 is a cpu arch, same as
> armv5. 

Is that really true?  I thought e500v2 was an actual core and the
equivalent architecture was Power v2.03 or something.

>For future reference, should all ppc related CPU-arch specific stuff 
>be directed to the FSL bsp ?

Since we do ship tune-ppce500.inc in oe-core, I think it makes sense to
also include the overrides necessary to make compilation work with SPE
enabled.  It doesn't seem unreasonable that users would expect that the
recipes in oe-core would work with a configuration that comes as
standard.

However, adding overrides for e500v2 specifically does seem like a bad
plan since we will then end up having to duplicate the thing for every
other non-altivec configuration.  Can you come up with some override
that will apply to all SPE configurations, or failing that a patch for
the configure script that will disable the altivec assembler if it
notices that ${CC} is configured for SPE?

p.
David Nyström - Feb. 21, 2014, 9:45 p.m.
On 2014-02-21 22:20, Phil Blundell wrote:
> On Fri, 2014-02-21 at 20:40 +0100, David Nystrom wrote:
>> Perhaps you know this already, but e500v2 is a cpu arch, same as
>> armv5.
> Is that really true?  I thought e500v2 was an actual core and the
> equivalent architecture was Power v2.03 or something.

Yes, good point.
I understand Sauls concern.
>
>> For future reference, should all ppc related CPU-arch specific stuff
>> be directed to the FSL bsp ?
> Since we do ship tune-ppce500.inc in oe-core, I think it makes sense to
> also include the overrides necessary to make compilation work with SPE
> enabled.  It doesn't seem unreasonable that users would expect that the
> recipes in oe-core would work with a configuration that comes as
> standard.
>
> However, adding overrides for e500v2 specifically does seem like a bad
> plan since we will then end up having to duplicate the thing for every
> other non-altivec configuration.  Can you come up with some override
> that will apply to all SPE configurations,
Yep, will return with V2.

> or failing that a patch for
> the configure script that will disable the altivec assembler if it
> notices that ${CC} is configured for SPE?
>
> p.
>
>

Patch

diff --git a/meta/recipes-multimedia/x264/x264_git.bb b/meta/recipes-multimedia/x264/x264_git.bb
index 5ebf885..782c26e 100644
--- a/meta/recipes-multimedia/x264/x264_git.bb
+++ b/meta/recipes-multimedia/x264/x264_git.bb
@@ -23,6 +23,7 @@  inherit lib_package pkgconfig
 X264_DISABLE_ASM = ""
 X264_DISABLE_ASM_armv4 = "--disable-asm"
 X264_DISABLE_ASM_armv5 = "--disable-asm"
+X264_DISABLE_ASM_e500v2 = "--disable-asm"
 
 EXTRA_OECONF = '--prefix=${prefix} \
                 --host=${HOST_SYS} \