[OE-core] [PATCH] eglibc_2.13: Add support for handling sqrt & sqrtf on powerpc

Mark Hatle mark.hatle at windriver.com
Mon Aug 1 15:03:02 UTC 2011


On 8/1/11 9:56 AM, Kumar Gala wrote:
> 
> On Aug 1, 2011, at 9:53 AM, Mark Hatle wrote:
> 
>> On 8/1/11 9:26 AM, Kumar Gala wrote:
>>> Some of powerpc's dont support the fsqrt[s] instructions so we need an
>>> implementation of the library functions for those processors.
>>>
>>> Signed-off-by: Kumar Gala <galak at kernel.crashing.org>
>>> ---
>>> .../recipes-core/eglibc/eglibc-2.13/ppc-sqrt.patch |  538 ++++++++++++++++++++
>>> meta/recipes-core/eglibc/eglibc_2.13.bb            |    3 +-
>>> 2 files changed, 540 insertions(+), 1 deletions(-)
>>> create mode 100644 meta/recipes-core/eglibc/eglibc-2.13/ppc-sqrt.patch
>>
>> I'm a bit confused by this patch.  Why is gcc even generating (or trying to) the
>> fsqrt instructions on cores that don't have it?  Did someone optimize gcc so
>> that it always requires fsqrt on PPC?  (This is horribly broken behavior BTW...)
>>
>> As for the patch, I don't object.. but this just looks like the wrong solution
>> to me..
>>
> 
> Mark, the issue isn't w/gcc but glibc itself.  It will use inline asm of the 'fsqrt[s]' instruction.  This we need a different implementation of the math lib functions for these cores.

Ahh, I see.  That makes a lot more sense.

--Mark

> - k
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core at lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core





More information about the Openembedded-core mailing list