[FFmpeg-devel] [PATCH 2/2] postproc/postprocess_template: remove FF_REG_sp from clobber list

[James Almer]

On 12/20/2018 8:56 PM, Michael Niedermayer wrote:
> On Thu, Dec 20, 2018 at 06:54:12PM -0300, James Almer wrote:
>> On 12/20/2018 6:40 PM, Michael Niedermayer wrote:
>>> Future gcc may no longer support this
>>>
>>> Signed-off-by: Michael Niedermayer <michael at niedermayer.cc>
>>> ---
>>>  libpostproc/postprocess_template.c | 2 +-
>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/libpostproc/postprocess_template.c b/libpostproc/postprocess_template.c
>>> index 485eb5cfc0..b0adfd168c 100644
>>> --- a/libpostproc/postprocess_template.c
>>> +++ b/libpostproc/postprocess_template.c
>>> @@ -1317,7 +1317,7 @@ DERING_CORE((%0, %1, 8)       ,(%%FF_REGd, %1, 4),%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,
>>>          "1:                        \n\t"
>>>          : : "r" (src), "r" ((x86_reg)stride), "m" (c->pQPb), "m"(c->pQPb2), "q"(tmp)
>>>            NAMED_CONSTRAINTS_ADD(deringThreshold,b00,b02,b08)
>>> -        : "%"FF_REG_a, "%"FF_REG_d, "%"FF_REG_sp
>>> +        : "%"FF_REG_a, "%"FF_REG_d
>>>      );
>>>  #else // HAVE_7REGS && (TEMPLATE_PP_MMXEXT || TEMPLATE_PP_3DNOW)
>>>      int y;
>>
>> The set compiles with gcc 9 now, so LGTM if tested. Should be backported
>> to supported releases as well.
> 
> ive tested it locally on x86 32&64 but not with gcc 9, it would be ideal
> if someone tests this with gcc 9 beyond compilation
> 
> this should use the code, the "31" can be varied to test different "strength"
> -vf 'pp=de/fq|31'
> 
> thx

Tested it with gcc 9. Doesn't crash and the output is bitexact to what i
get with gcc 8.