Hi,

2011/5/24 Måns Rullgård <[email protected]>:
> "Ronald S. Bultje" <[email protected]> writes:
>> On Tue, May 24, 2011 at 12:15 PM, Luca Barbato <[email protected]> wrote:
>>> On 05/24/2011 05:33 PM, Ronald S. Bultje wrote:
>>>> ---
>>>>  libswscale/x86/swscale_template.c |    3 ++-
>>>>  1 files changed, 2 insertions(+), 1 deletions(-)
>>>
>>> Ok.
>>
>> Slightly updated patch attached.
>>
>> Ronald
>>
>> From c9d8c222aaa159b3c4307048a9fa3ed8140c96ca Mon Sep 17 00:00:00 2001
>> From: Ronald S. Bultje <[email protected]>
>> Date: Tue, 24 May 2011 13:04:46 -0400
>> Subject: [PATCH 2/5] swscale: use emms_c().
>>
>> ---
>>  libswscale/utils.c                |    3 +--
>>  libswscale/x86/swscale_template.c |    2 +-
>>  2 files changed, 2 insertions(+), 3 deletions(-)
>>
>> diff --git a/libswscale/utils.c b/libswscale/utils.c
>> index b49ec89..09b56a2 100644
>> --- a/libswscale/utils.c
>> +++ b/libswscale/utils.c
>> @@ -193,8 +193,7 @@ static int initFilter(int16_t **outFilter, int16_t 
>> **filterPos, int *outFilterSi
>>      const int64_t fone= 1LL<<54;
>>      int ret= -1;
>>
>> -    if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX)
>> -        __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not 
>> be required but it IS (even for non-MMX versions)
>> +    emms_c(); //FIXME this should not be required but it IS (even for 
>> non-MMX versions)
>
> What's with that FIXME?

Good question. I'm basically not exactly sure what the comment is
doing there. Which is why I'm keeping it around.

I can remove if preferred.

> There's also one more emms instance in that file, at line 761.

Added.

Ronald

Attachment: 0002-swscale-use-emms_c.patch
Description: Binary data

_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to