[FFmpeg-devel] Does i686 have MMX?

[Måns Rullgård]

M?ns Rullg?rd <mans at mansr.com> writes:

> "Ronald S. Bultje" <rsbultje at gmail.com> writes:
>
>> Hi,
>>
>> On Thu, Aug 26, 2010 at 9:46 AM, mru <subversion at mplayerhq.hu> wrote:
>>> Author: mru
>>> Date: Thu Aug 26 15:46:45 2010
>>> New Revision: 24946
>>>
>>> Log:
>>> Disable MMX for i686 and pentiumpro
>>>
>>> Modified:
>>> ? trunk/configure
>>>
>>> Modified: trunk/configure
>>> ==============================================================================
>>> --- trunk/configure ? ? Thu Aug 26 14:52:10 2010 ? ? ? ?(r24945)
>>> +++ trunk/configure ? ? Thu Aug 26 15:46:45 2010 ? ? ? ?(r24946)
>>> @@ -2105,8 +2105,14 @@ elif enabled x86; then
>>> ? ? ? ? ? ? cpuflags="-march=$cpu"
>>> ? ? ? ? ? ? disable cmov
>>> ? ? ? ? ;;
>>> + ? ? ? ?i686|pentiumpro)
>>> + ? ? ? ? ? ?cpuflags="-march=$cpu"
>>> + ? ? ? ? ? ?enable cmov
>>> + ? ? ? ? ? ?enable fast_cmov
>>> + ? ? ? ? ? ?disable mmx
>>> + ? ? ? ?;;
>>
>> Most distros will be able to continue using their compile flags and
>> silently create a crippled build. Why not error out unless mmx is
>> disabled using --disable-mmx? It has the same effect and requires a
>> little more explicit confirmation that they really want to cripple
>> their build...
>
> To me saying --cpu=i686 includes requesting no mmx.  Why should users
> be required to disable it twice?  If they want mmx, they should
> specify a CPU that has it.

Some people are screaming bloody murder over this change, citing
distro builders as the victims.  Curiously enough, none of the
pitchfork mob is involved in a distro, and none of the distro people
have objected.

Reinhart, Dominik, can you please speak up?

-- 
M?ns Rullg?rd
mans at mansr.com