Hi,

On Fri, May 11, 2012 at 8:35 AM, Ronald S. Bultje <[email protected]> wrote:
> Hi,
>
> On Fri, May 11, 2012 at 8:31 AM, Christophe Gisquet
> <[email protected]> wrote:
>> Hi,
>>
>> 2012/5/11 Diego Biurrun <[email protected]>:
>> [...]
>>
>> Not limited to that patch: could the reporter state instead on what
>> instruction the sigill occurred? It may be possible to do something
>> else.
>>
>> Here, a cause could be that the movhps insn is sse, so the patch looks
>> either wrong (it should be marked as sse) or incomplete (another insn
>> is indeed mmx2, but this does not account for the sse insn).
>
> He means mmx2. I can look up what instruction it occurred on.

The problem is the pshufw emitted by SPLATW (see
libavutil/x86/x86util.asm). It should use punpcklwd+punpckldq on mmx.

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

Reply via email to