[FFmpeg-user] ffmpeg is crashing with using x264 codec
James Darnley
james.darnley at gmail.com
Thu May 22 13:05:41 CEST 2014
On 2014-05-22 10:37, Tom Evans wrote:
> On Wed, May 21, 2014 at 6:11 PM, James Darnley <james.darnley at gmail.com> wrote:
>> On 2014-05-21 18:04, Tom Evans wrote:
>>> x264 has been built requiring SSE3 and SSE4 support, your CPU does not
>>> support SSE3 or SSE4, so x264 dies.
>>>
>>> Rebuild x264 without SSE3/4 support and rebuild ffmpeg.
>>
>> No. No. No!
>>
>> x264 does not *require* anything newer than SSE when built with asm.
>
> Yes! Yes! Yes!
>
> You are arguing that white is black.
>
> 1) He has a compiled version of x264
> 2) His compiled version of x264 contains an assertion
> 3) The assertion asserts that his CPU supports SSE3 or SSE4.
> 4) His CPU does not support SSE3 or SSE4
> 5) He needs to recompile x264 such that it no longer asserts that his
> CPU supports SSE3 or SSE4.
Have you even read the code around that assert? It is in a block under
an if which should be detecting that he has a Pentium M or Core1
processor, none of which should feature SSSE3 or SSE4 instructions. I
asked Loren why he added the assert and he said that it was to guard
against the event that these flags are somehow set. x264 disables SSE2
and SSE3 on these CPUs but doesn't mask against newer instruction sets
which, if they really exist on the same processor design, would be just
as slow as the SSE2 and SSE3 that get disabled.
Something has caused the SSSE3 and/or SSE4 flags to be set. I am
interested in finding out why rather than telling him to cripple x264.
I will say it again. x264 does not require that a user have a CPU with
more than SSE instructions ("SSE1" that is, that came before SSE2) when
it was built with asm.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 683 bytes
Desc: OpenPGP digital signature
URL: <https://ffmpeg.org/pipermail/ffmpeg-user/attachments/20140522/9c6d0d6e/attachment.asc>
More information about the ffmpeg-user
mailing list