[FFmpeg-devel] [PATCH 2/4] x86: xvid_idct: port MMX IDCT to yasm

Michael Niedermayer michaelni at gmx.at
Fri Mar 13 14:34:34 CET 2015


On Wed, Mar 11, 2015 at 01:06:23PM -0300, James Almer wrote:
> On 11/03/15 11:29 AM, Christophe Gisquet wrote:
> > Hi,
> > 
> > 2015-03-11 15:20 GMT+01:00 Michael Niedermayer <michaelni at gmx.at>:
> >> personally iam in favor of more things being tested, but iam fine
> >> with either
> > 
> > I'll see what the opinions are in ~30H then. It looks to be around
> > 4.5KB more data, which isn't huge, and doesn't cause a noticeable
> > increase in compile time.
> 
> As discussed before, mmx, mmxext and sse functions that have also an sse2 version, and 
> 3dnow/3dnowext functions that have also an sse version, have no reason to exist in an 
> x86_64 build.

> They will not be used in any real world scenario.

well it could be usefull in debuging in same cases
especially if all mmx code is disabled on x86_64, for any single
function its not too likely

consider this hypothetical scenario, theres a bug
testing reveals it is reproduceable
testing with cpuflags 0 shows its likely in the ASM
testing with cpuflags mmxext shows it works
testing with cpuflags sse2 shows it fails limiting this down to sse2
specific code

vs.
testing reveals it is reproduceable
testing with cpuflags 0 shows its likely in the ASM
rebuilding for x86_32
testing with cpuflags mmxext shows it works
testing with cpuflags sse2 shows it works
testing with cpuflags all shows it works
we realize the bug is not reproduceable on x86_32
rebuilding for x86_64
the bug now can be in any asm code


[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Concerning the gods, I have no means of knowing whether they exist or not
or of what sort they may be, because of the obscurity of the subject, and
the brevity of human life -- Protagoras
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 181 bytes
Desc: Digital signature
URL: <https://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20150313/8313d535/attachment.asc>


More information about the ffmpeg-devel mailing list