[FFmpeg-devel] [PATCH] EA TGQ video fixes

Reimar Döffinger Reimar.Doeffinger
Fri Apr 17 19:01:17 CEST 2009


On Fri, Apr 17, 2009 at 07:18:43AM -0700, Mike Melanson wrote:
> Reimar D?ffinger wrote:
> > I guess aliasing violations can be handled very differently.
> > Anyway I committed one patch, it makes the PPC64/gcc4.1.2 results match
> > with the test specification (well, I only compared the first and last
> > few values I admit).
> > Maybe you could just enable the test? I think we will be taking care of
> > it in the near future if there are still issues left...
> 
> I am still seeing 6 unique sets of data. I don't want to enable a 
> bad/inconsistent test:

Well, but that makes it really hard to find out which systems have
issues and which kind of issues.
If you don't want to enable them IMO you really need to add a way that
they will at least be run anyway and the results are somehow possible to
access, at least if you want it to be a help for bug fixing beyond only
regressions.

> The following configurations agreed on the same stdout:
>    Linux / x86_32 / gcc 2.95.3
> 
> The following configurations agreed on the same stdout:
>    Linux / x86_32 / icc
>    Linux / x86_32 / gcc 4.1.2
>    Linux / x86_32 / gcc 4.2.4
>    Linux / x86_32 / gcc 4.3.2
>    Linux / x86_32 / gcc svn
> 
> The following configurations agreed on the same stdout:
>    Linux / x86_32 / gcc 3.4.6
>    Linux / x86_32 / gcc 4.0.4
>    Linux / x86_64 / gcc 4.0.4
>    Mac OS X / x86_64 / gcc 4.0.1
> 
> The following configurations agreed on the same stdout:
>    Linux / x86_64 / gcc 4.1.2
>    Linux / x86_64 / gcc 4.2.4
>    Linux / x86_64 / gcc 4.3.3
>    Linux / x86_64 / gcc svn
> 
> The following configurations agreed on the same stdout:
>    Linux / PPC / gcc 4.1.2
> 
> The following configurations agreed on the same stdout:
>    Linux / PPC / gcc 4.0.4
>    Linux / PPC / gcc 4.2.4
>    Linux / PPC / gcc 4.3.3
>    Linux / PPC / gcc svn
>    Mac OS X / PPC / gcc 4.0.1

Which of these match the test spec? How do they differ (e.g. do only
some frames differ or all, how does the difference look between the
frames)? Does it look like the applied patch improved things? Much?
You results seem strange since since for me the results of
Linux / PPC64 / gcc 4.1.2
and
Linux / x86_64 / gcc 4.3.3
seem to match and also match the test spec, both when compiled with
shared or dynamic libs.
Maybe you can also test with the alignment spec?
Does anyone else get results that do not match
http://fate.multimedia.cx/index.php?test_spec=262




More information about the ffmpeg-devel mailing list