[FFmpeg-devel] [PATCH] x86inc: support stack mem allocation and re-alignment in PROLOGUE.

[Ronald S. Bultje]

Hi,

On Thu, Dec 13, 2012 at 12:09 PM, Ronald S. Bultje <rsbultje at gmail.com> wrote:
> Hi,
>
> On Thu, Dec 13, 2012 at 10:20 AM, Alexander Strasser <eclipse7 at gmx.net> wrote:
>> Hi Ronald, Hi Anton!
>>
>> Ronald S. Bultje wrote:
>>> On Wed, Dec 12, 2012 at 7:03 PM, Michael Niedermayer <michaelni at gmx.at> wrote:
>>> > On Sat, Dec 08, 2012 at 04:12:38PM -0800, Ronald S. Bultje wrote:
>>> >> From: "Ronald S. Bultje" <rsbultje at gmail.com>
>>> >>
>>> >> Use this in VP8/H264-8bit loopfilter functions so they can be used if
>>> >> there is no aligned stack (e.g. MSVC 32bit or ICC 10.x).
>>> >> ---
>>> >>  libavcodec/x86/h264_deblock.asm |  27 ++----
>>> >>  libavcodec/x86/h264dsp_init.c   |   4 +-
>>> >>  libavcodec/x86/vp8dsp.asm       |  68 ++++++++-------
>>> >>  libavcodec/x86/vp8dsp_init.c    |   8 --
>>> >>  libavutil/x86/x86inc.asm        | 185 ++++++++++++++++++++++++++++++++--------
>>> >>  5 files changed, 191 insertions(+), 101 deletions(-)
>>> >
>>> > This patch breaks mingw (fate-h264-conformance-ba1_ft_c amongth many
>>> > others)
>>> > http://fate.ffmpeg.org/report.cgi?time=20121212232000&slot=x86_64-debian-mingw32-gcc-4.6
>>> >
>>> > It also seems to break MSVC
>>> > http://fate.ffmpeg.org/history.cgi?slot=x86_64-msvc10-windows-native
>>> >
>>> > please fix these
>>>
>>> http://lists.libav.org/pipermail/libav-devel/2012-December/040286.html
>>>
>>> Patch isn't actually mine, it's from Anton Mitrofanov
>>> <BugMaster at narod.ru> on #x264.
>>
>>   I guess it also broke compilation on Cygwin. I didn't investigate closely yet.
>>
>>   http://fate.ffmpeg.org/log.cgi?time=20121213170718&log=compile&slot=x86_64-win7-cygwin-gcc-4.5.3
>
> That's not Anton's fault, he fixed a bug in my original patch. It
> seems nasm doesn't like the latest changes to cglobal. I don't know
> why. Maybe it's related to DEFINE_ARGS_INTERNAL inlining not working
> correctly with varargs in nasm? Can you try manually unrolling
> DEFINE_ARGS_INTERNAL in each of its 3 callers and see if that fixes
> it?

http://lists.libav.org/pipermail/libav-devel/2012-December/040350.html

Fixes it.

Ronald