[FFmpeg-devel] [PATCH] Electronic Arts MAD Decoder

[Mike Melanson]

Peter Ross wrote:
> On Sat, Jun 13, 2009 at 09:03:26AM -0700, Mike Melanson wrote:
>> Michael Niedermayer wrote:
>>> looks ok
>> Problem: It doesn't work.
> 
> SVN-r19182 works flawlessly here on linux/ppc (gcc-4.3.0) and linux/x86
> (gcc-4.4.0). The patch has been tested on many MAD samples, including
> those on samples.mplayerhq.hu. My reference '-f crc' values are below.

Interesting. My results line up with yours, at least for those 2 
configurations. Check this out:

> #FILENAME, MD5SUM, FFMPEG-CRC
> NFS4T0_00.mad, 395ed76dc3c160f4ab29990434a2c91b, CRC=0x2bc41169

$ ./rpc-dist-test.py "FFMPEG -i SAMPLES_PATH/ea-mad/NFS4T0_00.mad -f crc -"

There were 6 unique stdout blobs collected

The following configurations agreed on the same stdout:
   Linux / x86_64 / icc 11.0
   Linux / x86_64 / icc 10.1
stdout:
CRC=0xfbe011a2


The following configurations agreed on the same stdout:
   Linux / x86_32 / llvm-svn
stdout:
CRC=0x5a9e0feb


The following configurations agreed on the same stdout:
   Linux / PPC / gcc 4.1.2
stdout:
CRC=0x919b709d


The following configurations agreed on the same stdout:
   Linux / x86_32 / gcc 4.3.3
   Linux / x86_32 / gcc 4.4.0
   Linux / x86_64 / gcc 4.3.3
   Linux / x86_64 / gcc 4.4.0
   Linux / x86_64 / gcc svn
   Linux / PPC / gcc 4.3.3
   Linux / PPC / gcc 4.4.0
   Linux / PPC / gcc svn
stdout:
CRC=0x2bc41169


The following configurations agreed on the same stdout:
   Linux / x86_32 / gcc 4.2.4
stdout:
CRC=0xda18113f


The following configurations agreed on the same stdout:
   Linux / x86_32 / icc 11.0
   Linux / x86_32 / icc 10.1
stdout:
CRC=0xd2c67e11


So, the largest agreement included the 2 configurations you tested.

There were also 14 more configurations that failed. I haven't 
investigated why yet.

-- 
     -Mike Melanson