Recherche avancée

Sur d’autres sites (12976)

• ARM compiler shoot-out, round 2

       
  20 août 2009, par Mans — ARM, Compilers

  In my recent test of ARM compilers, I had to leave out Texas Instrument’s compiler since it failed to build FFmpeg. Since then, the TI compiler team has been busy fixing bugs, and a snapshot I was given to test was able to build enough of a somewhat patched FFmpeg that I can now present round two in this shoot-out.

  The contenders this time were the fastest GCC variant from round one, ARM RVCT, and newcomer TI TMS470. With the same rules as last time, the exact versions and optimisation options were like this :

  • CodeSourcery GCC 2009q1 (based on 4.3.3), -mfpu=neon -mfloat-abi=softfp -mcpu=cortex-a8 -std=c99 -fomit-frame-pointer -O3 -fno-math-errno -fno-signed-zeros -fno-tree-vectorize
  • ARM RVCT 4.0 Build 591, -mfpu=neon -mfloat-abi=softfp -mcpu=cortex-a8 -std=c99 -fomit-frame-pointer -O3 -fno-math-errno -fno-signed-zeros
  • TI TMS470 4.7.0-a9229, --float_support=vfpv3 -mv=7a8 -O3 -mf=5

  
  To keep things fair, I left the vectoriser off also with the TI compiler. The table below lists the decoding times for the sample files, this time normalised against the participating GCC compiler. Remember, smaller numbers are better. Also keep in mind that this test was done with a development snapshot of TMS470, not an approved release.

  Sample name	Codec	Code type	GCC	RVCT	TI
  cathedral	H.264 CABAC	integer	1.00	0.95	1.02
  NeroAVC	H.264 CABAC	integer	1.00	0.96	1.05
  indiana_jones_4	H.264 CAVLC	integer	1.00	0.92	1.02
  NeroRecodeSample	MPEG-4 ASP	integer	1.00	1.01	1.08
  Silent_Light	MP3	64-bit integer	1.00	0.48	0.72
  When_I_Grow_Up	FLAC	integer	1.00	0.87	0.93
  Lumme-Badloop	Vorbis	float	1.00	0.94	1.05
  Canyon	AC-3	float	1.00	0.88	1.01
  lotr	DTS	float	1.00	1.00	1.08

  Overall, the TI TMS470 compiler comes off slightly worse than GCC. In two cases, however, it was significantly better than GCC, but not as good as RVCT. Incidentally, those were also the ones where RVCT scored the biggest win over GCC.

  My conclusions from this test are twofold :

  • ARM’s own compiler is very hard to beat. They do seem to know how their chips work.
  • GCC is incredibly bad at 64-bit arithmetic on 32-bit machines.

  The logical next step is to test these compilers with vectorisation enabled. FFmpeg should offer plenty of opportunities for this feature to shine. Unfortunately, that test will have to wait until the RVCT vectoriser is fixed. The current release does not compile FFmpeg with vectorisation enabled.

• mov : Export geotag metadata fields

  3 juin 2014, par Martin Storsjö

  mov : Export geotag metadata fields
  The ’ ?xyz’ form is used by android devices (and according to apple
  
  mailing list archives, also by older iOS devices). The ’loci’ field
  
  (defined in 3GPP 26.244) is used by recent iOS devices.
  Even though the loci field can contain an altitude, it was plain
  
  0 in my sample. Just export longitude and latitude, in a string
  
  format matching the one used by the ’ ?xyz’ metadata field.
  Signed-off-by : Martin Storsjö <martin@martin.st>
  

  • [DH] libavformat/mov.c

• Brute Force Dimensional Analysis

  15 juillet 2010, par Multimedia Mike — Game Hacking, Python

  I was poking at the data files of a really bad (is there any other kind ?) interactive movie video game known simply by one letter : D. The Sega Saturn version of the game is comprised primarily of Sega FILM/CPK files, about which I wrote the book. The second most prolific file type bears the extension ’.dg2’. Cursory examination of sample files revealed an apparently headerless format. Many of the video files are 288x144 in resolution. Multiplying that width by that height and then doubling it (as in, 2 bytes/pixel) yields 82944, which happens to be the size of a number of these DG2 files. Now, if only I had a tool that could take a suspected raw RGB file and convert it to a more standard image format.

  Here’s the FFmpeg conversion recipe I used :

   ffmpeg -f rawvideo -pix_fmt rgb555 -s 288x144 -i raw_file -y output.png
  

  So that covers the files that are suspected to be 288x144 in dimension. But what about other file sizes ? My brute force approach was to try all possible dimensions that would yield a particular file size. The Python code for performing this operation is listed at the end of this post.

  It’s interesting to view the progression as the script compresses to different sizes :

  
  
  

  That ’D’ is supposed to be red. So right away, we see that rgb555(le) is not the correct input format. Annoyingly, FFmpeg cannot handle rgb555be as a raw input format. But this little project worked well enough as a proof of concept.

  If you want to toy around with these files (and I know you do), I have uploaded a selection at : http://multimedia.cx/dg2/.

  Here is my quick Python script for converting one of these files to every acceptable resolution.

  work-out-resolution.py :
  

  PLAIN TEXT

  PYTHON :

  1. # !/usr/bin/python
  2.  
  3. import commands
  4. import math
  5. import os
  6. import sys
  7.  
  8. FFMPEG = "/path/to/ffmpeg"
  9.  
  10. def convert_file(width, height, filename) :
  11.  outfile = "%s-%dx%d.png" % (filename, width, height)
  12.  command = "%s -f rawvideo -pix_fmt rgb555 -s %dx%d -i %s -y %s" % (FFMPEG, width, height, filename, outfile)
  13.  commands.getstatusoutput(command)
  14.  
  15. if len(sys.argv) <2 :
  16.  print "USAGE : work-out-resolution.py <file>"
  17.  sys.exit(1)
  18.  
  19. filename = sys.argv[1]
  20. if not os.path.exists(filename) :
  21.  print filename + " does not exist"
  22.  sys.exit(1)
  23.  
  24. filesize = os.path.getsize(filename) / 2
  25.  
  26. limit = int(math.sqrt(filesize)) + 1
  27. for i in xrange(1, limit) :
  28.  if filesize % i == 0 and filesize & 1 == 0 :
  29.   convert_file(i, filesize / i, filename)
  30.   convert_file(filesize / i, i, filename)