Recherche avancée

Sur d’autres sites (7870)

• avcodec/aac_tablegen : speed up table initialization

  26 novembre 2015, par Ganesh Ajjanagadde

  avcodec/aac_tablegen : speed up table initialization
  This speeds up aac_tablegen to a ludicruous degree ( 97%), i.e to the point
  
  where it can be argued that runtime initialization can always be done instead of
  
  hard-coded tables. The only cost is essentially a trivial increase in
  
  the stack size.
  Even if one does not care about this, the patch also improves accuracy
  
  as detailed below.
  Performance :
  
  Benchmark obtained by looping 10^4 times over ff_aac_tableinit.
  Sample benchmark (x86-64, Haswell, GNU/Linux) :
  
  old :
  
  1295292 decicycles in ff_aac_tableinit,     512 runs,      0 skips
  
  1275981 decicycles in ff_aac_tableinit,    1024 runs,      0 skips
  
  1272932 decicycles in ff_aac_tableinit,    2048 runs,      0 skips
  
  1262164 decicycles in ff_aac_tableinit,    4096 runs,      0 skips
  
  1256720 decicycles in ff_aac_tableinit,    8192 runs,      0 skips
  new :
  
  21112 decicycles in ff_aac_tableinit,     511 runs,      1 skips
  
  21269 decicycles in ff_aac_tableinit,    1023 runs,      1 skips
  
  21352 decicycles in ff_aac_tableinit,    2043 runs,      5 skips
  
  21386 decicycles in ff_aac_tableinit,    4080 runs,     16 skips
  
  21299 decicycles in ff_aac_tableinit,    8173 runs,     19 skips
  Accuracy :
  
  The previous code was resulting in needless loss of
  
  accuracy due to the pow being called in succession. As an illustration
  
  of this :
  
  ff_aac_pow34sf_tab[3]
  
  old : 0.000000000007598092294225
  
  new : 0.000000000007598091426864
  
  real : 0.000000000007598091778545
  truncated to float
  
  old : 0.000000000007598092294225
  
  new : 0.000000000007598091426864
  
  real : 0.000000000007598091426864
  showing that the old value was not correctly rounded. This affects a
  
  large number of elements of the array.
  Patch tested with FATE.
  Reviewed-by : Rostislav Pehlivanov <atomnuker@gmail.com>
  
  Signed-off-by : Ganesh Ajjanagadde <gajjanagadde@gmail.com>
  

  • [DH] libavcodec/aac_tablegen.h

• Simulating MPEG1/2 transmission over a noisy channel [on hold]

  17 novembre 2015, par StepTNT

  The question may sound out of scope here but this is my last resource.

  I need to write a software that does :

  1. Get an uncompressed video from disk
  2. Compress it into MPEG-1 or MPEG-2 being able to change quantization matrix, GOP size and/or macroblock size for DCT/motion search
  3. Apply a repetition code to add redundancy
  4. Simulate transmission over a noisy channel with given error rate
  5. Reconstruct the original stream from the repetition code
  6. Decode the video and compare it with the original one by gathering stats like frame-by-frame difference, file size and stuff like that

  This should by done by a nice GUI to show the input and output videos, alongside their frame difference.

  Given what I need to do, I can write some requirements :

  • An encoder which allows me to change some of the parameters (needed for point 2)
  • A language that allows me to work at bit level (needed for points 3 and 5)
  • A language that allows me to build a nice GUI using a designer (GUI is not the core of the project so I can’t afford wasting time by writing one)

  So far my searches have led to mixed results that are not giving me enough resources to start.

  My first find was this MATLAB project which implements MPEG and has some parameters that can be tweaked (like quantization matrix and GOP pattern for example).
  The problem here is that I don’t know MATLAB at all, so I have no idea on how to link everything and build a GUI.

  So the next step was to move to JAVA, and I found a LOT of FFMPEG wrappers, but none seems to allow me to set the parameters that I need. My last try was with Xuggler but the Wiki is down and the documentation does not talk about what I need. Plus, JAVA doesn’t work at bit level so I’d have issues applying the repetition code.

  Failing with JAVA led me to C# and DirectShowNet, but the documentation is quite lacking and I don’t know how to start because I didn’t find anything related to setting the parameters that I need using Filters.

  The question now is : is there any language/framework/platform that allows me to do what I need without having to deal with pure C/C++ ?
  I’d expect a lot of stuff on this matter since we’re talking about well known codecs, still I’m having a hard time finding what I need.

• How do I create a 360 [stereo] panorama video of a virtual scene using OpenGL or DirectX ?

  14 novembre 2015, par Prashant Raina

  Like many of you, I have seen the magical 360 degree videos doing the rounds on Facebook right now. I would really like to know how to programmatically create such a video using OpenGL or DirectX, but I found surprisingly little information about it.
  There are several issues I am not clear on :

  1. How do I set up my vertex and fragment shaders to render a frame of 360 video ?
  2. Do I have to generate one image per frame, or a cubemap, or something else ?
  3. Once I have the frames, how do I combine them into a single video file for upload ? My Googling suggests that I should use ffmpeg, but I couldn’t figure out which parameters to use, or what sort of input it expects for 360 video.
  4. What additional steps are needed to make a stereo 360 video ?

  I am not asking how to render an existing video onto the screen, or how to use physical cameras for 360 video.

  Thanks in advance !