nengel@2: App: h264dec
nengel@2: 
nengel@2: This application decodes H.264 raw videos.
nengel@2: 
nengel@2: Build Sequential/Pthreads:
nengel@2: 
nengel@2: autoreconf -i -f
nengel@2: mkdir build
nengel@2: cd build
nengel@2: ../configure --enable-ssse3 --enable-sdl2
nengel@2: make
nengel@2: 
nengel@2: Build OmpSs:
nengel@2: 
nengel@2: autoreconf -i -f
nengel@2: mkdir build
nengel@2: cd build-ss
nengel@2: ../configure CC=sscc --enable-ssse3 --enable-sdl2
nengel@2: make
nengel@2: 
nengel@2: ssse3 enables assembler optimizations up to ssse3 (optional)
nengel@2: sdl enables a rudimentary viewing capability      (optional)
nengel@2: 
nengel@2: Usage Sequential/Pthreads:
nengel@2: ./h264dec -i $(INPUT_VIDEO) -s
nengel@2: ./h264dec -i $(INPUT_VIDEO) -t $(THREADS)
nengel@2: 
nengel@2: Usage OmpSs:
nengel@2: NX_PES=<numthreads> ./h264dec -i <inputfile> -e <num parallel entropy frames> -z <width> <height> --static-3d
nengel@2: 
nengel@2: -e specify the number of entropy decode pipeline buffers and should be ideally
nengel@2: the same as the number of threads.
nengel@2: 
nengel@2: -z allows to set the MB reconstruction grouped block size. A size between 6 by 6 to 10 by 10
nengel@2: was found to strike a good balance between overhead and parallelism, but is machine and input
nengel@2: dependent.
nengel@2: 
nengel@2: --static-3d performs overlapping wavefront decoding.
nengel@2: 
nengel@2: General usage:
nengel@2: -d 				displays output
nengel@2: -f 				fullscreen
nengel@2: -o $(OUT_FILE)  write raw YUV
nengel@2: -v  			show framerate
nengel@2: 
nengel@2: 
nengel@2: The INPUT_VIDEOs are in "inputs_encore", but should be able to decode any raw H.264 stream using
nengel@2: one slice per frame, non-interlaced, and CABAC, YUV420.
nengel@2: 
nengel@2: 
nengel@2: Integrated OmpSs player demo
nengel@2: ----------------------------
nengel@2: NOTE: for the player demo SDL2 must be installed.
nengel@2: 
nengel@2: 1. Go to the OmpSs build directory (/home/cchi/Projects/ffmpeg_smp/build-ss)
nengel@2: 
nengel@2: 2. Launch the H.264 decoder with the desired options:
nengel@2: 
nengel@2: NX_PES=<numthreads> ./h264dec <inputfile> -v (verbose) -e <num parallel entropy frames> -z <width> <height> -d (display) -f (fullscreen)
nengel@2: 
nengel@2: note that <num parallel entropy frames> should be equal or higher than <numthreads> for optimal performance
nengel@2: 
nengel@2: Examples:
nengel@2: 
nengel@2: NX_PES=7 ./h264dec -i ../../h264_movies/park_joy_2160px5.h264 -v -z 8 8 -df -e 9
nengel@2: NX_PES=7 ./h264dec -i ../../h264_movies/big_buck_bunny_1080p24.h264 -v -d  -z 6 6 -e 9
nengel@2: 
nengel@2: Interacting with the program
nengel@2: ----------------------------
nengel@2: <CTRL+F>    Fullscreen mode
nengel@2: <ESCAPE>    Window mode
nengel@2: <SPACE>     Pause/resume
nengel@2: <M>         Show/hide macroblock borders
nengel@2: <arrows>    When macroblock borders are shown resizes the macroblocks
nengel@2: <ALT+F4>    Close
nengel@2: 
nengel@2: Force close in case of lockup
nengel@2: -----------------------------
nengel@2: On a terminal: killall -9 h264dec