#Oggeyman From ogg/theora to OpenGL texture, with ARM speedup.

Theora video encoding has a reputation of being unfriendly to ARM architectures, used in all major mobile platforms. Several years ago Robin Watts provided a well needed speedup using pieces of native assembly code (Theorarm). Oggeyman takes that implementation and ties it together with mapping onto OpenGL Texture, in a piece of C++ code. Technically, Oggey provides only a BGRA formatted data buffer that can be used directly as an input for the OpenGL's glTexImage2D function or its first cousin [glTexSubImage2D] (https://www.opengl.org/sdk/docs/man3/xhtml/glTexSubImage2D.xml).

Java interface is also provided here, but without test code (see To Do below).

Test file credit: Polar_orbit.ogg from Wikimedia Commons.

GNU toolchain (gcc, g++, and make in particular; though I suspect it should not be too hard to move to other toolchains) for compilation of a generic executable.

For a cross-compiled version, you will need a cross-compiler, and an emulator. I.e. in a Ubuntu linux setting you would install the cross-compilation toolchain as

sudo apt-get install gcc-4.7-arm-linux-gnueabi libc6-dev-armel-cross

and for the (QEMU) emulator

sudo apt-get install qemu-user-static

Note this last step provides just the user-level emulation (execution of cross-compiled code) rather than the full emulated enviroment that QEMU can in principle provide.

In the armtest directory two makefiles are provided: plain_makefile, for compiling a regular C version of the code. arm_makefile provides and example of cross-compilation (on linux for arm platform). Run as

make -f plain makefile

or

make -f arm_makefile

For the generic case just type

./oggey

and for the cross-compiled one:

qemu-arm -L /usr/arm-linux-gnueabi oggeyARM

Make sure that the link path corresponds to the one where arm-linux-gnueabi resides in your setup. (Failing to do so results in /lib/ld-linux.so.3: No such file or directory error.)

In both cases, the output should be the time of each frame, printed on stdout, in the proper time sequence.

Note that Oggeyman.cpp can be used for threeway comparison/timing of the straightforward yuv2rgb implementation, the implementation using table lookup, and the ARM optimized implementation. See Oggeyman::yuv2bgra() in Oggeyman.cpp.

Oggeyman is not a full decoder - in particular it pulls out Theora (video) stream from the provided ogg file, and ignores the Vorbis (audio) or any other stream that might be present.

Oggeyman is not a player, either. It takes ogg file as an iput an returns a data buffer in BGRA format. The rest of the implementation is up to user.

Oggeyman was originally written to be used as a part of a larger platform, (LibGDX in particular), therefore comming up with some self-standing cross-compiled Java example is yet to bbe done. In the meantime, let it just be noted that a typical usage case would be to compile all C/C++ pieces into a library, to be called from Oggeyman.java. Oggeyman can then be used as a class in a bigger Java program.