c++ - Sharing Rendering Surfaces (FrameBuffers, Pbuffers, etc.) between OpenGL ES 2 Applications/Processes -

i'm designing opengl es 2.0-enabled applications on arm-based, embedded linux system using c++ , want share rendering surface (fbo or renderbuffer generated opengles2 or pbuffer created egl) of app "a" app "b".

since i'm beginner opengl es 2 (and, more importantly, since buffers on gpu's memory on have no control nor access), first thought of mapping shared memory object (e.g. uint8_t array of length width*height*colorcomponentcount, using posix shm_open() etc.) between 2 processes , periodically calling glreadpixels() on app "a" filling uploading in app "b" as, instance, 2d texture. of course, shared synchronization mechanism, such named semaphore, used well, not point of question:

i aware downloading buffer gpu (i.e. glreadpixels() in app "a") , uploading (i.e. texture 2d in app "b") not optimal solution sharing rendering surfaces between opengl es applications, why know if there better solution ?

the glreadpixels() approach way in portable, platform independent way opengl es 2.0 , incredibly slow. pixmaps allocated egl (not pbuffers) if platform implemented them.

i recommend either:

implement apps share common instance of opengl es somehow , use fbos texture attachments share images. if need portability.

or, customize linux platform provide support either @ kernel level or customized egl implements pixmap support. these articles fbos , egl image extension:

http://montgomery1.com/opengl/