performance - Why the data transfer is slow from GPU to CPU? -



performance - Why the data transfer is slow from GPU to CPU? -

today have figured out made me wondering. have samsung exynos 4412 arm9 cpu has gpu400(quadcore). tried texture gpu cpu known methods , slow. same scenario , slow speed happens in modern cpus , gpus in pc platform. wondering how happens , samsung exynos soc , both of them has same memory , should not care bus. why happens ?

the info gpu cpu transferred many methods have tried glreadpixels, gltexsubimage2d, glteximage2d, fbo. frame rate drops 40fps 7fps or 7fps while using of methods, on texture 1024*1024 24bits.

possible answers taken opengl forums:

latency: takes time read command reach hardware. opengl command buffering: reading info requires opengl driver finish outstanding commands. hardware buffering: hardware must empty gpu core pipelines before doing readback.

possible solution: - re-create info internally on gpu location , read number of frames after computing it. should allow writing location have completed before effort read it.

performance opengl-es embedded gpu

Comments

Popular posts from this blog

web services - java.lang.NoClassDefFoundError: Could not initialize class net.sf.cglib.proxy.Enhancer -

Accessing MATLAB's unicode strings from C -

javascript - mongodb won't find my schema method in nested container -