Zero-copy optimisation

[PoloLastik]

Hello,

I am a newbie in gpgpu stuff. I cloned you code for a personnal project with a mali GPU and would like if it is possible to handle zero-copy data transfer between CPU and GPU.
The app works well however, there is huge performance bottleneck in data transfer when increasing amount of data processed.