ability to use clCreateFromGLTexture with OpenGL textures with internal format GL_RGB* and GL_DEPTH_COMPONENT*, that is the hardware depth buffer format.

ability to read AND write the same image2d_t object in kernel.

Best regards,

oh, and the same goes for special formats like
GL_RGB9_E5 and

It also would be nice to have ability to pass array of images to kernel as parameter.

yep and also floating point math functions like mix and reflect.
according to the OGL specs mix is this function:
x * ( 1 - a ) + y * a
where x is the start of the interpolation y is the end and a is the weight. Is that so hard to implement? I mean the vendors have already done it in GLSL…

I stumbled upon another problem: when you do a syntax problem, the compiler doesn’t throw a proper
“syntax error on line x on column y” but something like this:
and the invalid kernel name error (probably due to the kernel not being compiled)
this was with the latest stable APP SDK and drivers.

recently I bumped into an issue where I badly needed the D24S8 format to be readable in OpenCL. Though using other formats solved the problem, still it’d be great to have that in OCL 1.3 :slight_smile: