Mapping OpenGL Render Buffer for in kernel use It is possible to map a OpenGL Render Buffer for use

Accelerated Computing CUDA CUDA Programming and Performance
1

I’m trying to make some post-process work with a GPU kernel. For this, I render some graphics into a framebuffer and then a Color Attachment is copied to a PBO in order to use inside a kernel. Something like

glGenRenderbuffers( 1, &this->depthBuffer );

glBindRenderbuffer( GL_RENDERBUFFER, this->depthBuffer );

glRenderbufferStorage( GL_RENDERBUFFER, GL_DEPTH_COMPONENT, scene->getCamera().getWidth(), scene->getCamera().getHeight() );

glBindRenderbuffer( GL_RENDERBUFFER, 0 );

glGenRenderbuffers( 1, &this->colorBuffer );

glBindRenderbuffer( GL_RENDERBUFFER, this->colorBuffer );

glRenderbufferStorage( GL_RENDERBUFFER, GL_RGBA32F, scene->getCamera().getWidth(), scene->getCamera().getHeight() );

glBindRenderbuffer( GL_RENDERBUFFER, 0 );

this->fbo->attachRenderBuffer( this->colorBuffer, GL_COLOR_ATTACHMENT0 );

this->fbo->attachRenderBuffer( this->depthBuffer, GL_DEPTH_ATTACHMENT );

My question is if it is possible to map the buffer attached to GL_COLOR_ATTACHMENT0 like this

cudaGraphicsGLRegisterBuffer( &this->cudaGraphicRes, this->colorBuffer, cudaGraphicsMapFlagsWriteDiscard );

or even the color buffer need to be registered using cudaGraphicsGLRegisterImage?

I wanna do this to avoid texture-to-buffer copies and vice-versa.