Hi there, I am currently developing a 3D game engine using a deferred renderer in OpenGL.
I have successfully implemented recursive portals (and mirrors) in my engine utilizing the stencil buffer to mask out regions of the screen. This solution is very favorable as I am able to have dozens of separate views drawn at once without needing to worry about requiring multiple G-buffers for each individual (sub)view. I also benefit with being able to perform post processing effects over all views, only needing to apply them over what is visible (one pass per-section with stencil masking for no risk of overdraw).
Now presently I am pondering ways of dealing with in-game camera displays (for an example think of the monitors from Half-Life 2). In the past I’ve handled these by rendering from the camera’s perspective onto separate render target, and then in the final shading pass applying it as a texture. However I was greatly disappointed with the performance and the inability to combine with post-processing effects (or at least the way I do presently with portals). Another concern being that I wish to have scenes containing several unique camera screens at once (such as a security CCTV room), without needing to worry about the associated vram usage of requiring several G-Buffers.
I wanted to ask more experienced members of this community if it would be possible to handle them in a similar fashion as I do with portals, but with the difference being for them to be transformed so they take on the appearance of a flat 2D surface. Would anybody with a more comprehensive understanding of matrix maths be able to tell me if this idea is feasible or not, and if so could come up with a possible solution?
I hope all this makes enough sense. Any possible insight would be greatly appreciated!
Thanks!