I am trying to implement one-pass stereo rendering in OpenGL for a VR application that I am building. I am implementing this through instancing, so I make one render call
count=2 and do clipping in the vertex shader but I am not seeing that much gain in performance.
My scene has just 2 massive meshes (rendered two times (2 instances), one for each eye totaling around 3.5m triangles for each eye or ~7m per frame). Obviously I can reduce the triangle count by doing some decimation but my question here is,
In theory, should I expect a significant gain in performance by using instancing in this scenario even though the number of draw calls and state changes in the entirety of my application is around 15 per frame For each pass and 30 per frame?
Or to Rephrase the question, Does OpenGL instancing improve performance mainly by reducing state changes or by making less calls to glDraw*?