Draft: gallium,mesa: new driver interface for faster display lists
New gallium interface for display lists. pipe_vertex_state is a screen-based CSO containing vertex buffers, vertex elements, and an index buffer. Draw merging in u_threaded_context will be done between draw_vertex_state calls.
This is just gallium, dd.h, and vbo_save.h changes for a rough idea.
The current prototype has ~2x better CPU-bound performance.
radeonsi implements both draw_vbo and draw_vertex_state with the same C++ template function to share code.