I am writing a small 2D SHMUP for Android at the moment. Pure 2D rendering with Bitmaps turned out to be a bit too slow, especially with a lot of alpha blending going on. So I decided to switch my rendering backend to OpenGL. Last time I used OpenGL there was still immediate mode. I now want to build a modern architecture with OpenGL ES 2.0.
Here are the details: Most of the time I am going to display a lot of Sprites, that are essentially texure mapped quads consisting of 2 triangles each. I am probably going to want to render a lot of them :) I am going to use Vertex Buffers because of the faulty VBO implementation in Froyo.
Here's the question: How would I manage the position data of the independent sprites? Do I need to modify my vertex buffers each time a sprite changes its position? Or is it faster to use a single draw call and translation matrix for each sprite?