ABSTRACT
No data buffers are required, making the shader simpler to setup, call, and reuse. There are no branches, and the required constants can be encoded directly in the shader’s load operations, thus requiring no trips through the data pipeline. This code can be instanced and used everywhere from full-screen quad generation to particle quad expansion. If your mesh can be encoded as a single strip with no restarts and no vertex reuse, like the simple case here, then there is no benefit to indexing the data in an attempt to take advantage of the post-transform cache. The GPU will only call the vertex shader once for each vertex.
