glamor could accelerate more complicated planemasks
At the moment
glamor_set_planemask only allows "solid" planemasks, where we end up writing all the color bitplanes. We could also allow planemasks that write only whole color channels if we used
Alternatively, if you have EXT_shader_framebuffer_fetch and bitwise ops in your shaders, you could accelerate pretty much any planemask. This is somewhat less useful at the moment, since (of Mesa drivers) only i965 implements that extension, and you could just as well use a big-GL context there.