rendering pipeline latency, dev docs?

[Dan B. Phung]

I'm trying to profile the latency of the rendering pipeline for both 
direct rendering and non-direct rendering contexts from the application 
down to the interrupts in the driver, and I'm having trouble following 
the path.  If there is documentation, please point me to that and I'll 
gladly read it.  I'm using the open source intel driver for my 
intel-945GM. 

Here's an example of a trace I'm trying to complete for atlantis:

draw_atlantis()
glxSwapbuffers()
glFinish()
_mesa_Finish()
...
...
...
[direct rendering case]
drm_ioctl()
drm_ioctl()
drm_ioctl()
i915_driver_vblank_do_wait()
drm_ioctl()
drm_ioctl()
i915_irq_emit()
i915_emit_irq()
i915_emit_irq()
i915_irq_emit()
drm_ioctl()
i915_irq_wait()
i915_wait_irq()

I'd like to fill in the blanks (ellipses above), and I'd also like to do 
this for the case where I don't have direct rendering. 

tia,
dan