[Intel-gfx] Operation ordering?

[Peter Clifton]

I'm unfamiliar with how this works, so please bear with me. I'm
attempting to track down an intermittent rendering bug in my application
which appears to be related to stencil information not being correct at
the time primitives are rendered.

The rendering builds up by rendering into the stencil buffer, changing
the stencil tests and doing more rendering on top.

Running with debug options set to always flush after each primitive
fixes the issue, as does a judicial insertion of glFlush into my
rendering code after changing the stencil test state, but before drawing
more geometry.

Is it possible that the rendering commands and the state-change are
getting their orders muddled up, or that the rendering before the state
change (affecting the stencil) hasn't hit the stencil buffer before
further rendering then uses the (not yet up to date) stencil buffer?
Some kind of stall missing / cache coherency problem?

Anyway.. you guys know more how this works than I do, so any pointers
would be greatly appreciated. I'll let you know if I narrow it down any
further.

-- 
Peter Clifton

Electrical Engineering Division,
Engineering Department,
University of Cambridge,
9, JJ Thomson Avenue,
Cambridge
CB3 0FA

Tel: +44 (0)7729 980173 - (No signal in the lab!)
Tel: +44 (0)1223 748328 - (Shared lab phone, ask for me)