pan/bi: Wire up shader-db
First commit is so I can do all this work from my Midgard laptop since I'm a creature of habit #shameless, next few commits fix some low-hanging crashes from shader-db stuff (outstanding issues with spilling TEXC and a hang during register allocation for t-rex), finally we wire in some statistics. We'll want a few more metrics once we have a better sense for what determines hardware performance. In particular, I imagine some metrics for:
- Code size
- Register file access (versus passthroughs)
- Latency estimations (based on scoreboarding), breakdowns for different operations maybe?
- Possibly divergence?
But for now this is better than nothing, and I'm out of time for the week.
WIP also because register counting is wrong with respect to staging registers.