Ping.
This has been broken for ages, and at least tesseract is carrying
hacks to work around it (extra color buffer containing depth values,
in their g-buffer). Getting rid of that wins us about 15%.
-- Chris
On Sat, Feb 8, 2014 at 3:54 PM, Chris Forbes wrote:
> Previously, we would program the
Previously, we would program the sample count based on the blorp
operation's dst sample count -- which is zero for a HiZ op; we'd also
ignore the difference between physical and logical surface size for
the depth surface.
That kindof worked -- the two errors almost cancel out -- but produced
stran
