The introduction and use of lp_build_clamp_zero_one_nanzero looks good.
I can't comment on the changes to the existing paths though, as they are too
subtle for me. As long there are no regressions I'm good.
Jose
- Original Message -
From: Roland Scheidegger srol...@vmware.com
d3d10
From: Roland Scheidegger srol...@vmware.com
d3d10 requires us to convert NaNs to zero for any float-int conversion.
We don't really do that but mostly seems to work. In particular I suspect the
very common float-unorm8 path only really passes because it relies on sse2
pack intrinsics which just
