What steps will reproduce the problem?
1. Translate a shader with assignment statements with precision emulation on.
What is the expected output? What do you see instead?
Optimal translated shader would not wrap assignment into angle_frm calls, like
it currently does. There are also other cases where the same value might be
rounded twice that could be improved. Fixing these cases would make precision
emulation more practical, right now it can make compile times prohibilitively
long.
Original issue reported on code.google.com by oetu...@nvidia.com on 20 Jan 2015 at 3:40
Original issue reported on code.google.com by
oetu...@nvidia.com
on 20 Jan 2015 at 3:40