On Apr 10, 2015 10:34 PM, "Matt Turner" <<a href="mailto:mattst88@gmail.com">mattst88@gmail.com</a>> wrote: > > On Fri, Apr 10, 2015 at 10:24 PM, Matt Turner <<a href="mailto:mattst88@gmail.com">mattst88@gmail.com</a>> wrote: > > On Fri, Apr 10, 2015 at 2:16 PM, Jason Ekstrand <<a href="mailto:jason@jlekstrand.net">jason@jlekstrand.net</a>> wrote: > >> Ping > > > > 1-2 are > > > > Reviewed-by: Matt Turner <<a href="mailto:mattst88@gmail.com">mattst88@gmail.com</a>> > > > > Thanks! > > Though I should mention that I do see some shaders that still do SEL > on UD types. Maybe we're not losing anything by doing that, but it is > still pretty annoying to figure out what the constant is in something > like > > (-f0) sel(8) g11<1>UD g10<8,8,1>UD 0x33d6bf95UD { align1 1Q }; Agreed. I did have a type-guessing patch that I kind of dropped on the floor. However no solution will be perfect. We could always do what we do in NIR and print both the integer and the float for integer constants (assuming the value is valid floating-point). --Jason