On Feb 3, 2015 10:39 PM, "Jason Ekstrand" <<a href="mailto:jason@jlekstrand.net">jason@jlekstrand.net</a>> wrote: > > > > On Tue, Feb 3, 2015 at 1:08 PM, Erik Faye-Lund <<a href="mailto:kusmabite@gmail.com">kusmabite@gmail.com</a>> wrote: >> >> On Tue, Feb 3, 2015 at 9:43 PM, Jason Ekstrand <<a href="mailto:jason@jlekstrand.net">jason@jlekstrand.net</a>> wrote: >> > --- >> > src/glsl/nir/nir_validate.c | 8 ++++++++ >> > 1 file changed, 8 insertions(+) >> > >> > diff --git a/src/glsl/nir/nir_validate.c b/src/glsl/nir/nir_validate.c >> > index 7c801b2..89dfdf8 100644 >> > --- a/src/glsl/nir/nir_validate.c >> > +++ b/src/glsl/nir/nir_validate.c >> > @@ -239,6 +239,14 @@ validate_alu_dest(nir_alu_dest *dest, validate_state *state) >> > * register/SSA value >> > */ >> > assert(is_packed || !(dest->write_mask & ~((1 << dest_size) - 1))); >> > + >> > + /* validate that saturate is only ever used on instructions with >> > + * destinations of type float >> > + */ >> > + nir_alu_instr *alu = nir_instr_as_alu(state->instr); >> > + assert(nir_op_infos[alu->op].output_type == nir_type_float || >> > + !dest->saturate); >> >> I think this can end up generating a warning on builds with asserts >> disabled due to "alu" being written but never read. Perhaps just do >> "nir_instr_as_alu(state->instr)->op" directly in the expression? It's >> a tad less readable, though :/ > > > It doesn't even get compiled if we have no asserts. There's a giant #ifdef DEBUG surrounding the entire file. Right, my bad.