On Wed, 28 Sep 2011 01:06:14 -0700, Kenneth Graunke <kenn...@whitecape.org>
wrote:
> On 09/27/2011 03:08 PM, Eric Anholt wrote:
> > They're both implemented the same in GLSL IR (since round() has
> > undefined behavior for N.5).
> >
> > Fixes glsl-1.30/compiler/built-in-functions/round*
> > ---
> > src/glsl/ir_constant_expression.cpp | 10 ++++++++++
> > 1 files changed, 10 insertions(+), 0 deletions(-)
> >
> > diff --git a/src/glsl/ir_constant_expression.cpp
> > b/src/glsl/ir_constant_expression.cpp
> > index b3fe6cf..59a7e67 100644
> > --- a/src/glsl/ir_constant_expression.cpp
> > +++ b/src/glsl/ir_constant_expression.cpp
> > @@ -196,6 +196,13 @@ ir_expression::constant_expression_value()
> > }
> > break;
> >
> > + case ir_unop_round_even:
> > + assert(op[0]->type->base_type == GLSL_TYPE_FLOAT);
> > + for (unsigned c = 0; c < op[0]->type->components(); c++) {
> > + data.f[c] = rint(op[0]->value.f[c]);
>
> It's not obvious to me that rint() rounds correctly.
>
> From the GLSL 1.30 spec:
> "Returns a value equal to the nearest integer to x. A fractional part
> of 0.5 will round toward the nearest even integer. (Both 3.5 and 4.5 for
> x will return 4.0.)"
>
> rint() does nearest integer, but probably not the mandated behavior for
> 0.5. At least, not unless we configure the rounding mode. Which I
> don't think we have...rint's manpage doesn't appear to mandate the behavior we want, but it does appear to provide the behavior we want according to spec/glsl-1.30/compiler/built-in-functions/roundEven-float.frag
pgp9bh7qhVQVU.pgp
Description: PGP signature
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
