On 11/16/2014 05:51 PM, Thomas Helland wrote: > Also handle undefined behaviour for sqrt(x) where x < 0 > and rsq(x) where x <= 0. > > This gives us some reduction in instruction count on some > Dungeon Defenders shaders as they are doing: max(exp(x), 0) > --- > src/glsl/opt_minmax.cpp | 17 +++++++++++++++++ > 1 file changed, 17 insertions(+) > > diff --git a/src/glsl/opt_minmax.cpp b/src/glsl/opt_minmax.cpp > index a48d4d8..1aa4611 100644 > --- a/src/glsl/opt_minmax.cpp > +++ b/src/glsl/opt_minmax.cpp > @@ -293,6 +293,23 @@ get_range(ir_rvalue *rval) > high = larger_constant(r0.high, > abs(r0.low)->constant_expression_value()); > return minmax_range(low, high); > > + case ir_unop_sqrt: > + case ir_unop_rsq: > + case ir_unop_exp: > + case ir_unop_exp2: > + r0 = get_range(expr->operands[0]); > + if (expr->operation == ir_unop_sqrt && > + is_less_than_zero(r0.high)) > + high = new(mem_ctx) ir_constant(0.0f);
A blank line here... > + if (expr->operation == ir_unop_rsq && > + is_less_than_or_equal_zero(r0.high)) > + high = new(mem_ctx) ir_constant(0.0f); ...and here makes the code much easier to read. These two blocks calculate high, but it's never used. Is that intentional? > + /* TODO: If we know, i.e, the lower range of the operand > + * we can calculate the lower range > + */ > + low = new(mem_ctx) ir_constant(0.0f); > + return minmax_range(low, NULL); > + > case ir_unop_sin: > case ir_unop_sin_reduced: > case ir_unop_cos: > _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
