[Mesa-dev] [PATCH] glsl: Clamp vector indices when lowering to swizzles
Kenneth Graunke
kenneth at whitecape.org
Tue Nov 8 10:37:36 PST 2011
On 11/07/2011 11:12 AM, Ian Romanick wrote:
> From: Ian Romanick<ian.d.romanick at intel.com>
>
> This prevents other code from seeing a swizzle of the 16th component
> of a vector, for example.
>
> NOTE: This is a candidate for the 7.11 branch.
>
> Signed-off-by: Ian Romanick<ian.d.romanick at intel.com>
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=42517
> ---
> src/glsl/lower_vec_index_to_swizzle.cpp | 17 +++++++++++++++--
> 1 files changed, 15 insertions(+), 2 deletions(-)
>
> diff --git a/src/glsl/lower_vec_index_to_swizzle.cpp b/src/glsl/lower_vec_index_to_swizzle.cpp
> index c7630c2..ebc0833 100644
> --- a/src/glsl/lower_vec_index_to_swizzle.cpp
> +++ b/src/glsl/lower_vec_index_to_swizzle.cpp
> @@ -33,6 +33,7 @@
> #include "ir_visitor.h"
> #include "ir_optimization.h"
> #include "glsl_types.h"
> +#include "main/macros.h"
>
> /**
> * Visitor class for replacing expressions with ir_constant values.
> @@ -76,8 +77,20 @@ ir_vec_index_to_swizzle_visitor::convert_vec_index_to_swizzle(ir_rvalue *ir)
>
> void *ctx = ralloc_parent(ir);
> this->progress = true;
> - return new(ctx) ir_swizzle(deref->array,
> - ir_constant->value.i[0], 0, 0, 0, 1);
> +
> + /* Page 40 of the GLSL 1.20 spec says:
> + *
> + * "When indexing with non-constant expressions, behavior is undefined
> + * if the index is negative, or greater than or equal to the size of
> + * the vector."
> + *
> + * The ir_swizzle constructor gets angry if the index is negative or too
> + * large. For simplicity sake, just clamp the index to [0, size-1].
> + */
> + const int i = MIN2(MAX2(ir_constant->value.i[0], 0),
> + (deref->array->type->vector_elements - 1));
> +
> + return new(ctx) ir_swizzle(deref->array, i, 0, 0, 0, 1);
> }
>
> ir_visitor_status
This seems like a sensible approach (since we can't generate an error at
this point.)
Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>
More information about the mesa-dev
mailing list