[Mesa-stable] [PATCH 2/2] glsl: Relax ir_quadop_bitfield_insert type restrictions.

[Emil Velikov]

Hi Ken,

On 5 January 2016 at 13:34, Kenneth Graunke <kenneth at whitecape.org> wrote:
> While GLSL restricts bitfieldInsert's offset and bits parameters to
> be scalars, we shouldn't require this in the IR.
>
> In particular, opt_vectorize() tries to combine
>
> result.x = bitfieldInsert(src0.x, src1.x, src2.x, src3.x);
> result.y = bitfieldInsert(src0.y, src1.y, src2.y, src3.y);
> result.z = bitfieldInsert(src0.z, src1.z, src2.z, src3.z);
> result.w = bitfieldInsert(src0.w, src1.w, src2.w, src3.w);
>
> into a single ivec4 bitfieldInsert operation.  This currently breaks,
> because the last two types become ivec4 rather than int.  It seems
> perfectly reasonable to allow this.
>
> i965 lowers ir_quadop_bitfield_insert to ir_binop_bfm and ir_triop_bfi,
> which already lift these restrictions.
>
> (I debated about using is_integer() or base_type == GLSL_TYPE_INT here;
> I ended up relaxing it to allow either int/uint because ir_binop_bfm
> and ir_triop_bfi do that already.)
>
> Fixes assertion failures when compiling Shadow of Mordor vertex shaders
> on i965 in vec4 mode (where OptimizeForAOS enables opt_vectorize()).
>
> Signed-off-by: Kenneth Graunke <kenneth at whitecape.org>
> Cc: Matt Turner <mattst88 at gmail.com>
> Cc: mesa-stable at lists.freedesktop.org

Seems like a tweaked version (addressing Jason's NIR concern) has
landed in master without the stable tag.
Just a reminder that I'll consider this as "self-rejected" and it
won't feature in 11.1. Do let me know if you dropped the tag by
mistake.

-Emil