[Mesa-dev] [PATCH 2/5] st/glsl_to_tgsi: fix conditional assignments to packed shader outputs

[Ian Romanick]

I'm hoping to land my patch series that removes ir_assignment::condition
either today or tomorrow.  I believe that series deletes all of this
code... but this patch would still be useful for stable.

On 09/26/2017 07:42 AM, Nicolai Hähnle wrote:
> From: Nicolai Hähnle <nicolai.haehnle at amd.com>
> 
> Overriding the default (no-op) swizzle is clearly counter-productive,
> since the whole point is putting the destination register as one of
> the source operands so that it remains unmodified when the assignment
> condition is false.
> 
> Fragment depth and stencil outputs are a special case due to how their
> source swizzles are manipulated in translate_src when compiling to
> TGSI.
> 
> Fixes dEQP-GLES2.functional.shaders.conditionals.if.*_vertex
> Cc: mesa-stable at lists.freedesktop.org
> ---
>  src/mesa/state_tracker/st_glsl_to_tgsi.cpp | 10 +++++++++-
>  1 file changed, 9 insertions(+), 1 deletion(-)
> 
> diff --git a/src/mesa/state_tracker/st_glsl_to_tgsi.cpp b/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
> index f4870a1c606..0daf5a14285 100644
> --- a/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
> +++ b/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
> @@ -2890,21 +2890,29 @@ glsl_to_tgsi_visitor::emit_block_mov(ir_assignment *ir, const struct glsl_type *
>        }
>        return;
>     }
>  
>     assert(type->is_scalar() || type->is_vector());
>  
>     l->type = type->base_type;
>     r->type = type->base_type;
>     if (cond) {
>        st_src_reg l_src = st_src_reg(*l);
> -      l_src.swizzle = swizzle_for_size(type->vector_elements);
> +
> +      if (l_src.file == PROGRAM_OUTPUT &&
> +          this->prog->Target == GL_FRAGMENT_PROGRAM_ARB &&
> +          (l_src.index == FRAG_RESULT_DEPTH || l_src.index == FRAG_RESULT_STENCIL)) {
> +         /* This is a special case because the source swizzles will be shifted
> +          * later to account for the difference between GLSL (where they're
> +          * plain floats) and TGSI (where they're Z and Y components). */
> +         l_src.swizzle = SWIZZLE_XXXX;
> +      }
>  
>        if (native_integers) {
>           emit_asm(ir, TGSI_OPCODE_UCMP, *l, *cond,
>                cond_swap ? l_src : *r,
>                cond_swap ? *r : l_src);
>        } else {
>           emit_asm(ir, TGSI_OPCODE_CMP, *l, *cond,
>                cond_swap ? l_src : *r,
>                cond_swap ? *r : l_src);
>        }
>