[Mesa-dev] [PATCH 4/4] glsl: Delete LRP_TO_ARITH lowering pass flag.

[Ian Romanick]

On 02/23/2014 07:19 PM, Kenneth Graunke wrote:
> Tt's kind of a trap---calling do_common_optimization() after
> lower_instructions() may cause opt_algebraic() to reintroduce
> ir_triop_lrp expressions that were lowered, effectively defeating the
> point.  Because of this, nobody uses it.
>
> Cc: "10.1" <mesa-stable at lists.freedesktop.org>
> Signed-off-by: Kenneth Graunke <kenneth at whitecape.org>
> ---
>   src/glsl/ir_optimization.h      |  5 ++---
>   src/glsl/lower_instructions.cpp | 10 ----------
>   2 files changed, 2 insertions(+), 13 deletions(-)
>
> diff --git a/src/glsl/ir_optimization.h b/src/glsl/ir_optimization.h
> index 055d655..1777b84 100644
> --- a/src/glsl/ir_optimization.h
> +++ b/src/glsl/ir_optimization.h
> @@ -36,9 +36,8 @@
>   #define LOG_TO_LOG2        0x10
>   #define MOD_TO_FRACT       0x20
>   #define INT_DIV_TO_MUL_RCP 0x40
> -#define LRP_TO_ARITH       0x80
> -#define BITFIELD_INSERT_TO_BFM_BFI 0x100
> -#define LDEXP_TO_ARITH     0x200
> +#define BITFIELD_INSERT_TO_BFM_BFI 0x80
> +#define LDEXP_TO_ARITH     0x100
>
>   /**
>    * \see class lower_packing_builtins_visitor
> diff --git a/src/glsl/lower_instructions.cpp b/src/glsl/lower_instructions.cpp
> index 44a6e80..9c5e7b3 100644
> --- a/src/glsl/lower_instructions.cpp
> +++ b/src/glsl/lower_instructions.cpp
> @@ -38,7 +38,6 @@
>    * - LOG_TO_LOG2
>    * - MOD_TO_FRACT
>    * - LDEXP_TO_ARITH
> - * - LRP_TO_ARITH
>    * - BITFIELD_INSERT_TO_BFM_BFI
>    *
>    * SUB_TO_ADD_NEG:
> @@ -87,10 +86,6 @@
>    * -------------
>    * Converts ir_binop_ldexp to arithmetic and bit operations.
>    *
> - * LRP_TO_ARITH:
> - * -------------
> - * Converts ir_triop_lrp to (op0 * (1.0f - op2)) + (op1 * op2).
> - *
>    * BITFIELD_INSERT_TO_BFM_BFI:
>    * ---------------------------
>    * Breaks ir_quadop_bitfield_insert into ir_binop_bfm (bitfield mask) and
> @@ -499,11 +494,6 @@ lower_instructions_visitor::visit_leave(ir_expression *ir)
>   	 pow_to_exp2(ir);
>         break;
>
> -   case ir_triop_lrp:
> -      if (lowering(LRP_TO_ARITH))
> -	 lrp_to_arith(ir);

Shouldn't the lrp_to_arith implementation also get removed?  This 
appears to be the only caller...

> -      break;
> -
>      case ir_quadop_bitfield_insert:
>         if (lowering(BITFIELD_INSERT_TO_BFM_BFI))
>            bitfield_insert_to_bfm_bfi(ir);
>