glsl: drop sub to add neg lowering in GLSL IR

NIR opt algebraic does this for us so no need to have it
implemented here also.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19112>

src/compiler/glsl/glsl_parser_extras.cpp

@@ -2381,8 +2381,6 @@ do_common_optimization(exec_list *ir, bool linked,
       }                                                                 \
    } while (false)
 
-   OPT(lower_instructions, ir, SUB_TO_ADD_NEG);
-
    if (linked) {
       OPT(do_function_inlining, ir);
       OPT(do_dead_functions, ir);

src/compiler/glsl/ir_optimization.h

@@ -34,7 +34,6 @@ struct gl_linked_shader;
 struct gl_shader_program;
 
 /* Operations for lower_instructions() */
-#define SUB_TO_ADD_NEG     0x01
 #define LDEXP_TO_ARITH     0x80
 #define CARRY_TO_ARITH     0x100
 #define BORROW_TO_ARITH    0x200

src/compiler/glsl/lower_instructions.cpp

@@ -30,22 +30,11 @@
  * rather than in each driver backend.
  *
  * Currently supported transformations:
- * - SUB_TO_ADD_NEG
  * - LDEXP_TO_ARITH
  * - CARRY_TO_ARITH
  * - BORROW_TO_ARITH
  * - DOPS_TO_DFRAC
  *
- * SUB_TO_ADD_NEG:
- * ---------------
- * Breaks an ir_binop_sub expression down to add(op0, neg(op1))
- *
- * This simplifies expression reassociation, and for many backends
- * there is no subtract operation separate from adding the negation.
- * For backends with native subtract operations, they will probably
- * want to recognize add(op0, neg(op1)) or the other way around to
- * produce a subtract anyway.
- *
  * LDEXP_TO_ARITH:
  * -------------
  * Converts ir_binop_ldexp to arithmetic and bit operations for float sources.
@@ -93,7 +82,6 @@ public:
 private:
    unsigned lower; /** Bitfield of which operations to lower */
 
-   void sub_to_add_neg(ir_expression *);
    void ldexp_to_arith(ir_expression *);
    void dldexp_to_arith(ir_expression *);
    void dfrexp_sig_to_arith(ir_expression *);
@@ -140,16 +128,6 @@ lower_instructions(exec_list *instructions, unsigned what_to_lower)
    return v.progress;
 }
 
-void
-lower_instructions_visitor::sub_to_add_neg(ir_expression *ir)
-{
-   ir->operation = ir_binop_add;
-   ir->init_num_operands();
-   ir->operands[1] = new(ir) ir_expression(ir_unop_neg, ir->operands[1]->type,
-					   ir->operands[1], NULL);
-   this->progress = true;
-}
-
 void
 lower_instructions_visitor::ldexp_to_arith(ir_expression *ir)
 {
@@ -1452,10 +1430,6 @@ lower_instructions_visitor::visit_leave(ir_expression *ir)
       if (ir->operands[0]->type->is_double())
          double_lrp(ir);
       break;
-   case ir_binop_sub:
-      if (lowering(SUB_TO_ADD_NEG))
-	 sub_to_add_neg(ir);
-      break;
 
    case ir_binop_ldexp:
       if (lowering(LDEXP_TO_ARITH) && ir->type->is_float())

src/compiler/glsl/tests/lower_precision_test.py

@@ -1318,7 +1318,7 @@ TESTS = [
                  color *= borrow;
          }
          """,
-         r'expression uint \+ \(var_ref x\) \(expression uint neg'),
+         r'expression uint \- \(var_ref x\) \(var_ref y'),
     Test("imulExtended",
          """
          #version 310 es