re PR tree-optimization/50031 (Sphinx3 has a 10% regression going from GCC 4.5 to GCC 4.6 on powerpc)

2012-02-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
	    Ira Rosen  <irar@il.ibm.com>

	PR tree-optimization/50031
	* targhooks.c (default_builtin_vectorization_cost): Handle
	vec_promote_demote.
	* target.h (enum vect_cost_for_stmt): Add vec_promote_demote.
	* tree-vect-loop.c (vect_get_single_scalar_iteraion_cost): Handle
	all types of reduction and pattern statements.
	(vect_estimate_min_profitable_iters): Likewise.
	* tree-vect-stmts.c (vect_model_promotion_demotion_cost): New function.
	(vect_get_load_cost): Use vec_perm for permutations; add dump logic
	for explicit realigns.
	(vectorizable_conversion): Call vect_model_promotion_demotion_cost.
	* config/spu/spu.c (spu_builtin_vectorization_cost): Handle
	vec_promote_demote.
	* config/i386/i386.c (ix86_builtin_vectorization_cost): Likewise.
	* config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Update
	vec_perm for VSX and handle vec_promote_demote.


Co-Authored-By: Ira Rosen <irar@il.ibm.com>

From-SVN: r184102

1 files changed, 56 insertions, 6 deletions

diff --git a/gcc/tree-vect-stmts.c b/gcc/tree-vect-stmts.c
index e854da5..20f10f3 100644
--- a/gcc/tree-vect-stmts.c
+++ b/gcc/tree-vect-stmts.c
@@ -811,6 +811,46 @@ vect_model_simple_cost (stmt_vec_info stmt_info, int ncopies,
 }
 
 
+/* Model cost for type demotion and promotion operations.  PWR is normally
+   zero for single-step promotions and demotions.  It will be one if 
+   two-step promotion/demotion is required, and so on.  Each additional
+   step doubles the number of instructions required.  */
+
+static void
+vect_model_promotion_demotion_cost (stmt_vec_info stmt_info,
+				    enum vect_def_type *dt, int pwr)
+{
+  int i, tmp;
+  int inside_cost = 0, outside_cost = 0, single_stmt_cost;
+
+  /* The SLP costs were already calculated during SLP tree build.  */
+  if (PURE_SLP_STMT (stmt_info))
+    return;
+
+  single_stmt_cost = vect_get_stmt_cost (vec_promote_demote);
+  for (i = 0; i < pwr + 1; i++)
+    {
+      tmp = (STMT_VINFO_TYPE (stmt_info) == type_promotion_vec_info_type) ?
+	(i + 1) : i;
+      inside_cost += vect_pow2 (tmp) * single_stmt_cost;
+    }
+
+  /* FORNOW: Assuming maximum 2 args per stmts.  */
+  for (i = 0; i < 2; i++)
+    {
+      if (dt[i] == vect_constant_def || dt[i] == vect_external_def)
+        outside_cost += vect_get_stmt_cost (vector_stmt);
+    }
+
+  if (vect_print_dump_info (REPORT_COST))
+    fprintf (vect_dump, "vect_model_promotion_demotion_cost: inside_cost = %d, "
+             "outside_cost = %d .", inside_cost, outside_cost);
+
+  /* Set the costs in STMT_INFO.  */
+  stmt_vinfo_set_inside_of_loop_cost (stmt_info, NULL, inside_cost);
+  stmt_vinfo_set_outside_of_loop_cost (stmt_info, NULL, outside_cost);
+}
+
 /* Function vect_cost_strided_group_size
 
    For strided load or store, return the group_size only if it is the first
@@ -887,7 +927,6 @@ vect_model_store_cost (stmt_vec_info stmt_info, int ncopies,
       if (vect_print_dump_info (REPORT_COST))
         fprintf (vect_dump, "vect_model_store_cost: strided group_size = %d .",
                  group_size);
-
     }
 
   /* Costs of the stores.  */
@@ -1049,7 +1088,7 @@ vect_get_load_cost (struct data_reference *dr, int ncopies,
     case dr_explicit_realign:
       {
         *inside_cost += ncopies * (2 * vect_get_stmt_cost (vector_load)
-           + vect_get_stmt_cost (vector_stmt));
+				   + vect_get_stmt_cost (vec_perm));
 
         /* FIXME: If the misalignment remains fixed across the iterations of
            the containing loop, the following cost should be added to the
@@ -1057,6 +1096,9 @@ vect_get_load_cost (struct data_reference *dr, int ncopies,
         if (targetm.vectorize.builtin_mask_for_load)
           *inside_cost += vect_get_stmt_cost (vector_stmt);
 
+        if (vect_print_dump_info (REPORT_COST))
+          fprintf (vect_dump, "vect_model_load_cost: explicit realign");
+
         break;
       }
     case dr_explicit_realign_optimized:
@@ -1080,7 +1122,12 @@ vect_get_load_cost (struct data_reference *dr, int ncopies,
           }
 
         *inside_cost += ncopies * (vect_get_stmt_cost (vector_load)
-          + vect_get_stmt_cost (vector_stmt));
+				   + vect_get_stmt_cost (vec_perm));
+
+        if (vect_print_dump_info (REPORT_COST))
+          fprintf (vect_dump,
+		   "vect_model_load_cost: explicit realign optimized");
+
         break;
       }
 
@@ -2392,16 +2439,19 @@ vectorizable_conversion (gimple stmt, gimple_stmt_iterator *gsi,
       if (vect_print_dump_info (REPORT_DETAILS))
 	fprintf (vect_dump, "=== vectorizable_conversion ===");
       if (code == FIX_TRUNC_EXPR || code == FLOAT_EXPR)
-	STMT_VINFO_TYPE (stmt_info) = type_conversion_vec_info_type;
+        {
+	  STMT_VINFO_TYPE (stmt_info) = type_conversion_vec_info_type;
+	  vect_model_simple_cost (stmt_info, ncopies, dt, NULL);
+	}
       else if (modifier == NARROW)
 	{
 	  STMT_VINFO_TYPE (stmt_info) = type_demotion_vec_info_type;
-	  vect_model_simple_cost (stmt_info, ncopies, dt, NULL);
+	  vect_model_promotion_demotion_cost (stmt_info, dt, multi_step_cvt);
 	}
       else
 	{
 	  STMT_VINFO_TYPE (stmt_info) = type_promotion_vec_info_type;
-	  vect_model_simple_cost (stmt_info, 2 * ncopies, dt, NULL);
+	  vect_model_promotion_demotion_cost (stmt_info, dt, multi_step_cvt);
 	}
       VEC_free (tree, heap, interm_types);
       return true;