Fold vector to scalar optimization

Index: gcc/testsuite/gcc.dg/vec-scal-opt2.c
===================================================================
--- gcc/testsuite/gcc.dg/vec-scal-opt2.c	(revision 0)
+++ gcc/testsuite/gcc.dg/vec-scal-opt2.c	(revision 0)
@@ -0,0 +1,19 @@ 
+/* { dg-do compile } */
+/* { dg-options "-O -fdump-tree-veclower" } */
+
+#define vidx(type, vec, idx) (*((type *) &(vec) + idx))
+#define vector(elcount, type)  \
+__attribute__((vector_size((elcount)*sizeof(type)))) type
+
+int main (int argc, char *argv[]) {
+   vector(8, short) v0 = {argc,1,2,3,4,5,6,7};
+   vector(8, short) v1 = {2,2,2,2,2,2,2,2};
+   vector(8, short) r1;
+
+   r1 = v0 >> v1;
+
+   return vidx(short, r1, 0);
+}
+
+/* { dg-final { scan-tree-dump-times ">> 2" 1 "veclower" } } */
+/* { dg-final { cleanup-tree-dump "veclower" } } */
Index: gcc/testsuite/gcc.dg/vec-scal-opt.c
===================================================================
--- gcc/testsuite/gcc.dg/vec-scal-opt.c	(revision 0)
+++ gcc/testsuite/gcc.dg/vec-scal-opt.c	(revision 0)
@@ -0,0 +1,22 @@ 
+/* { dg-do compile } */
+/* { dg-options "-O -fdump-tree-veclower" } */
+
+#define vidx(type, vec, idx) (*((type *) &(vec) + idx))
+#define vector(elcount, type)  \
+__attribute__((vector_size((elcount)*sizeof(type)))) type
+
+short k;
+
+int main (int argc, char *argv[]) {
+   k = argc;
+   vector(8, short) v0 = {argc,1,2,3,4,5,6,7};
+   vector(8, short) v2 = {k,   k,k,k,k,k,k,k};
+   vector(8, short) r1;
+
+   r1 = v0 >> v2;
+
+   return vidx(short, r1, 0);
+}
+
+/* { dg-final { scan-tree-dump-times ">> k.\[0-9_\]*" 1 "veclower" } } */
+/* { dg-final { cleanup-tree-dump "veclower" } } */
Index: gcc/testsuite/gcc.dg/vec-scal-opt1.c
===================================================================
--- gcc/testsuite/gcc.dg/vec-scal-opt1.c	(revision 0)
+++ gcc/testsuite/gcc.dg/vec-scal-opt1.c	(revision 0)
@@ -0,0 +1,20 @@ 
+/* { dg-do compile } */
+/* { dg-options "-O -fdump-tree-veclower" } */
+
+#define vidx(type, vec, idx) (*((type *) &(vec) + idx))
+#define vector(elcount, type)  \
+__attribute__((vector_size((elcount)*sizeof(type)))) type
+
+short k;
+
+int main (int argc, char *argv[]) {
+   vector(8, short) v0 = {argc,1,2,3,4,5,6,7};
+   vector(8, short) r1;
+
+   r1 = v0 >> (vector(8, short)){2,2,2,2,2,2,2,2};
+
+   return vidx(short, r1, 0);
+}
+
+/* { dg-final { scan-tree-dump-times ">> 2" 1 "veclower" } } */
+/* { dg-final { cleanup-tree-dump "veclower" } } */
Index: gcc/tree-vect-generic.c
===================================================================
--- gcc/tree-vect-generic.c	(revision 166249)
+++ gcc/tree-vect-generic.c	(working copy)
@@ -284,6 +284,62 @@  expand_vector_addition (gimple_stmt_iter
 				    a, b, code);
 }
 
+/* Check if vector VEC consists of all the equal elements and
+   that the number of elements corresponds to the type of VEC.
+   Function sets ELEMENT with the first element of the vector
+   or NULL_TREE if comparison failed.  */
+static bool uniform_vector_p (tree vec, tree *element)
+{
+  tree first, t, els;
+  bool eq = true;
+  unsigned HOST_WIDE_INT i;
+
+  *element = NULL_TREE;
+
+  if (vec == NULL_TREE)
+    return false;
+
+  if (TREE_CODE (vec) == VECTOR_CST)
+    {
+      els = TREE_VECTOR_CST_ELTS (vec);
+      first = TREE_VALUE (els);
+      els = TREE_CHAIN (els);
+
+      for (t = els, i = 0; t; t = TREE_CHAIN (t), i++)
+        {
+          eq = operand_equal_p (first, TREE_VALUE(t), 0);
+          if (!eq)
+            break;
+        }
+
+      *element = first;
+      return eq && (i == TYPE_VECTOR_SUBPARTS (TREE_TYPE (vec))-1);
+    }
+
+  else if (TREE_CODE (vec) == CONSTRUCTOR)
+    {
+      first = error_mark_node;
+
+      FOR_EACH_CONSTRUCTOR_VALUE (CONSTRUCTOR_ELTS (vec), i, t)
+        {
+          if (i == 0)
+            {
+              first = t;
+              continue;
+            }
+          eq = (first == t);
+          if (!eq)
+            break;
+        }
+      
+      *element = first;
+      return eq && (i == TYPE_VECTOR_SUBPARTS (TREE_TYPE (vec)));
+      
+    }
+  
+  return false;
+}
+
 static tree
 expand_vector_operation (gimple_stmt_iterator *gsi, tree type, tree compute_type,
 			 gimple assign, enum tree_code code)
@@ -392,7 +448,7 @@  expand_vector_operations_1 (gimple_stmt_
   tree lhs, rhs1, rhs2 = NULL, type, compute_type;
   enum tree_code code;
   enum machine_mode compute_mode;
-  optab op;
+  optab op = NULL;
   enum gimple_rhs_class rhs_class;
   tree new_rhs;
 
@@ -434,18 +490,45 @@  expand_vector_operations_1 (gimple_stmt_
       || code == LROTATE_EXPR
       || code == RROTATE_EXPR)
     {
-      /* If the 2nd argument is vector, we need a vector/vector shift */
+      bool vector_scalar_shift;
+      op = optab_for_tree_code (code, type, optab_scalar);
+      
+      /* Vector/Scalar shift is supported.  */
+      vector_scalar_shift = !(!op 
+                              || optab_handler (op, TYPE_MODE (type)) 
+                                 == CODE_FOR_nothing);
+
+      /* If the 2nd argument is vector, we need a vector/vector shift.
+         Except all the elements in the second vector are the same.  */
       if (VECTOR_MODE_P (TYPE_MODE (TREE_TYPE (rhs2))))
-	op = optab_for_tree_code (code, type, optab_vector);
-      else
-	{
-	  /* Try for a vector/scalar shift, and if we don't have one, see if we
-	     have a vector/vector shift */
-	  op = optab_for_tree_code (code, type, optab_scalar);
-	  if (!op
-	      || optab_handler (op, TYPE_MODE (type)) == CODE_FOR_nothing)
-	    op = optab_for_tree_code (code, type, optab_vector);
-	}
+        {
+          tree first, var = NULL_TREE;
+          gimple def_stmt;
+
+          /* Check whether we have vector <op> {x,x,x,x} where x
+             could be a scalar variable or a constant. Transform
+             vector <op> {x,x,x,x} ==> vector <op> scalar.  */
+          if (vector_scalar_shift 
+              && ((TREE_CODE (rhs2) == VECTOR_CST && (var = rhs2))
+                  ||
+                  (TREE_CODE (rhs2) == SSA_NAME 
+                   && (def_stmt = SSA_NAME_DEF_STMT (rhs2))
+                   && gimple_assign_single_p (def_stmt)
+                   && (var = gimple_assign_rhs1 (def_stmt))))
+              && uniform_vector_p (var, &first))
+            {
+              gimple_assign_set_rhs2 (stmt, first);
+              update_stmt (stmt);
+              rhs2 = first;
+            }
+          else
+            op = optab_for_tree_code (code, type, optab_vector);
+        }
+    
+      /* Try for a vector/scalar shift, and if we don't have one, see if we
+         have a vector/vector shift */
+      else if (!vector_scalar_shift)
+        op = optab_for_tree_code (code, type, optab_vector);
     }
   else
     op = optab_for_tree_code (code, type, optab_default);
Index: gcc/passes.c
===================================================================
--- gcc/passes.c	(revision 166249)
+++ gcc/passes.c	(working copy)
@@ -736,7 +736,6 @@  init_optimization_passes (void)
   NEXT_PASS (pass_refactor_eh);
   NEXT_PASS (pass_lower_eh);
   NEXT_PASS (pass_build_cfg);
-  NEXT_PASS (pass_lower_vector);
   NEXT_PASS (pass_warn_function_return);
   NEXT_PASS (pass_build_cgraph_edges);
   NEXT_PASS (pass_inline_parameters);
@@ -755,6 +754,7 @@  init_optimization_passes (void)
 
       NEXT_PASS (pass_referenced_vars);
       NEXT_PASS (pass_build_ssa);
+      NEXT_PASS (pass_lower_vector);
       NEXT_PASS (pass_early_warn_uninitialized);
       /* Note that it is not strictly necessary to schedule an early
 	 inline pass here.  However, some test cases (e.g.,