--- c-family/c-common.c	(revision 186176)
+++ c-family/c-common.c	(working copy)
@@ -4991,7 +4991,7 @@ c_common_nodes_and_builtins (void)
     uint8_type_node =
       TREE_TYPE (identifier_global_value (c_get_ident (UINT8_TYPE)));
   if (UINT16_TYPE)
-    uint16_type_node =
+    c_uint16_type_node =
       TREE_TYPE (identifier_global_value (c_get_ident (UINT16_TYPE)));
   if (UINT32_TYPE)
     c_uint32_type_node =
--- c-family/c-common.h	(revision 186176)
+++ c-family/c-common.h	(working copy)
@@ -390,7 +390,7 @@ extern const unsigned int num_c_common_r
 #define int32_type_node			c_global_trees[CTI_INT32_TYPE]
 #define int64_type_node			c_global_trees[CTI_INT64_TYPE]
 #define uint8_type_node			c_global_trees[CTI_UINT8_TYPE]
-#define uint16_type_node		c_global_trees[CTI_UINT16_TYPE]
+#define c_uint16_type_node		c_global_trees[CTI_UINT16_TYPE]
 #define c_uint32_type_node		c_global_trees[CTI_UINT32_TYPE]
 #define c_uint64_type_node		c_global_trees[CTI_UINT64_TYPE]
 #define int_least8_type_node		c_global_trees[CTI_INT_LEAST8_TYPE]
--- c-family/c-cppbuiltin.c	(revision 186176)
+++ c-family/c-cppbuiltin.c	(working copy)
@@ -448,8 +448,8 @@ builtin_define_stdint_macros (void)
     builtin_define_type_max ("__INT64_MAX__", int64_type_node);
   if (uint8_type_node)
     builtin_define_type_max ("__UINT8_MAX__", uint8_type_node);
-  if (uint16_type_node)
-    builtin_define_type_max ("__UINT16_MAX__", uint16_type_node);
+  if (c_uint16_type_node)
+    builtin_define_type_max ("__UINT16_MAX__", c_uint16_type_node);
   if (c_uint32_type_node)
     builtin_define_type_max ("__UINT32_MAX__", c_uint32_type_node);
   if (c_uint64_type_node)
--- optabs.c	(revision 186176)
+++ optabs.c	(working copy)
@@ -3030,6 +3030,47 @@ expand_unop (enum machine_mode mode, opt
   /* Widening (or narrowing) bswap needs special treatment.  */
   if (unoptab == bswap_optab)
     {
+      /* HImode is special because in this mode BSWAP is equivalent to ROTATE
+	 or ROTATERT.  First try these directly; if this fails, then try the
+	 obvious pair of shifts with allowed widening, as this will probably
+	 be always more efficient than the other fallback methods.  */
+      if (mode == HImode)
+	{
+	  rtx last, temp1, temp2;
+
+	  if (optab_handler (rotl_optab, mode) != CODE_FOR_nothing)
+	    {
+	      temp = expand_binop (mode, rotl_optab, op0, GEN_INT (8), target,
+				   unsignedp, OPTAB_DIRECT);
+	      if (temp)
+		return temp;
+	     }
+
+	  if (optab_handler (rotr_optab, mode) != CODE_FOR_nothing)
+	    {
+	      temp = expand_binop (mode, rotr_optab, op0, GEN_INT (8), target,
+				   unsignedp, OPTAB_DIRECT);
+	      if (temp)
+		return temp;
+	    }
+
+	  last = get_last_insn ();
+
+	  temp1 = expand_binop (mode, ashl_optab, op0, GEN_INT (8), NULL_RTX,
+			        unsignedp, OPTAB_WIDEN);
+	  temp2 = expand_binop (mode, lshr_optab, op0, GEN_INT (8), NULL_RTX,
+			        unsignedp, OPTAB_WIDEN);
+	  if (temp1 && temp2)
+	    {
+	      temp = expand_binop (mode, ior_optab, temp1, temp2, target,
+				   unsignedp, OPTAB_WIDEN);
+	      if (temp)
+		return temp;
+	    }
+
+	  delete_insns_since (last);
+	}
+
       temp = widen_bswap (mode, op0, target);
       if (temp)
 	return temp;
@@ -3222,10 +3263,10 @@ expand_unop (enum machine_mode mode, opt
 	      /* For certain operations, we need not actually extend
 		 the narrow operand, as long as we will truncate the
 		 results to the same narrowness.  */
-
 	      xop0 = widen_operand (xop0, wider_mode, mode, unsignedp,
 				    (unoptab == neg_optab
-				     || unoptab == one_cmpl_optab)
+				     || unoptab == one_cmpl_optab
+				     || unoptab == bswap_optab)
 				    && mclass == MODE_INT);
 
 	      temp = expand_unop (wider_mode, unoptab, xop0, NULL_RTX,
@@ -3240,6 +3281,20 @@ expand_unop (enum machine_mode mode, opt
 					      - GET_MODE_PRECISION (mode)),
 				     target, true, OPTAB_DIRECT);
 
+	      /* Likewise for bswap.  */
+	      if (unoptab == bswap_optab && temp != 0)
+		{
+		  gcc_assert (GET_MODE_PRECISION (wider_mode)
+			      == GET_MODE_BITSIZE (wider_mode)
+			      && GET_MODE_PRECISION (mode)
+				 == GET_MODE_BITSIZE (mode));
+
+		  temp = expand_shift (RSHIFT_EXPR, wider_mode, temp,
+				       GET_MODE_BITSIZE (wider_mode)
+				       - GET_MODE_BITSIZE (mode),
+				       NULL_RTX, true);
+		}
+
 	      if (temp)
 		{
 		  if (mclass != MODE_INT)
--- tree.c	(revision 186176)
+++ tree.c	(working copy)
@@ -9388,6 +9388,7 @@ build_common_tree_nodes (bool signed_cha
   integer_ptr_type_node = build_pointer_type (integer_type_node);
 
   /* Fixed size integer types.  */
+  uint16_type_node = build_nonstandard_integer_type (16, true);
   uint32_type_node = build_nonstandard_integer_type (32, true);
   uint64_type_node = build_nonstandard_integer_type (64, true);
 
--- tree.h	(revision 186176)
+++ tree.h	(working copy)
@@ -3747,6 +3747,7 @@ enum tree_index
   TI_UINTDI_TYPE,
   TI_UINTTI_TYPE,
 
+  TI_UINT16_TYPE,
   TI_UINT32_TYPE,
   TI_UINT64_TYPE,
 
@@ -3902,6 +3903,7 @@ extern GTY(()) tree global_trees[TI_MAX]
 #define unsigned_intDI_type_node	global_trees[TI_UINTDI_TYPE]
 #define unsigned_intTI_type_node	global_trees[TI_UINTTI_TYPE]
 
+#define uint16_type_node		global_trees[TI_UINT16_TYPE]
 #define uint32_type_node		global_trees[TI_UINT32_TYPE]
 #define uint64_type_node		global_trees[TI_UINT64_TYPE]
 
--- builtins.c	(revision 186176)
+++ builtins.c	(working copy)
@@ -4626,13 +4626,15 @@ expand_builtin_alloca (tree exp, bool ca
   return result;
 }
 
-/* Expand a call to a bswap builtin with argument ARG0.  MODE
-   is the mode to expand with.  */
+/* Expand a call to bswap builtin in EXP.
+   Return NULL_RTX if a normal call should be emitted rather than expanding the
+   function in-line.  If convenient, the result should be placed in TARGET.
+   SUBTARGET may be used as the target for computing one of EXP's operands.  */
 
 static rtx
-expand_builtin_bswap (tree exp, rtx target, rtx subtarget)
+expand_builtin_bswap (enum machine_mode target_mode, tree exp, rtx target,
+		      rtx subtarget)
 {
-  enum machine_mode mode;
   tree arg;
   rtx op0;
 
@@ -4640,14 +4642,18 @@ expand_builtin_bswap (tree exp, rtx targ
     return NULL_RTX;
 
   arg = CALL_EXPR_ARG (exp, 0);
-  mode = TYPE_MODE (TREE_TYPE (arg));
-  op0 = expand_expr (arg, subtarget, VOIDmode, EXPAND_NORMAL);
+  op0 = expand_expr (arg,
+		     subtarget && GET_MODE (subtarget) == target_mode
+		     ? subtarget : NULL_RTX,
+		     target_mode, EXPAND_NORMAL);
+  if (GET_MODE (op0) != target_mode)
+    op0 = convert_to_mode (target_mode, op0, 1);
 
-  target = expand_unop (mode, bswap_optab, op0, target, 1);
+  target = expand_unop (target_mode, bswap_optab, op0, target, 1);
 
   gcc_assert (target);
 
-  return convert_to_mode (mode, target, 0);
+  return convert_to_mode (target_mode, target, 1);
 }
 
 /* Expand a call to a unary builtin in EXP.
@@ -6077,10 +6083,10 @@ expand_builtin (tree exp, rtx target, rt
       expand_stack_restore (CALL_EXPR_ARG (exp, 0));
       return const0_rtx;
 
+    case BUILT_IN_BSWAP16:
     case BUILT_IN_BSWAP32:
     case BUILT_IN_BSWAP64:
-      target = expand_builtin_bswap (exp, target, subtarget);
-
+      target = expand_builtin_bswap (target_mode, exp, target, subtarget);
       if (target)
 	return target;
       break;
--- builtin-types.def	(revision 186176)
+++ builtin-types.def	(working copy)
@@ -76,6 +76,7 @@ DEF_PRIMITIVE_TYPE (BT_INT128, int128_in
 DEF_PRIMITIVE_TYPE (BT_UINT128, int128_unsigned_type_node)
 DEF_PRIMITIVE_TYPE (BT_INTMAX, intmax_type_node)
 DEF_PRIMITIVE_TYPE (BT_UINTMAX, uintmax_type_node)
+DEF_PRIMITIVE_TYPE (BT_UINT16, uint16_type_node)
 DEF_PRIMITIVE_TYPE (BT_UINT32, uint32_type_node)
 DEF_PRIMITIVE_TYPE (BT_UINT64, uint64_type_node)
 DEF_PRIMITIVE_TYPE (BT_WORD, (*lang_hooks.types.type_for_mode) (word_mode, 1))
@@ -226,6 +227,7 @@ DEF_FUNCTION_TYPE_1 (BT_FN_VOID_PTRPTR,
 DEF_FUNCTION_TYPE_1 (BT_FN_UINT_UINT, BT_UINT, BT_UINT)
 DEF_FUNCTION_TYPE_1 (BT_FN_ULONG_ULONG, BT_ULONG, BT_ULONG)
 DEF_FUNCTION_TYPE_1 (BT_FN_ULONGLONG_ULONGLONG, BT_ULONGLONG, BT_ULONGLONG)
+DEF_FUNCTION_TYPE_1 (BT_FN_UINT16_UINT16, BT_UINT16, BT_UINT16)
 DEF_FUNCTION_TYPE_1 (BT_FN_UINT32_UINT32, BT_UINT32, BT_UINT32)
 DEF_FUNCTION_TYPE_1 (BT_FN_UINT64_UINT64, BT_UINT64, BT_UINT64)
 
--- builtins.def	(revision 186176)
+++ builtins.def	(working copy)
@@ -628,6 +628,7 @@ DEF_GCC_BUILTIN        (BUILT_IN_AGGREGA
 DEF_EXT_LIB_BUILTIN    (BUILT_IN_ALLOCA, "alloca", BT_FN_PTR_SIZE, ATTR_MALLOC_NOTHROW_LEAF_LIST)
 DEF_GCC_BUILTIN        (BUILT_IN_APPLY, "apply", BT_FN_PTR_PTR_FN_VOID_VAR_PTR_SIZE, ATTR_NULL)
 DEF_GCC_BUILTIN        (BUILT_IN_APPLY_ARGS, "apply_args", BT_FN_PTR_VAR, ATTR_LEAF_LIST)
+DEF_GCC_BUILTIN        (BUILT_IN_BSWAP16, "bswap16", BT_FN_UINT16_UINT16, ATTR_CONST_NOTHROW_LEAF_LIST)
 DEF_GCC_BUILTIN        (BUILT_IN_BSWAP32, "bswap32", BT_FN_UINT32_UINT32, ATTR_CONST_NOTHROW_LEAF_LIST)
 DEF_GCC_BUILTIN        (BUILT_IN_BSWAP64, "bswap64", BT_FN_UINT64_UINT64, ATTR_CONST_NOTHROW_LEAF_LIST)
 DEF_EXT_LIB_BUILTIN    (BUILT_IN_CLEAR_CACHE, "__clear_cache", BT_FN_VOID_PTR_PTR, ATTR_NOTHROW_LEAF_LIST)