x86: enable __float128 also in 32-bit mode

["Jan Beulich" <JBeulich at novell dot com>]

It is not obvious why __float128 support, requiring XMM registers, was
designated for x86-64 only (but still partly enabled for x86). This
patch brings it to the same (incomplete) level of support for both 32
and
64 bits.

bootstrapped and tested on i686-pc-linux-gnu.

Jan

2004-07-08 Jan Beulich <jbeulich@novell.com>

	* config/i386/i386.c (ix86_function_regparm): Don't call
	cgraph_local_info for libcalls, which dont provide a
FUNCTION_DECL.
	(function_arg): Also pass TFmode through XMM registers.
	(ix86_return_in_memory): Don't return TFmode in memory when SSE
is
	enabled.
	(ix86_value_regno): Consolidate check for returning in xmm0, and
move
	it ahead of the check for floating point modes.
	(ix86_init_mmx_sse_builtins): Set up __float128 only when SSE
is
	enabled.
	* config/i386/i386.md (movtf): Also allow in 32-bit mode when
SSE is
	enabled.
	(*movtf_internal): New.

---
/home/jbeulich/src/gcc/mainline/2004-07-05.10.09/gcc/config/i386/i386.c	2004-07-02
15:20:42.000000000 +0200
+++ 2004-07-05.10.09/gcc/config/i386/i386.c	2004-07-07
15:39:12.884303648 +0200
@@ -1777,7 +1777,8 @@
 	}
 
       /* Use register calling convention for local functions when
possible.  */
-      if (!TARGET_64BIT && !user_convention && decl
+      if (!TARGET_64BIT && !user_convention
+	  && decl && TREE_CODE (decl) == FUNCTION_DECL
 	  && flag_unit_at_a_time && !profile_flag)
 	{
 	  struct cgraph_local_info *i = cgraph_local_info (decl);
@@ -2663,6 +2664,7 @@
 	  }
 	break;
       case TImode:
+      case TFmode:
       case V16QImode:
       case V8HImode:
       case V4SImode:
@@ -2891,7 +2893,7 @@
   if (MS_AGGREGATE_RETURN && AGGREGATE_TYPE_P (type) && size <= 8)
     return 0;
 
-  if (VECTOR_MODE_P (mode) || mode == TImode)
+  if (VECTOR_MODE_P (mode) || mode == TImode || mode == TFmode)
     {
       /* User-created vectors small enough to fit in EAX.  */
       if (size < 8)
@@ -2980,13 +2982,13 @@
 static int
 ix86_value_regno (enum machine_mode mode)
 {
-  /* Floating point return values in %st(0).  */
-  if (GET_MODE_CLASS (mode) == MODE_FLOAT &&
TARGET_FLOAT_RETURNS_IN_80387)
-    return FIRST_FLOAT_REG;
   /* 16-byte vector modes in %xmm0.  See ix86_return_in_memory for
where
      we prevent this case when sse is not available.  */
-  if (mode == TImode || (VECTOR_MODE_P (mode) && GET_MODE_SIZE (mode)
== 16))
+  if (SSE_REG_MODE_P (mode))
     return FIRST_SSE_REG;
+  /* Floating point return values in %st(0).  */
+  if (GET_MODE_CLASS (mode) == MODE_FLOAT &&
TARGET_FLOAT_RETURNS_IN_80387)
+    return FIRST_FLOAT_REG;
   /* Everything else in %eax.  */
   return 0;
 }
@@ -13317,9 +13319,6 @@
   tree v2di_ftype_v2di
     = build_function_type_list (V2DI_type_node, V2DI_type_node,
NULL_TREE);
 
-  tree float80_type;
-  tree float128_type;
-
   /* The __float80 type.  */
   if (TYPE_MODE (long_double_type_node) == XFmode)
     (*lang_hooks.types.register_builtin_type) (long_double_type_node,
@@ -13327,16 +13326,21 @@
   else
     {
       /* The __float80 type.  */
-      float80_type = make_node (REAL_TYPE);
+      tree float80_type = make_node (REAL_TYPE);
+
       TYPE_PRECISION (float80_type) = 96;
       layout_type (float80_type);
       (*lang_hooks.types.register_builtin_type) (float80_type,
"__float80");
     }
 
-  float128_type = make_node (REAL_TYPE);
-  TYPE_PRECISION (float128_type) = 128;
-  layout_type (float128_type);
-  (*lang_hooks.types.register_builtin_type) (float128_type,
"__float128");
+  if (TARGET_SSE)
+    {
+      tree float128_type = make_node (REAL_TYPE);
+
+      TYPE_PRECISION (float128_type) = 128;
+      layout_type (float128_type);
+      (*lang_hooks.types.register_builtin_type) (float128_type,
"__float128");
+    }
 
   /* Add all builtins that are more or less simple operations on two
      operands.  */
---
/home/jbeulich/src/gcc/mainline/2004-07-05.10.09/gcc/config/i386/i386.md	2004-07-02
15:20:47.000000000 +0200
+++ 2004-07-05.10.09/gcc/config/i386/i386.md	2004-07-07
15:39:12.932296352 +0200
@@ -19763,7 +19763,7 @@
 (define_expand "movtf"
   [(set (match_operand:TF 0 "nonimmediate_operand" "")
 	(match_operand:TF 1 "nonimmediate_operand" ""))]
-  "TARGET_64BIT"
+  "TARGET_SSE || TARGET_64BIT"
 {
   if (TARGET_64BIT)
     ix86_expand_move (TFmode, operands);
@@ -20157,6 +20157,37 @@
 		   (const_string "TI"))]
 	       (const_string "DI")))])
 
+(define_insn "*movtf_internal"
+  [(set (match_operand:TF 0 "nonimmediate_operand" "=x,x,m")
+	(match_operand:TF 1 "vector_move_operand" "C,xm,x"))]
+  "TARGET_SSE && !TARGET_64BIT
+   && (GET_CODE (operands[0]) != MEM || GET_CODE (operands[1]) !=
MEM)"
+{
+  switch (which_alternative)
+    {
+    case 0:
+      if (get_attr_mode (insn) == MODE_V4SF)
+	return "xorps\t%0, %0";
+      else
+	return "pxor\t%0, %0";
+    case 1:
+    case 2:
+      if (get_attr_mode (insn) == MODE_V4SF)
+	return "movaps\t{%1, %0|%0, %1}";
+      else
+	return "movdqa\t{%1, %0|%0, %1}";
+    default:
+      abort ();
+    }
+}
+  [(set_attr "type" "ssemov,ssemov,ssemov")
+   (set (attr "mode")
+	(if_then_else
+	  (ior (ne (symbol_ref "optimize_size") (const_int 0))
+	       (eq (symbol_ref "TARGET_SSE2") (const_int 0)))
+	  (const_string "V4SF")
+	  (const_string "TI")))])
+
 (define_insn "*movtf_rex64"
   [(set (match_operand:TF 0 "nonimmediate_operand" "=r,o,x,x,xm")
 	(match_operand:TF 1 "general_operand" "riFo,riF,C,xm,x"))]

Attachment: gcc-mainline-x86-TFmode.patch
Description: Binary data