i386.c (ix86_special_builtin_type): New.

2008-05-02  H.J. Lu  <hongjiu.lu@intel.com>

	* config/i386/i386.c (ix86_special_builtin_type): New.
	(bdesc_special_args): Likewise.
	(ix86_expand_special_args_builtin): Likewise.
	(ix86_init_mmx_sse_builtins): Updated.
	(ix86_expand_builtin): Updated.
	(ix86_expand_store_builtin): Removed.
	(ix86_expand_unop_builtin): Likewise.

	* config/i386/mm3dnow.h (__v2sf): Moved to ...
	* config/i386/mmintrin.h (__v2sf): Here.

	* config/i386/xmmintrin.h (_mm_loadh_pi): Replace __v2si with
	const __v2sf.
	(_mm_loadl_pi): Likewise.
	(_mm_storeh_pi): Replace __v2si with __v2sf.
	(_mm_storel_pi): Likewise.

	* doc/extend.texi: Correct __builtin_ia32_loadhps,
	__builtin_ia32_loadlps, __builtin_ia32_storehps,
	__builtin_ia32_storelps, __builtin_ia32_loadhpd and
	__builtin_ia32_loadlpd.

2008-05-02  H.J. Lu  <hongjiu.lu@intel.com>

	* config/i386/i386.c (ix86_builtin_type): Add FLOAT_FTYPE_FLOAT,
	V4SF_FTYPE_V4SF_VEC_MERGE and V2DF_FTYPE_V2DF_VEC_MERGE.
	(bdesc_args): Updated.  Add scalar SSE builtins with vec_merge.
	(ix86_init_mmx_sse_builtins): Updated.
	(ix86_expand_args_builtin): Likewise.
	(ix86_expand_builtin): Likewise.
	(ix86_expand_unop1_builtin): Renamed to ...
	(ix86_expand_unop_vec_merge_builtin): This.

From-SVN: r134886

gcc/ChangeLog

+2008-05-02  H.J. Lu  <hongjiu.lu@intel.com>
+
+	* config/i386/i386.c (ix86_special_builtin_type): New.
+	(bdesc_special_args): Likewise.
+	(ix86_expand_special_args_builtin): Likewise.
+	(ix86_init_mmx_sse_builtins): Updated.
+	(ix86_expand_builtin): Updated.
+	(ix86_expand_store_builtin): Removed.
+	(ix86_expand_unop_builtin): Likewise.
+
+	* config/i386/mm3dnow.h (__v2sf): Moved to ...
+	* config/i386/mmintrin.h (__v2sf): Here.
+
+	* config/i386/xmmintrin.h (_mm_loadh_pi): Replace __v2si with
+	const __v2sf.
+	(_mm_loadl_pi): Likewise.
+	(_mm_storeh_pi): Replace __v2si with __v2sf.
+	(_mm_storel_pi): Likewise.
+
+	* doc/extend.texi: Correct __builtin_ia32_loadhps,
+	__builtin_ia32_loadlps, __builtin_ia32_storehps,
+	__builtin_ia32_storelps, __builtin_ia32_loadhpd and
+	__builtin_ia32_loadlpd.
+
+2008-05-02  H.J. Lu  <hongjiu.lu@intel.com>
+
+	* config/i386/i386.c (ix86_builtin_type): Add FLOAT_FTYPE_FLOAT,
+	V4SF_FTYPE_V4SF_VEC_MERGE and V2DF_FTYPE_V2DF_VEC_MERGE.
+	(bdesc_args): Updated.  Add scalar SSE builtins with vec_merge.
+	(ix86_init_mmx_sse_builtins): Updated.
+	(ix86_expand_args_builtin): Likewise.
+	(ix86_expand_builtin): Likewise.
+	(ix86_expand_unop1_builtin): Renamed to ...
+	(ix86_expand_unop_vec_merge_builtin): This.
+
 2008-05-01  Jan Hubicka  <jh@suse.cz>
 
 	PR bootstrap/36100

gcc/config/i386/mm3dnow.h

@@ -34,9 +34,6 @@
 
 #include <mmintrin.h>
 
-/* Internal data types for implementing the intrinsics.  */
-typedef float __v2sf __attribute__ ((__vector_size__ (8)));
-
 extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _m_femms (void)
 {

gcc/config/i386/mmintrin.h

@@ -43,6 +43,7 @@ typedef int __v2si __attribute__ ((__vector_size__ (8)));
 typedef short __v4hi __attribute__ ((__vector_size__ (8)));
 typedef char __v8qi __attribute__ ((__vector_size__ (8)));
 typedef long long __v1di __attribute__ ((__vector_size__ (8)));
+typedef float __v2sf __attribute__ ((__vector_size__ (8)));
 
 /* Empty the multimedia state.  */
 extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))

gcc/config/i386/xmmintrin.h

@@ -745,14 +745,14 @@ _mm_unpacklo_ps (__m128 __A, __m128 __B)
 extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _mm_loadh_pi (__m128 __A, __m64 const *__P)
 {
-  return (__m128) __builtin_ia32_loadhps ((__v4sf)__A, (__v2si *)__P);
+  return (__m128) __builtin_ia32_loadhps ((__v4sf)__A, (const __v2sf *)__P);
 }
 
 /* Stores the upper two SPFP values of A into P.  */
 extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _mm_storeh_pi (__m64 *__P, __m128 __A)
 {
-  __builtin_ia32_storehps ((__v2si *)__P, (__v4sf)__A);
+  __builtin_ia32_storehps ((__v2sf *)__P, (__v4sf)__A);
 }
 
 /* Moves the upper two values of B into the lower two values of A.  */
@@ -774,14 +774,14 @@ _mm_movelh_ps (__m128 __A, __m128 __B)
 extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _mm_loadl_pi (__m128 __A, __m64 const *__P)
 {
-  return (__m128) __builtin_ia32_loadlps ((__v4sf)__A, (__v2si *)__P);
+  return (__m128) __builtin_ia32_loadlps ((__v4sf)__A, (const __v2sf *)__P);
 }
 
 /* Stores the lower two SPFP values of A into P.  */
 extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _mm_storel_pi (__m64 *__P, __m128 __A)
 {
-  __builtin_ia32_storelps ((__v2si *)__P, (__v4sf)__A);
+  __builtin_ia32_storelps ((__v2sf *)__P, (__v4sf)__A);
 }
 
 /* Creates a 4-bit mask from the most significant bits of the SPFP values.  */

gcc/doc/extend.texi

@@ -7647,13 +7647,13 @@ Generates the @code{movups} machine instruction as a store to memory.
 Generates the @code{movss} machine instruction as a load from memory.
 @item void __builtin_ia32_storess (float *, v4sf)
 Generates the @code{movss} machine instruction as a store to memory.
-@item v4sf __builtin_ia32_loadhps (v4sf, v2si *)
+@item v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)
 Generates the @code{movhps} machine instruction as a load from memory.
-@item v4sf __builtin_ia32_loadlps (v4sf, v2si *)
+@item v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)
 Generates the @code{movlps} machine instruction as a load from memory
-@item void __builtin_ia32_storehps (v4sf, v2si *)
+@item void __builtin_ia32_storehps (v2sf *, v4sf)
 Generates the @code{movhps} machine instruction as a store to memory.
-@item void __builtin_ia32_storelps (v4sf, v2si *)
+@item void __builtin_ia32_storelps (v2sf *, v4sf)
 Generates the @code{movlps} machine instruction as a store to memory.
 @end table
 
@@ -7755,8 +7755,8 @@ v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi)
 void __builtin_ia32_maskmovdqu (v16qi, v16qi)
 v2df __builtin_ia32_loadupd (double *)
 void __builtin_ia32_storeupd (double *, v2df)
-v2df __builtin_ia32_loadhpd (v2df, double *)
-v2df __builtin_ia32_loadlpd (v2df, double *)
+v2df __builtin_ia32_loadhpd (v2df, double const *)
+v2df __builtin_ia32_loadlpd (v2df, double const *)
 int __builtin_ia32_movmskpd (v2df)
 int __builtin_ia32_pmovmskb128 (v16qi)
 void __builtin_ia32_movnti (int *, int)