+2006-05-20 Uros Bizjak <ubizjak@gmail.com>
+
+ * config/i386/tmmintrin.h (_mm_alignr_epi32): Implement as always
+ inlined function, not as a macro.
+ (_mm_alignr_pi8): Ditto.
+ * config/i386/ammintrin.h (_mm_extracti_si64): Ditto.
+ (_mm_inserti_si64): Ditto.
+ * config/i386/emmintrin.h (_mm_shuffle_pd): Ditto.
+ (_mm_extract_epi16): Ditto.
+ (_mm_insert_epi16): Ditto.
+ (_mm_shufflehi_epi16): Ditto.
+ (_mm_shufflelo_epi16): Ditto.
+ (_mm_shuffle_epi32): Ditto.
+ * config/i386/xmmintrin.h (_mm_set_ss): Use 0.0f for float constant.
+ * config/386/mm3dnow.h: Add __attribute__((__always_inline__)) to
+ all functions.
+ (_m_from_float): Add __extension__ to conversion. Use 0.0f for
+ float constant.
+ (_m_to_float): Use C89 compatible assignment.
+
2007-04-20 Martin Michlmayr <tbm@cyrius.com>
PR target/32007
return (__m128i) __builtin_ia32_extrq ((__v2di) __X, (__v16qi) __Y);
}
-#define _mm_extracti_si64(X, I, L) \
-((__m128i) __builtin_ia32_extrqi ((__v2di)(X), I, L))
+static __inline __m128i __attribute__((__always_inline__))
+_mm_extracti_si64 (__m128i __X, unsigned const int __I, unsigned const int __L)
+{
+ return (__m128i) __builtin_ia32_extrqi ((__v2di) __X, __I, __L);
+}
static __inline __m128i __attribute__((__always_inline__))
_mm_insert_si64 (__m128i __X,__m128i __Y)
return (__m128i) __builtin_ia32_insertq ((__v2di)__X, (__v2di)__Y);
}
-#define _mm_inserti_si64(X, Y, I, L) \
-((__m128i) __builtin_ia32_insertqi ((__v2di)(X), (__v2di)(Y), I, L))
+static __inline __m128i __attribute__((__always_inline__))
+_mm_inserti_si64(__m128i __X, __m128i __Y, unsigned const int __I, unsigned const int __L)
+{
+ return (__m128i) __builtin_ia32_insertqi ((__v2di)__X, (__v2di)__Y, __I, __L);
+}
#endif /* __SSE4A__ */
static __inline __m128d __attribute__((__always_inline__))
_mm_set_sd (double __F)
{
- return __extension__ (__m128d){ __F, 0 };
+ return __extension__ (__m128d){ __F, 0.0 };
}
/* Create a vector with both elements equal to F. */
return (__m128d)__builtin_ia32_cvtss2sd ((__v2df) __A, (__v4sf)__B);
}
-#define _mm_shuffle_pd(__A, __B, __C) ((__m128d)__builtin_ia32_shufpd ((__v2df)__A, (__v2df)__B, (__C)))
+static __inline __m128d __attribute__((__always_inline__))
+_mm_shuffle_pd(__m128d __A, __m128d __B, const int __mask)
+{
+ return (__m128d)__builtin_ia32_shufpd ((__v2df)__A, (__v2df)__B, __mask);
+}
static __inline __m128d __attribute__((__always_inline__))
_mm_unpackhi_pd (__m128d __A, __m128d __B)
static __inline __m128i __attribute__((__always_inline__))
_mm_srli_si128 (__m128i __A, const int __B)
{
- return ((__m128i)__builtin_ia32_psrldqi128 (__A, __B * 8));
+ return (__m128i)__builtin_ia32_psrldqi128 (__A, __B * 8);
}
static __inline __m128i __attribute__((__always_inline__))
_mm_slli_si128 (__m128i __A, const int __B)
{
- return ((__m128i)__builtin_ia32_pslldqi128 (__A, __B * 8));
+ return (__m128i)__builtin_ia32_pslldqi128 (__A, __B * 8);
}
static __inline __m128i __attribute__((__always_inline__))
return (__m128i)__builtin_ia32_pcmpgtd128 ((__v4si)__A, (__v4si)__B);
}
-#if 0
static __inline int __attribute__((__always_inline__))
_mm_extract_epi16 (__m128i const __A, int const __N)
{
{
return (__m128i) __builtin_ia32_vec_set_v8hi ((__v8hi)__A, __D, __N);
}
-#else
-#define _mm_extract_epi16(A, N) \
- ((int) __builtin_ia32_vec_ext_v8hi ((__v8hi)(A), (N)))
-#define _mm_insert_epi16(A, D, N) \
- ((__m128i) __builtin_ia32_vec_set_v8hi ((__v8hi)(A), (D), (N)))
-#endif
static __inline __m128i __attribute__((__always_inline__))
_mm_max_epi16 (__m128i __A, __m128i __B)
return (__m128i)__builtin_ia32_pmulhuw128 ((__v8hi)__A, (__v8hi)__B);
}
-#define _mm_shufflehi_epi16(__A, __B) ((__m128i)__builtin_ia32_pshufhw ((__v8hi)__A, __B))
-#define _mm_shufflelo_epi16(__A, __B) ((__m128i)__builtin_ia32_pshuflw ((__v8hi)__A, __B))
-#define _mm_shuffle_epi32(__A, __B) ((__m128i)__builtin_ia32_pshufd ((__v4si)__A, __B))
+static __inline __m128i __attribute__((__always_inline__))
+_mm_shufflehi_epi16 (__m128i __A, const int __mask)
+{
+ return (__m128i)__builtin_ia32_pshufhw ((__v8hi)__A, __mask);
+}
+
+static __inline __m128i __attribute__((__always_inline__))
+_mm_shufflelo_epi16 (__m128i __A, const int __mask)
+{
+ return (__m128i)__builtin_ia32_pshuflw ((__v8hi)__A, __mask);
+}
+
+static __inline __m128i __attribute__((__always_inline__))
+_mm_shuffle_epi32 (__m128i __A, const int __mask)
+{
+ return (__m128i)__builtin_ia32_pshufd ((__v4si)__A, __mask);
+}
static __inline void __attribute__((__always_inline__))
_mm_maskmoveu_si128 (__m128i __A, __m128i __B, char *__C)
/* Internal data types for implementing the intrinsics. */
typedef float __v2sf __attribute__ ((__vector_size__ (8)));
-static __inline void
+static __inline void __attribute__((__always_inline__))
_m_femms (void)
{
__builtin_ia32_femms();
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pavgusb (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pavgusb ((__v8qi)__A, (__v8qi)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pf2id (__m64 __A)
{
return (__m64)__builtin_ia32_pf2id ((__v2sf)__A);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfacc (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfacc ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfadd (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfadd ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfcmpeq (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfcmpeq ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfcmpge (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfcmpge ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfcmpgt (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfcmpgt ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfmax (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfmax ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfmin (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfmin ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfmul (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfmul ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfrcp (__m64 __A)
{
return (__m64)__builtin_ia32_pfrcp ((__v2sf)__A);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfrcpit1 (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfrcpit1 ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfrcpit2 (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfrcpit2 ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfrsqrt (__m64 __A)
{
return (__m64)__builtin_ia32_pfrsqrt ((__v2sf)__A);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfrsqit1 (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfrsqit1 ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfsub (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfsub ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfsubr (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfsubr ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pi2fd (__m64 __A)
{
return (__m64)__builtin_ia32_pi2fd ((__v2si)__A);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pmulhrw (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pmulhrw ((__v4hi)__A, (__v4hi)__B);
}
-static __inline void
+static __inline void __attribute__((__always_inline__))
_m_prefetch (void *__P)
{
__builtin_prefetch (__P, 0, 3 /* _MM_HINT_T0 */);
}
-static __inline void
+static __inline void __attribute__((__always_inline__))
_m_prefetchw (void *__P)
{
__builtin_prefetch (__P, 1, 3 /* _MM_HINT_T0 */);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_from_float (float __A)
{
- return (__m64)(__v2sf){ __A, 0 };
+ return __extension__ (__m64)(__v2sf){ __A, 0.0f };
}
-static __inline float
+static __inline float __attribute__((__always_inline__))
_m_to_float (__m64 __A)
{
- union { __v2sf v; float a[2]; } __tmp = { (__v2sf)__A };
+ union { __v2sf v; float a[2]; } __tmp;
+ __tmp.v = (__v2sf)__A;
return __tmp.a[0];
}
#ifdef __3dNOW_A__
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pf2iw (__m64 __A)
{
return (__m64)__builtin_ia32_pf2iw ((__v2sf)__A);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfnacc (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfnacc ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pfpnacc (__m64 __A, __m64 __B)
{
return (__m64)__builtin_ia32_pfpnacc ((__v2sf)__A, (__v2sf)__B);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pi2fw (__m64 __A)
{
return (__m64)__builtin_ia32_pi2fw ((__v2si)__A);
}
-static __inline __m64
+static __inline __m64 __attribute__((__always_inline__))
_m_pswapd (__m64 __A)
{
return (__m64)__builtin_ia32_pswapdsf ((__v2sf)__A);
return (__m64) __builtin_ia32_psignd ((__v2si)__X, (__v2si)__Y);
}
-#define _mm_alignr_epi8(__X, __Y, __N) \
- ((__m128i)__builtin_ia32_palignr128 ((__v2di) __X, (__v2di) __Y, (__N) * 8))
+static __inline __m128i __attribute__((__always_inline__))
+_mm_alignr_epi8(__m128i __X, __m128i __Y, const int __N)
+{
+ return (__m128i)__builtin_ia32_palignr128 ((__v2di)__X, (__v2di)__Y, __N * 8);}
-#define _mm_alignr_pi8(__X, __Y, __N) \
- ((__m64)__builtin_ia32_palignr ((long long) (__X), (long long) (__Y), (__N) * 8))
+static __inline __m64 __attribute__((__always_inline__))
+_mm_alignr_pi8(__m64 __X, __m64 __Y, const int __N)
+{
+ return (__m64)__builtin_ia32_palignr ((long long)__X, (long long)__Y, __N * 8);
+}
static __inline __m128i __attribute__((__always_inline__))
_mm_abs_epi8 (__m128i __X)
static __inline __m128 __attribute__((__always_inline__))
_mm_set_ss (float __F)
{
- return __extension__ (__m128)(__v4sf){ __F, 0, 0, 0 };
+ return __extension__ (__m128)(__v4sf){ __F, 0.0f, 0.0f, 0.0f };
}
/* Create a vector with all four elements equal to F. */
+2006-05-20 Uros Bizjak <ubizjak@gmail.com>
+
+ * gcc.target/i386/sse-vect-types.c: Revert 'Use "-msse"
+ instead of "-msse2".'
+ * gcc.target/i386/sse-12.c: Use "-march=k8 -m3dnow -mssse3 -msse4a"
+ instead of "-msse3". Include only ammintrin.h, tmmintrin.h
+ and mm3dnow.h.
+ * gcc.target/i386/sse-13.c (__builtin_ia32_extrqi): Redefine
+ to test with immediate operand.
+ (__builtin_ia32_insertqi): Ditto.
+ (__builtin_ia32_palignr128): Ditto.
+ (__builtin_ia32_palignr): Ditto.
+ (__builtin_ia32_pshufhw): Ditto.
+ (__builtin_ia32_pshuflw): Ditto.
+ (__builtin_ia32_pshufd): Ditto.
+ (__builtin_ia32_vec_set_v8hi): Ditto.
+ (__builtin_ia32_vec_ext_v8hi): Ditto.
+ (__builtin_ia32_shufpd): Ditto.
+ * gcc.target/i386/sse-14.c: Same changes as sse-13.c.
+
2007-05-20 Daniel Franke <franke.daniel@gmail.org>
PR fortran/32001
-/* Test that {,x,e,p}mmintrin.h and mm_malloc.h are
+/* Test that {,x,e,p,t,a}mmintrin.h, mm3dnow.h and mm_malloc.h are
usable with -std=c89 -pedantic-errors. */
/* { dg-do compile { target i?86-*-* x86_64-*-* } } */
-/* { dg-options "-std=c89 -pedantic-errors -msse3" } */
+/* { dg-options "-std=c89 -pedantic-errors -march=k8 -m3dnow -mssse3 -msse4a" } */
-#include <mmintrin.h>
-#include <xmmintrin.h>
-#include <emmintrin.h>
-#include <pmmintrin.h>
-#include <mm_malloc.h>
+#include <ammintrin.h>
+#include <tmmintrin.h>
+#include <mm3dnow.h>
int dummy;
/* Following intrinsics require immediate arguments. */
+/* ammintrin.h */
+#define __builtin_ia32_extrqi(X, I, L) __builtin_ia32_extrqi(X, 1, 1)
+#define __builtin_ia32_insertqi(X, Y, I, L) __builtin_ia32_insertqi(X, Y, 1, 1)
+
+/* tmmintrin.h */
+#define __builtin_ia32_palignr128(X, Y, N) __builtin_ia32_palignr128(X, Y, 8)
+#define __builtin_ia32_palignr(X, Y, N) __builtin_ia32_palignr(X, Y, 8)
+
/* emmintrin.h */
#define __builtin_ia32_psllwi128(A, B) __builtin_ia32_psllwi128(A, 1)
#define __builtin_ia32_psrlqi128(A, B) __builtin_ia32_psrlqi128(A, 1)
#define __builtin_ia32_psradi128(A, B) __builtin_ia32_psradi128(A, 1)
#define __builtin_ia32_psllqi128(A, B) __builtin_ia32_psllqi128(A, 1)
#define __builtin_ia32_pslldi128(A, B) __builtin_ia32_pslldi128(A, 1)
+#define __builtin_ia32_pshufhw(A, N) __builtin_ia32_pshufhw(A, 0)
+#define __builtin_ia32_pshuflw(A, N) __builtin_ia32_pshuflw(A, 0)
+#define __builtin_ia32_pshufd(A, N) __builtin_ia32_pshufd(A, 0)
+#define __builtin_ia32_vec_set_v8hi(A, D, N) \
+ __builtin_ia32_vec_set_v8hi(A, D, 0)
+#define __builtin_ia32_vec_ext_v8hi(A, N) __builtin_ia32_vec_ext_v8hi(A, 0)
+#define __builtin_ia32_shufpd(A, B, N) __builtin_ia32_shufpd(A, B, 0)
/* xmmintrin.h */
#define __builtin_prefetch(P, A, I) __builtin_prefetch(P, A, _MM_HINT_NTA)
#define __builtin_ia32_vec_set_v4hi(A, D, N) \
__builtin_ia32_vec_set_v4hi(A, D, 0)
#define __builtin_ia32_vec_ext_v4hi(A, N) __builtin_ia32_vec_ext_v4hi(A, 0)
-#define __builtin_ia32_shufps(A, B, C) __builtin_ia32_shufps(A, B, 0)
+#define __builtin_ia32_shufps(A, B, N) __builtin_ia32_shufps(A, B, 0)
#include <ammintrin.h>
#include <tmmintrin.h>
/* Following intrinsics require immediate arguments. */
+/* ammintrin.h */
+#define __builtin_ia32_extrqi(X, I, L) __builtin_ia32_extrqi(X, 1, 1)
+#define __builtin_ia32_insertqi(X, Y, I, L) __builtin_ia32_insertqi(X, Y, 1, 1)
+
+/* tmmintrin.h */
+#define __builtin_ia32_palignr128(X, Y, N) __builtin_ia32_palignr128(X, Y, 8)
+#define __builtin_ia32_palignr(X, Y, N) __builtin_ia32_palignr(X, Y, 8)
+
/* emmintrin.h */
#define __builtin_ia32_psllwi128(A, B) __builtin_ia32_psllwi128(A, 1)
#define __builtin_ia32_psrlqi128(A, B) __builtin_ia32_psrlqi128(A, 1)
#define __builtin_ia32_psradi128(A, B) __builtin_ia32_psradi128(A, 1)
#define __builtin_ia32_psllqi128(A, B) __builtin_ia32_psllqi128(A, 1)
#define __builtin_ia32_pslldi128(A, B) __builtin_ia32_pslldi128(A, 1)
+#define __builtin_ia32_pshufhw(A, N) __builtin_ia32_pshufhw(A, 0)
+#define __builtin_ia32_pshuflw(A, N) __builtin_ia32_pshuflw(A, 0)
+#define __builtin_ia32_pshufd(A, N) __builtin_ia32_pshufd(A, 0)
+#define __builtin_ia32_vec_set_v8hi(A, D, N) \
+ __builtin_ia32_vec_set_v8hi(A, D, 0)
+#define __builtin_ia32_vec_ext_v8hi(A, N) __builtin_ia32_vec_ext_v8hi(A, 0)
+#define __builtin_ia32_shufpd(A, B, N) __builtin_ia32_shufpd(A, B, 0)
/* xmmintrin.h */
#define __builtin_prefetch(P, A, I) __builtin_prefetch(P, A, _MM_HINT_NTA)
#define __builtin_ia32_vec_set_v4hi(A, D, N) \
__builtin_ia32_vec_set_v4hi(A, D, 0)
#define __builtin_ia32_vec_ext_v4hi(A, N) __builtin_ia32_vec_ext_v4hi(A, 0)
-#define __builtin_ia32_shufps(A, B, C) __builtin_ia32_shufps(A, B, 0)
+#define __builtin_ia32_shufps(A, B, N) __builtin_ia32_shufps(A, B, 0)
#include <ammintrin.h>
#include <tmmintrin.h>
/* { dg-do compile { target i?86-*-* x86_64-*-* } } */
-/* { dg-options "-O0 -msse" } */
+/* { dg-options "-O0 -msse2" } */
#include <xmmintrin.h>
projects / platform / upstream / gcc.git / commitdiff