Add alternate names for x86 SIMD intrinsics. These aren't as common, but show

up enough to be worth supporting properly. Fixes PR7674.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@109086 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Headers/mmintrin.h b/lib/Headers/mmintrin.h
index 401d8a7aaededeec96ed68e7ed23e13e3514d031..a44b63d1a48d9c60d128b214738f5c3c1fce30a7 100644 (file)
--- a/lib/Headers/mmintrin.h
+++ b/lib/Headers/mmintrin.h
@@ -443,6 +443,64 @@ _mm_setr_pi8(char __b7, char __b6, char __b5, char __b4, char __b3, char __b2,
     return (__m64)(__v8qi){ __b7, __b6, __b5, __b4, __b3, __b2, __b1, __b0 };
 }
 
+
+/* Aliases for compatibility. */
+#define _m_empty _mm_empty
+#define _m_from_int _mm_cvtsi32_si64
+#define _m_to_int _mm_cvtsi64_si32
+#define _m_packsswb _mm_packs_pi16
+#define _m_packssdw _mm_packs_pi32
+#define _m_packuswb _mm_packs_pu16
+#define _m_punpckhbw _mm_unpackhi_pi8
+#define _m_punpckhwd _mm_unpackhi_pi16
+#define _m_punpckhdq _mm_unpackhi_pi32
+#define _m_punpcklbw _mm_unpacklo_pi8
+#define _m_punpcklwd _mm_unpacklo_pi16
+#define _m_punpckldq _mm_unpacklo_pi32
+#define _m_paddb _mm_add_pi8
+#define _m_paddw _mm_add_pi16
+#define _m_paddd _mm_add_pi32
+#define _m_paddsb _mm_adds_pi8
+#define _m_paddsw _mm_adds_pi16
+#define _m_paddusb _mm_adds_pu8
+#define _m_paddusw _mm_adds_pu16
+#define _m_psubb _mm_sub_pi8
+#define _m_psubw _mm_sub_pi16
+#define _m_psubd _mm_sub_pi32
+#define _m_psubsb _mm_subs_pi8
+#define _m_psubsw _mm_subs_pi16
+#define _m_psubusb _mm_subs_pu8
+#define _m_psubusw _mm_subs_pu16
+#define _m_pmaddwd _mm_madd_pi16
+#define _m_pmulhw _mm_mulhi_pi16
+#define _m_pmullw _mm_mullo_pi16
+#define _m_psllw _mm_sll_pi16
+#define _m_psllwi _mm_slli_pi16
+#define _m_pslld _mm_sll_pi32
+#define _m_pslldi _mm_slli_pi32
+#define _m_psllq _mm_sll_pi64
+#define _m_psllqi _mm_slli_pi64
+#define _m_psraw _mm_sra_pi16
+#define _m_psrawi _mm_srai_pi16
+#define _m_psrad _mm_sra_pi32
+#define _m_psradi _mm_srai_pi32
+#define _m_psrlw _mm_srl_pi16
+#define _m_psrlwi _mm_srli_pi16
+#define _m_psrld _mm_srl_pi32
+#define _m_psrldi _mm_srli_pi32
+#define _m_psrlq _mm_srl_pi64
+#define _m_psrlqi _mm_srli_pi64
+#define _m_pand _mm_and_si64
+#define _m_pandn _mm_andnot_si64
+#define _m_por _mm_or_si64
+#define _m_pxor _mm_xor_si64
+#define _m_pcmpeqb _mm_cmpeq_pi8
+#define _m_pcmpeqw _mm_cmpeq_pi16
+#define _m_pcmpeqd _mm_cmpeq_pi32
+#define _m_pcmpgtb _mm_cmpgt_pi8
+#define _m_pcmpgtw _mm_cmpgt_pi16
+#define _m_pcmpgtd _mm_cmpgt_pi32
+
 #endif /* __MMX__ */
 
 #endif /* __MMINTRIN_H */

diff --git a/lib/Headers/xmmintrin.h b/lib/Headers/xmmintrin.h
index 75e06b5f975470e5347b6b9c98490e0dd4d5f1b0..f0d252c4566930a3b47f6e0a4416220e1830084f 100644 (file)
--- a/lib/Headers/xmmintrin.h
+++ b/lib/Headers/xmmintrin.h
@@ -416,6 +416,12 @@ _mm_cvtps_pi32(__m128 a)
   return (__m64)__builtin_ia32_cvtps2pi(a);
 }
 
+static __inline__ __m64 __attribute__((__always_inline__, __nodebug__))
+_mm_cvt_ps2pi(__m128 a)
+{
+  return _mm_cvtps_pi32(a);
+}
+
 static __inline__ int __attribute__((__always_inline__, __nodebug__))
 _mm_cvttss_si32(__m128 a)
 {
@@ -440,6 +446,12 @@ _mm_cvttps_pi32(__m128 a)
   return (__m64)__builtin_ia32_cvttps2pi(a);
 }
 
+static __inline__ __m64 __attribute__((__always_inline__, __nodebug__))
+_mm_cvtt_ps2pi(__m128 a)
+{
+  return _mm_cvttps_pi32(a);
+}
+
 static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
 _mm_cvtsi32_ss(__m128 a, int b)
 {
@@ -447,6 +459,12 @@ _mm_cvtsi32_ss(__m128 a, int b)
   return a;
 }
 
+static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
+_mm_cvtt_si2ss(__m128 a, int b)
+{
+  return _mm_cvtsi32_ss(a, b);
+}
+
 #ifdef __x86_64__
 
 static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
@@ -464,6 +482,12 @@ _mm_cvtpi32_ps(__m128 a, __m64 b)
   return __builtin_ia32_cvtpi2ps(a, (__v2si)b);
 }
 
+static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
+_mm_cvtt_pi2ps(__m128 a, __m64 b)
+{
+  return _mm_cvtpi32_ps(a, b);
+}
+
 static __inline__ float __attribute__((__always_inline__, __nodebug__))
 _mm_cvtss_f32(__m128 a)
 {
@@ -589,6 +613,12 @@ _mm_store1_ps(float *p, __m128 a)
   _mm_storeu_ps(p, a);
 }
 
+static __inline__ void __attribute__((__always_inline__, __nodebug__))
+_mm_store_ps1(float *p, __m128 a)
+{
+    return _mm_store1_ps(p, a);
+}
+
 static __inline__ void __attribute__((__always_inline__, __nodebug__))
 _mm_store_ps(float *p, __m128 a)
 {
@@ -908,6 +938,23 @@ do { \
   (row3) = _mm_movehl_ps(tmp3, tmp1); \
 } while (0)
 
+/* Aliases for compatibility. */
+#define _m_pextrw _mm_extract_pi16
+#define _m_pinsrw _mm_insert_pi16
+#define _m_pmaxsw _mm_max_pi16
+#define _m_pmaxub _mm_max_pu8
+#define _m_pminsw _mm_min_pi16
+#define _m_pminub _mm_min_pu8
+#define _m_pmovmskb _mm_movemask_pi8
+#define _m_pmulhuw _mm_mulhi_pu16
+#define _m_pshufw _mm_shuffle_pi16
+#define _m_maskmovq _mm_maskmove_si64
+#define _m_pavgb _mm_avg_pu8
+#define _m_pavgw _mm_avg_pu16
+#define _m_psadbw _mm_sad_pu8
+#define _m_ _mm_
+#define _m_ _mm_
+
 /* Ugly hack for backwards-compatibility (compatible with gcc) */
 #ifdef __SSE2__
 #include <emmintrin.h>