[Clang][AVX512][BUILTIN] Adding intrinsics support to VEXTRACT{I|F} and VINSERT{I|F} instruction set

[delena]

include/clang/Basic/BuiltinsX86.def

1978 ↗

(On Diff #53664)

Why the predicate is empty?

[m_zuckerman]

Ping

[m_zuckerman]

ping

cfe/trunk/include/clang/Basic/BuiltinsX86.def

 TARGET_BUILTIN(__builtin_ia32_pmovqd128_mask, "V4iV2LLiV4iUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_pmovqd128mem_mask, "vV4i*V2LLiUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_pmovqd256_mask, "V4iV4LLiV4iUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_pmovqd256mem_mask, "vV4i*V4LLiUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_pmovqw128_mask, "V8sV2LLiV8sUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_pmovqw128mem_mask, "vV8s*V2LLiUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_pmovqw256_mask, "V8sV4LLiV8sUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_pmovqw256mem_mask, "vV8s*V4LLiUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_extractf32x8_mask, "V8fV16fIiV8fUc","","avx512dq")
+TARGET_BUILTIN(__builtin_ia32_extractf64x2_512_mask, "V2dV8dIiV2dUc","","avx512dq")
+TARGET_BUILTIN(__builtin_ia32_extracti32x8_mask, "V8iV16iIiV8iUc","","avx512dq")
+TARGET_BUILTIN(__builtin_ia32_extracti64x2_512_mask, "V2LLiV8LLiIiV2LLiUc","","avx512dq")
+TARGET_BUILTIN(__builtin_ia32_extracti32x4_mask, "V4iV16iIiV4iUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_extracti64x4_mask, "V4LLiV8LLiIiV4LLiUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_extractf64x2_256_mask, "V2dV4dIiV2dUc","","avx512dq,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_extracti64x2_256_mask, "V2LLiV4LLiIiV2LLiUc","","avx512dq,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_extractf32x4_256_mask, "V4fV8fIiV4fUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_extracti32x4_256_mask, "V4iV8iIiV4iUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_insertf64x2_512_mask, "V8dV8dV2dIiV8dUc","","avx512dq")
+TARGET_BUILTIN(__builtin_ia32_inserti32x8_mask, "V16iV16iV8iIiV16iUs","","avx512dq")
+TARGET_BUILTIN(__builtin_ia32_inserti64x2_512_mask, "V8LLiV8LLiV2LLiIiV8LLiUc","","avx512dq")
+TARGET_BUILTIN(__builtin_ia32_insertf64x4_mask, "V8dV8dV4dIiV8dUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_inserti64x4_mask, "V8LLiV8LLiV4LLiIiV8LLiUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_insertf64x2_256_mask, "V4dV4dV2dIiV4dUc","","avx512dq,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_inserti64x2_256_mask, "V4LLiV4LLiV2LLiIiV4LLiUc","","avx512dq,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_insertf32x4_256_mask, "V8fV8fV4fIiV8fUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_inserti32x4_256_mask, "V8iV8iV4iIiV8iUc","","avx512vl")
 
 #undef BUILTIN
 #undef TARGET_BUILTIN

cfe/trunk/lib/Headers/avx512dqintrin.h

 static __inline__ __m512i __DEFAULT_FN_ATTRS
 _mm512_maskz_broadcast_i64x2 (__mmask8 __M, __m128i __A)
 {
   return (__m512i) __builtin_ia32_broadcasti64x2_512_mask ((__v2di) __A,
                  (__v8di)_mm512_setzero_si512 (),
                  __M);
 }
 
+#define _mm512_extractf32x8_ps( __A, __imm) __extension__ ({ \
+__builtin_ia32_extractf32x8_mask ((__v16sf)( __A),\
+                ( __imm),\
+                (__v8sf) _mm256_setzero_ps (),\
+                (__mmask8) -1);\
+})
+
+#define _mm512_mask_extractf32x8_ps( __W, __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extractf32x8_mask ((__v16sf)( __A),\
+                ( __imm),\
+                (__v8sf)( __W),\
+                (__mmask8)( __U));\
+})
+
+#define _mm512_maskz_extractf32x8_ps( __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extractf32x8_mask ((__v16sf)( __A),\
+                ( __imm),\
+                (__v8sf) _mm256_setzero_ps (),\
+                (__mmask8)( __U));\
+})
+
+#define _mm512_extractf64x2_pd( __A, __imm) __extension__ ({ \
+__builtin_ia32_extractf64x2_512_mask ((__v8df)( __A),\
+               ( __imm),\
+               (__v2df) _mm_setzero_pd (),\
+               (__mmask8) -1);\
+})
+
+#define _mm512_mask_extractf64x2_pd( __W, __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extractf64x2_512_mask ((__v8df)( __A),\
+               ( __imm),\
+               (__v2df)( __W),\
+               (__mmask8) ( __U));\
+})
+
+#define _mm512_maskz_extractf64x2_pd( __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extractf64x2_512_mask ((__v8df)( __A),\
+               ( __imm),\
+               (__v2df) _mm_setzero_pd (),\
+               (__mmask8) ( __U));\
+})
+
+#define _mm512_extracti32x8_epi32( __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti32x8_mask ((__v16si)( __A),\
+                ( __imm),\
+                (__v8si) _mm256_setzero_si256 (),\
+                (__mmask8) -1);\
+})
+
+#define _mm512_mask_extracti32x8_epi32( __W, __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti32x8_mask ((__v16si)( __A),\
+                ( __imm),\
+                (__v8si)( __W),\
+                (__mmask8)( __U));\
+})
+
+#define _mm512_maskz_extracti32x8_epi32( __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti32x8_mask ((__v16si)( __A),\
+                ( __imm),\
+                (__v8si) _mm256_setzero_si256 (),\
+                (__mmask8)( __U));\
+})
+
+#define _mm512_extracti64x2_epi64( __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti64x2_512_mask ((__v8di)( __A),\
+                ( __imm),\
+                (__v2di) _mm_setzero_di (),\
+                (__mmask8) -1);\
+})
+
+#define _mm512_mask_extracti64x2_epi64( __W, __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti64x2_512_mask ((__v8di)( __A),\
+                ( __imm),\
+                (__v2di)( __W),\
+                (__mmask8) ( __U));\
+})
+
+#define _mm512_maskz_extracti64x2_epi64( __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti64x2_512_mask ((__v8di)( __A),\
+                ( __imm),\
+                (__v2di) _mm_setzero_di (),\
+                (__mmask8) ( __U));\
+})
+
+#define _mm512_insertf64x2( __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_insertf64x2_512_mask ((__v8df)( __A),\
+                (__v2df)( __B),\
+                ( __imm),\
+                (__v8df) _mm512_setzero_pd (),\
+                (__mmask8) -1);\
+})
+
+#define _mm512_mask_insertf64x2( __W, __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_insertf64x2_512_mask ((__v8df)( __A),\
+                (__v2df)( __B),\
+                ( __imm),\
+                (__v8df)( __W),\
+                (__mmask8) ( __U));\
+})
+
+#define _mm512_maskz_insertf64x2( __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_insertf64x2_512_mask ((__v8df)( __A),\
+                (__v2df)( __B),\
+                ( __imm),\
+                (__v8df) _mm512_setzero_pd (),\
+                (__mmask8) ( __U));\
+})
+
+#define _mm512_inserti32x8( __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti32x8_mask ((__v16si)( __A),\
+                (__v8si)( __B),\
+                ( __imm),\
+                (__v16si) _mm512_setzero_si512 (),\
+                (__mmask16) -1);\
+})
+
+#define _mm512_mask_inserti32x8( __W, __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti32x8_mask ((__v16si)( __A),\
+                (__v8si)( __B),\
+                ( __imm),\
+                (__v16si)( __W),\
+                (__mmask16)( __U));\
+})
+
+#define _mm512_maskz_inserti32x8( __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti32x8_mask ((__v16si)( __A),\
+                (__v8si)( __B),\
+                ( __imm),\
+                (__v16si) _mm512_setzero_si512 (),\
+                (__mmask16)( __U));\
+})
+
+#define _mm512_inserti64x2( __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti64x2_512_mask ((__v8di)( __A),\
+                (__v2di)( __B),\
+                ( __imm),\
+                (__v8di) _mm512_setzero_si512 (),\
+                (__mmask8) -1);\
+})
+
+#define _mm512_mask_inserti64x2( __W, __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti64x2_512_mask ((__v8di)( __A),\
+                (__v2di)( __B),\
+                ( __imm),\
+                (__v8di)( __W),\
+                (__mmask8) ( __U));\
+})
+
+#define _mm512_maskz_inserti64x2( __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti64x2_512_mask ((__v8di)( __A),\
+                (__v2di)( __B),\
+                ( __imm),\
+                (__v8di) _mm512_setzero_si512 (),\
+                (__mmask8) ( __U));\
+})
+
 #undef __DEFAULT_FN_ATTRS
 
 #endif

cfe/trunk/lib/Headers/avx512fintrin.h

 }
 
 static __inline__ void __DEFAULT_FN_ATTRS
 _mm512_mask_cvtepi64_storeu_epi16 (void *__P, __mmask8 __M, __m512i __A)
 {
   __builtin_ia32_pmovqw512mem_mask ((__v8hi *) __P, (__v8di) __A, __M);
 }
 
+#define _mm512_extracti32x4_epi32( __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti32x4_mask ((__v16si)( __A),\
+                 (__imm),\
+                 (__v4si) _mm_undefined_si128 (),\
+                 (__mmask8) -1);\
+})
+
+#define _mm512_mask_extracti32x4_epi32( __W, __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti32x4_mask ((__v16si)( __A),\
+                 ( __imm),\
+                 (__v4si)( __W),\
+                 (__mmask8)( __U));\
+})
+
+#define _mm512_maskz_extracti32x4_epi32( __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti32x4_mask ((__v16si)( __A),\
+                 ( __imm),\
+                 (__v4si) _mm_setzero_si128 (),\
+                 (__mmask8)( __U));\
+})
+
+#define _mm512_extracti64x4_epi64( __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti64x4_mask ((__v8di)( __A),\
+                 ( __imm),\
+                 (__v4di) _mm256_undefined_si256 (),\
+                 (__mmask8) -1);\
+})
+
+#define _mm512_mask_extracti64x4_epi64( __W, __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti64x4_mask ((__v8di)( __A),\
+                 ( __imm),\
+                 (__v4di)( __W),\
+                 (__mmask8)( __U));\
+})
+
+#define _mm512_maskz_extracti64x4_epi64( __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti64x4_mask ((__v8di)( __A),\
+                 ( __imm),\
+                 (__v4di) _mm256_setzero_si256 (),\
+                 (__mmask8)( __U));\
+})
+
+#define _mm512_insertf64x4( __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_insertf64x4_mask ((__v8df)( __A),\
+                (__v4df)( __B),\
+                ( __imm),\
+                (__v8df) _mm512_undefined_pd (),\
+                (__mmask8) -1);\
+})
+
+#define _mm512_mask_insertf64x4( __W, __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_insertf64x4_mask ((__v8df)( __A),\
+                (__v4df)( __B),\
+                ( __imm),\
+                (__v8df)( __W),\
+                (__mmask8)( __U));\
+})
+
+#define _mm512_maskz_insertf64x4( __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_insertf64x4_mask ((__v8df)( __A),\
+                (__v4df)( __B),\
+                ( __imm),\
+                (__v8df) _mm512_setzero_pd (),\
+                (__mmask8)( __U));\
+})
+
+#define _mm512_inserti64x4( __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti64x4_mask ((__v8di)( __A),\
+                (__v4di)( __B),\
+                ( __imm),\
+                (__v8di) _mm512_setzero_si512 (),\
+                (__mmask8) -1);\
+})
+
+#define _mm512_mask_inserti64x4( __W, __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti64x4_mask ((__v8di)( __A),\
+                (__v4di)( __B),\
+                ( __imm),\
+                (__v8di)( __W),\
+                (__mmask8)( __U));\
+})
+
+#define _mm512_maskz_inserti64x4( __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti64x4_mask ((__v8di)( __A),\
+                (__v4di)( __B),\
+                ( __imm),\
+                (__v8di) _mm512_setzero_si512 (),\
+                (__mmask8)( __U));\
+})
+
 #undef __DEFAULT_FN_ATTRS
 
 #endif // __AVX512FINTRIN_H

cfe/trunk/lib/Headers/avx512vldqintrin.h

 static __inline__ __m256i __DEFAULT_FN_ATTRS
 _mm256_maskz_broadcast_i64x2 (__mmask8 __M, __m128i __A)
 {
   return (__m256i) __builtin_ia32_broadcasti64x2_256_mask ((__v2di) __A,
                  (__v4di) _mm256_setzero_si256 (),
                  __M);
 }
 
+#define _mm256_extractf64x2_pd( __A, __imm) __extension__ ({ \
+__builtin_ia32_extractf64x2_256_mask ((__v4df)( __A),\
+               ( __imm),\
+               (__v2df) _mm_setzero_pd (),\
+               (__mmask8) -1);\
+})
+
+#define _mm256_mask_extractf64x2_pd( __W, __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extractf64x2_256_mask ((__v4df)( __A),\
+               ( __imm),\
+               (__v2df)( __W),\
+               (__mmask8) ( __U));\
+})
+
+#define _mm256_maskz_extractf64x2_pd( __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extractf64x2_256_mask ((__v4df)( __A),\
+               ( __imm),\
+               (__v2df) _mm_setzero_pd (),\
+               (__mmask8) ( __U));\
+})
+
+#define _mm256_extracti64x2_epi64( __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti64x2_256_mask ((__v4di)( __A),\
+               ( __imm),\
+               (__v2di) _mm_setzero_di (),\
+               (__mmask8) -1);\
+})
+
+#define _mm256_mask_extracti64x2_epi64( __W, __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti64x2_256_mask ((__v4di)( __A),\
+               ( __imm),\
+               (__v2di)( __W),\
+               (__mmask8) ( __U));\
+})
+
+#define _mm256_maskz_extracti64x2_epi64( __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti64x2_256_mask ((__v4di)( __A),\
+               ( __imm),\
+               (__v2di) _mm_setzero_di (),\
+               (__mmask8) ( __U));\
+})
+
+#define _mm256_insertf64x2( __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_insertf64x2_256_mask ((__v4df)( __A),\
+              (__v2df)( __B),\
+              ( __imm),\
+              (__v4df) _mm256_setzero_pd (),\
+              (__mmask8) -1);\
+})
+
+#define _mm256_mask_insertf64x2( __W, __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_insertf64x2_256_mask ((__v4df)( __A),\
+              (__v2df)( __B),\
+              ( __imm),\
+              (__v4df)( __W),\
+              (__mmask8) ( __U));\
+})
+
+#define _mm256_maskz_insertf64x2( __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_insertf64x2_256_mask ((__v4df)( __A),\
+              (__v2df)( __B),\
+              ( __imm),\
+              (__v4df) _mm256_setzero_pd (),\
+              (__mmask8) ( __U));\
+})
+
+#define _mm256_inserti64x2( __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti64x2_256_mask ((__v4di)( __A),\
+              (__v2di)( __B),\
+              ( __imm),\
+              (__v4di) _mm256_setzero_si256 (),\
+              (__mmask8) -1);\
+})
+
+#define _mm256_mask_inserti64x2( __W, __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti64x2_256_mask ((__v4di)( __A),\
+              (__v2di)( __B),\
+              ( __imm),\
+              (__v4di)( __W),\
+              (__mmask8) ( __U));\
+})
+
+#define _mm256_maskz_inserti64x2( __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti64x2_256_mask ((__v4di)( __A),\
+              (__v2di)( __B),\
+              ( __imm),\
+              (__v4di) _mm256_setzero_si256 (),\
+              (__mmask8) ( __U));\
+})
+
 #undef __DEFAULT_FN_ATTRS
 
 #endif

cfe/trunk/lib/Headers/avx512vlintrin.h

 }
 
 static __inline__ void __DEFAULT_FN_ATTRS
 _mm256_mask_cvtepi64_storeu_epi16 (void * __P, __mmask8 __M, __m256i __A)
 {
   __builtin_ia32_pmovqw256mem_mask ((__v8hi *) __P, (__v4di) __A, __M);
 }
 
+#define _mm256_extractf32x4_ps( __A, __imm) __extension__ ({ \
+__builtin_ia32_extractf32x4_256_mask ((__v8sf)( __A),\
+              ( __imm),\
+              (__v4sf) _mm_setzero_ps (),\
+              (__mmask8) -1);\
+})
+
+#define _mm256_mask_extractf32x4_ps( __W, __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extractf32x4_256_mask ((__v8sf)( __A),\
+                ( __imm),\
+                (__v4sf)( __W),\
+                (__mmask8) ( __U));\
+})
+
+#define _mm256_maskz_extractf32x4_ps( __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extractf32x4_256_mask ((__v8sf)( __A),\
+                ( __imm),\
+                (__v4sf) _mm_setzero_ps (),\
+                (__mmask8) ( __U));\
+})
+
+#define _mm256_extracti32x4_epi32( __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti32x4_256_mask ((__v8si)( __A),\
+                ( __imm),\
+                (__v4si) _mm_setzero_si128 (),\
+                (__mmask8) -1);\
+})
+
+#define _mm256_mask_extracti32x4_epi32( __W, __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti32x4_256_mask ((__v8si)( __A),\
+                ( __imm),\
+                (__v4si)( __W),\
+                (__mmask8)( __U));\
+})
+
+#define _mm256_maskz_extracti32x4_epi32( __U, __A, __imm) __extension__ ({ \
+__builtin_ia32_extracti32x4_256_mask ((__v8si)( __A),\
+               ( __imm),\
+               (__v4si) _mm_setzero_si128 (),\
+               (__mmask8) ( __U));\
+})
+
+#define _mm256_insertf32x4( __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_insertf32x4_256_mask ((__v8sf)( __A),\
+                (__v4sf)( __B),\
+                ( __imm),\
+                (__v8sf) _mm256_setzero_ps (),\
+                (__mmask8) -1);\
+})
+
+#define _mm256_mask_insertf32x4( __W, __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_insertf32x4_256_mask ((__v8sf)( __A),\
+                (__v4sf)( __B),\
+                ( __imm),\
+                (__v8sf)( __W),\
+                (__mmask8)( __U));\
+})
+
+#define _mm256_maskz_insertf32x4( __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_insertf32x4_256_mask ((__v8sf)( __A),\
+                (__v4sf)( __B),\
+                ( __imm),\
+                (__v8sf) _mm256_setzero_ps (),\
+                (__mmask8)( __U));\
+})
+
+#define _mm256_inserti32x4( __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti32x4_256_mask ((__v8si)( __A),\
+                (__v4si)( __B),\
+                ( __imm),\
+                (__v8si) _mm256_setzero_si256 (),\
+                (__mmask8) -1);\
+})
+
+#define _mm256_mask_inserti32x4( __W, __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti32x4_256_mask ((__v8si)( __A),\
+                (__v4si)( __B),\
+                ( __imm),\
+                (__v8si)( __W),\
+                (__mmask8) ( __U));\
+})
+
+#define _mm256_maskz_inserti32x4( __U, __A, __B, __imm) __extension__ ({ \
+__builtin_ia32_inserti32x4_256_mask ((__v8si)( __A),\
+                (__v4si)( __B),\
+                ( __imm),\
+                (__v8si) _mm256_setzero_si256 (),\
+                (__mmask8) ( __U));\
+})
+
 #undef __DEFAULT_FN_ATTRS
 #undef __DEFAULT_FN_ATTRS_BOTH
 
 #endif /* __AVX512VLINTRIN_H */

cfe/trunk/test/CodeGen/avx512dq-builtins.c

-// RUN: %clang_cc1 %s -triple=x86_64-apple-darwin -target-feature +avx512dq -emit-llvm -o - -Werror | FileCheck %s
+// RUN: %clang_cc1 %s -triple=x86_64-apple-darwin -target-feature +avx512dq -target-feature +avx512vl -emit-llvm -o - -Werror | FileCheck %s
 
 // Don't include mm_malloc.h, it's system specific.
 #define __MM_MALLOC_H
 
 #include <immintrin.h>
 
 __m512i test_mm512_mullo_epi64 (__m512i __A, __m512i __B) {
   // CHECK-LABEL: @test_mm512_mullo_epi64
__m512i test_mm512_mask_broadcast_i64x2(__m512i __O, __mmask8 __M, __m128i __A) {
   return _mm512_mask_broadcast_i64x2(__O, __M, __A); 
 }
 
 __m512i test_mm512_maskz_broadcast_i64x2(__mmask8 __M, __m128i __A) {
   // CHECK-LABEL: @test_mm512_maskz_broadcast_i64x2
   // CHECK: @llvm.x86.avx512.mask.broadcasti64x2
   return _mm512_maskz_broadcast_i64x2(__M, __A); 
 }
+__m256 test_mm512_extractf32x8_ps(__m512 __A) {
+  // CHECK-LABEL: @test_mm512_extractf32x8_ps
+  // CHECK: @llvm.x86.avx512.mask.vextractf32x8
+  return _mm512_extractf32x8_ps(__A, 1); 
+}
+
+__m256 test_mm512_mask_extractf32x8_ps(__m256 __W, __mmask8 __U, __m512 __A) {
+  // CHECK-LABEL: @test_mm512_mask_extractf32x8_ps
+  // CHECK: @llvm.x86.avx512.mask.vextractf32x8
+  return _mm512_mask_extractf32x8_ps(__W, __U, __A, 1); 
+}
+
+__m256 test_mm512_maskz_extractf32x8_ps(__mmask8 __U, __m512 __A) {
+  // CHECK-LABEL: @test_mm512_maskz_extractf32x8_ps
+  // CHECK: @llvm.x86.avx512.mask.vextractf32x8
+  return _mm512_maskz_extractf32x8_ps(__U, __A, 1); 
+}
+
+__m128d test_mm512_extractf64x2_pd(__m512d __A) {
+  // CHECK-LABEL: @test_mm512_extractf64x2_pd
+  // CHECK: @llvm.x86.avx512.mask.vextractf64x2
+  return _mm512_extractf64x2_pd(__A, 3); 
+}
+
+__m128d test_mm512_mask_extractf64x2_pd(__m128d __W, __mmask8 __U, __m512d __A) {
+  // CHECK-LABEL: @test_mm512_mask_extractf64x2_pd
+  // CHECK: @llvm.x86.avx512.mask.vextractf64x2
+  return _mm512_mask_extractf64x2_pd(__W, __U, __A, 3); 
+}
+
+__m128d test_mm512_maskz_extractf64x2_pd(__mmask8 __U, __m512d __A) {
+  // CHECK-LABEL: @test_mm512_maskz_extractf64x2_pd
+  // CHECK: @llvm.x86.avx512.mask.vextractf64x2
+  return _mm512_maskz_extractf64x2_pd(__U, __A, 3); 
+}
+
+__m256i test_mm512_extracti32x8_epi32(__m512i __A) {
+  // CHECK-LABEL: @test_mm512_extracti32x8_epi32
+  // CHECK: @llvm.x86.avx512.mask.vextracti32x8
+  return _mm512_extracti32x8_epi32(__A, 1); 
+}
+
+__m256i test_mm512_mask_extracti32x8_epi32(__m256i __W, __mmask8 __U, __m512i __A) {
+  // CHECK-LABEL: @test_mm512_mask_extracti32x8_epi32
+  // CHECK: @llvm.x86.avx512.mask.vextracti32x8
+  return _mm512_mask_extracti32x8_epi32(__W, __U, __A, 1); 
+}
+
+__m256i test_mm512_maskz_extracti32x8_epi32(__mmask8 __U, __m512i __A) {
+  // CHECK-LABEL: @test_mm512_maskz_extracti32x8_epi32
+  // CHECK: @llvm.x86.avx512.mask.vextracti32x8
+  return _mm512_maskz_extracti32x8_epi32(__U, __A, 1); 
+}
+
+__m128i test_mm512_extracti64x2_epi64(__m512i __A) {
+  // CHECK-LABEL: @test_mm512_extracti64x2_epi64
+  // CHECK: @llvm.x86.avx512.mask.vextracti64x2
+  return _mm512_extracti64x2_epi64(__A, 3); 
+}
+
+__m128i test_mm512_mask_extracti64x2_epi64(__m128i __W, __mmask8 __U, __m512i __A) {
+  // CHECK-LABEL: @test_mm512_mask_extracti64x2_epi64
+  // CHECK: @llvm.x86.avx512.mask.vextracti64x2
+  return _mm512_mask_extracti64x2_epi64(__W, __U, __A, 3); 
+}
+
+__m128i test_mm512_maskz_extracti64x2_epi64(__mmask8 __U, __m512i __A) {
+  // CHECK-LABEL: @test_mm512_maskz_extracti64x2_epi64
+  // CHECK: @llvm.x86.avx512.mask.vextracti64x2
+  return _mm512_maskz_extracti64x2_epi64(__U, __A, 3); 
+}
+
+__m512d test_mm512_insertf64x2(__m512d __A, __m128d __B) {
+  // CHECK-LABEL: @test_mm512_insertf64x2
+  // CHECK: @llvm.x86.avx512.mask.insertf64x2
+  return _mm512_insertf64x2(__A, __B, 3); 
+}
+
+__m512d test_mm512_mask_insertf64x2(__m512d __W, __mmask8 __U, __m512d __A, __m128d __B) {
+  // CHECK-LABEL: @test_mm512_mask_insertf64x2
+  // CHECK: @llvm.x86.avx512.mask.insertf64x2
+  return _mm512_mask_insertf64x2(__W, __U, __A, __B, 3); 
+}
+
+__m512d test_mm512_maskz_insertf64x2(__mmask8 __U, __m512d __A, __m128d __B) {
+  // CHECK-LABEL: @test_mm512_maskz_insertf64x2
+  // CHECK: @llvm.x86.avx512.mask.insertf64x2
+  return _mm512_maskz_insertf64x2(__U, __A, __B, 3); 
+}
+
+__m512i test_mm512_inserti32x8(__m512i __A, __m256i __B) {
+  // CHECK-LABEL: @test_mm512_inserti32x8
+  // CHECK: @llvm.x86.avx512.mask.inserti32x8
+  return _mm512_inserti32x8(__A, __B, 1); 
+}
+
+__m512i test_mm512_mask_inserti32x8(__m512i __W, __mmask16 __U, __m512i __A, __m256i __B) {
+  // CHECK-LABEL: @test_mm512_mask_inserti32x8
+  // CHECK: @llvm.x86.avx512.mask.inserti32x8
+  return _mm512_mask_inserti32x8(__W, __U, __A, __B, 1); 
+}
+
+__m512i test_mm512_maskz_inserti32x8(__mmask16 __U, __m512i __A, __m256i __B) {
+  // CHECK-LABEL: @test_mm512_maskz_inserti32x8
+  // CHECK: @llvm.x86.avx512.mask.inserti32x8
+  return _mm512_maskz_inserti32x8(__U, __A, __B, 1); 
+}
+
+__m512i test_mm512_inserti64x2(__m512i __A, __m128i __B) {
+  // CHECK-LABEL: @test_mm512_inserti64x2
+  // CHECK: @llvm.x86.avx512.mask.inserti64x2
+  return _mm512_inserti64x2(__A, __B, 1); 
+}
+
+__m512i test_mm512_mask_inserti64x2(__m512i __W, __mmask8 __U, __m512i __A, __m128i __B) {
+  // CHECK-LABEL: @test_mm512_mask_inserti64x2
+  // CHECK: @llvm.x86.avx512.mask.inserti64x2
+  return _mm512_mask_inserti64x2(__W, __U, __A, __B, 1); 
+}
+
+__m512i test_mm512_maskz_inserti64x2(__mmask8 __U, __m512i __A, __m128i __B) {
+  // CHECK-LABEL: @test_mm512_maskz_inserti64x2
+  // CHECK: @llvm.x86.avx512.mask.inserti64x2
+  return _mm512_maskz_inserti64x2(__U, __A, __B, 1); 
+}

cfe/trunk/test/CodeGen/avx512f-builtins.c

__m128i test_mm512_maskz_cvtepi64_epi16(__mmask8 __M, __m512i __A) {
   return _mm512_maskz_cvtepi64_epi16(__M, __A); 
 }
 
 void test_mm512_mask_cvtepi64_storeu_epi16(void *__P, __mmask8 __M, __m512i __A) {
   // CHECK-LABEL: @test_mm512_mask_cvtepi64_storeu_epi16
   // CHECK: @llvm.x86.avx512.mask.pmov.qw.mem.512
   return _mm512_mask_cvtepi64_storeu_epi16(__P, __M, __A); 
 }
+
+__m128i test_mm512_extracti32x4_epi32(__m512i __A) {
+  // CHECK-LABEL: @test_mm512_extracti32x4_epi32
+  // CHECK: @llvm.x86.avx512.mask.vextracti32x4
+  return _mm512_extracti32x4_epi32(__A, 3); 
+}
+
+__m128i test_mm512_mask_extracti32x4_epi32(__m128i __W, __mmask8 __U, __m512i __A) {
+  // CHECK-LABEL: @test_mm512_mask_extracti32x4_epi32
+  // CHECK: @llvm.x86.avx512.mask.vextracti32x4
+  return _mm512_mask_extracti32x4_epi32(__W, __U, __A, 3); 
+}
+
+__m128i test_mm512_maskz_extracti32x4_epi32(__mmask8 __U, __m512i __A) {
+  // CHECK-LABEL: @test_mm512_maskz_extracti32x4_epi32
+  // CHECK: @llvm.x86.avx512.mask.vextracti32x4
+  return _mm512_maskz_extracti32x4_epi32(__U, __A, 3); 
+}
+
+__m256i test_mm512_extracti64x4_epi64(__m512i __A) {
+  // CHECK-LABEL: @test_mm512_extracti64x4_epi64
+  // CHECK: @llvm.x86.avx512.mask.vextracti64x4
+  return _mm512_extracti64x4_epi64(__A, 1); 
+}
+
+__m256i test_mm512_mask_extracti64x4_epi64(__m256i __W, __mmask8 __U, __m512i __A) {
+  // CHECK-LABEL: @test_mm512_mask_extracti64x4_epi64
+  // CHECK: @llvm.x86.avx512.mask.vextracti64x4
+  return _mm512_mask_extracti64x4_epi64(__W, __U, __A, 1); 
+}
+
+__m256i test_mm512_maskz_extracti64x4_epi64(__mmask8 __U, __m512i __A) {
+  // CHECK-LABEL: @test_mm512_maskz_extracti64x4_epi64
+  // CHECK: @llvm.x86.avx512.mask.vextracti64x4
+  return _mm512_maskz_extracti64x4_epi64(__U, __A, 1); 
+}
+
+__m512d test_mm512_insertf64x4(__m512d __A, __m256d __B) {
+  // CHECK-LABEL: @test_mm512_insertf64x4
+  // CHECK: @llvm.x86.avx512.mask.insertf64x4
+  return _mm512_insertf64x4(__A, __B, 1);
+}
+
+__m512d test_mm512_mask_insertf64x4(__m512d __W, __mmask8 __U, __m512d __A, __m256d __B) {
+  // CHECK-LABEL: @test_mm512_mask_insertf64x4
+  // CHECK: @llvm.x86.avx512.mask.insertf64x4
+  return _mm512_mask_insertf64x4(__W, __U, __A, __B, 1); 
+}
+
+__m512d test_mm512_maskz_insertf64x4(__mmask8 __U, __m512d __A, __m256d __B) {
+  // CHECK-LABEL: @test_mm512_maskz_insertf64x4
+  // CHECK: @llvm.x86.avx512.mask.insertf64x4
+  return _mm512_maskz_insertf64x4(__U, __A, __B, 1); 
+}
+
+__m512i test_mm512_inserti64x4(__m512i __A, __m256i __B) {
+  // CHECK-LABEL: @test_mm512_inserti64x4
+  // CHECK: @llvm.x86.avx512.mask.inserti64x4
+  return _mm512_inserti64x4(__A, __B, 1); 
+}
+
+__m512i test_mm512_mask_inserti64x4(__m512i __W, __mmask8 __U, __m512i __A, __m256i __B) {
+  // CHECK-LABEL: @test_mm512_mask_inserti64x4
+  // CHECK: @llvm.x86.avx512.mask.inserti64x4
+  return _mm512_mask_inserti64x4(__W, __U, __A, __B, 1); 
+}
+
+__m512i test_mm512_maskz_inserti64x4(__mmask8 __U, __m512i __A, __m256i __B) {
+  // CHECK-LABEL: @test_mm512_maskz_inserti64x4
+  // CHECK: @llvm.x86.avx512.mask.inserti64x4
+  return _mm512_maskz_inserti64x4(__U, __A, __B, 1); 
+}

cfe/trunk/test/CodeGen/avx512vl-builtins.c

__m128i test_mm256_maskz_cvtepi64_epi16(__mmask8 __M, __m256i __A) {
   return _mm256_maskz_cvtepi64_epi16(__M, __A); 
 }
 
 void test_mm256_mask_cvtepi64_storeu_epi16(void * __P, __mmask8 __M, __m256i __A) {
   // CHECK-LABEL: @test_mm256_mask_cvtepi64_storeu_epi16
   // CHECK: @llvm.x86.avx512.mask.pmov.qw.mem.256
   return _mm256_mask_cvtepi64_storeu_epi16(__P, __M, __A); 
 }
+
+__m128 test_mm256_extractf32x4_ps(__m256 __A) {
+  // CHECK-LABEL: @test_mm256_extractf32x4_ps
+  // CHECK: @llvm.x86.avx512.mask.vextractf32x4
+  return _mm256_extractf32x4_ps(__A, 1); 
+}
+
+__m128 test_mm256_mask_extractf32x4_ps(__m128 __W, __mmask8 __U, __m256 __A) {
+  // CHECK-LABEL: @test_mm256_mask_extractf32x4_ps
+  // CHECK: @llvm.x86.avx512.mask.vextractf32x4
+  return _mm256_mask_extractf32x4_ps(__W, __U, __A, 1); 
+}
+
+__m128 test_mm256_maskz_extractf32x4_ps(__mmask8 __U, __m256 __A) {
+  // CHECK-LABEL: @test_mm256_maskz_extractf32x4_ps
+  // CHECK: @llvm.x86.avx512.mask.vextractf32x4
+  return _mm256_maskz_extractf32x4_ps(__U, __A, 1); 
+}
+
+__m128i test_mm256_extracti32x4_epi32(__m256i __A) {
+  // CHECK-LABEL: @test_mm256_extracti32x4_epi32
+  // CHECK: @llvm.x86.avx512.mask.vextracti32x4
+  return _mm256_extracti32x4_epi32(__A, 1); 
+}
+
+__m128i test_mm256_mask_extracti32x4_epi32(__m128i __W, __mmask8 __U, __m256i __A) {
+  // CHECK-LABEL: @test_mm256_mask_extracti32x4_epi32
+  // CHECK: @llvm.x86.avx512.mask.vextracti32x4
+  return _mm256_mask_extracti32x4_epi32(__W, __U, __A, 1); 
+}
+
+__m128i test_mm256_maskz_extracti32x4_epi32(__mmask8 __U, __m256i __A) {
+  // CHECK-LABEL: @test_mm256_maskz_extracti32x4_epi32
+  // CHECK: @llvm.x86.avx512.mask.vextracti32x4
+  return _mm256_maskz_extracti32x4_epi32(__U, __A, 1); 
+}
+
+__m256 test_mm256_insertf32x4(__m256 __A, __m128 __B) {
+  // CHECK-LABEL: @test_mm256_insertf32x4
+  // CHECK: @llvm.x86.avx512.mask.insertf32x4
+  return _mm256_insertf32x4(__A, __B, 1); 
+}
+
+__m256 test_mm256_mask_insertf32x4(__m256 __W, __mmask8 __U, __m256 __A, __m128 __B) {
+  // CHECK-LABEL: @test_mm256_mask_insertf32x4
+  // CHECK: @llvm.x86.avx512.mask.insertf32x4
+  return _mm256_mask_insertf32x4(__W, __U, __A, __B, 1); 
+}
+
+__m256 test_mm256_maskz_insertf32x4(__mmask8 __U, __m256 __A, __m128 __B) {
+  // CHECK-LABEL: @test_mm256_maskz_insertf32x4
+  // CHECK: @llvm.x86.avx512.mask.insertf32x4
+  return _mm256_maskz_insertf32x4(__U, __A, __B, 1); 
+}
+
+__m256i test_mm256_inserti32x4(__m256i __A, __m128i __B) {
+  // CHECK-LABEL: @test_mm256_inserti32x4
+  // CHECK: @llvm.x86.avx512.mask.inserti32x4
+  return _mm256_inserti32x4(__A, __B, 1); 
+}
+
+__m256i test_mm256_mask_inserti32x4(__m256i __W, __mmask8 __U, __m256i __A, __m128i __B) {
+  // CHECK-LABEL: @test_mm256_mask_inserti32x4
+  // CHECK: @llvm.x86.avx512.mask.inserti32x4
+  return _mm256_mask_inserti32x4(__W, __U, __A, __B, 1); 
+}
+
+__m256i test_mm256_maskz_inserti32x4(__mmask8 __U, __m256i __A, __m128i __B) {
+  // CHECK-LABEL: @test_mm256_maskz_inserti32x4
+  // CHECK: @llvm.x86.avx512.mask.inserti32x4
+  return _mm256_maskz_inserti32x4(__U, __A, __B, 1); 
+}

cfe/trunk/test/CodeGen/avx512vldq-builtins.c

__m256i test_mm256_mask_broadcast_i64x2(__m256i __O, __mmask8 __M, __m128i __A) {
   return _mm256_mask_broadcast_i64x2(__O, __M, __A); 
 }
 
 __m256i test_mm256_maskz_broadcast_i64x2(__mmask8 __M, __m128i __A) {
   // CHECK-LABEL: @test_mm256_maskz_broadcast_i64x2
   // CHECK: @llvm.x86.avx512.mask.broadcasti64x2
   return _mm256_maskz_broadcast_i64x2(__M, __A); 
 }
+
+__m128d test_mm256_extractf64x2_pd(__m256d __A) {
+  // CHECK-LABEL: @test_mm256_extractf64x2_pd
+  // CHECK: @llvm.x86.avx512.mask.vextractf64x2
+  return _mm256_extractf64x2_pd(__A, 1); 
+}
+
+__m128d test_mm256_mask_extractf64x2_pd(__m128d __W, __mmask8 __U, __m256d __A) {
+  // CHECK-LABEL: @test_mm256_mask_extractf64x2_pd
+  // CHECK: @llvm.x86.avx512.mask.vextractf64x2
+  return _mm256_mask_extractf64x2_pd(__W, __U, __A, 1); 
+}
+
+__m128d test_mm256_maskz_extractf64x2_pd(__mmask8 __U, __m256d __A) {
+  // CHECK-LABEL: @test_mm256_maskz_extractf64x2_pd
+  // CHECK: @llvm.x86.avx512.mask.vextractf64x2
+  return _mm256_maskz_extractf64x2_pd(__U, __A, 1); 
+}
+
+__m128i test_mm256_extracti64x2_epi64(__m256i __A) {
+  // CHECK-LABEL: @test_mm256_extracti64x2_epi64
+  // CHECK: @llvm.x86.avx512.mask.vextracti64x2
+  return _mm256_extracti64x2_epi64(__A, 1); 
+}
+
+__m128i test_mm256_mask_extracti64x2_epi64(__m128i __W, __mmask8 __U, __m256i __A) {
+  // CHECK-LABEL: @test_mm256_mask_extracti64x2_epi64
+  // CHECK: @llvm.x86.avx512.mask.vextracti64x2
+  return _mm256_mask_extracti64x2_epi64(__W, __U, __A, 1); 
+}
+
+__m128i test_mm256_maskz_extracti64x2_epi64(__mmask8 __U, __m256i __A) {
+  // CHECK-LABEL: @test_mm256_maskz_extracti64x2_epi64
+  // CHECK: @llvm.x86.avx512.mask.vextracti64x2
+  return _mm256_maskz_extracti64x2_epi64(__U, __A, 1); 
+}
+
+__m256d test_mm256_insertf64x2(__m256d __A, __m128d __B) {
+  // CHECK-LABEL: @test_mm256_insertf64x2
+  // CHECK: @llvm.x86.avx512.mask.insertf64x2
+  return _mm256_insertf64x2(__A, __B, 1); 
+}
+
+__m256d test_mm256_mask_insertf64x2(__m256d __W, __mmask8 __U, __m256d __A, __m128d __B) {
+  // CHECK-LABEL: @test_mm256_mask_insertf64x2
+  // CHECK: @llvm.x86.avx512.mask.insertf64x2
+  return _mm256_mask_insertf64x2(__W, __U, __A, __B, 1); 
+}
+
+__m256d test_mm256_maskz_insertf64x2(__mmask8 __U, __m256d __A, __m128d __B) {
+  // CHECK-LABEL: @test_mm256_maskz_insertf64x2
+  // CHECK: @llvm.x86.avx512.mask.insertf64x2
+  return _mm256_maskz_insertf64x2(__U, __A, __B, 1); 
+}
+
+__m256i test_mm256_inserti64x2(__m256i __A, __m128i __B) {
+  // CHECK-LABEL: @test_mm256_inserti64x2
+  // CHECK: @llvm.x86.avx512.mask.inserti64x2
+  return _mm256_inserti64x2(__A, __B, 1); 
+}
+
+__m256i test_mm256_mask_inserti64x2(__m256i __W, __mmask8 __U, __m256i __A, __m128i __B) {
+  // CHECK-LABEL: @test_mm256_mask_inserti64x2
+  // CHECK: @llvm.x86.avx512.mask.inserti64x2
+  return _mm256_mask_inserti64x2(__W, __U, __A, __B, 1); 
+}
+
+__m256i test_mm256_maskz_inserti64x2(__mmask8 __U, __m256i __A, __m128i __B) {
+  // CHECK-LABEL: @test_mm256_maskz_inserti64x2
+  // CHECK: @llvm.x86.avx512.mask.inserti64x2
+  return _mm256_maskz_inserti64x2(__U, __A, __B, 1); 
+}