[CLANG][avx512][BUILTIN] Adding fixupimm{pd|ps|sd|ss} getexp{sd|ss} getmant{sd|ss} kunpck{di|si} loada{pd|ps} loaddqu{di|hi|qi|si} max{sd|ss} min{sd|ss} kmov16 builtins to clang

[m_zuckerman]

Summary

builtin_ia32_fixupimmpd128_mask
builtin_ia32_fixupimmpd128_maskz
builtin_ia32_fixupimmpd256_mask
builtin_ia32_fixupimmpd256_maskz
builtin_ia32_fixupimmpd512_mask
builtin_ia32_fixupimmpd512_maskz
builtin_ia32_fixupimmps128_mask
builtin_ia32_fixupimmps128_maskz
builtin_ia32_fixupimmps256_mask
builtin_ia32_fixupimmps256_maskz
builtin_ia32_fixupimmps512_mask
builtin_ia32_fixupimmps512_maskz
builtin_ia32_fixupimmsd_mask
builtin_ia32_fixupimmsd_maskz
builtin_ia32_fixupimmss_mask
builtin_ia32_fixupimmss_maskz
builtin_ia32_getexpsd128_mask
builtin_ia32_getexpsd128_round
builtin_ia32_getexpss128_mask
builtin_ia32_getexpss128_round
builtin_ia32_getmantsd_round
builtin_ia32_getmantss_round
builtin_ia32_kunpckdi
builtin_ia32_kunpcksi
builtin_ia32_loadapd128_mask
builtin_ia32_loadapd256_mask
builtin_ia32_loadaps128_mask
builtin_ia32_kmov16
builtin_ia32_loadaps256_mask
builtin_ia32_loaddqudi128_mask
builtin_ia32_loaddqudi256_mask
builtin_ia32_loaddquhi128_mask
builtin_ia32_loaddquhi256_mask
builtin_ia32_loaddquhi512_mask
builtin_ia32_loaddquqi128_mask
builtin_ia32_loaddquqi256_mask
builtin_ia32_loaddquqi512_mask
builtin_ia32_loaddqusi128_mask
builtin_ia32_loaddqusi256_mask
builtin_ia32_loadupd128_mask
builtin_ia32_loadupd256_mask
builtin_ia32_loadups128_mask
builtin_ia32_loadups256_mask
builtin_ia32_maxsd_round
builtin_ia32_maxss_round
builtin_ia32_minsd_round
__builtin_ia32_minss_round

[AsafBadouh]

lib/Headers/avx512fintrin.h

54 ↗

(On Diff #50837)

move the whole block to line 48, right after the rounding mode macros.

75 ↗

(On Diff #50837)

Not sure there is need for these two intrinsics.
you can use _mm_setzero_pd/ps.

78 ↗

(On Diff #50837)

return type should be __m128d

lib/Headers/avx512vlbwintrin.h

2968 ↗

(On Diff #50837)

can you use _mm_setzero_si128() instead?

3003 ↗

(On Diff #50837)

can you use _mm_setzero_si128() instead?

lib/Headers/avx512vlintrin.h

6249 ↗

(On Diff #50837)

please use _mm_setzero_si128()

[AsafBadouh]

LGTM

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

 TARGET_BUILTIN(__builtin_ia32_vpermi2varqi128_mask, "V16cV16cV16cV16cUs","","avx512vbmi,avx512vl")
 TARGET_BUILTIN(__builtin_ia32_vpermi2varqi256_mask, "V32cV32cV32cV32cUi","","avx512vbmi,avx512vl")
 TARGET_BUILTIN(__builtin_ia32_vpermt2varqi128_mask, "V16cV16cV16cV16cUs","","avx512vbmi,avx512vl")
 TARGET_BUILTIN(__builtin_ia32_vpermt2varqi128_maskz, "V16cV16cV16cV16cUs","","avx512vbmi,avx512vl")
 TARGET_BUILTIN(__builtin_ia32_vpermt2varqi256_mask, "V32cV32cV32cV32cUi","","avx512vbmi,avx512vl")
 TARGET_BUILTIN(__builtin_ia32_vpermt2varqi256_maskz, "V32cV32cV32cV32cUi","","avx512vbmi,avx512vl")
 TARGET_BUILTIN(__builtin_ia32_vcomisd, "iV2dV2dIiIi","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_vcomiss, "iV4fV4fIiIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_kunpckdi, "ULLiULLiULLi","","avx512bw")
+TARGET_BUILTIN(__builtin_ia32_kunpcksi, "UiUiUi","","avx512bw")
+TARGET_BUILTIN(__builtin_ia32_loaddquhi512_mask, "V32sV32s*V32sUi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_loaddquqi512_mask, "V64cV64c*V64cULLi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_fixupimmpd512_mask, "V8dV8dV8dV8LLiIiUcIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_fixupimmpd512_maskz, "V8dV8dV8dV8LLiIiUcIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_fixupimmps512_mask, "V16fV16fV16fV16iIiUsIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_fixupimmps512_maskz, "V16fV16fV16fV16iIiUsIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_fixupimmsd_mask, "V2dV2dV2dV2LLiIiUcIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_fixupimmsd_maskz, "V2dV2dV2dV2LLiIiUcIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_fixupimmss_mask, "V4fV4fV4fV4iIiUcIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_fixupimmss_maskz, "V4fV4fV4fV4iIiUcIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_getexpsd128_round_mask, "V2dV2dV2dV2dUcIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_getexpss128_round_mask, "V4fV4fV4fV4fUcIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_getmantsd_round_mask, "V2dV2dV2dIiV2dUcIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_getmantss_round_mask, "V4fV4fV4fIiV4fUcIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_kmov16, "UsUs","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_maxsd_round, "V2dV2dV2dIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_maxss_round, "V4fV4fV4fIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_minsd_round, "V2dV2dV2dIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_minss_round, "V4fV4fV4fIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_loaddquhi128_mask, "V8sV8s*V8sUc","","avx512bw,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_loaddquhi256_mask, "V16sV16s*V16sUs","","avx512bw,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_loaddquqi128_mask, "V16cV16c*V16cUs","","avx512bw,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_loaddquqi256_mask, "V32cV32c*V32cUi","","avx512bw,avx512vl")
+TARGET_BUILTIN(__builtin_ia32_fixupimmpd128_mask, "V2dV2dV2dV2LLiIiUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_fixupimmpd128_maskz, "V2dV2dV2dV2LLiIiUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_fixupimmpd256_mask, "V4dV4dV4dV4LLiIiUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_fixupimmpd256_maskz, "V4dV4dV4dV4LLiIiUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_fixupimmps128_mask, "V4fV4fV4fV4iIiUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_fixupimmps128_maskz, "V4fV4fV4fV4iIiUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_fixupimmps256_mask, "V8fV8fV8fV8iIiUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_fixupimmps256_maskz, "V8fV8fV8fV8iIiUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_loadapd128_mask, "V2dV2d*V2dUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_loadapd256_mask, "V4dV4d*V4dUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_loadaps128_mask, "V4fV4f*V4fUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_loadaps256_mask, "V8fV8f*V8fUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_loaddqudi128_mask, "V2LLiV2LLi*V2LLiUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_loaddqudi256_mask, "V4LLiV4LLi*V4LLiUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_loaddqusi128_mask, "V4iV4i*V4iUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_loaddqusi256_mask, "V8iV8i*V8iUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_loadupd128_mask, "V2dV2d*V2dUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_loadupd256_mask, "V4dV4d*V4dUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_loadups128_mask, "V4fV4f*V4fUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_loadups256_mask, "V8fV8f*V8fUc","","avx512vl")
 
 #undef BUILTIN
 #undef TARGET_BUILTIN

cfe/trunk/lib/Headers/avx512bwintrin.h

 _mm512_maskz_set1_epi8 (__mmask64 __M, char __A)
 {
   return (__m512i) __builtin_ia32_pbroadcastb512_gpr_mask (__A,
                  (__v64qi)
                  _mm512_setzero_qi(),
                  __M);
 }
 
+static __inline__ __mmask64 __DEFAULT_FN_ATTRS
+_mm512_kunpackd (__mmask64 __A, __mmask64 __B)
+{
+  return (__mmask64) __builtin_ia32_kunpckdi ((__mmask64) __A,
+                (__mmask64) __B);
+}
+
+static __inline__ __mmask32 __DEFAULT_FN_ATTRS
+_mm512_kunpackw (__mmask32 __A, __mmask32 __B)
+{
+  return (__mmask32) __builtin_ia32_kunpcksi ((__mmask32) __A,
+                (__mmask32) __B);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_mask_loadu_epi16 (__m512i __W, __mmask32 __U, void const *__P)
+{
+  return (__m512i) __builtin_ia32_loaddquhi512_mask ((__v32hi *) __P,
+                 (__v32hi) __W,
+                 (__mmask32) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_maskz_loadu_epi16 (__mmask32 __U, void const *__P)
+{
+  return (__m512i) __builtin_ia32_loaddquhi512_mask ((__v32hi *) __P,
+                 (__v32hi)
+                 _mm512_setzero_hi (),
+                 (__mmask32) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_mask_loadu_epi8 (__m512i __W, __mmask64 __U, void const *__P)
+{
+  return (__m512i) __builtin_ia32_loaddquqi512_mask ((__v64qi *) __P,
+                 (__v64qi) __W,
+                 (__mmask64) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_maskz_loadu_epi8 (__mmask64 __U, void const *__P)
+{
+  return (__m512i) __builtin_ia32_loaddquqi512_mask ((__v64qi *) __P,
+                 (__v64qi)
+                 _mm512_setzero_hi (),
+                 (__mmask64) __U);
+}
 #undef __DEFAULT_FN_ATTRS
 
 #endif

cfe/trunk/lib/Headers/avx512fintrin.h

 
 /* Rounding mode macros.  */
 #define _MM_FROUND_TO_NEAREST_INT   0x00
 #define _MM_FROUND_TO_NEG_INF       0x01
 #define _MM_FROUND_TO_POS_INF       0x02
 #define _MM_FROUND_TO_ZERO          0x03
 #define _MM_FROUND_CUR_DIRECTION    0x04
 
+typedef enum
+{
+  _MM_MANT_NORM_1_2,    /* interval [1, 2)      */
+  _MM_MANT_NORM_p5_2,   /* interval [0.5, 2)    */
+  _MM_MANT_NORM_p5_1,   /* interval [0.5, 1)    */
+  _MM_MANT_NORM_p75_1p5   /* interval [0.75, 1.5) */
+} _MM_MANTISSA_NORM_ENUM;
+
+typedef enum
+{
+  _MM_MANT_SIGN_src,    /* sign = sign(SRC)     */
+  _MM_MANT_SIGN_zero,   /* sign = 0             */
+  _MM_MANT_SIGN_nan   /* DEST = NaN if sign(SRC) = 1 */
+} _MM_MANTISSA_SIGN_ENUM;
+
 /* Define the default attributes for the functions in this file. */
 #define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("avx512f")))
 
 /* Create vectors with repeated elements */
 
 static  __inline __m512i __DEFAULT_FN_ATTRS
 _mm512_setzero_si512(void)
 {

 _mm512_maskz_movedup_pd (__mmask8 __U, __m512d __A)
 {
   return (__m512d) __builtin_ia32_movddup512_mask ((__v8df) __A,
                (__v8df)
                _mm512_setzero_pd (),
                (__mmask8) __U);
 }
 
+#define _mm512_fixupimm_round_pd( __A, __B, __C, __imm, __R) __extension__ ({ \
+__builtin_ia32_fixupimmpd512_mask ((__v8df)( __A),\
+                  (__v8df)( __B),\
+                  (__v8di)( __C),\
+                  (__imm),\
+                  (__mmask8) -1, (__R));\
+})
+
+#define _mm512_mask_fixupimm_round_pd( __A, __U, __B, __C, __imm, __R) __extension__ ({ \
+__builtin_ia32_fixupimmpd512_mask ((__v8df)( __A),\
+                  (__v8df)( __B),\
+                  (__v8di)( __C),\
+                  (__imm),\
+                  (__mmask8)( __U), (__R));\
+})
+
+#define _mm512_fixupimm_pd( __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmpd512_mask ((__v8df)( __A),\
+                  (__v8df)( __B),\
+                  (__v8di)( __C),\
+                 ( __imm),\
+                  (__mmask8) -1,\
+                  _MM_FROUND_CUR_DIRECTION);\
+})
+
+#define _mm512_mask_fixupimm_pd( __A, __U, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmpd512_mask ((__v8df)( __A),\
+                  (__v8df)( __B),\
+                  (__v8di)( __C),\
+                 ( __imm),\
+                  (__mmask8)( __U),\
+                  _MM_FROUND_CUR_DIRECTION);\
+})
+
+#define _mm512_maskz_fixupimm_round_pd( __U, __A, __B, __C, __imm, __R) __extension__ ({ \
+__builtin_ia32_fixupimmpd512_maskz ((__v8df)( __A),\
+                   (__v8df)( __B),\
+                   (__v8di)( __C),\
+                   (__imm),\
+                   (__mmask8)( __U), (__R));\
+})
+
+#define _mm512_maskz_fixupimm_pd( __U, __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmpd512_maskz ((__v8df)( __A),\
+                   (__v8df)( __B),\
+                   (__v8di)( __C),\
+                  ( __imm),\
+                   (__mmask8)( __U),\
+                   _MM_FROUND_CUR_DIRECTION);\
+})
+
+#define _mm512_fixupimm_round_ps( __A, __B, __C, __imm, __R) __extension__ ({ \
+__builtin_ia32_fixupimmps512_mask ((__v16sf)( __A),\
+                 (__v16sf)( __B),\
+                 (__v16si)( __C),\
+                 (__imm),\
+                 (__mmask16) -1, (__R));\
+})
+
+#define _mm512_mask_fixupimm_round_ps( __A, __U, __B, __C, __imm, __R) __extension__ ({ \
+__builtin_ia32_fixupimmps512_mask ((__v16sf)( __A),\
+                 (__v16sf)( __B),\
+                 (__v16si)( __C),\
+                 (__imm),\
+                 (__mmask16)( __U), (__R));\
+})
+
+#define _mm512_fixupimm_ps( __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmps512_mask ((__v16sf)( __A),\
+                 (__v16sf)( __B),\
+                 (__v16si)( __C),\
+                ( __imm),\
+                 (__mmask16) -1,\
+                 _MM_FROUND_CUR_DIRECTION);\
+})
+
+#define _mm512_mask_fixupimm_ps( __A, __U, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmps512_mask ((__v16sf)( __A),\
+                 (__v16sf)( __B),\
+                 (__v16si)( __C),\
+                ( __imm),\
+                 (__mmask16)( __U),\
+                 _MM_FROUND_CUR_DIRECTION);\
+})
+
+#define _mm512_maskz_fixupimm_round_ps( __U, __A, __B, __C, __imm, __R) __extension__ ({ \
+__builtin_ia32_fixupimmps512_maskz ((__v16sf)( __A),\
+                  (__v16sf)( __B),\
+                  (__v16si)( __C),\
+                  (__imm),\
+                  (__mmask16)( __U), (__R));\
+})
+
+#define _mm512_maskz_fixupimm_ps( __U, __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmps512_maskz ((__v16sf)( __A),\
+                  (__v16sf)( __B),\
+                  (__v16si)( __C),\
+                 ( __imm),\
+                  (__mmask16)( __U),\
+                  _MM_FROUND_CUR_DIRECTION);\
+})
+
+#define _mm_fixupimm_round_sd( __A, __B, __C, __imm, __R) __extension__ ({ \
+__builtin_ia32_fixupimmsd_mask ((__v2df)( __A),\
+               (__v2df)( __B),\
+               (__v2di)( __C), __imm,\
+               (__mmask8) -1, (__R));\
+})
+
+#define _mm_mask_fixupimm_round_sd( __A, __U, __B, __C, __imm, __R) __extension__ ({ \
+__builtin_ia32_fixupimmsd_mask ((__v2df)( __A),\
+               (__v2df)( __B),\
+               (__v2di)( __C), __imm,\
+               (__mmask8)( __U), (__R));\
+})
+
+#define _mm_fixupimm_sd( __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmsd_mask ((__v2df)( __A),\
+               (__v2df)( __B),\
+               (__v2di)( __C),( __imm),\
+               (__mmask8) -1,\
+               _MM_FROUND_CUR_DIRECTION);\
+})
+
+#define _mm_mask_fixupimm_sd( __A, __U, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmsd_mask ((__v2df)( __A),\
+               (__v2df)( __B),\
+               (__v2di)( __C),( __imm),\
+               (__mmask8)( __U),\
+               _MM_FROUND_CUR_DIRECTION);\
+})
+
+#define _mm_maskz_fixupimm_round_sd( __U, __A, __B, __C, __imm, __R) __extension__ ({ \
+__builtin_ia32_fixupimmsd_maskz ((__v2df)( __A),\
+                (__v2df)( __B),\
+                (__v2di)( __C),\
+                __imm,\
+                (__mmask8)( __U), (__R));\
+})
+
+#define _mm_maskz_fixupimm_sd( __U, __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmsd_maskz ((__v2df)( __A),\
+                (__v2df)( __B),\
+                (__v2di)( __C),\
+               ( __imm),\
+                (__mmask8)( __U),\
+                _MM_FROUND_CUR_DIRECTION);\
+})
+
+#define _mm_fixupimm_round_ss( __A, __B, __C, __imm, __R) __extension__ ({ \
+__builtin_ia32_fixupimmss_mask ((__v4sf)( __A),\
+              (__v4sf)( __B),\
+              (__v4si)( __C), (__imm),\
+              (__mmask8) -1, (__R));\
+})
+
+#define _mm_mask_fixupimm_round_ss( __A, __U, __B, __C, __imm, __R) __extension__ ({ \
+__builtin_ia32_fixupimmss_mask ((__v4sf)( __A),\
+              (__v4sf)( __B),\
+              (__v4si)( __C), (__imm),\
+              (__mmask8)( __U), (__R));\
+})
+
+#define _mm_fixupimm_ss( __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmss_mask ((__v4sf)( __A),\
+              (__v4sf)( __B),\
+              (__v4si)( __C),( __imm),\
+              (__mmask8) -1,\
+              _MM_FROUND_CUR_DIRECTION);\
+})
+
+#define _mm_mask_fixupimm_ss( __A, __U, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmss_mask ((__v4sf)( __A),\
+              (__v4sf)( __B),\
+              (__v4si)( __C),( __imm),\
+              (__mmask8)( __U),\
+              _MM_FROUND_CUR_DIRECTION);\
+})
+
+#define _mm_maskz_fixupimm_round_ss( __U, __A, __B, __C, __imm, __R) __extension__ ({ \
+__builtin_ia32_fixupimmss_maskz ((__v4sf)( __A),\
+               (__v4sf)( __B),\
+               (__v4si)( __C), (__imm),\
+               (__mmask8)( __U), (__R));\
+})
+
+#define _mm_maskz_fixupimm_ss( __U, __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmss_maskz ((__v4sf)( __A),\
+               (__v4sf)( __B),\
+               (__v4si)( __C),( __imm),\
+               (__mmask8)( __U),\
+               _MM_FROUND_CUR_DIRECTION);\
+})
+
+#define _mm_getexp_round_sd( __A, __B ,__R) __extension__ ({ \
+__builtin_ia32_getexpsd128_round_mask ((__v2df)(__A),\
+                 (__v2df)( __B), (__v2df) _mm_setzero_pd(), (__mmask8) -1,\
+                 ( __R));\
+})
+
+
+static __inline__ __m128d __DEFAULT_FN_ATTRS
+_mm_getexp_sd (__m128d __A, __m128d __B)
+{
+  return (__m128d) __builtin_ia32_getexpsd128_round_mask ((__v2df) __A,
+                 (__v2df) __B, (__v2df) _mm_setzero_pd(), (__mmask8) -1, _MM_FROUND_CUR_DIRECTION);
+}
+
+#define _mm_getexp_round_ss( __A, __B, __R) __extension__ ({ \
+__builtin_ia32_getexpss128_round_mask ((__v4sf)( __A),\
+                (__v4sf)( __B), (__v4sf)  _mm_setzero_ps(), (__mmask8) -1,\
+                ( __R));\
+})
+
+static __inline__ __m128 __DEFAULT_FN_ATTRS
+_mm_getexp_ss (__m128 __A, __m128 __B)
+{
+  return (__m128) __builtin_ia32_getexpss128_round_mask ((__v4sf) __A,
+                (__v4sf) __B, (__v4sf)  _mm_setzero_ps(), (__mmask8) -1, _MM_FROUND_CUR_DIRECTION);
+}
+
+#define _mm_getmant_round_sd( __A, __B, __C, __D, __R) __extension__ ({ \
+__builtin_ia32_getmantsd_round_mask ((__v2df)( __A),\
+              (__v2df)( __B),\
+              (( __D) << 2) |( __C), (__v2df) _mm_setzero_pd(), (__mmask8) -1,\
+              ( __R));\
+})
+
+#define _mm_getmant_sd( __A,  __B,  __C, __D)  __extension__ ({ \
+__builtin_ia32_getmantsd_round_mask ((__v2df)( __A),\
+              (__v2df)( __B),\
+              (( __D) << 2) |( __C), (__v2df) _mm_setzero_pd(), (__mmask8) -1,\
+              _MM_FROUND_CUR_DIRECTION);\
+})
+
+#define _mm_getmant_round_ss( __A, __B, __C, __D, __R) __extension__ ({ \
+__builtin_ia32_getmantss_round_mask ((__v4sf)( __A),\
+              (__v4sf)( __B),\
+              ((__D) << 2) |( __C), (__v4sf)  _mm_setzero_ps(), (__mmask8) -1,\
+              ( __R));\
+})
+
+#define  _mm_getmant_ss(__A,  __B,  __C, __D) __extension__ ({ \
+__builtin_ia32_getmantss_round_mask ((__v4sf)( __A),\
+              (__v4sf)( __B),\
+              ((__D) << 2) |( __C), (__v4sf)  _mm_setzero_ps(), (__mmask8) -1,\
+              _MM_FROUND_CUR_DIRECTION);\
+})
+
+
+static __inline__ __mmask16 __DEFAULT_FN_ATTRS
+_mm512_kmov (__mmask16 __A)
+{
+  return  __A;
+}
+
 #define _mm_comi_round_sd(__A, __B, __P, __R) __extension__ ({\
 __builtin_ia32_vcomisd ((__v2df) (__A), (__v2df) (__B), ( __P), (  __R));\
 })
 
 #define _mm_comi_round_ss( __A,  __B, __P, __R) __extension__ ({\
 __builtin_ia32_vcomiss ((__v4sf) (__A), (__v4sf) (__B), ( __P), ( __R));\
 })
 
 
 #undef __DEFAULT_FN_ATTRS
 
 #endif // __AVX512FINTRIN_H

cfe/trunk/lib/Headers/avx512vlbwintrin.h

 static __inline__ __m256i __DEFAULT_FN_ATTRS
 _mm256_maskz_set1_epi8 (__mmask32 __M, char __A)
 {
   return (__m256i) __builtin_ia32_pbroadcastb256_gpr_mask (__A,
                  (__v32qi)
                  _mm256_setzero_si256 (),
                  __M);
 }
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_mask_loadu_epi16 (__m128i __W, __mmask8 __U, void const *__P)
+{
+  return (__m128i) __builtin_ia32_loaddquhi128_mask ((__v8hi *) __P,
+                 (__v8hi) __W,
+                 (__mmask8) __U);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_maskz_loadu_epi16 (__mmask8 __U, void const *__P)
+{
+  return (__m128i) __builtin_ia32_loaddquhi128_mask ((__v8hi *) __P,
+                 (__v8hi)
+                 _mm_setzero_hi (),
+                 (__mmask8) __U);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_mask_loadu_epi16 (__m256i __W, __mmask16 __U, void const *__P)
+{
+  return (__m256i) __builtin_ia32_loaddquhi256_mask ((__v16hi *) __P,
+                 (__v16hi) __W,
+                 (__mmask16) __U);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_maskz_loadu_epi16 (__mmask16 __U, void const *__P)
+{
+  return (__m256i) __builtin_ia32_loaddquhi256_mask ((__v16hi *) __P,
+                 (__v16hi)
+                 _mm256_setzero_si256 (),
+                 (__mmask16) __U);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_mask_loadu_epi8 (__m128i __W, __mmask16 __U, void const *__P)
+{
+  return (__m128i) __builtin_ia32_loaddquqi128_mask ((__v16qi *) __P,
+                 (__v16qi) __W,
+                 (__mmask16) __U);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_maskz_loadu_epi8 (__mmask16 __U, void const *__P)
+{
+  return (__m128i) __builtin_ia32_loaddquqi128_mask ((__v16qi *) __P,
+                 (__v16qi)
+                 _mm_setzero_si128 (),
+                 (__mmask16) __U);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_mask_loadu_epi8 (__m256i __W, __mmask32 __U, void const *__P)
+{
+  return (__m256i) __builtin_ia32_loaddquqi256_mask ((__v32qi *) __P,
+                 (__v32qi) __W,
+                 (__mmask32) __U);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_maskz_loadu_epi8 (__mmask32 __U, void const *__P)
+{
+  return (__m256i) __builtin_ia32_loaddquqi256_mask ((__v32qi *) __P,
+                 (__v32qi)
+                 _mm256_setzero_si256 (),
+                 (__mmask32) __U);
+}
 #undef __DEFAULT_FN_ATTRS
 
 #endif /* __AVX512VLBWINTRIN_H */

cfe/trunk/lib/Headers/avx512vlintrin.h

 static __inline__ __m256i __DEFAULT_FN_ATTRS
 _mm256_maskz_set1_epi64 (__mmask8 __M, long long __A)
 {
   return (__m256i) __builtin_ia32_pbroadcastq256_gpr_mask (__A,
                  (__v4di)
                  _mm256_setzero_si256 (),
                  __M);
 }
+
+#define _mm_fixupimm_pd( __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmpd128_mask ((__v2df)( __A),\
+                  (__v2df)( __B),\
+                  (__v2di)( __C),\
+                 ( __imm),\
+                  (__mmask8) -1);\
+})
+
+#define _mm_mask_fixupimm_pd( __A, __U, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmpd128_mask ((__v2df)( __A),\
+                  (__v2df)( __B),\
+                  (__v2di)( __C),\
+                 ( __imm),\
+                  (__mmask8)( __U));\
+})
+
+#define _mm_maskz_fixupimm_pd( __U, __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmpd128_maskz ((__v2df)( __A),\
+                   (__v2df)( __B),\
+                   (__v2di)( __C),\
+                  ( __imm),\
+                   (__mmask8)( __U));\
+})
+
+#define _mm256_fixupimm_pd( __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmpd256_mask ((__v4df)( __A),\
+                  (__v4df)( __B),\
+                  (__v4di)( __C),\
+                 ( __imm),\
+                  (__mmask8) -1);\
+})
+
+#define _mm256_mask_fixupimm_pd( __A, __U, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmpd256_mask ((__v4df)( __A),\
+                  (__v4df)( __B),\
+                  (__v4di)( __C),\
+                 ( __imm),\
+                  (__mmask8)( __U));\
+})
+
+#define _mm256_maskz_fixupimm_pd( __U, __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmpd256_maskz ((__v4df)( __A),\
+                   (__v4df)( __B),\
+                   (__v4di)( __C),\
+                  ( __imm),\
+                   (__mmask8)( __U));\
+})
+
+#define _mm_fixupimm_ps( __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmps128_mask ((__v4sf)( __A),\
+                 (__v4sf)( __B),\
+                 (__v4si)( __C),\
+                ( __imm),\
+                 (__mmask8) -1);\
+})
+
+#define _mm_mask_fixupimm_ps( __A, __U, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmps128_mask ((__v4sf)( __A),\
+                 (__v4sf)( __B),\
+                 (__v4si)( __C),\
+                ( __imm),\
+                 (__mmask8)( __U));\
+})
+
+#define _mm_maskz_fixupimm_ps( __U, __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmps128_maskz ((__v4sf)( __A),\
+                  (__v4sf)( __B),\
+                  (__v4si)( __C),\
+                 ( __imm),\
+                  (__mmask8)( __U));\
+})
+
+#define _mm256_fixupimm_ps( __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmps256_mask ((__v8sf)( __A),\
+                 (__v8sf)( __B),\
+                 (__v8si)( __C),\
+                ( __imm),\
+                 (__mmask8) -1);\
+})
+
+#define _mm256_mask_fixupimm_ps( __A, __U, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmps256_mask ((__v8sf)( __A),\
+                 (__v8sf)( __B),\
+                 (__v8si)( __C),\
+                ( __imm),\
+                 (__mmask8)( __U));\
+})
+
+#define _mm256_maskz_fixupimm_ps( __U, __A, __B, __C, __imm) __extension__ ({ \
+__builtin_ia32_fixupimmps256_maskz ((__v8sf)( __A),\
+                  (__v8sf)( __B),\
+                  (__v8si)( __C),\
+                 ( __imm),\
+                  (__mmask8)( __U));\
+})
+
+static __inline__ __m128d __DEFAULT_FN_ATTRS
+_mm_mask_load_pd (__m128d __W, __mmask8 __U, void const *__P)
+{
+  return (__m128d) __builtin_ia32_loadapd128_mask ((__v2df *) __P,
+               (__v2df) __W,
+               (__mmask8) __U);
+}
+
+static __inline__ __m128d __DEFAULT_FN_ATTRS
+_mm_maskz_load_pd (__mmask8 __U, void const *__P)
+{
+  return (__m128d) __builtin_ia32_loadapd128_mask ((__v2df *) __P,
+               (__v2df)
+               _mm_setzero_pd (),
+               (__mmask8) __U);
+}
+
+static __inline__ __m256d __DEFAULT_FN_ATTRS
+_mm256_mask_load_pd (__m256d __W, __mmask8 __U, void const *__P)
+{
+  return (__m256d) __builtin_ia32_loadapd256_mask ((__v4df *) __P,
+               (__v4df) __W,
+               (__mmask8) __U);
+}
+
+static __inline__ __m256d __DEFAULT_FN_ATTRS
+_mm256_maskz_load_pd (__mmask8 __U, void const *__P)
+{
+  return (__m256d) __builtin_ia32_loadapd256_mask ((__v4df *) __P,
+               (__v4df)
+               _mm256_setzero_pd (),
+               (__mmask8) __U);
+}
+
+static __inline__ __m128 __DEFAULT_FN_ATTRS
+_mm_mask_load_ps (__m128 __W, __mmask8 __U, void const *__P)
+{
+  return (__m128) __builtin_ia32_loadaps128_mask ((__v4sf *) __P,
+              (__v4sf) __W,
+              (__mmask8) __U);
+}
+
+static __inline__ __m128 __DEFAULT_FN_ATTRS
+_mm_maskz_load_ps (__mmask8 __U, void const *__P)
+{
+  return (__m128) __builtin_ia32_loadaps128_mask ((__v4sf *) __P,
+              (__v4sf)
+              _mm_setzero_ps (),
+              (__mmask8) __U);
+}
+
+static __inline__ __m256 __DEFAULT_FN_ATTRS
+_mm256_mask_load_ps (__m256 __W, __mmask8 __U, void const *__P)
+{
+  return (__m256) __builtin_ia32_loadaps256_mask ((__v8sf *) __P,
+              (__v8sf) __W,
+              (__mmask8) __U);
+}
+
+static __inline__ __m256 __DEFAULT_FN_ATTRS
+_mm256_maskz_load_ps (__mmask8 __U, void const *__P)
+{
+  return (__m256) __builtin_ia32_loadaps256_mask ((__v8sf *) __P,
+              (__v8sf)
+              _mm256_setzero_ps (),
+              (__mmask8) __U);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_mask_loadu_epi64 (__m128i __W, __mmask8 __U, void const *__P)
+{
+  return (__m128i) __builtin_ia32_loaddqudi128_mask ((__v2di *) __P,
+                 (__v2di) __W,
+                 (__mmask8) __U);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_maskz_loadu_epi64 (__mmask8 __U, void const *__P)
+{
+  return (__m128i) __builtin_ia32_loaddqudi128_mask ((__v2di *) __P,
+                 (__v2di)
+                 _mm_setzero_si128 (),
+                 (__mmask8) __U);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_mask_loadu_epi64 (__m256i __W, __mmask8 __U, void const *__P)
+{
+  return (__m256i) __builtin_ia32_loaddqudi256_mask ((__v4di *) __P,
+                 (__v4di) __W,
+                 (__mmask8) __U);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_maskz_loadu_epi64 (__mmask8 __U, void const *__P)
+{
+  return (__m256i) __builtin_ia32_loaddqudi256_mask ((__v4di *) __P,
+                 (__v4di)
+                 _mm256_setzero_si256 (),
+                 (__mmask8) __U);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_mask_loadu_epi32 (__m128i __W, __mmask8 __U, void const *__P)
+{
+  return (__m128i) __builtin_ia32_loaddqusi128_mask ((__v4si *) __P,
+                 (__v4si) __W,
+                 (__mmask8) __U);
+}
+
+static __inline__ __m128i __DEFAULT_FN_ATTRS
+_mm_maskz_loadu_epi32 (__mmask8 __U, void const *__P)
+{
+  return (__m128i) __builtin_ia32_loaddqusi128_mask ((__v4si *) __P,
+                 (__v4si)
+                 _mm_setzero_si128 (),
+                 (__mmask8) __U);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_mask_loadu_epi32 (__m256i __W, __mmask8 __U, void const *__P)
+{
+  return (__m256i) __builtin_ia32_loaddqusi256_mask ((__v8si *) __P,
+                 (__v8si) __W,
+                 (__mmask8) __U);
+}
+
+static __inline__ __m256i __DEFAULT_FN_ATTRS
+_mm256_maskz_loadu_epi32 (__mmask8 __U, void const *__P)
+{
+  return (__m256i) __builtin_ia32_loaddqusi256_mask ((__v8si *) __P,
+                 (__v8si)
+                 _mm256_setzero_si256 (),
+                 (__mmask8) __U);
+}
+
+static __inline__ __m128d __DEFAULT_FN_ATTRS
+_mm_mask_loadu_pd (__m128d __W, __mmask8 __U, void const *__P)
+{
+  return (__m128d) __builtin_ia32_loadupd128_mask ((__v2df *) __P,
+               (__v2df) __W,
+               (__mmask8) __U);
+}
+
+static __inline__ __m128d __DEFAULT_FN_ATTRS
+_mm_maskz_loadu_pd (__mmask8 __U, void const *__P)
+{
+  return (__m128d) __builtin_ia32_loadupd128_mask ((__v2df *) __P,
+               (__v2df)
+               _mm_setzero_pd (),
+               (__mmask8) __U);
+}
+
+static __inline__ __m256d __DEFAULT_FN_ATTRS
+_mm256_mask_loadu_pd (__m256d __W, __mmask8 __U, void const *__P)
+{
+  return (__m256d) __builtin_ia32_loadupd256_mask ((__v4df *) __P,
+               (__v4df) __W,
+               (__mmask8) __U);
+}
+
+static __inline__ __m256d __DEFAULT_FN_ATTRS
+_mm256_maskz_loadu_pd (__mmask8 __U, void const *__P)
+{
+  return (__m256d) __builtin_ia32_loadupd256_mask ((__v4df *) __P,
+               (__v4df)
+               _mm256_setzero_pd (),
+               (__mmask8) __U);
+}
+
+static __inline__ __m128 __DEFAULT_FN_ATTRS
+_mm_mask_loadu_ps (__m128 __W, __mmask8 __U, void const *__P)
+{
+  return (__m128) __builtin_ia32_loadups128_mask ((__v4sf *) __P,
+              (__v4sf) __W,
+              (__mmask8) __U);
+}
+
+static __inline__ __m128 __DEFAULT_FN_ATTRS
+_mm_maskz_loadu_ps (__mmask8 __U, void const *__P)
+{
+  return (__m128) __builtin_ia32_loadups128_mask ((__v4sf *) __P,
+              (__v4sf)
+              _mm_setzero_ps (),
+              (__mmask8) __U);
+}
+
+static __inline__ __m256 __DEFAULT_FN_ATTRS
+_mm256_mask_loadu_ps (__m256 __W, __mmask8 __U, void const *__P)
+{
+  return (__m256) __builtin_ia32_loadups256_mask ((__v8sf *) __P,
+              (__v8sf) __W,
+              (__mmask8) __U);
+}
+
+static __inline__ __m256 __DEFAULT_FN_ATTRS
+_mm256_maskz_loadu_ps (__mmask8 __U, void const *__P)
+{
+  return (__m256) __builtin_ia32_loadups256_mask ((__v8sf *) __P,
+              (__v8sf)
+              _mm256_setzero_ps (),
+              (__mmask8) __U);
+}
 #undef __DEFAULT_FN_ATTRS
 #undef __DEFAULT_FN_ATTRS_BOTH
 
 #endif /* __AVX512VLINTRIN_H */

cfe/trunk/test/CodeGen/avx512bw-builtins.c

 }
 
 __m512i test_mm512_maskz_set1_epi8(__mmask64 __M, char __A) {
   // CHECK-LABEL: @test_mm512_maskz_set1_epi8
   // CHECK: @llvm.x86.avx512.mask.pbroadcast.b.gpr.512
   return _mm512_maskz_set1_epi8(__M, __A); 
 }
 
+__mmask64 test_mm512_kunpackd(__mmask64 __A, __mmask64 __B) {
+  // CHECK-LABEL: @test_mm512_kunpackd
+  // CHECK: @llvm.x86.avx512.kunpck.dq
+  return _mm512_kunpackd(__A, __B); 
+}
+
+__mmask32 test_mm512_kunpackw(__mmask32 __A, __mmask32 __B) {
+  // CHECK-LABEL: @test_mm512_kunpackw
+  // CHECK: @llvm.x86.avx512.kunpck.wd
+  return _mm512_kunpackw(__A, __B); 
+}
+
+__m512i test_mm512_mask_loadu_epi16(__m512i __W, __mmask32 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm512_mask_loadu_epi16
+  // CHECK: @llvm.x86.avx512.mask.loadu.w.512
+  return _mm512_mask_loadu_epi16(__W, __U, __P); 
+}
+
+__m512i test_mm512_maskz_loadu_epi16(__mmask32 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm512_maskz_loadu_epi16
+  // CHECK: @llvm.x86.avx512.mask.loadu.w.512
+  return _mm512_maskz_loadu_epi16(__U, __P); 
+}
+
+__m512i test_mm512_mask_loadu_epi8(__m512i __W, __mmask64 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm512_mask_loadu_epi8
+  // CHECK: @llvm.x86.avx512.mask.loadu.b.512
+  return _mm512_mask_loadu_epi8(__W, __U, __P); 
+}
+
+__m512i test_mm512_maskz_loadu_epi8(__mmask64 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm512_maskz_loadu_epi8
+  // CHECK: @llvm.x86.avx512.mask.loadu.b.512
+  return _mm512_maskz_loadu_epi8(__U, __P); 
+}
+
 

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

int test_mm_comi_round_sd(__m128d __A, __m128d __B) {
   return _mm_comi_round_sd(__A, __B, 5, 3); 
 }
 
 int test_mm_comi_round_ss(__m128 __A, __m128 __B) {
   // CHECK-LABEL: @test_mm_comi_round_ss
   // CHECK: @llvm.x86.avx512.vcomi.ss
   return _mm_comi_round_ss(__A, __B, 5, 3); 
 }
+
+__m512d test_mm512_fixupimm_round_pd(__m512d __A, __m512d __B, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_fixupimm_round_pd
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.pd.512
+  return _mm512_fixupimm_round_pd(__A, __B, __C, 5, 8); 
+}
+
+__m512d test_mm512_mask_fixupimm_round_pd(__m512d __A, __mmask8 __U, __m512d __B, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_mask_fixupimm_round_pd
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.pd.512
+  return _mm512_mask_fixupimm_round_pd(__A, __U, __B, __C, 5, 8); 
+}
+
+__m512d test_mm512_fixupimm_pd(__m512d __A, __m512d __B, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_fixupimm_pd
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.pd.512
+  return _mm512_fixupimm_pd(__A, __B, __C, 5); 
+}
+
+__m512d test_mm512_mask_fixupimm_pd(__m512d __A, __mmask8 __U, __m512d __B, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_mask_fixupimm_pd
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.pd.512
+  return _mm512_mask_fixupimm_pd(__A, __U, __B, __C, 5); 
+}
+
+__m512d test_mm512_maskz_fixupimm_round_pd(__mmask8 __U, __m512d __A, __m512d __B, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_maskz_fixupimm_round_pd
+  // CHECK: @llvm.x86.avx512.maskz.fixupimm.pd.512
+  return _mm512_maskz_fixupimm_round_pd(__U, __A, __B, __C, 5, 8); 
+}
+
+__m512d test_mm512_maskz_fixupimm_pd(__mmask8 __U, __m512d __A, __m512d __B, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_maskz_fixupimm_pd
+  // CHECK: @llvm.x86.avx512.maskz.fixupimm.pd.512
+  return _mm512_maskz_fixupimm_pd(__U, __A, __B, __C, 5); 
+}
+
+__m512 test_mm512_fixupimm_round_ps(__m512 __A, __m512 __B, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_fixupimm_round_ps
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.ps.512
+  return _mm512_fixupimm_round_ps(__A, __B, __C, 5, 8); 
+}
+
+__m512 test_mm512_mask_fixupimm_round_ps(__m512 __A, __mmask16 __U, __m512 __B, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_mask_fixupimm_round_ps
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.ps.512
+  return _mm512_mask_fixupimm_round_ps(__A, __U, __B, __C, 5, 8); 
+}
+
+__m512 test_mm512_fixupimm_ps(__m512 __A, __m512 __B, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_fixupimm_ps
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.ps.512
+  return _mm512_fixupimm_ps(__A, __B, __C, 5); 
+}
+
+__m512 test_mm512_mask_fixupimm_ps(__m512 __A, __mmask16 __U, __m512 __B, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_mask_fixupimm_ps
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.ps.512
+  return _mm512_mask_fixupimm_ps(__A, __U, __B, __C, 5); 
+}
+
+__m512 test_mm512_maskz_fixupimm_round_ps(__mmask16 __U, __m512 __A, __m512 __B, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_maskz_fixupimm_round_ps
+  // CHECK: @llvm.x86.avx512.maskz.fixupimm.ps.512
+  return _mm512_maskz_fixupimm_round_ps(__U, __A, __B, __C, 5, 8); 
+}
+
+__m512 test_mm512_maskz_fixupimm_ps(__mmask16 __U, __m512 __A, __m512 __B, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_maskz_fixupimm_ps
+  // CHECK: @llvm.x86.avx512.maskz.fixupimm.ps.512
+  return _mm512_maskz_fixupimm_ps(__U, __A, __B, __C, 5); 
+}
+
+__m128d test_mm_fixupimm_round_sd(__m128d __A, __m128d __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_fixupimm_round_sd
+  // CHECK: @llvm.x86.avx512.mask.fixupimm
+  return _mm_fixupimm_round_sd(__A, __B, __C, 5, 8); 
+}
+
+__m128d test_mm_mask_fixupimm_round_sd(__m128d __A, __mmask8 __U, __m128d __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_mask_fixupimm_round_sd
+  // CHECK: @llvm.x86.avx512.mask.fixupimm
+  return _mm_mask_fixupimm_round_sd(__A, __U, __B, __C, 5, 8); 
+}
+
+__m128d test_mm_fixupimm_sd(__m128d __A, __m128d __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_fixupimm_sd
+  // CHECK: @llvm.x86.avx512.mask.fixupimm
+  return _mm_fixupimm_sd(__A, __B, __C, 5); 
+}
+
+__m128d test_mm_mask_fixupimm_sd(__m128d __A, __mmask8 __U, __m128d __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_mask_fixupimm_sd
+  // CHECK: @llvm.x86.avx512.mask.fixupimm
+  return _mm_mask_fixupimm_sd(__A, __U, __B, __C, 5); 
+}
+
+__m128d test_mm_maskz_fixupimm_round_sd(__mmask8 __U, __m128d __A, __m128d __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_maskz_fixupimm_round_sd
+  // CHECK: @llvm.x86.avx512.maskz.fixupimm
+  return _mm_maskz_fixupimm_round_sd(__U, __A, __B, __C, 5, 8); 
+}
+
+__m128d test_mm_maskz_fixupimm_sd(__mmask8 __U, __m128d __A, __m128d __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_maskz_fixupimm_sd
+  // CHECK: @llvm.x86.avx512.maskz.fixupimm
+  return _mm_maskz_fixupimm_sd(__U, __A, __B, __C, 5); 
+}
+
+__m128 test_mm_fixupimm_round_ss(__m128 __A, __m128 __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_fixupimm_round_ss
+  // CHECK: @llvm.x86.avx512.mask.fixupimm
+  return _mm_fixupimm_round_ss(__A, __B, __C, 5, 8); 
+}
+
+__m128 test_mm_mask_fixupimm_round_ss(__m128 __A, __mmask8 __U, __m128 __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_mask_fixupimm_round_ss
+  // CHECK: @llvm.x86.avx512.mask.fixupimm
+  return _mm_mask_fixupimm_round_ss(__A, __U, __B, __C, 5, 8); 
+}
+
+__m128 test_mm_fixupimm_ss(__m128 __A, __m128 __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_fixupimm_ss
+  // CHECK: @llvm.x86.avx512.mask.fixupimm
+  return _mm_fixupimm_ss(__A, __B, __C, 5); 
+}
+
+__m128 test_mm_mask_fixupimm_ss(__m128 __A, __mmask8 __U, __m128 __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_mask_fixupimm_ss
+  // CHECK: @llvm.x86.avx512.mask.fixupimm
+  return _mm_mask_fixupimm_ss(__A, __U, __B, __C, 5); 
+}
+
+__m128 test_mm_maskz_fixupimm_round_ss(__mmask8 __U, __m128 __A, __m128 __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_maskz_fixupimm_round_ss
+  // CHECK: @llvm.x86.avx512.maskz.fixupimm
+  return _mm_maskz_fixupimm_round_ss(__U, __A, __B, __C, 5, 8); 
+}
+
+__m128 test_mm_maskz_fixupimm_ss(__mmask8 __U, __m128 __A, __m128 __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_maskz_fixupimm_ss
+  // CHECK: @llvm.x86.avx512.maskz.fixupimm
+  return _mm_maskz_fixupimm_ss(__U, __A, __B, __C, 5); 
+}
+
+__m128d test_mm_getexp_round_sd(__m128d __A, __m128d __B) {
+  // CHECK-LABEL: @test_mm_getexp_round_sd
+  // CHECK: @llvm.x86.avx512.mask.getexp.sd
+  return _mm_getexp_round_sd(__A, __B, 8); 
+}
+
+__m128d test_mm_getexp_sd(__m128d __A, __m128d __B) {
+  // CHECK-LABEL: @test_mm_getexp_sd
+  // CHECK: @llvm.x86.avx512.mask.getexp.sd
+  return _mm_getexp_sd(__A, __B); 
+}
+
+__m128 test_mm_getexp_round_ss(__m128 __A, __m128 __B) {
+  // CHECK-LABEL: @test_mm_getexp_round_ss
+  // CHECK: @llvm.x86.avx512.mask.getexp.ss
+  return _mm_getexp_round_ss(__A, __B, 8); 
+}
+
+__m128 test_mm_getexp_ss(__m128 __A, __m128 __B) {
+  // CHECK-LABEL: @test_mm_getexp_ss
+  // CHECK: @llvm.x86.avx512.mask.getexp.ss
+  return _mm_getexp_ss(__A, __B); 
+}
+
+__m128d test_mm_getmant_round_sd(__m128d __A, __m128d __B) {
+  // CHECK-LABEL: @test_mm_getmant_round_sd
+  // CHECK: @llvm.x86.avx512.mask.getmant.sd
+  return _mm_getmant_round_sd(__A, __B, _MM_MANT_NORM_1_2, _MM_MANT_SIGN_src, 8); 
+}
+
+__m128d test_mm_getmant_sd(__m128d __A, __m128d __B) {
+  // CHECK-LABEL: @test_mm_getmant_sd
+  // CHECK: @llvm.x86.avx512.mask.getmant.sd
+  return _mm_getmant_sd(__A, __B, _MM_MANT_NORM_1_2, _MM_MANT_SIGN_src); 
+}
+
+
+
+__m128 test_mm_getmant_round_ss(__m128 __A, __m128 __B) {
+  // CHECK-LABEL: @test_mm_getmant_round_ss
+  // CHECK: @llvm.x86.avx512.mask.getmant.ss
+  return _mm_getmant_round_ss(__A, __B, _MM_MANT_NORM_1_2, _MM_MANT_SIGN_src, 8); 
+}
+
+__m128 test_mm_getmant_ss(__m128 __A, __m128 __B) {
+  // CHECK-LABEL: @test_mm_getmant_ss
+  // CHECK: @llvm.x86.avx512.mask.getmant.ss
+  return _mm_getmant_ss(__A, __B, _MM_MANT_NORM_1_2, _MM_MANT_SIGN_src); 
+}
+
+__mmask16 test_mm512_kmov(__mmask16 __A) {
+  // CHECK-LABEL: @test_mm512_kmov
+  // CHECK: load i16, i16* %__A.addr.i, align 2
+  return _mm512_kmov(__A); 
+}
+
+
+

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

 }
 
 __m256i test_mm256_maskz_set1_epi64(__mmask8 __M, long long __A) {
   // CHECK-LABEL: @test_mm256_maskz_set1_epi64
   // CHECK: @llvm.x86.avx512.mask.pbroadcast.q.gpr.256
   return _mm256_maskz_set1_epi64(__M, __A); 
 }
 
+__m128d test_mm_fixupimm_pd(__m128d __A, __m128d __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_fixupimm_pd
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.pd.128
+  return _mm_fixupimm_pd(__A, __B, __C, 5); 
+}
+
+__m128d test_mm_mask_fixupimm_pd(__m128d __A, __mmask8 __U, __m128d __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_mask_fixupimm_pd
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.pd.128
+  return _mm_mask_fixupimm_pd(__A, __U, __B, __C, 5); 
+}
+
+__m128d test_mm_maskz_fixupimm_pd(__mmask8 __U, __m128d __A, __m128d __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_maskz_fixupimm_pd
+  // CHECK: @llvm.x86.avx512.maskz.fixupimm.pd.128
+  return _mm_maskz_fixupimm_pd(__U, __A, __B, __C, 5); 
+}
+
+__m256d test_mm256_fixupimm_pd(__m256d __A, __m256d __B, __m256i __C) {
+  // CHECK-LABEL: @test_mm256_fixupimm_pd
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.pd.256
+  return _mm256_fixupimm_pd(__A, __B, __C, 5); 
+}
+
+__m256d test_mm256_mask_fixupimm_pd(__m256d __A, __mmask8 __U, __m256d __B, __m256i __C) {
+  // CHECK-LABEL: @test_mm256_mask_fixupimm_pd
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.pd.256
+  return _mm256_mask_fixupimm_pd(__A, __U, __B, __C, 5); 
+}
+
+__m256d test_mm256_maskz_fixupimm_pd(__mmask8 __U, __m256d __A, __m256d __B, __m256i __C) {
+  // CHECK-LABEL: @test_mm256_maskz_fixupimm_pd
+  // CHECK: @llvm.x86.avx512.maskz.fixupimm.pd.256
+  return _mm256_maskz_fixupimm_pd(__U, __A, __B, __C, 5); 
+}
+
+__m128 test_mm_fixupimm_ps(__m128 __A, __m128 __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_fixupimm_ps
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.ps.128
+  return _mm_fixupimm_ps(__A, __B, __C, 5); 
+}
+
+__m128 test_mm_mask_fixupimm_ps(__m128 __A, __mmask8 __U, __m128 __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_mask_fixupimm_ps
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.ps.128
+  return _mm_mask_fixupimm_ps(__A, __U, __B, __C, 5); 
+}
+
+__m128 test_mm_maskz_fixupimm_ps(__mmask8 __U, __m128 __A, __m128 __B, __m128i __C) {
+  // CHECK-LABEL: @test_mm_maskz_fixupimm_ps
+  // CHECK: @llvm.x86.avx512.maskz.fixupimm.ps.128
+  return _mm_maskz_fixupimm_ps(__U, __A, __B, __C, 5); 
+}
+
+__m256 test_mm256_fixupimm_ps(__m256 __A, __m256 __B, __m256i __C) {
+  // CHECK-LABEL: @test_mm256_fixupimm_ps
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.ps.256
+  return _mm256_fixupimm_ps(__A, __B, __C, 5); 
+}
+
+__m256 test_mm256_mask_fixupimm_ps(__m256 __A, __mmask8 __U, __m256 __B, __m256i __C) {
+  // CHECK-LABEL: @test_mm256_mask_fixupimm_ps
+  // CHECK: @llvm.x86.avx512.mask.fixupimm.ps.256
+  return _mm256_mask_fixupimm_ps(__A, __U, __B, __C, 5); 
+}
+
+__m256 test_mm256_maskz_fixupimm_ps(__mmask8 __U, __m256 __A, __m256 __B, __m256i __C) {
+  // CHECK-LABEL: @test_mm256_maskz_fixupimm_ps
+  // CHECK: @llvm.x86.avx512.maskz.fixupimm.ps.256
+  return _mm256_maskz_fixupimm_ps(__U, __A, __B, __C, 5); 
+}
+
+__m128d test_mm_mask_load_pd(__m128d __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_mask_load_pd
+  // CHECK: @llvm.x86.avx512.mask.load.pd.128
+  return _mm_mask_load_pd(__W, __U, __P); 
+}
+
+__m128d test_mm_maskz_load_pd(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_maskz_load_pd
+  // CHECK: @llvm.x86.avx512.mask.load.pd.128
+  return _mm_maskz_load_pd(__U, __P); 
+}
+
+__m256d test_mm256_mask_load_pd(__m256d __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_mask_load_pd
+  // CHECK: @llvm.x86.avx512.mask.load.pd.256
+  return _mm256_mask_load_pd(__W, __U, __P); 
+}
+
+__m256d test_mm256_maskz_load_pd(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_maskz_load_pd
+  // CHECK: @llvm.x86.avx512.mask.load.pd.256
+  return _mm256_maskz_load_pd(__U, __P); 
+}
+
+__m128 test_mm_mask_load_ps(__m128 __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_mask_load_ps
+  // CHECK: @llvm.x86.avx512.mask.load.ps.128
+  return _mm_mask_load_ps(__W, __U, __P); 
+}
+
+__m128 test_mm_maskz_load_ps(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_maskz_load_ps
+  // CHECK: @llvm.x86.avx512.mask.load.ps.128
+  return _mm_maskz_load_ps(__U, __P); 
+}
+
+__m256 test_mm256_mask_load_ps(__m256 __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_mask_load_ps
+  // CHECK: @llvm.x86.avx512.mask.load.ps.256
+  return _mm256_mask_load_ps(__W, __U, __P); 
+}
+
+__m256 test_mm256_maskz_load_ps(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_maskz_load_ps
+  // CHECK: @llvm.x86.avx512.mask.load.ps.256
+  return _mm256_maskz_load_ps(__U, __P); 
+}
+
+__m128i test_mm_mask_loadu_epi64(__m128i __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_mask_loadu_epi64
+  // CHECK: @llvm.x86.avx512.mask.loadu.q.128
+  return _mm_mask_loadu_epi64(__W, __U, __P); 
+}
+
+__m128i test_mm_maskz_loadu_epi64(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_maskz_loadu_epi64
+  // CHECK: @llvm.x86.avx512.mask.loadu.q.128
+  return _mm_maskz_loadu_epi64(__U, __P); 
+}
+
+__m256i test_mm256_mask_loadu_epi64(__m256i __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_mask_loadu_epi64
+  // CHECK: @llvm.x86.avx512.mask.loadu.q.256
+  return _mm256_mask_loadu_epi64(__W, __U, __P); 
+}
+
+__m256i test_mm256_maskz_loadu_epi64(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_maskz_loadu_epi64
+  // CHECK: @llvm.x86.avx512.mask.loadu.q.256
+  return _mm256_maskz_loadu_epi64(__U, __P); 
+}
+
+__m128i test_mm_mask_loadu_epi32(__m128i __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_mask_loadu_epi32
+  // CHECK: @llvm.x86.avx512.mask.loadu.d.128
+  return _mm_mask_loadu_epi32(__W, __U, __P); 
+}
+
+__m128i test_mm_maskz_loadu_epi32(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_maskz_loadu_epi32
+  // CHECK: @llvm.x86.avx512.mask.loadu.d.128
+  return _mm_maskz_loadu_epi32(__U, __P); 
+}
+
+__m256i test_mm256_mask_loadu_epi32(__m256i __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_mask_loadu_epi32
+  // CHECK: @llvm.x86.avx512.mask.loadu.d.256
+  return _mm256_mask_loadu_epi32(__W, __U, __P); 
+}
+
+__m256i test_mm256_maskz_loadu_epi32(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_maskz_loadu_epi32
+  // CHECK: @llvm.x86.avx512.mask.loadu.d.256
+  return _mm256_maskz_loadu_epi32(__U, __P); 
+}
+
+__m128d test_mm_mask_loadu_pd(__m128d __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_mask_loadu_pd
+  // CHECK: @llvm.x86.avx512.mask.loadu.pd.128
+  return _mm_mask_loadu_pd(__W, __U, __P); 
+}
+
+__m128d test_mm_maskz_loadu_pd(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_maskz_loadu_pd
+  // CHECK: @llvm.x86.avx512.mask.loadu.pd.128
+  return _mm_maskz_loadu_pd(__U, __P); 
+}
+
+__m256d test_mm256_mask_loadu_pd(__m256d __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_mask_loadu_pd
+  // CHECK: @llvm.x86.avx512.mask.loadu.pd.256
+  return _mm256_mask_loadu_pd(__W, __U, __P); 
+}
+
+__m256d test_mm256_maskz_loadu_pd(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_maskz_loadu_pd
+  // CHECK: @llvm.x86.avx512.mask.loadu.pd.256
+  return _mm256_maskz_loadu_pd(__U, __P); 
+}
+
+__m128 test_mm_mask_loadu_ps(__m128 __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_mask_loadu_ps
+  // CHECK: @llvm.x86.avx512.mask.loadu.ps.128
+  return _mm_mask_loadu_ps(__W, __U, __P); 
+}
+
+__m128 test_mm_maskz_loadu_ps(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_maskz_loadu_ps
+  // CHECK: @llvm.x86.avx512.mask.loadu.ps.128
+  return _mm_maskz_loadu_ps(__U, __P); 
+}
+
+__m256 test_mm256_mask_loadu_ps(__m256 __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_mask_loadu_ps
+  // CHECK: @llvm.x86.avx512.mask.loadu.ps.256
+  return _mm256_mask_loadu_ps(__W, __U, __P); 
+}
+
+__m256 test_mm256_maskz_loadu_ps(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_maskz_loadu_ps
+  // CHECK: @llvm.x86.avx512.mask.loadu.ps.256
+  return _mm256_maskz_loadu_ps(__U, __P); 
+}
+

cfe/trunk/test/CodeGen/avx512vlbw-builtins.c

 }
 
 __m256i test_mm256_maskz_mov_epi8(__mmask32 __U, __m256i __A) {
   // CHECK-LABEL: @test_mm256_maskz_mov_epi8
   // CHECK: @llvm.x86.avx512.mask.movu.b.256
   return _mm256_maskz_mov_epi8(__U, __A); 
 }
 
+__m128i test_mm_mask_loadu_epi16(__m128i __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_mask_loadu_epi16
+  // CHECK: @llvm.x86.avx512.mask.loadu.w.128
+  return _mm_mask_loadu_epi16(__W, __U, __P); 
+}
+
+__m128i test_mm_maskz_loadu_epi16(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_maskz_loadu_epi16
+  // CHECK: @llvm.x86.avx512.mask.loadu.w.128
+  return _mm_maskz_loadu_epi16(__U, __P); 
+}
+
+__m256i test_mm256_mask_loadu_epi16(__m256i __W, __mmask16 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_mask_loadu_epi16
+  // CHECK: @llvm.x86.avx512.mask.loadu.w.256
+  return _mm256_mask_loadu_epi16(__W, __U, __P); 
+}
+
+__m256i test_mm256_maskz_loadu_epi16(__mmask16 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_maskz_loadu_epi16
+  // CHECK: @llvm.x86.avx512.mask.loadu.w.256
+  return _mm256_maskz_loadu_epi16(__U, __P); 
+}
+
+__m128i test_mm_mask_loadu_epi8(__m128i __W, __mmask16 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_mask_loadu_epi8
+  // CHECK: @llvm.x86.avx512.mask.loadu.b.128
+  return _mm_mask_loadu_epi8(__W, __U, __P); 
+}
+
+__m128i test_mm_maskz_loadu_epi8(__mmask16 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm_maskz_loadu_epi8
+  // CHECK: @llvm.x86.avx512.mask.loadu.b.128
+  return _mm_maskz_loadu_epi8(__U, __P); 
+}
+
+__m256i test_mm256_mask_loadu_epi8(__m256i __W, __mmask32 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_mask_loadu_epi8
+  // CHECK: @llvm.x86.avx512.mask.loadu.b.256
+  return _mm256_mask_loadu_epi8(__W, __U, __P); 
+}
+
+__m256i test_mm256_maskz_loadu_epi8(__mmask32 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm256_maskz_loadu_epi8
+  // CHECK: @llvm.x86.avx512.mask.loadu.b.256
+  return _mm256_maskz_loadu_epi8(__U, __P); 
+}
+