[Clang][AVX512][BuiltIn] Adding avx512 ( vperm{i|t}2var, vpermil{var}{ps|pd}{256|512} ) builtin to clang.

[delena]

test/CodeGen/avx512f-builtins.c

2606 ↗

(On Diff #53138)

What is 4? Current?

test/CodeGen/avx512vl-builtins.c

4302 ↗

(On Diff #53138)

I don't understand round mode in shuffles. What instruction is generated with non-current mode?

[m_zuckerman]

test/CodeGen/avx512f-builtins.c

2606 ↗

(On Diff #53138)

Not is an imm8 . All permute function gets imm8 argument as the last argument .

test/CodeGen/avx512vl-builtins.c

4302 ↗

(On Diff #53153)

This is not a round mode, this function get imm8

"Shuffle double-precision (64-bit) floating-point elements in a using the control in imm8, and store the results in dst using writemask k (elements are copied from src when the corresponding mask bit is not set)."

This was a mistake in replacing 4 in non-current mode.

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

 TARGET_BUILTIN(__builtin_ia32_vcvttsd2si32, "iV2dIi","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_vcvttsd2si64, "LLiV2dIi","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_vcvttsd2usi32, "UiV2dIi","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_vcvttsd2usi64, "ULLiV2dIi","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_vcvttss2si32, "iV4fIi","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_vcvttss2si64, "LLiV4fIi","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_vcvttss2usi32, "UiV4fIi","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_vcvttss2usi64, "ULLiV4fIi","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_vpermi2vard512_mask, "V16iV16iV16iV16iUs","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_vpermi2varpd512_mask, "V8dV8dV8LLiV8dUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_vpermi2varps512_mask, "V16fV16fV16iV16fUs","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_vpermi2varq512_mask, "V8LLiV8LLiV8LLiV8LLiUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_vpermilpd512_mask, "V8dV8dIiV8dUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_vpermilps512_mask, "V16fV16fIiV16fUs","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_vpermilvarpd512_mask, "V8dV8dV8LLiV8dUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_vpermilvarps512_mask, "V16fV16fV16iV16fUs","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_vpermt2vard512_maskz, "V16iV16iV16iV16iUs","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_vpermt2varpd512_maskz, "V8dV8LLiV8dV8dUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_vpermt2varps512_maskz, "V16fV16iV16fV16fUs","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_vpermt2varq512_maskz, "V8LLiV8LLiV8LLiV8LLiUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_vpermilpd_mask, "V2dV2dIiV2dUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_vpermilpd256_mask, "V4dV4dIiV4dUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_vpermilps_mask, "V4fV4fIiV4fUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_vpermilps256_mask, "V8fV8fIiV8fUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_vpermilvarpd_mask, "V2dV2dV2LLiV2dUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_vpermilvarpd256_mask, "V4dV4dV4LLiV4dUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_vpermilvarps_mask, "V4fV4fV4iV4fUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_vpermilvarps256_mask, "V8fV8fV8iV8fUc","","avx512vl")
 
 #undef BUILTIN
 #undef TARGET_BUILTIN

cfe/trunk/lib/Headers/avx512fintrin.h

_mm512_mask_unpackhi_pd (__m512d __W, __mmask8 __U, __m512d __A, __m512d __B)
   return (__m512d) __builtin_ia32_unpckhpd512_mask ((__v8df) __A,
                 (__v8df) __B,
                 (__v8df) __W,
                 (__mmask8) __U);
 }
 #define _mm_cvt_roundsd_si64( __A, __R) __extension__ ({ \
 __builtin_ia32_vcvtsd2si64 ((__v2df)( __A),( __R));\
 })
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_mask2_permutex2var_epi32 (__m512i __A, __m512i __I,
+         __mmask16 __U, __m512i __B)
+{
+  return (__m512i) __builtin_ia32_vpermi2vard512_mask ((__v16si) __A,
+                   (__v16si) __I
+                   /* idx */ ,
+                   (__v16si) __B,
+                   (__mmask16) __U);
+}
 
 static __inline__ __m512d __DEFAULT_FN_ATTRS
 _mm512_maskz_unpackhi_pd (__mmask8 __U, __m512d __A, __m512d __B)
 {
   return (__m512d) __builtin_ia32_unpckhpd512_mask ((__v8df) __A,
                 (__v8df) __B,
                 (__v8df)
                 _mm512_setzero_pd (),

 static __inline__ unsigned long long __DEFAULT_FN_ATTRS
 _mm_cvttss_u64 (__m128 __A)
 {
   return (unsigned long long) __builtin_ia32_vcvttss2usi64 ((__v4sf)
                   __A,
                   _MM_FROUND_CUR_DIRECTION);
 }
 
+static __inline__ __m512d __DEFAULT_FN_ATTRS
+_mm512_mask2_permutex2var_pd (__m512d __A, __m512i __I, __mmask8 __U,
+            __m512d __B)
+{
+  return (__m512d) __builtin_ia32_vpermi2varpd512_mask ((__v8df) __A,
+              (__v8di) __I
+              /* idx */ ,
+              (__v8df) __B,
+              (__mmask8) __U);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS
+_mm512_mask2_permutex2var_ps (__m512 __A, __m512i __I, __mmask16 __U,
+            __m512 __B)
+{
+  return (__m512) __builtin_ia32_vpermi2varps512_mask ((__v16sf) __A,
+                   (__v16si) __I
+                   /* idx */ ,
+                   (__v16sf) __B,
+                   (__mmask16) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_mask2_permutex2var_epi64 (__m512i __A, __m512i __I,
+         __mmask8 __U, __m512i __B)
+{
+  return (__m512i) __builtin_ia32_vpermi2varq512_mask ((__v8di) __A,
+                   (__v8di) __I
+                   /* idx */ ,
+                   (__v8di) __B,
+                   (__mmask8) __U);
+}
+
+#define _mm512_permute_pd( __X, __C) __extension__ ({ \
+__builtin_ia32_vpermilpd512_mask ((__v8df)( __X),( __C),\
+                 (__v8df)\
+                 _mm512_undefined_pd (),\
+                 (__mmask8) -1);\
+})
+
+#define _mm512_mask_permute_pd( __W, __U, __X, __C) __extension__ ({ \
+__builtin_ia32_vpermilpd512_mask ((__v8df)( __X),( __C),\
+                 (__v8df)( __W),\
+                 (__mmask8)( __U));\
+})
+
+#define _mm512_maskz_permute_pd( __U, __X, __C) __extension__ ({ \
+__builtin_ia32_vpermilpd512_mask ((__v8df)( __X),( __C),\
+                 (__v8df)\
+                 _mm512_setzero_pd (),\
+                 (__mmask8)( __U));\
+})
+
+#define _mm512_permute_ps( __X, __C) __extension__ ({ \
+__builtin_ia32_vpermilps512_mask ((__v16sf)( __X),( __C),\
+                (__v16sf)\
+                _mm512_undefined_ps (),\
+                (__mmask16) -1);\
+})
+
+#define _mm512_mask_permute_ps( __W, __U, __X, __C) __extension__ ({ \
+__builtin_ia32_vpermilps512_mask ((__v16sf)( __X),( __C),\
+                (__v16sf)( __W),\
+                (__mmask16)( __U));\
+})
+
+#define _mm512_maskz_permute_ps( __U, __X, __C) __extension__ ({ \
+__builtin_ia32_vpermilps512_mask ((__v16sf)( __X),( __C),\
+                (__v16sf)\
+                _mm512_setzero_ps (),\
+                (__mmask16)( __U));\
+})
+
+static __inline__ __m512d __DEFAULT_FN_ATTRS
+_mm512_permutevar_pd (__m512d __A, __m512i __C)
+{
+  return (__m512d) __builtin_ia32_vpermilvarpd512_mask ((__v8df) __A,
+              (__v8di) __C,
+              (__v8df)
+              _mm512_undefined_pd (),
+              (__mmask8) -1);
+}
+
+static __inline__ __m512d __DEFAULT_FN_ATTRS
+_mm512_mask_permutevar_pd (__m512d __W, __mmask8 __U, __m512d __A, __m512i __C)
+{
+  return (__m512d) __builtin_ia32_vpermilvarpd512_mask ((__v8df) __A,
+              (__v8di) __C,
+              (__v8df) __W,
+              (__mmask8) __U);
+}
+
+static __inline__ __m512d __DEFAULT_FN_ATTRS
+_mm512_maskz_permutevar_pd (__mmask8 __U, __m512d __A, __m512i __C)
+{
+  return (__m512d) __builtin_ia32_vpermilvarpd512_mask ((__v8df) __A,
+              (__v8di) __C,
+              (__v8df)
+              _mm512_setzero_pd (),
+              (__mmask8) __U);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS
+_mm512_permutevar_ps (__m512 __A, __m512i __C)
+{
+  return (__m512) __builtin_ia32_vpermilvarps512_mask ((__v16sf) __A,
+                   (__v16si) __C,
+                   (__v16sf)
+                   _mm512_undefined_ps (),
+                   (__mmask16) -1);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS
+_mm512_mask_permutevar_ps (__m512 __W, __mmask16 __U, __m512 __A, __m512i __C)
+{
+  return (__m512) __builtin_ia32_vpermilvarps512_mask ((__v16sf) __A,
+                   (__v16si) __C,
+                   (__v16sf) __W,
+                   (__mmask16) __U);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS
+_mm512_maskz_permutevar_ps (__mmask16 __U, __m512 __A, __m512i __C)
+{
+  return (__m512) __builtin_ia32_vpermilvarps512_mask ((__v16sf) __A,
+                   (__v16si) __C,
+                   (__v16sf)
+                   _mm512_setzero_ps (),
+                   (__mmask16) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_maskz_permutex2var_epi32 (__mmask16 __U, __m512i __A,
+         __m512i __I, __m512i __B)
+{
+  return (__m512i) __builtin_ia32_vpermt2vard512_maskz ((__v16si) __I
+              /* idx */ ,
+              (__v16si) __A,
+              (__v16si) __B,
+              (__mmask16) __U);
+}
+
+static __inline__ __m512d __DEFAULT_FN_ATTRS
+_mm512_maskz_permutex2var_pd (__mmask8 __U, __m512d __A, __m512i __I,
+            __m512d __B)
+{
+  return (__m512d) __builtin_ia32_vpermt2varpd512_maskz ((__v8di) __I
+               /* idx */ ,
+               (__v8df) __A,
+               (__v8df) __B,
+               (__mmask8) __U);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS
+_mm512_maskz_permutex2var_ps (__mmask16 __U, __m512 __A, __m512i __I,
+            __m512 __B)
+{
+  return (__m512) __builtin_ia32_vpermt2varps512_maskz ((__v16si) __I
+              /* idx */ ,
+              (__v16sf) __A,
+              (__v16sf) __B,
+              (__mmask16) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_maskz_permutex2var_epi64 (__mmask8 __U, __m512i __A,
+         __m512i __I, __m512i __B)
+{
+  return (__m512i) __builtin_ia32_vpermt2varq512_maskz ((__v8di) __I
+              /* idx */ ,
+              (__v8di) __A,
+              (__v8di) __B,
+              (__mmask8) __U);
+}
+
 #undef __DEFAULT_FN_ATTRS
 
 #endif // __AVX512FINTRIN_H

cfe/trunk/lib/Headers/avx512vlintrin.h

 _mm256_maskz_lzcnt_epi64 (__mmask8 __U, __m256i __A)
 {
   return (__m256i) __builtin_ia32_vplzcntq_256_mask ((__v4di) __A,
                  (__v4di)
                  _mm256_setzero_si256 (),
                  (__mmask8) __U);
 }
 
+
+#define _mm_mask_permute_pd( __W, __U, __X, __C) __extension__ ({ \
+__builtin_ia32_vpermilpd_mask ((__v2df)( __X),( __C),\
+              (__v2df)( __W),\
+              (__mmask8)( __U));\
+})
+
+#define _mm_maskz_permute_pd( __U, __X, __C) __extension__ ({ \
+__builtin_ia32_vpermilpd_mask ((__v2df)( __X),( __C),\
+              (__v2df)\
+              _mm_setzero_pd (),\
+              (__mmask8)( __U));\
+})
+
+#define _mm256_mask_permute_pd( __W, __U, __X, __C) __extension__ ({ \
+__builtin_ia32_vpermilpd256_mask ((__v4df)( __X),( __C),\
+                 (__v4df)( __W),\
+                 (__mmask8)( __U));\
+})
+
+#define _mm256_maskz_permute_pd( __U, __X, __C) __extension__ ({ \
+__builtin_ia32_vpermilpd256_mask ((__v4df)( __X),( __C),\
+                 (__v4df)\
+                 _mm256_setzero_pd (),\
+                 (__mmask8)( __U));\
+})
+
+#define _mm_mask_permute_ps( __W, __U, __X, __C) __extension__ ({ \
+__builtin_ia32_vpermilps_mask ((__v4sf)( __X),( __C),\
+             (__v4sf)( __W),\
+             (__mmask8)( __U));\
+})
+
+#define _mm_maskz_permute_ps( __U, __X, __C) __extension__ ({ \
+__builtin_ia32_vpermilps_mask ((__v4sf)( __X),( __C),\
+             (__v4sf)\
+             _mm_setzero_ps (),\
+             (__mmask8)( __U));\
+})
+
+#define _mm256_mask_permute_ps( __W, __U, __X, __C) __extension__ ({ \
+__builtin_ia32_vpermilps256_mask ((__v8sf)( __X),( __C),\
+                (__v8sf)( __W),\
+                (__mmask8)( __U));\
+})
+
+#define _mm256_maskz_permute_ps( __U, __X, __C) __extension__ ({ \
+__builtin_ia32_vpermilps256_mask ((__v8sf)( __X),( __C),\
+                (__v8sf)\
+                _mm256_setzero_ps (),\
+                (__mmask8)( __U));\
+})
+
+static __inline__ __m128d __DEFAULT_FN_ATTRS
+_mm_mask_permutevar_pd (__m128d __W, __mmask8 __U, __m128d __A,
+      __m128i __C)
+{
+  return (__m128d) __builtin_ia32_vpermilvarpd_mask ((__v2df) __A,
+                 (__v2di) __C,
+                 (__v2df) __W,
+                 (__mmask8) __U);
+}
+
+static __inline__ __m128d __DEFAULT_FN_ATTRS
+_mm_maskz_permutevar_pd (__mmask8 __U, __m128d __A, __m128i __C)
+{
+  return (__m128d) __builtin_ia32_vpermilvarpd_mask ((__v2df) __A,
+                 (__v2di) __C,
+                 (__v2df)
+                 _mm_setzero_pd (),
+                 (__mmask8) __U);
+}
+
+static __inline__ __m256d __DEFAULT_FN_ATTRS
+_mm256_mask_permutevar_pd (__m256d __W, __mmask8 __U, __m256d __A,
+         __m256i __C)
+{
+  return (__m256d) __builtin_ia32_vpermilvarpd256_mask ((__v4df) __A,
+              (__v4di) __C,
+              (__v4df) __W,
+              (__mmask8)
+              __U);
+}
+
+static __inline__ __m256d __DEFAULT_FN_ATTRS
+_mm256_maskz_permutevar_pd (__mmask8 __U, __m256d __A, __m256i __C)
+{
+  return (__m256d) __builtin_ia32_vpermilvarpd256_mask ((__v4df) __A,
+              (__v4di) __C,
+              (__v4df)
+              _mm256_setzero_pd (),
+              (__mmask8)
+              __U);
+}
+
+static __inline__ __m128 __DEFAULT_FN_ATTRS
+_mm_mask_permutevar_ps (__m128 __W, __mmask8 __U, __m128 __A,
+      __m128i __C)
+{
+  return (__m128) __builtin_ia32_vpermilvarps_mask ((__v4sf) __A,
+                (__v4si) __C,
+                (__v4sf) __W,
+                (__mmask8) __U);
+}
+
+static __inline__ __m128 __DEFAULT_FN_ATTRS
+_mm_maskz_permutevar_ps (__mmask8 __U, __m128 __A, __m128i __C)
+{
+  return (__m128) __builtin_ia32_vpermilvarps_mask ((__v4sf) __A,
+                (__v4si) __C,
+                (__v4sf)
+                _mm_setzero_ps (),
+                (__mmask8) __U);
+}
+
+static __inline__ __m256 __DEFAULT_FN_ATTRS
+_mm256_mask_permutevar_ps (__m256 __W, __mmask8 __U, __m256 __A,
+         __m256i __C)
+{
+  return (__m256) __builtin_ia32_vpermilvarps256_mask ((__v8sf) __A,
+                   (__v8si) __C,
+                   (__v8sf) __W,
+                   (__mmask8) __U);
+}
+
+static __inline__ __m256 __DEFAULT_FN_ATTRS
+_mm256_maskz_permutevar_ps (__mmask8 __U, __m256 __A, __m256i __C)
+{
+  return (__m256) __builtin_ia32_vpermilvarps256_mask ((__v8sf) __A,
+                   (__v8si) __C,
+                   (__v8sf)
+                   _mm256_setzero_ps (),
+                   (__mmask8) __U);
+}
+
 #undef __DEFAULT_FN_ATTRS
 #undef __DEFAULT_FN_ATTRS_BOTH
 
 #endif /* __AVX512VLINTRIN_H */

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

__m512d test_mm512_mask_unpackhi_pd(__m512d __W, __mmask8 __U, __m512d __A, __m512d __B) {
   // CHECK: @llvm.x86.avx512.mask.unpckh.pd.512
   return _mm512_mask_unpackhi_pd(__W, __U, __A, __B); 
 }
 unsigned long long test_mm_cvt_roundsd_si64(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvt_roundsd_si64
   // CHECK: @llvm.x86.avx512.vcvtsd2si64
   return _mm_cvt_roundsd_si64(__A, _MM_FROUND_CUR_DIRECTION); 
 }
+__m512i test_mm512_mask2_permutex2var_epi32(__m512i __A, __m512i __I, __mmask16 __U, __m512i __B) {
+  // CHECK-LABEL: @test_mm512_mask2_permutex2var_epi32
+  // CHECK: @llvm.x86.avx512.mask.vpermi2var.d.512
+  return _mm512_mask2_permutex2var_epi32(__A, __I, __U, __B); 
+}
 
 __m512d test_mm512_maskz_unpackhi_pd(__mmask8 __U, __m512d __A, __m512d __B) {
   // CHECK-LABEL: @test_mm512_maskz_unpackhi_pd
   // CHECK: @llvm.x86.avx512.mask.unpckh.pd.512
   return _mm512_maskz_unpackhi_pd(__U, __A, __B); 
 }
 long long test_mm_cvt_roundsd_i64(__m128d __A) {
   // CHECK-LABEL: @test_mm_cvt_roundsd_i64
   // CHECK: @llvm.x86.avx512.vcvtsd2si64
   return _mm_cvt_roundsd_i64(__A, _MM_FROUND_CUR_DIRECTION); 
 }
+__m512d test_mm512_mask2_permutex2var_pd(__m512d __A, __m512i __I, __mmask8 __U, __m512d __B) {
+  // CHECK-LABEL: @test_mm512_mask2_permutex2var_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermi2var.pd.512
+  return _mm512_mask2_permutex2var_pd(__A, __I, __U, __B); 
+}
 
 __m512 test_mm512_mask_unpackhi_ps(__m512 __W, __mmask16 __U, __m512 __A, __m512 __B) {
   // CHECK-LABEL: @test_mm512_mask_unpackhi_ps
   // CHECK: @llvm.x86.avx512.mask.unpckh.ps.512
   return _mm512_mask_unpackhi_ps(__W, __U, __A, __B); 
 }
 
 __m512 test_mm512_maskz_unpackhi_ps(__mmask16 __U, __m512 __A, __m512 __B) {
unsigned long long test_mm_cvtt_roundss_u64(__m128 __A) {
   return _mm_cvtt_roundss_u64(__A, _MM_FROUND_CUR_DIRECTION); 
 }
 
 unsigned long long test_mm_cvttss_u64(__m128 __A) {
   // CHECK-LABEL: @test_mm_cvttss_u64
   // CHECK: @llvm.x86.avx512.cvttss2usi64
   return _mm_cvttss_u64(__A); 
 }
+__m512 test_mm512_mask2_permutex2var_ps(__m512 __A, __m512i __I, __mmask16 __U, __m512 __B) {
+  // CHECK-LABEL: @test_mm512_mask2_permutex2var_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermi2var.ps.512
+  return _mm512_mask2_permutex2var_ps(__A, __I, __U, __B); 
+}
+
+__m512i test_mm512_mask2_permutex2var_epi64(__m512i __A, __m512i __I, __mmask8 __U, __m512i __B) {
+  // CHECK-LABEL: @test_mm512_mask2_permutex2var_epi64
+  // CHECK: @llvm.x86.avx512.mask.vpermi2var.q.512
+  return _mm512_mask2_permutex2var_epi64(__A, __I, __U, __B); 
+}
+
+__m512d test_mm512_permute_pd(__m512d __X) {
+  // CHECK-LABEL: @test_mm512_permute_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermil.pd.512
+  return _mm512_permute_pd(__X, 2); 
+}
+
+__m512d test_mm512_mask_permute_pd(__m512d __W, __mmask8 __U, __m512d __X) {
+  // CHECK-LABEL: @test_mm512_mask_permute_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermil.pd.512
+  return _mm512_mask_permute_pd(__W, __U, __X, 2); 
+}
+
+__m512d test_mm512_maskz_permute_pd(__mmask8 __U, __m512d __X) {
+  // CHECK-LABEL: @test_mm512_maskz_permute_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermil.pd.512
+  return _mm512_maskz_permute_pd(__U, __X, 2); 
+}
+
+__m512 test_mm512_permute_ps(__m512 __X) {
+  // CHECK-LABEL: @test_mm512_permute_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermil.ps.512
+  return _mm512_permute_ps(__X, 2); 
+}
+
+__m512 test_mm512_mask_permute_ps(__m512 __W, __mmask16 __U, __m512 __X) {
+  // CHECK-LABEL: @test_mm512_mask_permute_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermil.ps.512
+  return _mm512_mask_permute_ps(__W, __U, __X, 2); 
+}
+
+__m512 test_mm512_maskz_permute_ps(__mmask16 __U, __m512 __X) {
+  // CHECK-LABEL: @test_mm512_maskz_permute_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermil.ps.512
+  return _mm512_maskz_permute_ps(__U, __X, 2); 
+}
+
+__m512d test_mm512_permutevar_pd(__m512d __A, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_permutevar_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermilvar.pd.512
+  return _mm512_permutevar_pd(__A, __C); 
+}
+
+__m512d test_mm512_mask_permutevar_pd(__m512d __W, __mmask8 __U, __m512d __A, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_mask_permutevar_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermilvar.pd.512
+  return _mm512_mask_permutevar_pd(__W, __U, __A, __C); 
+}
+
+__m512d test_mm512_maskz_permutevar_pd(__mmask8 __U, __m512d __A, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_maskz_permutevar_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermilvar.pd.512
+  return _mm512_maskz_permutevar_pd(__U, __A, __C); 
+}
+
+__m512 test_mm512_permutevar_ps(__m512 __A, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_permutevar_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermilvar.ps.512
+  return _mm512_permutevar_ps(__A, __C); 
+}
+
+__m512 test_mm512_mask_permutevar_ps(__m512 __W, __mmask16 __U, __m512 __A, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_mask_permutevar_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermilvar.ps.512
+  return _mm512_mask_permutevar_ps(__W, __U, __A, __C); 
+}
+
+__m512 test_mm512_maskz_permutevar_ps(__mmask16 __U, __m512 __A, __m512i __C) {
+  // CHECK-LABEL: @test_mm512_maskz_permutevar_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermilvar.ps.512
+  return _mm512_maskz_permutevar_ps(__U, __A, __C); 
+}
+
+__m512i test_mm512_maskz_permutex2var_epi32(__mmask16 __U, __m512i __A, __m512i __I, __m512i __B) {
+  // CHECK-LABEL: @test_mm512_maskz_permutex2var_epi32
+  // CHECK: @llvm.x86.avx512.maskz.vpermt2var.d.512
+  return _mm512_maskz_permutex2var_epi32(__U, __A, __I, __B); 
+}
+
+__m512d test_mm512_maskz_permutex2var_pd(__mmask8 __U, __m512d __A, __m512i __I, __m512d __B) {
+  // CHECK-LABEL: @test_mm512_maskz_permutex2var_pd
+  // CHECK: @llvm.x86.avx512.maskz.vpermt2var.pd.512
+  return _mm512_maskz_permutex2var_pd(__U, __A, __I, __B); 
+}
+
+__m512 test_mm512_maskz_permutex2var_ps(__mmask16 __U, __m512 __A, __m512i __I, __m512 __B) {
+  // CHECK-LABEL: @test_mm512_maskz_permutex2var_ps
+  // CHECK: @llvm.x86.avx512.maskz.vpermt2var.ps.512
+  return _mm512_maskz_permutex2var_ps(__U, __A, __I, __B); 
+}
+
+__m512i test_mm512_maskz_permutex2var_epi64(__mmask8 __U, __m512i __A, __m512i __I, __m512i __B) {
+  // CHECK-LABEL: @test_mm512_maskz_permutex2var_epi64
+  // CHECK: @llvm.x86.avx512.maskz.vpermt2var.q.512
+  return _mm512_maskz_permutex2var_epi64(__U, __A, __I, __B);
+}

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

 
 __m256i test_mm256_maskz_lzcnt_epi64(__mmask8 __U, __m256i __A) {
   // CHECK-LABEL: @test_mm256_maskz_lzcnt_epi64
   // CHECK: @llvm.x86.avx512.mask.lzcnt.q
   return _mm256_maskz_lzcnt_epi64(__U, __A); 
 }
 
 
+__m128d test_mm_mask_permute_pd(__m128d __W, __mmask8 __U, __m128d __X) {
+  // CHECK-LABEL: @test_mm_mask_permute_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermil.pd
+  return _mm_mask_permute_pd(__W, __U, __X, 2); 
+}
+
+__m128d test_mm_maskz_permute_pd(__mmask8 __U, __m128d __X) {
+  // CHECK-LABEL: @test_mm_maskz_permute_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermil.pd
+  return _mm_maskz_permute_pd(__U, __X, 2); 
+}
+
+__m256d test_mm256_mask_permute_pd(__m256d __W, __mmask8 __U, __m256d __X) {
+  // CHECK-LABEL: @test_mm256_mask_permute_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermil.pd.256
+  return _mm256_mask_permute_pd(__W, __U, __X, 2); 
+}
+
+__m256d test_mm256_maskz_permute_pd(__mmask8 __U, __m256d __X) {
+  // CHECK-LABEL: @test_mm256_maskz_permute_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermil.pd.256
+  return _mm256_maskz_permute_pd(__U, __X, 2); 
+}
+
+__m128 test_mm_mask_permute_ps(__m128 __W, __mmask8 __U, __m128 __X) {
+  // CHECK-LABEL: @test_mm_mask_permute_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermil.ps
+  return _mm_mask_permute_ps(__W, __U, __X, 2); 
+}
+
+__m128 test_mm_maskz_permute_ps(__mmask8 __U, __m128 __X) {
+  // CHECK-LABEL: @test_mm_maskz_permute_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermil.ps
+  return _mm_maskz_permute_ps(__U, __X, 2); 
+}
+
+__m256 test_mm256_mask_permute_ps(__m256 __W, __mmask8 __U, __m256 __X) {
+  // CHECK-LABEL: @test_mm256_mask_permute_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermil.ps.256
+  return _mm256_mask_permute_ps(__W, __U, __X, 2); 
+}
+
+__m256 test_mm256_maskz_permute_ps(__mmask8 __U, __m256 __X) {
+  // CHECK-LABEL: @test_mm256_maskz_permute_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermil.ps.256
+  return _mm256_maskz_permute_ps(__U, __X, 2); 
+}
+
+__m128d test_mm_mask_permutevar_pd(__m128d __W, __mmask8 __U, __m128d __A, __m128i __C) {
+  // CHECK-LABEL: @test_mm_mask_permutevar_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermilvar.pd
+  return _mm_mask_permutevar_pd(__W, __U, __A, __C); 
+}
+
+__m128d test_mm_maskz_permutevar_pd(__mmask8 __U, __m128d __A, __m128i __C) {
+  // CHECK-LABEL: @test_mm_maskz_permutevar_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermilvar.pd
+  return _mm_maskz_permutevar_pd(__U, __A, __C); 
+}
+
+__m256d test_mm256_mask_permutevar_pd(__m256d __W, __mmask8 __U, __m256d __A, __m256i __C) {
+  // CHECK-LABEL: @test_mm256_mask_permutevar_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermilvar.pd.256
+  return _mm256_mask_permutevar_pd(__W, __U, __A, __C); 
+}
+
+__m256d test_mm256_maskz_permutevar_pd(__mmask8 __U, __m256d __A, __m256i __C) {
+  // CHECK-LABEL: @test_mm256_maskz_permutevar_pd
+  // CHECK: @llvm.x86.avx512.mask.vpermilvar.pd.256
+  return _mm256_maskz_permutevar_pd(__U, __A, __C); 
+}
+
+__m128 test_mm_mask_permutevar_ps(__m128 __W, __mmask8 __U, __m128 __A, __m128i __C) {
+  // CHECK-LABEL: @test_mm_mask_permutevar_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermilvar.ps
+  return _mm_mask_permutevar_ps(__W, __U, __A, __C); 
+}
+
+__m128 test_mm_maskz_permutevar_ps(__mmask8 __U, __m128 __A, __m128i __C) {
+  // CHECK-LABEL: @test_mm_maskz_permutevar_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermilvar.ps
+  return _mm_maskz_permutevar_ps(__U, __A, __C); 
+}
+
+__m256 test_mm256_mask_permutevar_ps(__m256 __W, __mmask8 __U, __m256 __A, __m256i __C) {
+  // CHECK-LABEL: @test_mm256_mask_permutevar_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermilvar.ps.256
+  return _mm256_mask_permutevar_ps(__W, __U, __A, __C); 
+}
+
+__m256 test_mm256_maskz_permutevar_ps(__mmask8 __U, __m256 __A, __m256i __C) {
+  // CHECK-LABEL: @test_mm256_maskz_permutevar_ps
+  // CHECK: @llvm.x86.avx512.mask.vpermilvar.ps.256
+  return _mm256_maskz_permutevar_ps(__U, __A, __C); 
+}