[Clang][AVX512] completing missing intrinsics [load/store].

[AsafBadouh]

test/CodeGen/avx512f-builtins.c

94

please remove the "return" from all void tests

97

please add CHECK for the store

106

please add CHECK for the store

115

please add CHECK for the store

237

you have it already in line:189

266

you have it already in line:244

[AsafBadouh]

LGTM

lib/Headers/avx512fintrin.h

_mm512_test_epi64_mask(__m512i __A, __m512i __B)
   return (__mmask8) __builtin_ia32_ptestmq512 ((__v8di) __A,
                  (__v8di) __B,
                  (__mmask8) -1);
 }
 
 /* SIMD load ops */
 
 static __inline __m512i __DEFAULT_FN_ATTRS
+_mm512_loadu_si512 (void const *__P)
+{
+  return (__m512i) __builtin_ia32_loaddqusi512_mask ((const __v16si *) __P,
+                  (__v16si)
+                  _mm512_setzero_si512 (),
+                  (__mmask16) -1);
+}
+
+static __inline __m512i __DEFAULT_FN_ATTRS
+_mm512_mask_loadu_epi32 (__m512i __W, __mmask16 __U, void const *__P)
+{
+  return (__m512i) __builtin_ia32_loaddqusi512_mask ((const __v16si *) __P,
+                  (__v16si) __W,
+                  (__mmask16) __U);
+}
+
+
+static __inline __m512i __DEFAULT_FN_ATTRS
 _mm512_maskz_loadu_epi32(__mmask16 __U, void const *__P)
 {
   return (__m512i) __builtin_ia32_loaddqusi512_mask ((const __v16si *)__P,
                                                      (__v16si)
                                                      _mm512_setzero_si512 (),
                                                      (__mmask16) __U);
 }
 
 static __inline __m512i __DEFAULT_FN_ATTRS
+_mm512_mask_loadu_epi64 (__m512i __W, __mmask8 __U, void const *__P)
+{
+  return (__m512i) __builtin_ia32_loaddqudi512_mask ((const __v8di *) __P,
+                  (__v8di) __W,
+                  (__mmask8) __U);
+}
+
+static __inline __m512i __DEFAULT_FN_ATTRS
 _mm512_maskz_loadu_epi64(__mmask8 __U, void const *__P)
 {
   return (__m512i) __builtin_ia32_loaddqudi512_mask ((const __v8di *)__P,
                                                      (__v8di)
                                                      _mm512_setzero_si512 (),
                                                      (__mmask8) __U);
 }
 
 static __inline __m512 __DEFAULT_FN_ATTRS
+_mm512_mask_loadu_ps (__m512 __W, __mmask16 __U, void const *__P)
+{
+  return (__m512) __builtin_ia32_loadups512_mask ((const __v16sf *) __P,
+                   (__v16sf) __W,
+                   (__mmask16) __U);
+}
+
+static __inline __m512 __DEFAULT_FN_ATTRS
 _mm512_maskz_loadu_ps(__mmask16 __U, void const *__P)
 {
   return (__m512) __builtin_ia32_loadups512_mask ((const __v16sf *)__P,
                                                   (__v16sf)
                                                   _mm512_setzero_ps (),
                                                   (__mmask16) __U);
 }
 
 static __inline __m512d __DEFAULT_FN_ATTRS
-_mm512_maskz_loadu_pd(__mmask8 __U, void const *__P)
+_mm512_mask_loadu_pd (__m512d __W, __mmask8 __U, void const *__P)
 {
-  return (__m512d) __builtin_ia32_loadupd512_mask ((const __v8df *)__P,
-                                                   (__v8df)
-                                                   _mm512_setzero_pd (),
-                                                   (__mmask8) __U);
+  return (__m512d) __builtin_ia32_loadupd512_mask ((const __v8df *) __P,
+                (__v8df) __W,
 }
 
-static __inline __m512 __DEFAULT_FN_ATTRS
-_mm512_maskz_load_ps(__mmask16 __U, void const *__P)
-{
-  return (__m512) __builtin_ia32_loadaps512_mask ((const __v16sf *)__P,
-                                                  (__v16sf)
-                                                  _mm512_setzero_ps (),
-                                                  (__mmask16) __U);
-}
-
 static __inline __m512d __DEFAULT_FN_ATTRS
-_mm512_maskz_load_pd(__mmask8 __U, void const *__P)
+_mm512_maskz_loadu_pd(__mmask8 __U, void const *__P)
 {
-  return (__m512d) __builtin_ia32_loadapd512_mask ((const __v8df *)__P,
+  return (__m512d) __builtin_ia32_loadupd512_mask ((const __v8df *)__P,
                                                    (__v8df)
                                                    _mm512_setzero_pd (),
                                                    (__mmask8) __U);
 }
 
 static __inline __m512d __DEFAULT_FN_ATTRS
 _mm512_loadu_pd(double const *__p)
 {

 _mm512_load_ps(float const *__p)
 {
   return (__m512) __builtin_ia32_loadaps512_mask ((const __v16sf *)__p,
                                                   (__v16sf)
                                                   _mm512_setzero_ps (),
                                                   (__mmask16) -1);
 }
 
+static __inline __m512 __DEFAULT_FN_ATTRS
+_mm512_mask_load_ps (__m512 __W, __mmask16 __U, void const *__P)
+{
+  return (__m512) __builtin_ia32_loadaps512_mask ((const __v16sf *) __P,
+                   (__v16sf) __W,
+                   (__mmask16) __U);
+}
+
+static __inline __m512 __DEFAULT_FN_ATTRS
+_mm512_maskz_load_ps(__mmask16 __U, void const *__P)
+{
+  return (__m512) __builtin_ia32_loadaps512_mask ((const __v16sf *)__P,
+                                                  (__v16sf)
+                                                  _mm512_setzero_ps (),
+                                                  (__mmask16) __U);
+}
+
 static __inline __m512d __DEFAULT_FN_ATTRS
 _mm512_load_pd(double const *__p)
 {
   return (__m512d) __builtin_ia32_loadapd512_mask ((const __v8df *)__p,
                                                    (__v8df)
                                                    _mm512_setzero_pd (),
                                                    (__mmask8) -1);
 }
 
+static __inline __m512d __DEFAULT_FN_ATTRS
+_mm512_mask_load_pd (__m512d __W, __mmask8 __U, void const *__P)
+{
+  return (__m512d) __builtin_ia32_loadapd512_mask ((const __v8df *) __P,
+                           (__v8df) __W,
+                           (__mmask8) __U);
+}
+
+static __inline __m512d __DEFAULT_FN_ATTRS
+_mm512_maskz_load_pd(__mmask8 __U, void const *__P)
+{
+  return (__m512d) __builtin_ia32_loadapd512_mask ((const __v8df *)__P,
+                                                   (__v8df)
+                                                   _mm512_setzero_pd (),
+                                                   (__mmask8) __U);
+}
+
+static __inline __m512i __DEFAULT_FN_ATTRS
+_mm512_load_si512 (void const *__P)
+{
+  return *(__m512i *) __P;
+}
+
+static __inline __m512i __DEFAULT_FN_ATTRS
+_mm512_load_epi32 (void const *__P)
+{
+  return *(__m512i *) __P;
+}
+
+static __inline __m512i __DEFAULT_FN_ATTRS
+_mm512_load_epi64 (void const *__P)
+{
+  return *(__m512i *) __P;
+}
+
 /* SIMD store ops */
 
 static __inline void __DEFAULT_FN_ATTRS
 _mm512_mask_storeu_epi64(void *__P, __mmask8 __U, __m512i __A)
 {
   __builtin_ia32_storedqudi512_mask ((__v8di *)__P, (__v8di) __A,
                                      (__mmask8) __U);
 }
 
 static __inline void __DEFAULT_FN_ATTRS
+_mm512_storeu_si512 (void *__P, __m512i __A)
+{
+  __builtin_ia32_storedqusi512_mask ((__v16si *) __P, (__v16si) __A,
+             (__mmask16) -1);
+}
+
+static __inline void __DEFAULT_FN_ATTRS
 _mm512_mask_storeu_epi32(void *__P, __mmask16 __U, __m512i __A)
 {
   __builtin_ia32_storedqusi512_mask ((__v16si *)__P, (__v16si) __A,
                                      (__mmask16) __U);
 }
 
 static __inline void __DEFAULT_FN_ATTRS
 _mm512_mask_storeu_pd(void *__P, __mmask8 __U, __m512d __A)

 }
 
 static __inline void __DEFAULT_FN_ATTRS
 _mm512_store_ps(void *__P, __m512 __A)
 {
   *(__m512*)__P = __A;
 }
 
+static __inline void __DEFAULT_FN_ATTRS
+_mm512_store_si512 (void *__P, __m512i __A)
+{
+  *(__m512i *) __P = __A;
+}
+
+static __inline void __DEFAULT_FN_ATTRS
+_mm512_store_epi32 (void *__P, __m512i __A)
+{
+  *(__m512i *) __P = __A;
+}
+
+static __inline void __DEFAULT_FN_ATTRS
+_mm512_store_epi64 (void *__P, __m512i __A)
+{
+  *(__m512i *) __P = __A;
+}
+
 /* Mask ops */
 
 static __inline __mmask16 __DEFAULT_FN_ATTRS
 _mm512_knot(__mmask16 __M)
 {
   return __builtin_ia32_knothi(__M);
 }
 

test/CodeGen/avx512f-builtins.c

 
 __m512d test_mm512_mul_pd(__m512d a, __m512d b)
 {
   // CHECK-LABEL: @test_mm512_mul_pd
   // CHECK: fmul <8 x double>
   return _mm512_mul_pd(a, b);
 }
 
+void test_mm512_storeu_si512 (void *__P, __m512i __A)
+{
+  // CHECK-LABEL: @test_mm512_storeu_si512 
+  // CHECK: @llvm.x86.avx512.mask.storeu.d.512
+  return _mm512_storeu_si512 ( __P,__A);
+}
+
 void test_mm512_storeu_ps(void *p, __m512 a)
 {
   // CHECK-LABEL: @test_mm512_storeu_ps
   // CHECK: @llvm.x86.avx512.mask.storeu.ps.512
   _mm512_storeu_ps(p, a);
 }
 
 void test_mm512_storeu_pd(void *p, __m512d a)
 {
   // CHECK-LABEL: @test_mm512_storeu_pd
   // CHECK: @llvm.x86.avx512.mask.storeu.pd.512
   _mm512_storeu_pd(p, a);
 }
 
 void test_mm512_mask_store_ps(void *p, __m512 a, __mmask16 m)
 {
   // CHECK-LABEL: @test_mm512_mask_store_ps
   // CHECK: @llvm.x86.avx512.mask.store.ps.512
   _mm512_mask_store_ps(p, m, a);
 }
 
+void test_mm512_store_si512 (void *__P, __m512i __A)
+{
+  // CHECK-LABEL: @test_mm512_store_si512 
+  // CHECK: load <8 x i64>, <8 x i64>* %__A.addr.i, align 64
+  // CHECK: load i8*, i8** %__P.addr.i, align 8
+  // CHECK: bitcast i8* %3 to <8 x i64>*
+  return _mm512_store_si512 ( __P,__A);
+}
+
+void test_mm512_store_epi32 (void *__P, __m512i __A)
+{
+  // CHECK-LABEL: @test_mm512_store_epi32 
+  // CHECK: load <8 x i64>, <8 x i64>* %__A.addr.i, align 64
+  // CHECK: load i8*, i8** %__P.addr.i, align 8
+  // CHECK: bitcast i8* %3 to <8 x i64>*
+  return _mm512_store_epi32 ( __P,__A);
+}
+
+void test_mm512_store_epi64 (void *__P, __m512i __A)
+{
+  // CHECK-LABEL: @test_mm512_store_epi64 
+  // CHECK: load <8 x i64>, <8 x i64>* %__A.addr.i, align 64
+  // CHECK: load i8*, i8** %__P.addr.i, align 8
+  // CHECK: bitcast i8* %3 to <8 x i64>*
+  return _mm512_store_epi64 ( __P,__A);
+}
+
 void test_mm512_store_ps(void *p, __m512 a)
 {
   // CHECK-LABEL: @test_mm512_store_ps
   // CHECK: store <16 x float>
   _mm512_store_ps(p, a);
 }
 
+void test_mm512_store_pd(void *p, __m512d a)
+{
+  // CHECK-LABEL: @test_mm512_store_pd
+  // CHECK: store <8 x double>
+  _mm512_store_pd(p, a);
+}
+
 void test_mm512_mask_store_pd(void *p, __m512d a, __mmask8 m)
 {
   // CHECK-LABEL: @test_mm512_mask_store_pd
   // CHECK: @llvm.x86.avx512.mask.store.pd.512
   _mm512_mask_store_pd(p, m, a);
 }
 
-void test_mm512_store_pd(void *p, __m512d a)
+__m512i test_mm512_loadu_si512 (void *__P)
 {
-  // CHECK-LABEL: @test_mm512_store_pd
-  // CHECK: store <8 x double>
-  _mm512_store_pd(p, a);
+  // CHECK-LABEL: @test_mm512_loadu_si512 
+  // CHECK: @llvm.x86.avx512.mask.loadu.d.512
+  return _mm512_loadu_si512 ( __P);
+}
+
+__m512i test_mm512_mask_loadu_epi32 (__m512i __W, __mmask16 __U, void *__P)
+{
+  // CHECK-LABEL: @test_mm512_mask_loadu_epi32 
+  // CHECK: @llvm.x86.avx512.mask.loadu.d.512
+  return _mm512_mask_loadu_epi32 (__W,__U, __P);
+}
+
+__m512i test_mm512_mask_loadu_epi64 (__m512i __W, __mmask8 __U, void *__P)
+{
+  // CHECK-LABEL: @test_mm512_mask_loadu_epi64 
+  // CHECK: @llvm.x86.avx512.mask.loadu.q.512(
+  return _mm512_mask_loadu_epi64 (__W,__U, __P);
 }
 
 __m512 test_mm512_loadu_ps(void *p)
 {
   // CHECK-LABEL: @test_mm512_loadu_ps
   // CHECK: load <16 x float>, <16 x float>* {{.*}}, align 1{{$}}
   return _mm512_loadu_ps(p);
 }
 
+__m512 test_mm512_mask_loadu_ps (__m512 __W, __mmask16 __U, void *__P)
+{
+  // CHECK-LABEL: @test_mm512_mask_loadu_ps 
+  // CHECK: @llvm.x86.avx512.mask.loadu.ps.512
+  return _mm512_mask_loadu_ps (__W,__U, __P);
+}
+
 __m512d test_mm512_loadu_pd(void *p)
 {
   // CHECK-LABEL: @test_mm512_loadu_pd
   // CHECK: load <8 x double>, <8 x double>* {{.*}}, align 1{{$}}
   return _mm512_loadu_pd(p);
 }
 
+__m512d test_mm512_mask_loadu_pd (__m512d __W, __mmask8 __U, void *__P)
+{
+  // CHECK-LABEL: @test_mm512_mask_loadu_pd 
+  // CHECK: @llvm.x86.avx512.mask.loadu.pd.512
+  return _mm512_mask_loadu_pd (__W,__U, __P);
+}
+
 __m512 test_mm512_maskz_load_ps(void *p, __mmask16 m)
 {
   // CHECK-LABEL: @test_mm512_maskz_load_ps
   // CHECK: @llvm.x86.avx512.mask.load.ps.512
   return _mm512_maskz_load_ps(m, p);
 }
 
+__m512i test_mm512_load_si512 (void *__P)
+{
+  // CHECK-LABEL: @test_mm512_load_si512 
+  // CHECK: load i8*, i8** %__P.addr.i, align 8
+  // CHECK: bitcast i8* %1 to <8 x i64>*
+  // CHECK: load <8 x i64>, <8 x i64>* %2, align 64
+  return _mm512_load_si512 ( __P);
+}
+
+__m512i test_mm512_load_epi32 (void *__P)
+{
+  // CHECK-LABEL: @test_mm512_load_epi32 
+  // CHECK: load i8*, i8** %__P.addr.i, align 8
+  // CHECK: bitcast i8* %1 to <8 x i64>*
+  // CHECK: load <8 x i64>, <8 x i64>* %2, align 64
+  return _mm512_load_epi32 ( __P);
+}
+
+__m512i test_mm512_load_epi64 (void *__P)
+{
+  // CHECK-LABEL: @test_mm512_load_epi64 
+  // CHECK: load i8*, i8** %__P.addr.i, align 8
+  // CHECK: bitcast i8* %1 to <8 x i64>*
+  // CHECK: load <8 x i64>, <8 x i64>* %2, align 64
+  return _mm512_load_epi64 ( __P);
+}
+
 __m512 test_mm512_load_ps(void *p)
 {
   // CHECK-LABEL: @test_mm512_load_ps
   // CHECK: @llvm.x86.avx512.mask.load.ps.512
   return _mm512_load_ps(p);
 }
 
+__m512 test_mm512_mask_load_ps (__m512 __W, __mmask16 __U, void *__P)
+{
+  // CHECK-LABEL: @test_mm512_mask_load_ps 
+  // CHECK: @llvm.x86.avx512.mask.load.ps.512
+  return _mm512_mask_load_ps (__W,__U, __P);
+}
+
+__m512 test_mm512_maskz_load_ps(__mmask16 __U, void *__P)
+{
+  // CHECK-LABEL: @test_mm512_maskz_load_ps
+  // CHECK:  @llvm.x86.avx512.mask.load.ps.512
+  return _mm512_maskz_load_ps(__U, __P);
+}
+
 __m512d test_mm512_maskz_load_pd(void *p, __mmask8 m)
 {
   // CHECK-LABEL: @test_mm512_maskz_load_pd
   // CHECK: @llvm.x86.avx512.mask.load.pd.512
   return _mm512_maskz_load_pd(m, p);
 }
 
 __m512d test_mm512_load_pd(void *p)
 {
   // CHECK-LABEL: @test_mm512_load_pd
   // CHECK: @llvm.x86.avx512.mask.load.pd.512
   return _mm512_load_pd(p);
 }
 
+__m512d test_mm512_mask_load_pd (__m512d __W, __mmask8 __U, void *__P)
+{
+  // CHECK-LABEL: @test_mm512_mask_load_pd 
+  // CHECK: @llvm.x86.avx512.mask.load.pd.512
+  return _mm512_mask_load_pd (__W,__U, __P);
+}
+
+__m512d test_mm512_maskz_load_pd(__mmask8 __U, void *__P)
+{
+  // CHECK-LABEL: @test_mm512_maskz_load_pd
+  // CHECK: @llvm.x86.avx512.mask.load.pd.512
+  return _mm512_maskz_load_pd(__U, __P);
+}
+
 __m512d test_mm512_set1_pd(double d)
 {
   // CHECK-LABEL: @test_mm512_set1_pd
   // CHECK: insertelement <8 x double> {{.*}}, i32 0
   // CHECK: insertelement <8 x double> {{.*}}, i32 1
   // CHECK: insertelement <8 x double> {{.*}}, i32 2
   // CHECK: insertelement <8 x double> {{.*}}, i32 3
   // CHECK: insertelement <8 x double> {{.*}}, i32 4