[Clang][avx512][builtin] Adding intrinsics for vexpand{d|q|ps|pd} instrctuon set

[AsafBadouh]

intrinsics with zero masks (maskz) are missing.

[AsafBadouh]

LGTM

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

 TARGET_BUILTIN(__builtin_ia32_movsldup512_mask, "V16fV16fV16fUs","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_movshdup128_mask, "V4fV4fV4fUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_movshdup256_mask, "V8fV8fV8fUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_movsldup128_mask, "V4fV4fV4fUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_movsldup256_mask, "V8fV8fV8fUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_pshufd512_mask, "V16iV16iCsV16iUc","","avx512f")
 TARGET_BUILTIN(__builtin_ia32_pshufd256_mask, "V8iV8iCsV8iUc","","avx512vl")
 TARGET_BUILTIN(__builtin_ia32_pshufd128_mask, "V4iV4iCsV4iUc","","avx512vl")
+TARGET_BUILTIN(__builtin_ia32_expanddf512_mask, "V8dV8dV8dUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_expanddi512_mask, "V8LLiV8LLiV8LLiUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_expandloaddf512_mask, "V8dvC*V8dUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_expandloaddi512_mask, "V8LLivC*V8LLiUc","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_expandloadsf512_mask, "V16fvC*V16fUs","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_expandloadsi512_mask, "V16ivC*V16iUs","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_expandsf512_mask, "V16fV16fV16fUs","","avx512f")
+TARGET_BUILTIN(__builtin_ia32_expandsi512_mask, "V16iV16iV16iUs","","avx512f")
 
 #undef BUILTIN
 #undef TARGET_BUILTIN

cfe/trunk/lib/Headers/avx512fintrin.h

 
 #define _mm512_maskz_shuffle_epi32( __U, __A, __I) __extension__ ({ \
 __builtin_ia32_pshufd512_mask ((__v16si)( __A),\
               ( __I),\
               (__v16si) _mm512_setzero_si512 (),\
               (__mmask16)( __U));\
 })
 
+static __inline__ __m512d __DEFAULT_FN_ATTRS
+_mm512_mask_expand_pd (__m512d __W, __mmask8 __U, __m512d __A)
+{
+  return (__m512d) __builtin_ia32_expanddf512_mask ((__v8df) __A,
+                (__v8df) __W,
+                (__mmask8) __U);
+}
+
+static __inline__ __m512d __DEFAULT_FN_ATTRS
+_mm512_maskz_expand_pd (__mmask8 __U, __m512d __A)
+{
+  return (__m512d) __builtin_ia32_expanddf512_mask ((__v8df) __A,
+                (__v8df) _mm512_setzero_pd (),
+                (__mmask8) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_mask_expand_epi64 (__m512i __W, __mmask8 __U, __m512i __A)
+{
+  return (__m512i) __builtin_ia32_expanddi512_mask ((__v8di) __A,
+                (__v8di) __W,
+                (__mmask8) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_maskz_expand_epi64 ( __mmask8 __U, __m512i __A)
+{
+  return (__m512i) __builtin_ia32_expanddi512_mask ((__v8di) __A,
+                (__v8di) _mm512_setzero_pd (),
+                (__mmask8) __U);
+}
+
+static __inline__ __m512d __DEFAULT_FN_ATTRS
+_mm512_mask_expandloadu_pd(__m512d __W, __mmask8 __U, void const *__P)
+{
+  return (__m512d) __builtin_ia32_expandloaddf512_mask ((const __v8df *)__P,
+              (__v8df) __W,
+              (__mmask8) __U);
+}
+
+static __inline__ __m512d __DEFAULT_FN_ATTRS
+_mm512_maskz_expandloadu_pd(__mmask8 __U, void const *__P)
+{
+  return (__m512d) __builtin_ia32_expandloaddf512_mask ((const __v8df *)__P,
+              (__v8df) _mm512_setzero_pd(),
+              (__mmask8) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_mask_expandloadu_epi64(__m512i __W, __mmask8 __U, void const *__P)
+{
+  return (__m512i) __builtin_ia32_expandloaddi512_mask ((const __v8di *)__P,
+              (__v8di) __W,
+              (__mmask8) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_maskz_expandloadu_epi64(__mmask8 __U, void const *__P)
+{
+  return (__m512i) __builtin_ia32_expandloaddi512_mask ((const __v8di *)__P,
+              (__v8di) _mm512_setzero_pd(),
+              (__mmask8) __U);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS
+_mm512_mask_expandloadu_ps(__m512 __W, __mmask16 __U, void const *__P)
+{
+  return (__m512) __builtin_ia32_expandloadsf512_mask ((const __v16sf *)__P,
+                   (__v16sf) __W,
+                   (__mmask16) __U);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS
+_mm512_maskz_expandloadu_ps(__mmask16 __U, void const *__P)
+{
+  return (__m512) __builtin_ia32_expandloadsf512_mask ((const __v16sf *)__P,
+                   (__v16sf) _mm512_setzero_ps(),
+                   (__mmask16) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_mask_expandloadu_epi32(__m512i __W, __mmask16 __U, void const *__P)
+{
+  return (__m512i) __builtin_ia32_expandloadsi512_mask ((const __v16si *)__P,
+              (__v16si) __W,
+              (__mmask16) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_maskz_expandloadu_epi32(__mmask16 __U, void const *__P)
+{
+  return (__m512i) __builtin_ia32_expandloadsi512_mask ((const __v16si *)__P,
+              (__v16si) _mm512_setzero_ps(),
+              (__mmask16) __U);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS
+_mm512_mask_expand_ps (__m512 __W, __mmask16 __U, __m512 __A)
+{
+  return (__m512) __builtin_ia32_expandsf512_mask ((__v16sf) __A,
+               (__v16sf) __W,
+               (__mmask16) __U);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS
+_mm512_maskz_expand_ps (__mmask16 __U, __m512 __A)
+{
+  return (__m512) __builtin_ia32_expandsf512_mask ((__v16sf) __A,
+               (__v16sf) _mm512_setzero_ps(),
+               (__mmask16) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_mask_expand_epi32 (__m512i __W, __mmask16 __U, __m512i __A)
+{
+  return (__m512i) __builtin_ia32_expandsi512_mask ((__v16si) __A,
+                (__v16si) __W,
+                (__mmask16) __U);
+}
+
+static __inline__ __m512i __DEFAULT_FN_ATTRS
+_mm512_maskz_expand_epi32 (__mmask16 __U, __m512i __A)
+{
+  return (__m512i) __builtin_ia32_expandsi512_mask ((__v16si) __A,
+                (__v16si) _mm512_setzero_ps(),
+                (__mmask16) __U);
+}
+
 #undef __DEFAULT_FN_ATTRS
 
 #endif // __AVX512FINTRIN_H

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

 }
 
 __m512i test_mm512_maskz_shuffle_epi32(__mmask16 __U, __m512i __A) {
   // CHECK-LABEL: @test_mm512_maskz_shuffle_epi32
   // CHECK: @llvm.x86.avx512.mask.pshuf.d.512
   return _mm512_maskz_shuffle_epi32(__U, __A, 1); 
 }
 
+__m512d test_mm512_mask_expand_pd(__m512d __W, __mmask8 __U, __m512d __A) {
+  // CHECK-LABEL: @test_mm512_mask_expand_pd
+  // CHECK: @llvm.x86.avx512.mask.expand.pd.512
+  return _mm512_mask_expand_pd(__W, __U, __A); 
+}
+
+__m512d test_mm512_maskz_expand_pd(__mmask8 __U, __m512d __A) {
+  // CHECK-LABEL: @test_mm512_maskz_expand_pd
+  // CHECK: @llvm.x86.avx512.mask.expand.pd.512
+  return _mm512_maskz_expand_pd(__U, __A); 
+}
+
+__m512i test_mm512_mask_expand_epi64(__m512i __W, __mmask8 __U, __m512i __A) {
+  // CHECK-LABEL: @test_mm512_mask_expand_epi64
+  // CHECK: @llvm.x86.avx512.mask.expand.q.512
+  return _mm512_mask_expand_epi64(__W, __U, __A); 
+}
+
+__m512i test_mm512_maskz_expand_epi64(__mmask8 __U, __m512i __A) {
+  // CHECK-LABEL: @test_mm512_maskz_expand_epi64
+  // CHECK: @llvm.x86.avx512.mask.expand.q.512
+  return _mm512_maskz_expand_epi64(__U, __A); 
+}
+__m512i test_mm512_mask_expandloadu_epi64(__m512i __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm512_mask_expandloadu_epi64
+  // CHECK: @llvm.x86.avx512.mask.expand.load.q.512
+  return _mm512_mask_expandloadu_epi64(__W, __U, __P); 
+}
+
+__m512i test_mm512_maskz_expandloadu_epi64(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm512_maskz_expandloadu_epi64
+  // CHECK: @llvm.x86.avx512.mask.expand.load.q.512
+  return _mm512_maskz_expandloadu_epi64(__U, __P); 
+}
+
+__m512d test_mm512_mask_expandloadu_pd(__m512d __W, __mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm512_mask_expandloadu_pd
+  // CHECK: @llvm.x86.avx512.mask.expand.load.pd.512
+  return _mm512_mask_expandloadu_pd(__W, __U, __P); 
+}
+
+__m512d test_mm512_maskz_expandloadu_pd(__mmask8 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm512_maskz_expandloadu_pd
+  // CHECK: @llvm.x86.avx512.mask.expand.load.pd.512
+  return _mm512_maskz_expandloadu_pd(__U, __P); 
+}
+
+__m512i test_mm512_mask_expandloadu_epi32(__m512i __W, __mmask16 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm512_mask_expandloadu_epi32
+  // CHECK: @llvm.x86.avx512.mask.expand.load.d.512
+  return _mm512_mask_expandloadu_epi32(__W, __U, __P); 
+}
+
+__m512i test_mm512_maskz_expandloadu_epi32(__mmask16 __U, void const *__P) {
+  // CHECK-LABEL: @test_mm512_maskz_expandloadu_epi32
+  // CHECK: @llvm.x86.avx512.mask.expand.load.d.512
+  return _mm512_maskz_expandloadu_epi32(__U, __P); 
+}
+
+__m512 test_mm512_mask_expand_ps(__m512 __W, __mmask16 __U, __m512 __A) {
+  // CHECK-LABEL: @test_mm512_mask_expand_ps
+  // CHECK: @llvm.x86.avx512.mask.expand.ps.512
+  return _mm512_mask_expand_ps(__W, __U, __A); 
+}
+
+__m512 test_mm512_maskz_expand_ps(__mmask16 __U, __m512 __A) {
+  // CHECK-LABEL: @test_mm512_maskz_expand_ps
+  // CHECK: @llvm.x86.avx512.mask.expand.ps.512
+  return _mm512_maskz_expand_ps(__U, __A); 
+}
+
+__m512i test_mm512_mask_expand_epi32(__m512i __W, __mmask16 __U, __m512i __A) {
+  // CHECK-LABEL: @test_mm512_mask_expand_epi32
+  // CHECK: @llvm.x86.avx512.mask.expand.d.512
+  return _mm512_mask_expand_epi32(__W, __U, __A); 
+}
+
+__m512i test_mm512_maskz_expand_epi32(__mmask16 __U, __m512i __A) {
+  // CHECK-LABEL: @test_mm512_maskz_expand_epi32
+  // CHECK: @llvm.x86.avx512.mask.expand.d.512
+  return _mm512_maskz_expand_epi32(__U, __A); 
+}