From: Craig Topper <craig.topper@gmail.com>
Date: Sat, 24 Dec 2011 17:20:15 +0000 (+0000)
Subject: Add last of the AVX2 intrinsics except for gather.
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=5cbd751a2f9d73248c5336140d73680fcd4669a3;p=clang

Add last of the AVX2 intrinsics except for gather.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@147253 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/include/clang/Basic/BuiltinsX86.def b/include/clang/Basic/BuiltinsX86.def
index 5aa2c1bd5b..62ebec25a9 100644
--- a/include/clang/Basic/BuiltinsX86.def
+++ b/include/clang/Basic/BuiltinsX86.def
@@ -575,5 +575,25 @@ BUILTIN(__builtin_ia32_permdf256, "V4dV4dIc", "")
 BUILTIN(__builtin_ia32_permvarsf256, "V8fV8fV8f", "")
 BUILTIN(__builtin_ia32_permdi256, "V4LLiV4LLiIc", "")
 BUILTIN(__builtin_ia32_permti256, "V4LLiV4LLiV4LLiIc", "")
+BUILTIN(__builtin_ia32_extract128i256, "V2LLiV4LLiIc", "")
+BUILTIN(__builtin_ia32_insert128i256, "V4LLiV4LLiV2LLiIc", "")
+BUILTIN(__builtin_ia32_maskloadd256, "V8iV8iC*V8i", "")
+BUILTIN(__builtin_ia32_maskloadq256, "V4LLiV4LLiC*V4LLi", "")
+BUILTIN(__builtin_ia32_maskloadd, "V4iV4iC*V4i", "")
+BUILTIN(__builtin_ia32_maskloadq, "V2LLiV2LLiC*V2LLi", "")
+BUILTIN(__builtin_ia32_maskstored256, "vV8i*V8iV8i", "")
+BUILTIN(__builtin_ia32_maskstoreq256, "vV4LLi*V4LLiV4LLi", "")
+BUILTIN(__builtin_ia32_maskstored, "vV4i*V4iV4i", "")
+BUILTIN(__builtin_ia32_maskstoreq, "vV2LLi*V2LLiV2LLi", "")
+BUILTIN(__builtin_ia32_psllv8si, "V8iV8iV8i", "")
+BUILTIN(__builtin_ia32_psllv4si, "V4iV4iV4i", "")
+BUILTIN(__builtin_ia32_psllv4di, "V4LLiV4LLiV4LLi", "")
+BUILTIN(__builtin_ia32_psllv2di, "V2LLiV2LLiV2LLi", "")
+BUILTIN(__builtin_ia32_psrav8si, "V8iV8iV8i", "")
+BUILTIN(__builtin_ia32_psrav4si, "V4iV4iV4i", "")
+BUILTIN(__builtin_ia32_psrlv8si, "V8iV8iV8i", "")
+BUILTIN(__builtin_ia32_psrlv4si, "V4iV4iV4i", "")
+BUILTIN(__builtin_ia32_psrlv4di, "V4LLiV4LLiV4LLi", "")
+BUILTIN(__builtin_ia32_psrlv2di, "V2LLiV2LLiV2LLi", "")
 
 #undef BUILTIN
diff --git a/lib/Headers/avx2intrin.h b/lib/Headers/avx2intrin.h
index 283227fdcf..cc3ffe69b3 100644
--- a/lib/Headers/avx2intrin.h
+++ b/lib/Headers/avx2intrin.h
@@ -180,7 +180,7 @@ _mm256_cmpeq_epi32(__m256i a, __m256i b)
 static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
 _mm256_cmpeq_epi64(__m256i a, __m256i b)
 {
-  return (__m256i)((__v4di)a == (__v4di)b);
+  return (__m256i)(a == b);
 }
 
 static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
@@ -204,7 +204,7 @@ _mm256_cmpgt_epi32(__m256i a, __m256i b)
 static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
 _mm256_cmpgt_epi64(__m256i a, __m256i b)
 {
-  return (__m256i)((__v4di)a > (__v4di)b);
+  return (__m256i)(a > b);
 }
 
 static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
@@ -832,9 +832,126 @@ _mm256_permutevar8x32_ps(__m256 a, __m256 b)
 
 #define _mm256_permute4x64_epi64(V, M) __extension__ ({ \
   __m256i __V = (V); \
-  (__m256i)__builtin_ia32_permdi256((__v4di)__V, (M)); })
+  (__m256i)__builtin_ia32_permdi256(__V, (M)); })
 
 #define _mm256_permute2x128_si256(V1, V2, M) __extension__ ({ \
   __m256i __V1 = (V1); \
   __m256i __V2 = (V2); \
-  (__m256i)__builtin_ia32_permti256((__v4di)__V1, (__v4di)__V2, (M)); })
+  (__m256i)__builtin_ia32_permti256(__V1, __V2, (M)); })
+
+#define _mm256_extracti128_si256(A, O) __extension__ ({ \
+  __m256i __A = (A); \
+  (__m128i)__builtin_ia32_extract128i256(__A, (O)); })
+
+#define _mm256_inserti128_si256(V1, V2, O) __extension__ ({ \
+  __m256i __V1 = (V1); \
+  __m128i __V2 = (V2); \
+  (__m256i)__builtin_ia32_insert128i256(__V1, __V2, (O)); })
+
+static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
+_mm256_maskload_epi32(int const *__X, __m256i __M)
+{
+  return (__m256i)__builtin_ia32_maskloadd256((const __v8si *)__X, (__v8si)__M);
+}
+
+static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
+_mm256_maskload_epi64(long long const *__X, __m256i __M)
+{
+  return (__m256i)__builtin_ia32_maskloadq256((const __v4di *)__X, __M);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_maskload_epi32(int const *__X, __m128i __M)
+{
+  return (__m128i)__builtin_ia32_maskloadd((const __v4si *)__X, (__v4si)__M);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_maskload_epi64(long long const *__X, __m128i __M)
+{
+  return (__m128i)__builtin_ia32_maskloadq((const __v2di *)__X, (__v2di)__M);
+}
+
+static __inline__ void __attribute__((__always_inline__, __nodebug__))
+_mm256_maskstore_epi32(int *__X, __m256i __M, __m256i __Y)
+{
+  __builtin_ia32_maskstored256((__v8si *)__X, (__v8si)__M, (__v8si)__Y);
+}
+
+static __inline__ void __attribute__((__always_inline__, __nodebug__))
+_mm256_maskstore_epi64(long long *__X, __m256i __M, __m256i __Y)
+{
+  __builtin_ia32_maskstoreq256((__v4di *)__X, __M, __Y);
+}
+
+static __inline__ void __attribute__((__always_inline__, __nodebug__))
+_mm_maskstore_epi32(int *__X, __m128i __M, __m128i __Y)
+{
+  __builtin_ia32_maskstored((__v4si *)__X, (__v4si)__M, (__v4si)__Y);
+}
+
+static __inline__ void __attribute__((__always_inline__, __nodebug__))
+_mm_maskstore_epi64(long long *__X, __m128i __M, __m128i __Y)
+{
+  __builtin_ia32_maskstoreq(( __v2di *)__X, __M, __Y);
+}
+
+static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
+_mm256_sllv_epi32(__m256i __X, __m256i __Y)
+{
+  return (__m256i)__builtin_ia32_psllv8si((__v8si)__X, (__v8si)__Y);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_sllv_epi32(__m128i __X, __m128i __Y)
+{
+  return (__m128i)__builtin_ia32_psllv4si((__v4si)__X, (__v4si)__Y);
+}
+
+static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
+_mm256_sllv_epi64(__m256i __X, __m256i __Y)
+{
+  return (__m256i)__builtin_ia32_psllv4di(__X, __Y);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_sllv_epi64(__m128i __X, __m128i __Y)
+{
+  return (__m128i)__builtin_ia32_psllv2di(__X, __Y);
+}
+
+static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
+_mm256_srav_epi32(__m256i __X, __m256i __Y)
+{
+  return (__m256i)__builtin_ia32_psrav8si((__v8si)__X, (__v8si)__Y);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_srav_epi32(__m128i __X, __m128i __Y)
+{
+  return (__m128i)__builtin_ia32_psrav4si((__v4si)__X, (__v4si)__Y);
+}
+
+static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
+_mm256_srlv_epi32(__m256i __X, __m256i __Y)
+{
+  return (__m256i)__builtin_ia32_psrlv8si((__v8si)__X, (__v8si)__Y);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_srlv_epi32(__m128i __X, __m128i __Y)
+{
+  return (__m128i)__builtin_ia32_psrlv4si((__v4si)__X, (__v4si)__Y);
+}
+
+static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
+_mm256_srlv_epi64(__m256i __X, __m256i __Y)
+{
+  return (__m256i)__builtin_ia32_psrlv4di(__X, __Y);
+}
+
+static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
+_mm_srlv_epi64(__m128i __X, __m128i __Y)
+{
+  return (__m128i)__builtin_ia32_psrlv2di(__X, __Y);
+}
diff --git a/test/CodeGen/avx2-builtins.c b/test/CodeGen/avx2-builtins.c
index caa27f639d..ba0338e931 100644
--- a/test/CodeGen/avx2-builtins.c
+++ b/test/CodeGen/avx2-builtins.c
@@ -680,3 +680,103 @@ __m256i test_mm256_permute2x128_si256(__m256i a, __m256i b) {
   // CHECK: @llvm.x86.avx2.vperm2i128
   return _mm256_permute2x128_si256(a, b, 94);
 }
+
+__m128i test_mm256_extracti128_si256(__m256i a) {
+  // CHECK: @llvm.x86.avx2.vextracti128
+  return _mm256_extracti128_si256(a, 1);
+}
+
+__m256i test_mm256_inserti128_si256(__m256i a, __m128i b) {
+  // CHECK: @llvm.x86.avx2.vinserti128
+  return _mm256_inserti128_si256(a, b, 1);
+}
+
+__m256i test_mm256_maskload_epi32(int const *a, __m256i m) {
+  // CHECK: @llvm.x86.avx2.maskload.d.256
+  return _mm256_maskload_epi32(a, m);
+}
+
+__m256i test_mm256_maskload_epi64(long long const *a, __m256i m) {
+  // CHECK: @llvm.x86.avx2.maskload.q.256
+  return _mm256_maskload_epi64(a, m);
+}
+
+__m128i test_mm_maskload_epi32(int const *a, __m128i m) {
+  // CHECK: @llvm.x86.avx2.maskload.d
+  return _mm_maskload_epi32(a, m);
+}
+
+__m128i test_mm_maskload_epi64(long long const *a, __m128i m) {
+  // CHECK: @llvm.x86.avx2.maskload.q
+  return _mm_maskload_epi64(a, m);
+}
+
+void test_mm256_maskstore_epi32(int *a, __m256i m, __m256i b) {
+  // CHECK: @llvm.x86.avx2.maskstore.d.256
+  _mm256_maskstore_epi32(a, m, b);
+}
+
+void test_mm256_maskstore_epi64(long long *a, __m256i m, __m256i b) {
+  // CHECK: @llvm.x86.avx2.maskstore.q.256
+  _mm256_maskstore_epi64(a, m, b);
+}
+
+void test_mm_maskstore_epi32(int *a, __m128i m, __m128i b) {
+  // CHECK: @llvm.x86.avx2.maskstore.d
+  _mm_maskstore_epi32(a, m, b);
+}
+
+void test_mm_maskstore_epi64(long long *a, __m128i m, __m128i b) {
+  // CHECK: @llvm.x86.avx2.maskstore.q
+  _mm_maskstore_epi64(a, m, b);
+}
+
+__m256i test_mm256_sllv_epi32(__m256i a, __m256i b) {
+  // CHECK: @llvm.x86.avx2.psllv.d.256
+  return _mm256_sllv_epi32(a, b);
+}
+
+__m128i test_mm_sllv_epi32(__m128i a, __m128i b) {
+  // CHECK: @llvm.x86.avx2.psllv.d
+  return _mm_sllv_epi32(a, b);
+}
+
+__m256i test_mm256_sllv_epi64(__m256i a, __m256i b) {
+  // CHECK: @llvm.x86.avx2.psllv.q.256
+  return _mm256_sllv_epi64(a, b);
+}
+
+__m128i test_mm_sllv_epi64(__m128i a, __m128i b) {
+  // CHECK: @llvm.x86.avx2.psllv.q
+  return _mm_sllv_epi64(a, b);
+}
+
+__m256i test_mm256_srav_epi32(__m256i a, __m256i b) {
+  // CHECK: @llvm.x86.avx2.psrav.d.256
+  return _mm256_srav_epi32(a, b);
+}
+
+__m128i test_mm_srav_epi32(__m128i a, __m128i b) {
+  // CHECK: @llvm.x86.avx2.psrav.d
+  return _mm_srav_epi32(a, b);
+}
+
+__m256i test_mm256_srlv_epi32(__m256i a, __m256i b) {
+  // CHECK: @llvm.x86.avx2.psrlv.d.256
+  return _mm256_srlv_epi32(a, b);
+}
+
+__m128i test_mm_srlv_epi32(__m128i a, __m128i b) {
+  // CHECK: @llvm.x86.avx2.psrlv.d
+  return _mm_srlv_epi32(a, b);
+}
+
+__m256i test_mm256_srlv_epi64(__m256i a, __m256i b) {
+  // CHECK: @llvm.x86.avx2.psrlv.q.256
+  return _mm256_srlv_epi64(a, b);
+}
+
+__m128i test_mm_srlv_epi64(__m128i a, __m128i b) {
+  // CHECK: @llvm.x86.avx2.psrlv.q
+  return _mm_srlv_epi64(a, b);
+}