Add fma3 intrinsic header file.

llvm-svn: 157913

4 files changed, 400 insertions, 0 deletions

diff --git a/clang/lib/Headers/CMakeLists.txt b/clang/lib/Headers/CMakeLists.txt
index 3c58a82d050..3043e03d7f5 100644
--- a/clang/lib/Headers/CMakeLists.txt
+++ b/clang/lib/Headers/CMakeLists.txt
@@ -8,6 +8,7 @@ set(files
   emmintrin.h
   float.h
   fma4intrin.h
+  fmaintrin.h
   immintrin.h
   iso646.h
   limits.h
diff --git a/clang/lib/Headers/fmaintrin.h b/clang/lib/Headers/fmaintrin.h
new file mode 100644
index 00000000000..6bfd5a88b3c
--- /dev/null
+++ b/clang/lib/Headers/fmaintrin.h
@@ -0,0 +1,229 @@
+/*===---- fma4intrin.h - FMA4 intrinsics -----------------------------------===
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ *
+ *===-----------------------------------------------------------------------===
+ */
+
+#ifndef __IMMINTRIN_H
+#error "Never use <fmaintrin.h> directly; include <immintrin.h> instead."
+#endif
+
+#ifndef __FMAINTRIN_H
+#define __FMAINTRIN_H
+
+#ifndef __FMA__
+# error "FMA instruction set is not enabled"
+#else
+
+static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
+_mm_fmadd_ps(__m128 __A, __m128 __B, __m128 __C)
+{
+  return (__m128)__builtin_ia32_vfmaddps(__A, __B, __C);
+}
+
+static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
+_mm_fmadd_pd(__m128d __A, __m128d __B, __m128d __C)
+{
+  return (__m128d)__builtin_ia32_vfmaddpd(__A, __B, __C);
+}
+
+static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
+_mm_fmadd_ss(__m128 __A, __m128 __B, __m128 __C)
+{
+  return (__m128)__builtin_ia32_vfmaddss(__A, __B, __C);
+}
+
+static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
+_mm_fmadd_sd(__m128d __A, __m128d __B, __m128d __C)
+{
+  return (__m128d)__builtin_ia32_vfmaddsd(__A, __B, __C);
+}
+
+static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
+_mm_fmsub_ps(__m128 __A, __m128 __B, __m128 __C)
+{
+  return (__m128)__builtin_ia32_vfmsubps(__A, __B, __C);
+}
+
+static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
+_mm_fmsub_pd(__m128d __A, __m128d __B, __m128d __C)
+{
+  return (__m128d)__builtin_ia32_vfmsubpd(__A, __B, __C);
+}
+
+static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
+_mm_fmsub_ss(__m128 __A, __m128 __B, __m128 __C)
+{
+  return (__m128)__builtin_ia32_vfmsubss(__A, __B, __C);
+}
+
+static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
+_mm_fmsub_sd(__m128d __A, __m128d __B, __m128d __C)
+{
+  return (__m128d)__builtin_ia32_vfmsubsd(__A, __B, __C);
+}
+
+static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
+_mm_fnmadd_ps(__m128 __A, __m128 __B, __m128 __C)
+{
+  return (__m128)__builtin_ia32_vfnmaddps(__A, __B, __C);
+}
+
+static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
+_mm_fnmadd_pd(__m128d __A, __m128d __B, __m128d __C)
+{
+  return (__m128d)__builtin_ia32_vfnmaddpd(__A, __B, __C);
+}
+
+static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
+_mm_fnmadd_ss(__m128 __A, __m128 __B, __m128 __C)
+{
+  return (__m128)__builtin_ia32_vfnmaddss(__A, __B, __C);
+}
+
+static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
+_mm_fnmadd_sd(__m128d __A, __m128d __B, __m128d __C)
+{
+  return (__m128d)__builtin_ia32_vfnmaddsd(__A, __B, __C);
+}
+
+static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
+_mm_fnmsub_ps(__m128 __A, __m128 __B, __m128 __C)
+{
+  return (__m128)__builtin_ia32_vfnmsubps(__A, __B, __C);
+}
+
+static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
+_mm_fnmsub_pd(__m128d __A, __m128d __B, __m128d __C)
+{
+  return (__m128d)__builtin_ia32_vfnmsubpd(__A, __B, __C);
+}
+
+static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
+_mm_fnmsub_ss(__m128 __A, __m128 __B, __m128 __C)
+{
+  return (__m128)__builtin_ia32_vfnmsubss(__A, __B, __C);
+}
+
+static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
+_mm_fnmsub_sd(__m128d __A, __m128d __B, __m128d __C)
+{
+  return (__m128d)__builtin_ia32_vfnmsubsd(__A, __B, __C);
+}
+
+static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
+_mm_fmaddsub_ps(__m128 __A, __m128 __B, __m128 __C)
+{
+  return (__m128)__builtin_ia32_vfmaddsubps(__A, __B, __C);
+}
+
+static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
+_mm_fmaddsub_pd(__m128d __A, __m128d __B, __m128d __C)
+{
+  return (__m128d)__builtin_ia32_vfmaddsubpd(__A, __B, __C);
+}
+
+static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
+_mm_fmsubadd_ps(__m128 __A, __m128 __B, __m128 __C)
+{
+  return (__m128)__builtin_ia32_vfmsubaddps(__A, __B, __C);
+}
+
+static __inline__ __m128d __attribute__((__always_inline__, __nodebug__))
+_mm_fmsubadd_pd(__m128d __A, __m128d __B, __m128d __C)
+{
+  return (__m128d)__builtin_ia32_vfmsubaddpd(__A, __B, __C);
+}
+
+static __inline__ __m256 __attribute__((__always_inline__, __nodebug__))
+_mm256_fmadd_ps(__m256 __A, __m256 __B, __m256 __C)
+{
+  return (__m256)__builtin_ia32_vfmaddps256(__A, __B, __C);
+}
+
+static __inline__ __m256d __attribute__((__always_inline__, __nodebug__))
+_mm256_fmadd_pd(__m256d __A, __m256d __B, __m256d __C)
+{
+  return (__m256d)__builtin_ia32_vfmaddpd256(__A, __B, __C);
+}
+
+static __inline__ __m256 __attribute__((__always_inline__, __nodebug__))
+_mm256_fmsub_ps(__m256 __A, __m256 __B, __m256 __C)
+{
+  return (__m256)__builtin_ia32_vfmsubps256(__A, __B, __C);
+}
+
+static __inline__ __m256d __attribute__((__always_inline__, __nodebug__))
+_mm256_fmsub_pd(__m256d __A, __m256d __B, __m256d __C)
+{
+  return (__m256d)__builtin_ia32_vfmsubpd256(__A, __B, __C);
+}
+
+static __inline__ __m256 __attribute__((__always_inline__, __nodebug__))
+_mm256_fnmadd_ps(__m256 __A, __m256 __B, __m256 __C)
+{
+  return (__m256)__builtin_ia32_vfnmaddps256(__A, __B, __C);
+}
+
+static __inline__ __m256d __attribute__((__always_inline__, __nodebug__))
+_mm256_fnmadd_pd(__m256d __A, __m256d __B, __m256d __C)
+{
+  return (__m256d)__builtin_ia32_vfnmaddpd256(__A, __B, __C);
+}
+
+static __inline__ __m256 __attribute__((__always_inline__, __nodebug__))
+_mm256_fnmsub_ps(__m256 __A, __m256 __B, __m256 __C)
+{
+  return (__m256)__builtin_ia32_vfnmsubps256(__A, __B, __C);
+}
+
+static __inline__ __m256d __attribute__((__always_inline__, __nodebug__))
+_mm256_fnmsub_pd(__m256d __A, __m256d __B, __m256d __C)
+{
+  return (__m256d)__builtin_ia32_vfnmsubpd256(__A, __B, __C);
+}
+
+static __inline__ __m256 __attribute__((__always_inline__, __nodebug__))
+_mm256_fmaddsub_ps(__m256 __A, __m256 __B, __m256 __C)
+{
+  return (__m256)__builtin_ia32_vfmaddsubps256(__A, __B, __C);
+}
+
+static __inline__ __m256d __attribute__((__always_inline__, __nodebug__))
+_mm256_fmaddsub_pd(__m256d __A, __m256d __B, __m256d __C)
+{
+  return (__m256d)__builtin_ia32_vfmaddsubpd256(__A, __B, __C);
+}
+
+static __inline__ __m256 __attribute__((__always_inline__, __nodebug__))
+_mm256_fmsubadd_ps(__m256 __A, __m256 __B, __m256 __C)
+{
+  return (__m256)__builtin_ia32_vfmsubaddps256(__A, __B, __C);
+}
+
+static __inline__ __m256d __attribute__((__always_inline__, __nodebug__))
+_mm256_fmsubadd_pd(__m256d __A, __m256d __B, __m256d __C)
+{
+  return (__m256d)__builtin_ia32_vfmsubaddpd256(__A, __B, __C);
+}
+
+#endif /* __FMA__ */
+
+#endif /* __FMAINTRIN_H */
diff --git a/clang/lib/Headers/immintrin.h b/clang/lib/Headers/immintrin.h
index 16055251dd0..f4258dc28ed 100644
--- a/clang/lib/Headers/immintrin.h
+++ b/clang/lib/Headers/immintrin.h
@@ -72,4 +72,8 @@
 #include <lzcntintrin.h>
 #endif
 
+#ifdef __FMA__
+#include <fmaintrin.h>
+#endif
+
 #endif /* __IMMINTRIN_H */
diff --git a/clang/test/CodeGen/fma-builtins.c b/clang/test/CodeGen/fma-builtins.c
new file mode 100644
index 00000000000..3424616b123
--- /dev/null
+++ b/clang/test/CodeGen/fma-builtins.c
@@ -0,0 +1,166 @@
+// RUN: %clang_cc1 %s -O3 -triple=x86_64-apple-darwin -target-feature +fma -emit-llvm -o - | FileCheck %s
+
+// Don't include mm_malloc.h, it's system specific.
+#define __MM_MALLOC_H
+
+#include <immintrin.h>
+
+__m128 test_mm_fmadd_ps(__m128 a, __m128 b, __m128 c) {
+  // CHECK: @llvm.x86.fma.vfmadd.ps
+  return _mm_fmadd_ps(a, b, c);
+}
+
+__m128d test_mm_fmadd_pd(__m128d a, __m128d b, __m128d c) {
+  // CHECK: @llvm.x86.fma.vfmadd.pd
+  return _mm_fmadd_pd(a, b, c);
+}
+
+__m128 test_mm_fmadd_ss(__m128 a, __m128 b, __m128 c) {
+  // CHECK: @llvm.x86.fma.vfmadd.ss
+  return _mm_fmadd_ss(a, b, c);
+}
+
+__m128d test_mm_fmadd_sd(__m128d a, __m128d b, __m128d c) {
+  // CHECK: @llvm.x86.fma.vfmadd.sd
+  return _mm_fmadd_sd(a, b, c);
+}
+
+__m128 test_mm_fmsub_ps(__m128 a, __m128 b, __m128 c) {
+  // CHECK: @llvm.x86.fma.vfmsub.ps
+  return _mm_fmsub_ps(a, b, c);
+}
+
+__m128d test_mm_fmsub_pd(__m128d a, __m128d b, __m128d c) {
+  // CHECK: @llvm.x86.fma.vfmsub.pd
+  return _mm_fmsub_pd(a, b, c);
+}
+
+__m128 test_mm_fmsub_ss(__m128 a, __m128 b, __m128 c) {
+  // CHECK: @llvm.x86.fma.vfmsub.ss
+  return _mm_fmsub_ss(a, b, c);
+}
+
+__m128d test_mm_fmsub_sd(__m128d a, __m128d b, __m128d c) {
+  // CHECK: @llvm.x86.fma.vfmsub.sd
+  return _mm_fmsub_sd(a, b, c);
+}
+
+__m128 test_mm_fnmadd_ps(__m128 a, __m128 b, __m128 c) {
+  // CHECK: @llvm.x86.fma.vfnmadd.ps
+  return _mm_fnmadd_ps(a, b, c);
+}
+
+__m128d test_mm_fnmadd_pd(__m128d a, __m128d b, __m128d c) {
+  // CHECK: @llvm.x86.fma.vfnmadd.pd
+  return _mm_fnmadd_pd(a, b, c);
+}
+
+__m128 test_mm_fnmadd_ss(__m128 a, __m128 b, __m128 c) {
+  // CHECK: @llvm.x86.fma.vfnmadd.ss
+  return _mm_fnmadd_ss(a, b, c);
+}
+
+__m128d test_mm_fnmadd_sd(__m128d a, __m128d b, __m128d c) {
+  // CHECK: @llvm.x86.fma.vfnmadd.sd
+  return _mm_fnmadd_sd(a, b, c);
+}
+
+__m128 test_mm_fnmsub_ps(__m128 a, __m128 b, __m128 c) {
+  // CHECK: @llvm.x86.fma.vfnmsub.ps
+  return _mm_fnmsub_ps(a, b, c);
+}
+
+__m128d test_mm_fnmsub_pd(__m128d a, __m128d b, __m128d c) {
+  // CHECK: @llvm.x86.fma.vfnmsub.pd
+  return _mm_fnmsub_pd(a, b, c);
+}
+
+__m128 test_mm_fnmsub_ss(__m128 a, __m128 b, __m128 c) {
+  // CHECK: @llvm.x86.fma.vfnmsub.ss
+  return _mm_fnmsub_ss(a, b, c);
+}
+
+__m128d test_mm_fnmsub_sd(__m128d a, __m128d b, __m128d c) {
+  // CHECK: @llvm.x86.fma.vfnmsub.sd
+  return _mm_fnmsub_sd(a, b, c);
+}
+
+__m128 test_mm_fmaddsub_ps(__m128 a, __m128 b, __m128 c) {
+  // CHECK: @llvm.x86.fma.vfmaddsub.ps
+  return _mm_fmaddsub_ps(a, b, c);
+}
+
+__m128d test_mm_fmaddsub_pd(__m128d a, __m128d b, __m128d c) {
+  // CHECK: @llvm.x86.fma.vfmaddsub.pd
+  return _mm_fmaddsub_pd(a, b, c);
+}
+
+__m128 test_mm_fmsubadd_ps(__m128 a, __m128 b, __m128 c) {
+  // CHECK: @llvm.x86.fma.vfmsubadd.ps
+  return _mm_fmsubadd_ps(a, b, c);
+}
+
+__m128d test_mm_fmsubadd_pd(__m128d a, __m128d b, __m128d c) {
+  // CHECK: @llvm.x86.fma.vfmsubadd.pd
+  return _mm_fmsubadd_pd(a, b, c);
+}
+
+__m256 test_mm256_fmadd_ps(__m256 a, __m256 b, __m256 c) {
+  // CHECK: @llvm.x86.fma.vfmadd.ps.256
+  return _mm256_fmadd_ps(a, b, c);
+}
+
+__m256d test_mm256_fmadd_pd(__m256d a, __m256d b, __m256d c) {
+  // CHECK: @llvm.x86.fma.vfmadd.pd.256
+  return _mm256_fmadd_pd(a, b, c);
+}
+
+__m256 test_mm256_fmsub_ps(__m256 a, __m256 b, __m256 c) {
+  // CHECK: @llvm.x86.fma.vfmsub.ps.256
+  return _mm256_fmsub_ps(a, b, c);
+}
+
+__m256d test_mm256_fmsub_pd(__m256d a, __m256d b, __m256d c) {
+  // CHECK: @llvm.x86.fma.vfmsub.pd.256
+  return _mm256_fmsub_pd(a, b, c);
+}
+
+__m256 test_mm256_fnmadd_ps(__m256 a, __m256 b, __m256 c) {
+  // CHECK: @llvm.x86.fma.vfnmadd.ps.256
+  return _mm256_fnmadd_ps(a, b, c);
+}
+
+__m256d test_mm256_fnmadd_pd(__m256d a, __m256d b, __m256d c) {
+  // CHECK: @llvm.x86.fma.vfnmadd.pd.256
+  return _mm256_fnmadd_pd(a, b, c);
+}
+
+__m256 test_mm256_fnmsub_ps(__m256 a, __m256 b, __m256 c) {
+  // CHECK: @llvm.x86.fma.vfnmsub.ps.256
+  return _mm256_fnmsub_ps(a, b, c);
+}
+
+__m256d test_mm256_fnmsub_pd(__m256d a, __m256d b, __m256d c) {
+  // CHECK: @llvm.x86.fma.vfnmsub.pd.256
+  return _mm256_fnmsub_pd(a, b, c);
+}
+
+__m256 test_mm256_fmaddsub_ps(__m256 a, __m256 b, __m256 c) {
+  // CHECK: @llvm.x86.fma.vfmaddsub.ps.256
+  return _mm256_fmaddsub_ps(a, b, c);
+}
+
+__m256d test_mm256_fmaddsub_pd(__m256d a, __m256d b, __m256d c) {
+  // CHECK: @llvm.x86.fma.vfmaddsub.pd.256
+  return _mm256_fmaddsub_pd(a, b, c);
+}
+
+__m256 test_mm256_fmsubadd_ps(__m256 a, __m256 b, __m256 c) {
+  // CHECK: @llvm.x86.fma.vfmsubadd.ps.256
+  return _mm256_fmsubadd_ps(a, b, c);
+}
+
+__m256d test_mm256_fmsubadd_pd(__m256d a, __m256d b, __m256d c) {
+  // CHECK: @llvm.x86.fma.vfmsubadd.pd.256
+  return _mm256_fmsubadd_pd(a, b, c);
+}