[AVX512] Added VPCMPEQ intrinisics to headers.

Added tests.

Patch by Maxim Blumenthal <maxim.blumenthal@intel.com>

llvm-svn: 219319

5 files changed, 264 insertions, 0 deletions

diff --git a/clang/lib/Headers/avx512bwintrin.h b/clang/lib/Headers/avx512bwintrin.h
new file mode 100644
index 00000000000..bc4d4ac6afd
--- /dev/null
+++ b/clang/lib/Headers/avx512bwintrin.h
@@ -0,0 +1,60 @@
+/*===------------- avx512bwintrin.h - AVX512BW 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 __AVX512BWINTRIN_H
+#define __AVX512BWINTRIN_H
+
+typedef unsigned int __mmask32;
+typedef unsigned long long __mmask64;
+typedef char __v64qi __attribute__ ((vector_size (64)));
+typedef short __v32hi __attribute__ ((__vector_size__ (64)));
+
+
+/* Integer compare */
+
+static __inline__ __mmask64 __attribute__((__always_inline__, __nodebug__))
+_mm512_cmpeq_epi8_mask(__m512i __a, __m512i __b) {
+  return (__mmask64)__builtin_ia32_pcmpeqb512_mask((__v64qi)__a, (__v64qi)__b,
+                                                   (__mmask64)-1);
+}
+
+static __inline__ __mmask64 __attribute__((__always_inline__, __nodebug__))
+_mm512_mask_cmpeq_epi8_mask(__mmask64 __u, __m512i __a, __m512i __b) {
+  return (__mmask64)__builtin_ia32_pcmpeqb512_mask((__v64qi)__a, (__v64qi)__b,
+                                                   __u);
+}
+
+static __inline__ __mmask32 __attribute__((__always_inline__, __nodebug__))
+_mm512_cmpeq_epi16_mask(__m512i __a, __m512i __b) {
+  return (__mmask32)__builtin_ia32_pcmpeqw512_mask((__v32hi)__a, (__v32hi)__b,
+                                                   (__mmask32)-1);
+}
+
+static __inline__ __mmask32 __attribute__((__always_inline__, __nodebug__))
+_mm512_mask_cmpeq_epi16_mask(__mmask32 __u, __m512i __a, __m512i __b) {
+  return (__mmask32)__builtin_ia32_pcmpeqw512_mask((__v32hi)__a, (__v32hi)__b,
+                                                   __u);
+}
+
+#endif
diff --git a/clang/lib/Headers/avx512fintrin.h b/clang/lib/Headers/avx512fintrin.h
index 5736fbe555d..9591dcf37a5 100644
--- a/clang/lib/Headers/avx512fintrin.h
+++ b/clang/lib/Headers/avx512fintrin.h
@@ -974,4 +974,30 @@ _mm512_knot(__mmask16 __M)
   return __builtin_ia32_knothi(__M);
 }
 
+/* Integer compare */
+
+static __inline__ __mmask16 __attribute__((__always_inline__, __nodebug__))
+_mm512_cmpeq_epi32_mask(__m512i __a, __m512i __b) {
+  return (__mmask16)__builtin_ia32_pcmpeqd512_mask((__v16si)__a, (__v16si)__b,
+                                                   (__mmask16)-1);
+}
+
+static __inline__ __mmask16 __attribute__((__always_inline__, __nodebug__))
+_mm512_mask_cmpeq_epi32_mask(__mmask16 __u, __m512i __a, __m512i __b) {
+  return (__mmask16)__builtin_ia32_pcmpeqd512_mask((__v16si)__a, (__v16si)__b,
+                                                   __u);
+}
+
+static __inline__ __mmask8 __attribute__((__always_inline__, __nodebug__))
+_mm512_mask_cmpeq_epi64_mask(__mmask8 __u, __m512i __a, __m512i __b) {
+  return (__mmask8)__builtin_ia32_pcmpeqq512_mask((__v8di)__a, (__v8di)__b,
+                                                  __u);
+}
+
+static __inline__ __mmask8 __attribute__((__always_inline__, __nodebug__))
+_mm512_cmpeq_epi64_mask(__m512i __a, __m512i __b) {
+  return (__mmask8)__builtin_ia32_pcmpeqq512_mask((__v8di)__a, (__v8di)__b,
+                                                  (__mmask8)-1);
+}
+
 #endif // __AVX512FINTRIN_H
diff --git a/clang/lib/Headers/avx512vlbwintrin.h b/clang/lib/Headers/avx512vlbwintrin.h
new file mode 100644
index 00000000000..11333f85175
--- /dev/null
+++ b/clang/lib/Headers/avx512vlbwintrin.h
@@ -0,0 +1,83 @@
+/*===---- avx512vlbwintrin.h - AVX512VL and AVX512BW 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 <avx512vlbwintrin.h> directly; include <immintrin.h> instead."
+#endif
+
+#ifndef __AVX512VLBWINTRIN_H
+#define __AVX512VLBWINTRIN_H
+
+/* Integer compare */
+
+static __inline__ __mmask16 __attribute__((__always_inline__, __nodebug__))
+_mm_cmpeq_epi8_mask(__m128i __a, __m128i __b) {
+  return (__mmask16)__builtin_ia32_pcmpeqb128_mask((__v16qi)__a, (__v16qi)__b,
+                                                   (__mmask16)-1);
+}
+
+static __inline__ __mmask16 __attribute__((__always_inline__, __nodebug__))
+_mm_mask_cmpeq_epi8_mask(__mmask16 __u, __m128i __a, __m128i __b) {
+  return (__mmask16)__builtin_ia32_pcmpeqb128_mask((__v16qi)__a, (__v16qi)__b,
+                                                   __u);
+}
+
+
+static __inline__ __mmask32 __attribute__((__always_inline__, __nodebug__))
+_mm256_cmpeq_epi8_mask(__m256i __a, __m256i __b) {
+  return (__mmask32)__builtin_ia32_pcmpeqb256_mask((__v32qi)__a, (__v32qi)__b,
+                                                   (__mmask32)-1);
+}
+
+static __inline__ __mmask32 __attribute__((__always_inline__, __nodebug__))
+_mm256_mask_cmpeq_epi8_mask(__mmask32 __u, __m256i __a, __m256i __b) {
+  return (__mmask32)__builtin_ia32_pcmpeqb256_mask((__v32qi)__a, (__v32qi)__b,
+                                                   __u);
+}
+
+static __inline__ __mmask8 __attribute__((__always_inline__, __nodebug__))
+_mm_cmpeq_epi16_mask(__m128i __a, __m128i __b) {
+  return (__mmask8)__builtin_ia32_pcmpeqw128_mask((__v8hi)__a, (__v8hi)__b,
+                                                  (__mmask8)-1);
+}
+
+static __inline__ __mmask8 __attribute__((__always_inline__, __nodebug__))
+_mm_mask_cmpeq_epi16_mask(__mmask8 __u, __m128i __a, __m128i __b) {
+  return (__mmask8)__builtin_ia32_pcmpeqw128_mask((__v8hi)__a, (__v8hi)__b,
+                                                  __u);
+}
+
+
+static __inline__ __mmask16 __attribute__((__always_inline__, __nodebug__))
+_mm256_cmpeq_epi16_mask(__m256i __a, __m256i __b) {
+  return (__mmask16)__builtin_ia32_pcmpeqw256_mask((__v16hi)__a, (__v16hi)__b,
+                                                   (__mmask16)-1);
+}
+
+static __inline__ __mmask16 __attribute__((__always_inline__, __nodebug__))
+_mm256_mask_cmpeq_epi16_mask(__mmask16 __u, __m256i __a, __m256i __b) {
+  return (__mmask16)__builtin_ia32_pcmpeqw256_mask((__v16hi)__a, (__v16hi)__b,
+                                                   __u);
+}
+
+#endif /* __AVX512VLBWINTRIN_H */
diff --git a/clang/lib/Headers/avx512vlintrin.h b/clang/lib/Headers/avx512vlintrin.h
new file mode 100644
index 00000000000..8a374b10267
--- /dev/null
+++ b/clang/lib/Headers/avx512vlintrin.h
@@ -0,0 +1,83 @@
+/*===---- avx512vlintrin.h - AVX512VL 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 <avx512vlintrin.h> directly; include <immintrin.h> instead."
+#endif
+
+#ifndef __AVX512VLINTRIN_H
+#define __AVX512VLINTRIN_H
+
+/* Integer compare */
+
+static __inline__ __mmask8 __attribute__((__always_inline__, __nodebug__))
+_mm_cmpeq_epi32_mask(__m128i __a, __m128i __b) {
+  return (__mmask8)__builtin_ia32_pcmpeqd128_mask((__v4si)__a, (__v4si)__b,
+                                                  (__mmask8)-1);
+}
+
+static __inline__ __mmask8 __attribute__((__always_inline__, __nodebug__))
+_mm_mask_cmpeq_epi32_mask(__mmask8 __u, __m128i __a, __m128i __b) {
+  return (__mmask8)__builtin_ia32_pcmpeqd128_mask((__v4si)__a, (__v4si)__b,
+                                                  __u);
+}
+
+
+static __inline__ __mmask8 __attribute__((__always_inline__, __nodebug__))
+_mm256_cmpeq_epi32_mask(__m256i __a, __m256i __b) {
+  return (__mmask8)__builtin_ia32_pcmpeqd256_mask((__v8si)__a, (__v8si)__b,
+                                                  (__mmask8)-1);
+}
+
+static __inline__ __mmask8 __attribute__((__always_inline__, __nodebug__))
+_mm256_mask_cmpeq_epi32_mask(__mmask8 __u, __m256i __a, __m256i __b) {
+  return (__mmask8)__builtin_ia32_pcmpeqd256_mask((__v8si)__a, (__v8si)__b,
+                                                  __u);
+}
+
+static __inline__ __mmask8 __attribute__((__always_inline__, __nodebug__))
+_mm_cmpeq_epi64_mask(__m128i __a, __m128i __b) {
+  return (__mmask8)__builtin_ia32_pcmpeqq128_mask((__v2di)__a, (__v2di)__b,
+                                                  (__mmask8)-1);
+}
+
+static __inline__ __mmask8 __attribute__((__always_inline__, __nodebug__))
+_mm_mask_cmpeq_epi64_mask(__mmask8 __u, __m128i __a, __m128i __b) {
+  return (__mmask8)__builtin_ia32_pcmpeqq128_mask((__v2di)__a, (__v2di)__b,
+                                                  __u);
+}
+
+
+static __inline__ __mmask8 __attribute__((__always_inline__, __nodebug__))
+_mm256_cmpeq_epi64_mask(__m256i __a, __m256i __b) {
+  return (__mmask8)__builtin_ia32_pcmpeqq256_mask((__v4di)__a, (__v4di)__b,
+                                                  (__mmask8)-1);
+}
+
+static __inline__ __mmask8 __attribute__((__always_inline__, __nodebug__))
+_mm256_mask_cmpeq_epi64_mask(__mmask8 __u, __m256i __a, __m256i __b) {
+  return (__mmask8)__builtin_ia32_pcmpeqq256_mask((__v4di)__a, (__v4di)__b,
+                                                  __u);
+}
+
+#endif /* __AVX512VLINTRIN_H */
diff --git a/clang/lib/Headers/immintrin.h b/clang/lib/Headers/immintrin.h
index 38c020604a9..295c6630424 100644
--- a/clang/lib/Headers/immintrin.h
+++ b/clang/lib/Headers/immintrin.h
@@ -80,6 +80,18 @@
 #include <avx512fintrin.h>
 #endif
 
+#ifdef __AVX512VL__
+#include <avx512vlintrin.h>
+#endif
+
+#ifdef __AVX512BW__
+#include <avx512bwintrin.h>
+#endif
+
+#if defined (__AVX512VL__) && defined (__AVX512BW__)
+#include <avx512vlbwintrin.h>
+#endif
+
 #ifdef __AVX512ER__
 #include <avx512erintrin.h>
 #endif