From 1b6fd15322ca8431cd390d66c671f6ca60fbe59a Mon Sep 17 00:00:00 2001
From: Robert Khasanov <rob.khasanov@gmail.com>
Date: Wed, 8 Oct 2014 17:18:13 +0000
Subject: [PATCH] [AVX512] Added VPCMPEQ intrinisics to headers. Added tests.
Patch by Maxim Blumenthal <maxim.blumenthal@intel.com>
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@219319 91177308-0d34-0410-b5e6-96231b3b80d8
---
include/clang/Basic/BuiltinsX86.def | 12 +++++
lib/Headers/avx512bwintrin.h | 60 +++++++++++++++++++++
lib/Headers/avx512fintrin.h | 26 +++++++++
lib/Headers/avx512vlbwintrin.h | 83 +++++++++++++++++++++++++++++
lib/Headers/avx512vlintrin.h | 83 +++++++++++++++++++++++++++++
lib/Headers/immintrin.h | 12 +++++
test/CodeGen/avx512bw-builtins.c | 27 ++++++++++
test/CodeGen/avx512f-builtins.c | 24 +++++++++
test/CodeGen/avx512vl-builtins.c | 51 ++++++++++++++++++
test/CodeGen/avx512vlbw-builtins.c | 51 ++++++++++++++++++
test/Headers/x86intrin.c | 6 +++
11 files changed, 435 insertions(+)
create mode 100644 lib/Headers/avx512bwintrin.h
create mode 100644 lib/Headers/avx512vlbwintrin.h
create mode 100644 lib/Headers/avx512vlintrin.h
create mode 100644 test/CodeGen/avx512bw-builtins.c
create mode 100644 test/CodeGen/avx512vl-builtins.c
create mode 100644 test/CodeGen/avx512vlbw-builtins.c
diff --git a/include/clang/Basic/BuiltinsX86.def b/include/clang/Basic/BuiltinsX86.def
index a571fca96f5..b0f6becce20 100644
--- a/include/clang/Basic/BuiltinsX86.def
+++ b/include/clang/Basic/BuiltinsX86.def
@@ -801,6 +801,18 @@ BUILTIN(__builtin_ia32_cvttps2udq512_mask, "V16iV16fV16iUsCi", "")
BUILTIN(__builtin_ia32_cvttpd2dq512_mask, "V8iV8dV8iUcCi", "")
BUILTIN(__builtin_ia32_cvttpd2udq512_mask, "V8iV8dV8iUcCi", "")
BUILTIN(__builtin_ia32_cmpps512_mask, "UsV16fV16fCiUsi", "")
+BUILTIN(__builtin_ia32_pcmpeqb512_mask, "LLiV64cV64cLLi", "")
+BUILTIN(__builtin_ia32_pcmpeqd512_mask, "sV16iV16is", "")
+BUILTIN(__builtin_ia32_pcmpeqq512_mask, "cV8LLiV8LLic", "")
+BUILTIN(__builtin_ia32_pcmpeqw512_mask, "iV32sV32si", "")
+BUILTIN(__builtin_ia32_pcmpeqb256_mask, "iV32cV32ci", "")
+BUILTIN(__builtin_ia32_pcmpeqd256_mask, "cV8iV8ic", "")
+BUILTIN(__builtin_ia32_pcmpeqq256_mask, "cV4LLiV4LLic", "")
+BUILTIN(__builtin_ia32_pcmpeqw256_mask, "sV16sV16ss", "")
+BUILTIN(__builtin_ia32_pcmpeqb128_mask, "sV16cV16cs", "")
+BUILTIN(__builtin_ia32_pcmpeqd128_mask, "cV4iV4ic", "")
+BUILTIN(__builtin_ia32_pcmpeqq128_mask, "cV2LLiV2LLic", "")
+BUILTIN(__builtin_ia32_pcmpeqw128_mask, "cV8sV8sc", "")
BUILTIN(__builtin_ia32_cmppd512_mask, "UcV8dV8dCiUci", "")
BUILTIN(__builtin_ia32_rndscaleps_mask, "V16fV16fCiV16fUsCi", "")
BUILTIN(__builtin_ia32_rndscalepd_mask, "V8dV8dCiV8dUcCi", "")
diff --git a/lib/Headers/avx512bwintrin.h b/lib/Headers/avx512bwintrin.h
new file mode 100644
index 00000000000..bc4d4ac6afd
--- /dev/null
+++ b/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/lib/Headers/avx512fintrin.h b/lib/Headers/avx512fintrin.h
index 5736fbe555d..9591dcf37a5 100644
--- a/lib/Headers/avx512fintrin.h
+++ b/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/lib/Headers/avx512vlbwintrin.h b/lib/Headers/avx512vlbwintrin.h
new file mode 100644
index 00000000000..11333f85175
--- /dev/null
+++ b/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/lib/Headers/avx512vlintrin.h b/lib/Headers/avx512vlintrin.h
new file mode 100644
index 00000000000..8a374b10267
--- /dev/null
+++ b/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/lib/Headers/immintrin.h b/lib/Headers/immintrin.h
index 38c020604a9..295c6630424 100644
--- a/lib/Headers/immintrin.h
+++ b/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
diff --git a/test/CodeGen/avx512bw-builtins.c b/test/CodeGen/avx512bw-builtins.c
new file mode 100644
index 00000000000..ada84657a60
--- /dev/null
+++ b/test/CodeGen/avx512bw-builtins.c
@@ -0,0 +1,27 @@
+// RUN: %clang_cc1 %s -O0 -triple=x86_64-apple-darwin -ffreestanding -target-feature +avx512bw -emit-llvm -o - -Werror | FileCheck %s
+
+#include <immintrin.h>
+
+__mmask64 test_mm512_cmpeq_epi8_mask(__m512i __a, __m512i __b) {
+ // CHECK-LABEL: @test_mm512_cmpeq_epi8_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.b.512
+ return (__mmask64)_mm512_cmpeq_epi8_mask(__a, __b);
+}
+
+__mmask64 test_mm512_mask_cmpeq_epi8_mask(__mmask64 __u, __m512i __a, __m512i __b) {
+ // CHECK-LABEL: @test_mm512_mask_cmpeq_epi8_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.b.512
+ return (__mmask64)_mm512_mask_cmpeq_epi8_mask(__u, __a, __b);
+}
+
+__mmask32 test_mm512_cmpeq_epi16_mask(__m512i __a, __m512i __b) {
+ // CHECK-LABEL: @test_mm512_cmpeq_epi16_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.w.512
+ return (__mmask32)_mm512_cmpeq_epi16_mask(__a, __b);
+}
+
+__mmask32 test_mm512_mask_cmpeq_epi16_mask(__mmask32 __u, __m512i __a, __m512i __b) {
+ // CHECK-LABEL: @test_mm512_mask_cmpeq_epi16_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.w.512
+ return (__mmask32)_mm512_mask_cmpeq_epi16_mask(__u, __a, __b);
+}
diff --git a/test/CodeGen/avx512f-builtins.c b/test/CodeGen/avx512f-builtins.c
index bde93fb271f..ba05f4abdc8 100644
--- a/test/CodeGen/avx512f-builtins.c
+++ b/test/CodeGen/avx512f-builtins.c
@@ -158,3 +158,27 @@ __m512i test_mm512_fmadd_pd(__m512d a, __m512d b, __m512d c)
// CHECK: @llvm.x86.fma.mask.vfmadd.pd.512
return _mm512_fmadd_pd(a, b, c);
}
+
+__mmask16 test_mm512_cmpeq_epi32_mask(__m512i __a, __m512i __b) {
+ // CHECK-LABEL: @test_mm512_cmpeq_epi32_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.d.512
+ return (__mmask16)_mm512_cmpeq_epi32_mask(__a, __b);
+}
+
+__mmask16 test_mm512_mask_cmpeq_epi32_mask(__mmask16 __u, __m512i __a, __m512i __b) {
+ // CHECK-LABEL: @test_mm512_mask_cmpeq_epi32_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.d.512
+ return (__mmask16)_mm512_mask_cmpeq_epi32_mask(__u, __a, __b);
+}
+
+__mmask8 test_mm512_mask_cmpeq_epi64_mask(__mmask8 __u, __m512i __a, __m512i __b) {
+ // CHECK-LABEL: @test_mm512_mask_cmpeq_epi64_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.q.512
+ return (__mmask8)_mm512_mask_cmpeq_epi64_mask(__u, __a, __b);
+}
+
+__mmask8 test_mm512_cmpeq_epi64_mask(__m512i __a, __m512i __b) {
+ // CHECK-LABEL: @test_mm512_cmpeq_epi64_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.q.512
+ return (__mmask8)_mm512_cmpeq_epi64_mask(__a, __b);
+}
diff --git a/test/CodeGen/avx512vl-builtins.c b/test/CodeGen/avx512vl-builtins.c
new file mode 100644
index 00000000000..e4b45173bb1
--- /dev/null
+++ b/test/CodeGen/avx512vl-builtins.c
@@ -0,0 +1,51 @@
+// RUN: %clang_cc1 %s -O0 -triple=x86_64-apple-darwin -ffreestanding -target-feature +avx512f -target-feature +avx512vl -emit-llvm -o - -Werror | FileCheck %s
+
+#include <immintrin.h>
+
+__mmask8 test_mm256_cmpeq_epi32_mask(__m256i __a, __m256i __b) {
+ // CHECK-LABEL: @test_mm256_cmpeq_epi32_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.d.256
+ return (__mmask8)_mm256_cmpeq_epi32_mask(__a, __b);
+}
+
+__mmask8 test_mm256_mask_cmpeq_epi32_mask(__mmask8 __u, __m256i __a, __m256i __b) {
+ // CHECK-LABEL: @test_mm256_mask_cmpeq_epi32_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.d.256
+ return (__mmask8)_mm256_mask_cmpeq_epi32_mask(__u, __a, __b);
+}
+
+__mmask8 test_mm_cmpeq_epi32_mask(__m128i __a, __m128i __b) {
+ // CHECK-LABEL: @test_mm_cmpeq_epi32_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.d.128
+ return (__mmask8)_mm_cmpeq_epi32_mask(__a, __b);
+}
+
+__mmask8 test_mm_mask_cmpeq_epi32_mask(__mmask8 __u, __m128i __a, __m128i __b) {
+ // CHECK-LABEL: @test_mm_mask_cmpeq_epi32_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.d.128
+ return (__mmask8)_mm_mask_cmpeq_epi32_mask(__u, __a, __b);
+}
+
+__mmask8 test_mm256_cmpeq_epi64_mask(__m256i __a, __m256i __b) {
+ // CHECK-LABEL: @test_mm256_cmpeq_epi64_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.q.256
+ return (__mmask8)_mm256_cmpeq_epi64_mask(__a, __b);
+}
+
+__mmask8 test_mm256_mask_cmpeq_epi64_mask(__mmask8 __u, __m256i __a, __m256i __b) {
+ // CHECK-LABEL: @test_mm256_mask_cmpeq_epi64_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.q.256
+ return (__mmask8)_mm256_mask_cmpeq_epi64_mask(__u, __a, __b);
+}
+
+__mmask8 test_mm_cmpeq_epi64_mask(__m128i __a, __m128i __b) {
+ // CHECK-LABEL: @test_mm_cmpeq_epi64_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.q.128
+ return (__mmask8)_mm_cmpeq_epi64_mask(__a, __b);
+}
+
+__mmask8 test_mm_mask_cmpeq_epi64_mask(__mmask8 __u, __m128i __a, __m128i __b) {
+ // CHECK-LABEL: @test_mm_mask_cmpeq_epi64_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.q.128
+ return (__mmask8)_mm_mask_cmpeq_epi64_mask(__u, __a, __b);
+}
diff --git a/test/CodeGen/avx512vlbw-builtins.c b/test/CodeGen/avx512vlbw-builtins.c
new file mode 100644
index 00000000000..a304f7b3d36
--- /dev/null
+++ b/test/CodeGen/avx512vlbw-builtins.c
@@ -0,0 +1,51 @@
+// RUN: %clang_cc1 %s -O0 -triple=x86_64-apple-darwin -ffreestanding -target-feature +avx512bw -target-feature +avx512vl -emit-llvm -o - -Werror | FileCheck %s
+
+#include <immintrin.h>
+
+__mmask32 test_mm256_cmpeq_epi8_mask(__m256i __a, __m256i __b) {
+ // CHECK-LABEL: @test_mm256_cmpeq_epi8_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.b.256
+ return (__mmask32)_mm256_cmpeq_epi8_mask(__a, __b);
+}
+
+__mmask32 test_mm256_mask_cmpeq_epi8_mask(__mmask32 __u, __m256i __a, __m256i __b) {
+ // CHECK-LABEL: @test_mm256_mask_cmpeq_epi8_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.b.256
+ return (__mmask32)_mm256_mask_cmpeq_epi8_mask(__u, __a, __b);
+}
+
+__mmask16 test_mm_cmpeq_epi8_mask(__m128i __a, __m128i __b) {
+ // CHECK-LABEL: @test_mm_cmpeq_epi8_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.b.128
+ return (__mmask16)_mm_cmpeq_epi8_mask(__a, __b);
+}
+
+__mmask16 test_mm_mask_cmpeq_epi8_mask(__mmask16 __u, __m128i __a, __m128i __b) {
+ // CHECK-LABEL: @test_mm_mask_cmpeq_epi8_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.b.128
+ return (__mmask16)_mm_mask_cmpeq_epi8_mask(__u, __a, __b);
+}
+
+__mmask16 test_mm256_cmpeq_epi16_mask(__m256i __a, __m256i __b) {
+ // CHECK-LABEL: @test_mm256_cmpeq_epi16_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.w.256
+ return (__mmask16)_mm256_cmpeq_epi16_mask(__a, __b);
+}
+
+__mmask16 test_mm256_mask_cmpeq_epi16_mask(__mmask16 __u, __m256i __a, __m256i __b) {
+ // CHECK-LABEL: @test_mm256_mask_cmpeq_epi16_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.w.256
+ return (__mmask16)_mm256_mask_cmpeq_epi16_mask(__u, __a, __b);
+}
+
+__mmask8 test_mm_cmpeq_epi16_mask(__m128i __a, __m128i __b) {
+ // CHECK-LABEL: @test_mm_cmpeq_epi16_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.w.128
+ return (__mmask8)_mm_cmpeq_epi16_mask(__a, __b);
+}
+
+__mmask8 test_mm_mask_cmpeq_epi16_mask(__mmask8 __u, __m128i __a, __m128i __b) {
+ // CHECK-LABEL: @test_mm_mask_cmpeq_epi16_mask
+ // CHECK: @llvm.x86.avx512.mask.pcmpeq.w.128
+ return (__mmask8)_mm_mask_cmpeq_epi16_mask(__u, __a, __b);
+}
diff --git a/test/Headers/x86intrin.c b/test/Headers/x86intrin.c
index 577e4cd0e86..4c9665bf286 100644
--- a/test/Headers/x86intrin.c
+++ b/test/Headers/x86intrin.c
@@ -87,6 +87,12 @@
#ifndef __AVX512F__
#define __AVX512F__
#endif
+#ifndef __AVX512VL__
+#define __AVX512VL__
+#endif
+#ifndef __AVX512BW__
+#define __AVX512BW__
+#endif
#ifndef __AVX512ER__
#define __AVX512ER__
#endif
--
GitLab