https://github.com/ningxinr updated
https://github.com/llvm/llvm-project/pull/153088
>From fcb1db6b8b0d13bfc84321a0e5bcd31db646524e Mon Sep 17 00:00:00 2001
From: ningxinr <ningx...@live.cn>
Date: Mon, 11 Aug 2025 13:36:48 -0700
Subject: [PATCH 1/4] [Headers][X86] Allow MMX integer saturated arithmetic
intrinsics to be used in constexpr
---
clang/lib/Headers/mmintrin.h | 28 ++++++++++++---------------
clang/test/CodeGen/X86/mmx-builtins.c | 4 ++++
2 files changed, 16 insertions(+), 16 deletions(-)
diff --git a/clang/lib/Headers/mmintrin.h b/clang/lib/Headers/mmintrin.h
index 3961b790cea74..afb93ccba781c 100644
--- a/clang/lib/Headers/mmintrin.h
+++ b/clang/lib/Headers/mmintrin.h
@@ -448,10 +448,9 @@ _mm_add_pi32(__m64 __m1, __m64 __m2)
/// A 64-bit integer vector of [8 x i8].
/// \returns A 64-bit integer vector of [8 x i8] containing the saturated sums
/// of both parameters.
-static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
-_mm_adds_pi8(__m64 __m1, __m64 __m2)
-{
- return (__m64)__builtin_elementwise_add_sat((__v8qs)__m1, (__v8qs)__m2);
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
+_mm_adds_pi8(__m64 __m1, __m64 __m2) {
+ return (__m64)__builtin_elementwise_add_sat((__v8qs)__m1, (__v8qs)__m2);
}
/// Adds, with saturation, each 16-bit signed integer element of the first
@@ -472,10 +471,9 @@ _mm_adds_pi8(__m64 __m1, __m64 __m2)
/// A 64-bit integer vector of [4 x i16].
/// \returns A 64-bit integer vector of [4 x i16] containing the saturated sums
/// of both parameters.
-static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
-_mm_adds_pi16(__m64 __m1, __m64 __m2)
-{
- return (__m64)__builtin_elementwise_add_sat((__v4hi)__m1, (__v4hi)__m2);
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
+_mm_adds_pi16(__m64 __m1, __m64 __m2) {
+ return (__m64)__builtin_elementwise_add_sat((__v4hi)__m1, (__v4hi)__m2);
}
/// Adds, with saturation, each 8-bit unsigned integer element of the first
@@ -605,10 +603,9 @@ _mm_sub_pi32(__m64 __m1, __m64 __m2)
/// A 64-bit integer vector of [8 x i8] containing the subtrahends.
/// \returns A 64-bit integer vector of [8 x i8] containing the saturated
/// differences of both parameters.
-static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
-_mm_subs_pi8(__m64 __m1, __m64 __m2)
-{
- return (__m64)__builtin_elementwise_sub_sat((__v8qs)__m1, (__v8qs)__m2);
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
+_mm_subs_pi8(__m64 __m1, __m64 __m2) {
+ return (__m64)__builtin_elementwise_sub_sat((__v8qs)__m1, (__v8qs)__m2);
}
/// Subtracts, with saturation, each 16-bit signed integer element of the
@@ -629,10 +626,9 @@ _mm_subs_pi8(__m64 __m1, __m64 __m2)
/// A 64-bit integer vector of [4 x i16] containing the subtrahends.
/// \returns A 64-bit integer vector of [4 x i16] containing the saturated
/// differences of both parameters.
-static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
-_mm_subs_pi16(__m64 __m1, __m64 __m2)
-{
- return (__m64)__builtin_elementwise_sub_sat((__v4hi)__m1, (__v4hi)__m2);
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2_CONSTEXPR
+_mm_subs_pi16(__m64 __m1, __m64 __m2) {
+ return (__m64)__builtin_elementwise_sub_sat((__v4hi)__m1, (__v4hi)__m2);
}
/// Subtracts each 8-bit unsigned integer element of the second 64-bit
diff --git a/clang/test/CodeGen/X86/mmx-builtins.c
b/clang/test/CodeGen/X86/mmx-builtins.c
index d47368c207f4b..29d31f85f2bc5 100644
--- a/clang/test/CodeGen/X86/mmx-builtins.c
+++ b/clang/test/CodeGen/X86/mmx-builtins.c
@@ -62,12 +62,14 @@ __m64 test_mm_adds_pi8(__m64 a, __m64 b) {
// CHECK: call <8 x i8> @llvm.sadd.sat.v8i8(
return _mm_adds_pi8(a, b);
}
+TEST_CONSTEXPR(match_v8qi(_mm_adds_pi8(_mm_setr_pi8(+100, +50, -100, +20, +80,
-50, +120, -20), _mm_setr_pi8(+50, +80, -50, +110, +60, -30, +20, -10)), +127,
+127, -128, +127, +127, -80, +127, -30));
__m64 test_mm_adds_pi16(__m64 a, __m64 b) {
// CHECK-LABEL: test_mm_adds_pi16
// CHECK: call <4 x i16> @llvm.sadd.sat.v4i16(
return _mm_adds_pi16(a, b);
}
+TEST_CONSTEXPR(match_v4hi(_mm_adds_pi16((__m64)(__v4hi){+32000, -32000,
+32000, -32000}, (__m64)(__v4hi){+800, -800, -800, +800}), +32767, -32768,
+31200, -31200));
__m64 test_mm_adds_pu8(__m64 a, __m64 b) {
// CHECK-LABEL: test_mm_adds_pu8
@@ -718,12 +720,14 @@ __m64 test_mm_subs_pi8(__m64 a, __m64 b) {
// CHECK: call <8 x i8> @llvm.ssub.sat.v8i8(
return _mm_subs_pi8(a, b);
}
+TEST_CONSTEXPR(match_v8qi(_mm_subs_pi8(_mm_setr_pi8(+100, +50, -100, +20, +80,
-50, +120, -20), _mm_setr_pi8(-50, -80, +50, -110, -60, +30, -20, +10)), +127,
+127, -128, +127, +127, -80, +127, -30));
__m64 test_mm_subs_pi16(__m64 a, __m64 b) {
// CHECK-LABEL: test_mm_subs_pi16
// CHECK: call <4 x i16> @llvm.ssub.sat.v4i16(
return _mm_subs_pi16(a, b);
}
+TEST_CONSTEXPR(match_v4hi(_mm_subs_pi16((__m64)(__v4hi){+32000, -32000,
+32000, -32000}, (__m64)(__v4hi){-800, +800, +800, -800}), +32767, -32768,
+31200, -31200));
__m64 test_mm_subs_pu8(__m64 a, __m64 b) {
// CHECK-LABEL: test_mm_subs_pu8
>From 13322c6b50cbd63b2c9f8c948e4e45545634b064 Mon Sep 17 00:00:00 2001
From: ningxinr <ningx...@live.cn>
Date: Mon, 11 Aug 2025 13:38:46 -0700
Subject: [PATCH 2/4] [Headers][X86] Allow SSE2 integer saturated arithmetic
intrinsics to be used in constexpr
---
clang/lib/Headers/emmintrin.h | 16 ++++++++--------
clang/test/CodeGen/X86/sse2-builtins.c | 4 ++++
2 files changed, 12 insertions(+), 8 deletions(-)
diff --git a/clang/lib/Headers/emmintrin.h b/clang/lib/Headers/emmintrin.h
index 60d2000dfb809..8936f3110050f 100644
--- a/clang/lib/Headers/emmintrin.h
+++ b/clang/lib/Headers/emmintrin.h
@@ -2170,8 +2170,8 @@ _mm_add_epi64(__m128i __a, __m128i __b) {
/// A 128-bit signed [16 x i8] vector.
/// \returns A 128-bit signed [16 x i8] vector containing the saturated sums of
/// both parameters.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_adds_epi8(__m128i __a,
- __m128i __b) {
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_adds_epi8(__m128i __a, __m128i __b) {
return (__m128i)__builtin_elementwise_add_sat((__v16qs)__a, (__v16qs)__b);
}
@@ -2192,8 +2192,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_adds_epi8(__m128i __a,
/// A 128-bit signed [8 x i16] vector.
/// \returns A 128-bit signed [8 x i16] vector containing the saturated sums of
/// both parameters.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_adds_epi16(__m128i __a,
- __m128i __b) {
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_adds_epi16(__m128i __a, __m128i __b) {
return (__m128i)__builtin_elementwise_add_sat((__v8hi)__a, (__v8hi)__b);
}
@@ -2597,8 +2597,8 @@ _mm_sub_epi64(__m128i __a, __m128i __b) {
/// A 128-bit integer vector containing the subtrahends.
/// \returns A 128-bit integer vector containing the differences of the values
/// in the operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_subs_epi8(__m128i __a,
- __m128i __b) {
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_subs_epi8(__m128i __a, __m128i __b) {
return (__m128i)__builtin_elementwise_sub_sat((__v16qs)__a, (__v16qs)__b);
}
@@ -2619,8 +2619,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_subs_epi8(__m128i __a,
/// A 128-bit integer vector containing the subtrahends.
/// \returns A 128-bit integer vector containing the differences of the values
/// in the operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_subs_epi16(__m128i __a,
- __m128i __b) {
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_subs_epi16(__m128i __a, __m128i __b) {
return (__m128i)__builtin_elementwise_sub_sat((__v8hi)__a, (__v8hi)__b);
}
diff --git a/clang/test/CodeGen/X86/sse2-builtins.c
b/clang/test/CodeGen/X86/sse2-builtins.c
index 23013dd75d641..04b357b88817c 100644
--- a/clang/test/CodeGen/X86/sse2-builtins.c
+++ b/clang/test/CodeGen/X86/sse2-builtins.c
@@ -63,12 +63,14 @@ __m128i test_mm_adds_epi8(__m128i A, __m128i B) {
// CHECK: call <16 x i8> @llvm.sadd.sat.v16i8(<16 x i8> %{{.*}}, <16 x i8>
%{{.*}})
return _mm_adds_epi8(A, B);
}
+TEST_CONSTEXPR(match_v16qi(_mm_adds_epi8(_mm_setr_epi8(+100, +50, -100, +20,
+80, -50, +120, -20, -100, -50, +100, -20, -80, +50, -120, +20),
_mm_setr_epi8(+50, +80, -50, +110, +60, -30, +20, -10, +50, +80, -50, +110,
+60, -30, +20, -10)), +127, +127, -128, +127, +127, -80, +127, -30, -50, +30,
+50, +90, -20, +20, -100, +10));
__m128i test_mm_adds_epi16(__m128i A, __m128i B) {
// CHECK-LABEL: test_mm_adds_epi16
// CHECK: call <8 x i16> @llvm.sadd.sat.v8i16(<8 x i16> %{{.*}}, <8 x i16>
%{{.*}})
return _mm_adds_epi16(A, B);
}
+TEST_CONSTEXPR(match_v8hi(_mm_adds_epi16(_mm_setr_epi16(+32000, -32000,
+32000, -32000, +80, -50, +120, -20), _mm_setr_epi16(+800, -800, -800, +800,
+60, -30, +20, -10)), +32767, -32768, +31200, -31200, +140, -80, +140, -30));
__m128i test_mm_adds_epu8(__m128i A, __m128i B) {
// CHECK-LABEL: test_mm_adds_epu8
@@ -1691,12 +1693,14 @@ __m128i test_mm_subs_epi8(__m128i A, __m128i B) {
// CHECK: call <16 x i8> @llvm.ssub.sat.v16i8(<16 x i8> %{{.*}}, <16 x i8>
%{{.*}})
return _mm_subs_epi8(A, B);
}
+TEST_CONSTEXPR(match_v16qi(_mm_subs_epi8(_mm_setr_epi8(+100, +50, -100, +20,
+80, -50, +120, -20, -100, -50, +100, -20, -80, +50, -120, +20),
_mm_setr_epi8(-50, -80, +50, -110, -60, +30, -20, +10, -50, -80, +50, -110,
-60, +30, -20, +10)), +127, +127, -128, +127, +127, -80, +127, -30, -50, +30,
+50, +90, -20, +20, -100, +10));
__m128i test_mm_subs_epi16(__m128i A, __m128i B) {
// CHECK-LABEL: test_mm_subs_epi16
// CHECK: call <8 x i16> @llvm.ssub.sat.v8i16(<8 x i16> %{{.*}}, <8 x i16>
%{{.*}})
return _mm_subs_epi16(A, B);
}
+TEST_CONSTEXPR(match_v8hi(_mm_subs_epi16(_mm_setr_epi16(+32000, -32000,
+32000, -32000, +80, -50, +120, -20), _mm_setr_epi16(-800, +800, +800, -800,
-60, +30, -20, +10)), +32767, -32768, +31200, -31200, +140, -80, +140, -30));
__m128i test_mm_subs_epu8(__m128i A, __m128i B) {
// CHECK-LABEL: test_mm_subs_epu8
>From 36cc2b3dbf5d2db1693299e1446e5a1ab1f637e3 Mon Sep 17 00:00:00 2001
From: ningxinr <ningx...@live.cn>
Date: Mon, 11 Aug 2025 13:39:38 -0700
Subject: [PATCH 3/4] [Headers][X86] Allow AVX2 integer saturated arithmetic
intrinsics to be used in constexpr
---
clang/lib/Headers/avx2intrin.h | 20 ++++++++------------
clang/test/CodeGen/X86/avx2-builtins.c | 4 ++++
2 files changed, 12 insertions(+), 12 deletions(-)
diff --git a/clang/lib/Headers/avx2intrin.h b/clang/lib/Headers/avx2intrin.h
index f00a8a516ecfe..b2e0b701c95ec 100644
--- a/clang/lib/Headers/avx2intrin.h
+++ b/clang/lib/Headers/avx2intrin.h
@@ -367,9 +367,8 @@ _mm256_add_epi64(__m256i __a, __m256i __b)
/// \param __b
/// A 256-bit integer vector containing one of the source operands.
/// \returns A 256-bit integer vector containing the sums.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_adds_epi8(__m256i __a, __m256i __b)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_adds_epi8(__m256i __a, __m256i __b) {
return (__m256i)__builtin_elementwise_add_sat((__v32qs)__a, (__v32qs)__b);
}
@@ -385,9 +384,8 @@ _mm256_adds_epi8(__m256i __a, __m256i __b)
/// \param __b
/// A 256-bit vector of [16 x i16] containing one of the source operands.
/// \returns A 256-bit vector of [16 x i16] containing the sums.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_adds_epi16(__m256i __a, __m256i __b)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_adds_epi16(__m256i __a, __m256i __b) {
return (__m256i)__builtin_elementwise_add_sat((__v16hi)__a, (__v16hi)__b);
}
@@ -2607,9 +2605,8 @@ _mm256_sub_epi64(__m256i __a, __m256i __b)
/// \param __b
/// A 256-bit integer vector containing the subtrahends.
/// \returns A 256-bit integer vector containing the differences.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_subs_epi8(__m256i __a, __m256i __b)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_subs_epi8(__m256i __a, __m256i __b) {
return (__m256i)__builtin_elementwise_sub_sat((__v32qs)__a, (__v32qs)__b);
}
@@ -2633,9 +2630,8 @@ _mm256_subs_epi8(__m256i __a, __m256i __b)
/// \param __b
/// A 256-bit vector of [16 x i16] containing the subtrahends.
/// \returns A 256-bit vector of [16 x i16] containing the differences.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_subs_epi16(__m256i __a, __m256i __b)
-{
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_subs_epi16(__m256i __a, __m256i __b) {
return (__m256i)__builtin_elementwise_sub_sat((__v16hi)__a, (__v16hi)__b);
}
diff --git a/clang/test/CodeGen/X86/avx2-builtins.c
b/clang/test/CodeGen/X86/avx2-builtins.c
index 8790485f00a8c..ce279b65a399d 100644
--- a/clang/test/CodeGen/X86/avx2-builtins.c
+++ b/clang/test/CodeGen/X86/avx2-builtins.c
@@ -60,12 +60,14 @@ __m256i test_mm256_adds_epi8(__m256i a, __m256i b) {
// CHECK: call <32 x i8> @llvm.sadd.sat.v32i8(<32 x i8> %{{.*}}, <32 x i8>
%{{.*}})
return _mm256_adds_epi8(a, b);
}
+TEST_CONSTEXPR(match_v32qi(_mm256_adds_epi8(_mm256_setr_epi8(0, +1, +2, +3,
+4, +5, +6, +7, +8, +9, +10, +11, +12, +13, +14, +15, +100, +50, -100, +20,
+80, -50, +120, -20, -100, -50, +100, -20, -80, +50, -120, +20),
_mm256_setr_epi8(0, +1, +2, +3, +4, +5, +6, +7, +8, +9, +10, +11, +12, +13,
+14, +15, +50, +80, -50, +110, +60, -30, +20, -10, +50, +80, -50, +110, +60,
-30, +20, -10)), 0, +2, +4, +6, +8, +10, +12, +14, +16, +18, +20, +22, +24,
+26, +28, +30, +127, +127, -128, +127, +127, -80, +127, -30, -50, +30, +50,
+90, -20, +20, -100, +10));
__m256i test_mm256_adds_epi16(__m256i a, __m256i b) {
// CHECK-LABEL: test_mm256_adds_epi16
// CHECK: call <16 x i16> @llvm.sadd.sat.v16i16(<16 x i16> %{{.*}}, <16 x
i16> %{{.*}})
return _mm256_adds_epi16(a, b);
}
+TEST_CONSTEXPR(match_v16hi(_mm256_adds_epi16(_mm256_setr_epi16(0, -1, -2, -3,
-4, -5, -6, -7, -8, -9, -10, -11, +32000, -32000, +32000, -32000),
_mm256_setr_epi16(0, -1, -2, -3, -4, -5, -6, -7, -8, -9, -10, -11, +800, -800,
-800, +800)),0, -2, -4, -6, -8, -10, -12, -14, -16, -18, -20, -22, +32767,
-32768, +31200, -31200));
__m256i test_mm256_adds_epu8(__m256i a, __m256i b) {
// CHECK-LABEL: test_mm256_adds_epu8
@@ -1293,12 +1295,14 @@ __m256i test_mm256_subs_epi8(__m256i a, __m256i b) {
// CHECK: call <32 x i8> @llvm.ssub.sat.v32i8(<32 x i8> %{{.*}}, <32 x i8>
%{{.*}})
return _mm256_subs_epi8(a, b);
}
+TEST_CONSTEXPR(match_v32qi(_mm256_subs_epi8(_mm256_setr_epi8(0, +1, +2, +3,
+4, +5, +6, +7, +8, +9, +10, +11, +12, +13, +14, +15, +100, +50, -100, +20,
+80, -50, +120, -20, -100, -50, +100, -20, -80, +50, -120, +20),
_mm256_setr_epi8(0, -1, -2, -3, -4, -5, -6, -7, -8, -9, -10, -11, -12, -13,
-14, -15, -50, -80, +50, -110, -60, +30, -20, +10, -50, -80, +50, -110, -60,
+30, -20, +10)), 0, +2, +4, +6, +8, +10, +12, +14, +16, +18, +20, +22, +24,
+26, +28, +30, +127, +127, -128, +127, +127, -80, +127, -30, -50, +30, +50,
+90, -20, +20, -100, +10));
__m256i test_mm256_subs_epi16(__m256i a, __m256i b) {
// CHECK-LABEL: test_mm256_subs_epi16
// CHECK: call <16 x i16> @llvm.ssub.sat.v16i16(<16 x i16> %{{.*}}, <16 x
i16> %{{.*}})
return _mm256_subs_epi16(a, b);
}
+TEST_CONSTEXPR(match_v16hi(_mm256_subs_epi16(_mm256_setr_epi16(0, -1, -2, -3,
-4, -5, -6, -7, -8, -9, -10, -11, +32000, -32000, +32000, -32000),
_mm256_setr_epi16(0, +1, +2, +3, +4, +5, +6, +7, +8, +9, +10, +11, -800, +800,
+800, -800)),0, -2, -4, -6, -8, -10, -12, -14, -16, -18, -20, -22, +32767,
-32768, +31200, -31200));
__m256i test_mm256_subs_epu8(__m256i a, __m256i b) {
// CHECK-LABEL: test_mm256_subs_epu8
>From c41317b2f64ae73348327eff10ed361bc1887e5e Mon Sep 17 00:00:00 2001
From: ningxinr <ningx...@live.cn>
Date: Mon, 11 Aug 2025 13:40:05 -0700
Subject: [PATCH 4/4] [Headers][X86] Allow AVX512BW integer saturated
arithmetic intrinsics to be used in constexpr
---
clang/lib/Headers/avx512bwintrin.h | 20 ++++++++------------
clang/test/CodeGen/X86/avx512bw-builtins.c | 7 +++++++
2 files changed, 15 insertions(+), 12 deletions(-)
diff --git a/clang/lib/Headers/avx512bwintrin.h
b/clang/lib/Headers/avx512bwintrin.h
index d7377e5d5196c..57a261f30eadb 100644
--- a/clang/lib/Headers/avx512bwintrin.h
+++ b/clang/lib/Headers/avx512bwintrin.h
@@ -613,9 +613,8 @@ _mm512_maskz_packus_epi16(__mmask64 __M, __m512i __A,
__m512i __B)
(__v64qi)_mm512_setzero_si512());
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_adds_epi8 (__m512i __A, __m512i __B)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_adds_epi8(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_add_sat((__v64qs)__A, (__v64qs)__B);
}
@@ -635,9 +634,8 @@ _mm512_maskz_adds_epi8 (__mmask64 __U, __m512i __A, __m512i
__B)
(__v64qi)_mm512_setzero_si512());
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_adds_epi16 (__m512i __A, __m512i __B)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_adds_epi16(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_add_sat((__v32hi)__A, (__v32hi)__B);
}
@@ -946,9 +944,8 @@ _mm512_maskz_shuffle_epi8(__mmask64 __U, __m512i __A,
__m512i __B)
(__v64qi)_mm512_setzero_si512());
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_subs_epi8 (__m512i __A, __m512i __B)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_subs_epi8(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_sub_sat((__v64qs)__A, (__v64qs)__B);
}
@@ -968,9 +965,8 @@ _mm512_maskz_subs_epi8 (__mmask64 __U, __m512i __A, __m512i
__B)
(__v64qi)_mm512_setzero_si512());
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
-_mm512_subs_epi16 (__m512i __A, __m512i __B)
-{
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_subs_epi16(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_sub_sat((__v32hi)__A, (__v32hi)__B);
}
diff --git a/clang/test/CodeGen/X86/avx512bw-builtins.c
b/clang/test/CodeGen/X86/avx512bw-builtins.c
index 2e3e13d2769c4..c2426dc0e6b9d 100644
--- a/clang/test/CodeGen/X86/avx512bw-builtins.c
+++ b/clang/test/CodeGen/X86/avx512bw-builtins.c
@@ -957,6 +957,8 @@ __m512i test_mm512_adds_epi8(__m512i __A, __m512i __B) {
// CHECK: @llvm.sadd.sat.v64i8
return _mm512_adds_epi8(__A,__B);
}
+TEST_CONSTEXPR(match_v64qi(_mm512_adds_epi8((__m512i)(__v64qi){0, +1, -2, +3,
-4, +5, -6, +7, -8, +9, -10, +11, -12, +13, -14, +15, -16, +17, -18, +19, -20,
+21, -22, +23, -24, +25, -26, +27, -28, +29, -30, +31, -32, +33, -34, +35, -36,
+37, -38, +39, -40, +41, -42, +43, -44, +45, -46, +47, +100, +50, -100, +20,
+80, -50, +120, -20, -100, -50, +100, -20, -80, +50, -120, +20},
(__m512i)(__v64qi){0, +1, -2, +3, -4, +5, -6, +7, -8, +9, -10, +11, -12, +13,
-14, +15, -16, +17, -18, +19, -20, +21, -22, +23, -24, +25, -26, +27, -28, +29,
-30, +31, -32, +33, -34, +35, -36, +37, -38, +39, -40, +41, -42, +43, -44, +45,
-46, +47, +50, +80, -50, +110, +60, -30, +20, -10, +50, +80, -50, +110, +60,
-30, +20, -10}), 0, +2, -4, +6, -8, +10, -12, +14, -16, +18, -20, +22, -24,
+26, -28, +30, -32, +34, -36, +38, -40, +42, -44, +46, -48, +50, -52, +54, -56,
+58, -60, +62, -64, +66, -68, +70, -72, +74, -76, +78, -80, +82, -84, +86, -88,
+90, -92, +94, +127, +127, -128, +127, +127, -80, +127, -30, -50, +30, +50,
+90, -20, +20, -100, +10));
+
__m512i test_mm512_mask_adds_epi8(__m512i __W, __mmask64 __U, __m512i __A,
__m512i __B) {
// CHECK-LABEL: @test_mm512_mask_adds_epi8
// CHECK: @llvm.sadd.sat.v64i8
@@ -974,6 +976,8 @@ __m512i test_mm512_adds_epi16(__m512i __A, __m512i __B) {
// CHECK: @llvm.sadd.sat.v32i16
return _mm512_adds_epi16(__A,__B);
}
+TEST_CONSTEXPR(match_v32hi(_mm512_adds_epi16((__m512i)(__v32hi){0, +1, -2, +3,
-4, +5, -6, +7, -8, +9, -10, +11, -12, +13, -14, +15, -16, +17, -18, +19, -20,
+21, -22, +23, -24, +25, -26, +27, +32000, -32000, +32000, -32000},
(__m512i)(__v32hi){0, +1, -2, +3, -4, +5, -6, +7, -8, +9, -10, +11, -12, +13,
-14, +15, -16, +17, -18, +19, -20, +21, -22, +23, -24, +25, -26, +27, +800,
-800, -800, +800}), 0, +2, -4, +6, -8, +10, -12, +14, -16, +18, -20, +22, -24,
+26, -28, +30, -32, +34, -36, +38, -40, +42, -44, +46, -48, +50, -52, +54,
+32767, -32768, +31200, -31200));
+
__m512i test_mm512_mask_adds_epi16(__m512i __W, __mmask32 __U, __m512i __A,
__m512i __B) {
// CHECK-LABEL: @test_mm512_mask_adds_epi16
// CHECK: @llvm.sadd.sat.v32i16
@@ -1218,6 +1222,8 @@ __m512i test_mm512_subs_epi8(__m512i __A, __m512i __B) {
// CHECK: @llvm.ssub.sat.v64i8
return _mm512_subs_epi8(__A,__B);
}
+TEST_CONSTEXPR(match_v64qi(_mm512_subs_epi8((__m512i)(__v64qi){0, +1, -2, +3,
-4, +5, -6, +7, -8, +9, -10, +11, -12, +13, -14, +15, -16, +17, -18, +19, -20,
+21, -22, +23, -24, +25, -26, +27, -28, +29, -30, +31, -32, +33, -34, +35, -36,
+37, -38, +39, -40, +41, -42, +43, -44, +45, -46, +47, +100, +50, -100, +20,
+80, -50, +120, -20, -100, -50, +100, -20, -80, +50, -120, +20},
(__m512i)(__v64qi){0, -1, +2, -3, +4, -5, +6, -7, +8, -9, +10, -11, +12, -13,
+14, -15, +16, -17, +18, -19, +20, -21, +22, -23, +24, -25, +26, -27, +28, -29,
+30, -31, +32, -33, +34, -35, +36, -37, +38, -39, +40, -41, +42, -43, +44, -45,
+46, -47, -50, -80, +50, -110, -60, +30, -20, +10, -50, -80, +50, -110, -60,
+30, -20, +10}), 0, +2, -4, +6, -8, +10, -12, +14, -16, +18, -20, +22, -24,
+26, -28, +30, -32, +34, -36, +38, -40, +42, -44, +46, -48, +50, -52, +54, -56,
+58, -60, +62, -64, +66, -68, +70, -72, +74, -76, +78, -80, +82, -84, +86, -88,
+90, -92, +94, +127, +127, -128, +127, +127, -80, +127, -30, -50, +30, +50,
+90, -20, +20, -100, +10));
+
__m512i test_mm512_mask_subs_epi8(__m512i __W, __mmask64 __U, __m512i __A,
__m512i __B) {
// CHECK-LABEL: @test_mm512_mask_subs_epi8
// CHECK: @llvm.ssub.sat.v64i8
@@ -1235,6 +1241,7 @@ __m512i test_mm512_subs_epi16(__m512i __A, __m512i __B) {
// CHECK: @llvm.ssub.sat.v32i16
return _mm512_subs_epi16(__A,__B);
}
+TEST_CONSTEXPR(match_v32hi(_mm512_subs_epi16((__m512i)(__v32hi){0, +1, -2, +3,
-4, +5, -6, +7, -8, +9, -10, +11, -12, +13, -14, +15, -16, +17, -18, +19, -20,
+21, -22, +23, -24, +25, -26, +27, +32000, -32000, +32000, -32000},
(__m512i)(__v32hi){0, -1, +2, -3, +4, -5, +6, -7, +8, -9, +10, -11, +12, -13,
+14, -15, +16, -17, +18, -19, +20, -21, +22, -23, +24, -25, +26, -27, -800,
+800, +800, -800}), 0, +2, -4, +6, -8, +10, -12, +14, -16, +18, -20, +22, -24,
+26, -28, +30, -32, +34, -36, +38, -40, +42, -44, +46, -48, +50, -52, +54,
+32767, -32768, +31200, -31200));
__m512i test_mm512_mask_subs_epi16(__m512i __W, __mmask32 __U, __m512i __A,
__m512i __B) {
// CHECK-LABEL: @test_mm512_mask_subs_epi16
// CHECK: @llvm.ssub.sat.v32i16
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
