[clang] [Headers][X86] Update SSE/AVX and/andnot/or/xor intrinsics to be used in constexpr (PR #152305)
via cfe-commits
cfe-commits at lists.llvm.org
Wed Aug 6 06:16:48 PDT 2025
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-clang
Author: Simon Pilgrim (RKSimon)
<details>
<summary>Changes</summary>
---
Patch is 32.02 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/152305.diff
10 Files Affected:
- (modified) clang/lib/Headers/avx2intrin.h (+14-4)
- (modified) clang/lib/Headers/avx512dqintrin.h (+18-8)
- (modified) clang/lib/Headers/avx512fintrin.h (+12-12)
- (modified) clang/lib/Headers/avxintrin.h (+8-8)
- (modified) clang/lib/Headers/emmintrin.h (+8-8)
- (modified) clang/test/CodeGen/X86/avx-builtins.c (+8)
- (modified) clang/test/CodeGen/X86/avx2-builtins.c (+5)
- (modified) clang/test/CodeGen/X86/avx512dq-builtins.c (+9)
- (modified) clang/test/CodeGen/X86/avx512f-builtins.c (+9)
- (modified) clang/test/CodeGen/X86/sse2-builtins.c (+4)
``````````diff
diff --git a/clang/lib/Headers/avx2intrin.h b/clang/lib/Headers/avx2intrin.h
index dc9fc07314323..3c3a3d1453934 100644
--- a/clang/lib/Headers/avx2intrin.h
+++ b/clang/lib/Headers/avx2intrin.h
@@ -31,6 +31,14 @@
__min_vector_width__(128)))
#endif
+#if defined(__cplusplus) && (__cplusplus >= 201103L)
+#define __DEFAULT_FN_ATTRS256_CONSTEXPR __DEFAULT_FN_ATTRS256 constexpr
+#define __DEFAULT_FN_ATTRS128_CONSTEXPR __DEFAULT_FN_ATTRS128 constexpr
+#else
+#define __DEFAULT_FN_ATTRS256_CONSTEXPR __DEFAULT_FN_ATTRS256
+#define __DEFAULT_FN_ATTRS128_CONSTEXPR __DEFAULT_FN_ATTRS128
+#endif
+
/* SSE4 Multiple Packed Sums of Absolute Difference. */
/// Computes sixteen sum of absolute difference (SAD) operations on sets of
/// four unsigned 8-bit integers from the 256-bit integer vectors \a X and
@@ -460,7 +468,7 @@ _mm256_adds_epu16(__m256i __a, __m256i __b)
/// \param __b
/// A 256-bit integer vector.
/// \returns A 256-bit integer vector containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_and_si256(__m256i __a, __m256i __b)
{
return (__m256i)((__v4du)__a & (__v4du)__b);
@@ -478,7 +486,7 @@ _mm256_and_si256(__m256i __a, __m256i __b)
/// \param __b
/// A 256-bit integer vector.
/// \returns A 256-bit integer vector containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_andnot_si256(__m256i __a, __m256i __b)
{
return (__m256i)(~(__v4du)__a & (__v4du)__b);
@@ -1822,7 +1830,7 @@ _mm256_mul_epu32(__m256i __a, __m256i __b)
/// \param __b
/// A 256-bit integer vector.
/// \returns A 256-bit integer vector containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_or_si256(__m256i __a, __m256i __b)
{
return (__m256i)((__v4du)__a | (__v4du)__b);
@@ -2974,7 +2982,7 @@ _mm256_unpacklo_epi64(__m256i __a, __m256i __b)
/// \param __b
/// A 256-bit integer vector.
/// \returns A 256-bit integer vector containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_xor_si256(__m256i __a, __m256i __b)
{
return (__m256i)((__v4du)__a ^ (__v4du)__b);
@@ -5289,5 +5297,7 @@ _mm_srlv_epi64(__m128i __X, __m128i __Y)
#undef __DEFAULT_FN_ATTRS256
#undef __DEFAULT_FN_ATTRS128
+#undef __DEFAULT_FN_ATTRS256_CONSTEXPR
+#undef __DEFAULT_FN_ATTRS128_CONSTEXPR
#endif /* __AVX2INTRIN_H */
diff --git a/clang/lib/Headers/avx512dqintrin.h b/clang/lib/Headers/avx512dqintrin.h
index 88b48e3a32070..62325b96f202b 100644
--- a/clang/lib/Headers/avx512dqintrin.h
+++ b/clang/lib/Headers/avx512dqintrin.h
@@ -20,6 +20,14 @@
__attribute__((__always_inline__, __nodebug__, \
__target__("avx512dq,no-evex512")))
+#if defined(__cplusplus) && (__cplusplus >= 201103L)
+#define __DEFAULT_FN_ATTRS512_CONSTEXPR __DEFAULT_FN_ATTRS512 constexpr
+#define __DEFAULT_FN_ATTRS_CONSTEXPR __DEFAULT_FN_ATTRS constexpr
+#else
+#define __DEFAULT_FN_ATTRS512_CONSTEXPR __DEFAULT_FN_ATTRS512
+#define __DEFAULT_FN_ATTRS_CONSTEXPR __DEFAULT_FN_ATTRS
+#endif
+
static __inline __mmask8 __DEFAULT_FN_ATTRS
_knot_mask8(__mmask8 __M)
{
@@ -167,7 +175,7 @@ _mm512_maskz_mullo_epi64(__mmask8 __U, __m512i __A, __m512i __B) {
(__v8di)_mm512_setzero_si512());
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_xor_pd(__m512d __A, __m512d __B) {
return (__m512d)((__v8du)__A ^ (__v8du)__B);
}
@@ -186,7 +194,7 @@ _mm512_maskz_xor_pd(__mmask8 __U, __m512d __A, __m512d __B) {
(__v8df)_mm512_setzero_pd());
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_xor_ps (__m512 __A, __m512 __B) {
return (__m512)((__v16su)__A ^ (__v16su)__B);
}
@@ -205,7 +213,7 @@ _mm512_maskz_xor_ps(__mmask16 __U, __m512 __A, __m512 __B) {
(__v16sf)_mm512_setzero_ps());
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_or_pd(__m512d __A, __m512d __B) {
return (__m512d)((__v8du)__A | (__v8du)__B);
}
@@ -224,7 +232,7 @@ _mm512_maskz_or_pd(__mmask8 __U, __m512d __A, __m512d __B) {
(__v8df)_mm512_setzero_pd());
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_or_ps(__m512 __A, __m512 __B) {
return (__m512)((__v16su)__A | (__v16su)__B);
}
@@ -243,7 +251,7 @@ _mm512_maskz_or_ps(__mmask16 __U, __m512 __A, __m512 __B) {
(__v16sf)_mm512_setzero_ps());
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_and_pd(__m512d __A, __m512d __B) {
return (__m512d)((__v8du)__A & (__v8du)__B);
}
@@ -262,7 +270,7 @@ _mm512_maskz_and_pd(__mmask8 __U, __m512d __A, __m512d __B) {
(__v8df)_mm512_setzero_pd());
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_and_ps(__m512 __A, __m512 __B) {
return (__m512)((__v16su)__A & (__v16su)__B);
}
@@ -281,7 +289,7 @@ _mm512_maskz_and_ps(__mmask16 __U, __m512 __A, __m512 __B) {
(__v16sf)_mm512_setzero_ps());
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_andnot_pd(__m512d __A, __m512d __B) {
return (__m512d)(~(__v8du)__A & (__v8du)__B);
}
@@ -300,7 +308,7 @@ _mm512_maskz_andnot_pd(__mmask8 __U, __m512d __A, __m512d __B) {
(__v8df)_mm512_setzero_pd());
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_andnot_ps(__m512 __A, __m512 __B) {
return (__m512)(~(__v16su)__A & (__v16su)__B);
}
@@ -1375,5 +1383,7 @@ _mm512_maskz_broadcast_i64x2(__mmask8 __M, __m128i __A)
#undef __DEFAULT_FN_ATTRS512
#undef __DEFAULT_FN_ATTRS
+#undef __DEFAULT_FN_ATTRS512_CONSTEXPR
+#undef __DEFAULT_FN_ATTRS_CONSTEXPR
#endif
diff --git a/clang/lib/Headers/avx512fintrin.h b/clang/lib/Headers/avx512fintrin.h
index 45e7eeb5327d0..21809a90d4d26 100644
--- a/clang/lib/Headers/avx512fintrin.h
+++ b/clang/lib/Headers/avx512fintrin.h
@@ -645,7 +645,7 @@ _mm512_zextsi256_si512(__m256i __a)
}
/* Bitwise operators */
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_and_epi32(__m512i __a, __m512i __b)
{
return (__m512i)((__v16su)__a & (__v16su)__b);
@@ -666,7 +666,7 @@ _mm512_maskz_and_epi32(__mmask16 __k, __m512i __a, __m512i __b)
__k, __a, __b);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_and_epi64(__m512i __a, __m512i __b)
{
return (__m512i)((__v8du)__a & (__v8du)__b);
@@ -687,13 +687,13 @@ _mm512_maskz_and_epi64(__mmask8 __k, __m512i __a, __m512i __b)
__k, __a, __b);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_andnot_si512 (__m512i __A, __m512i __B)
{
return (__m512i)(~(__v8du)__A & (__v8du)__B);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_andnot_epi32 (__m512i __A, __m512i __B)
{
return (__m512i)(~(__v16su)__A & (__v16su)__B);
@@ -714,7 +714,7 @@ _mm512_maskz_andnot_epi32(__mmask16 __U, __m512i __A, __m512i __B)
__U, __A, __B);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_andnot_epi64(__m512i __A, __m512i __B)
{
return (__m512i)(~(__v8du)__A & (__v8du)__B);
@@ -735,7 +735,7 @@ _mm512_maskz_andnot_epi64(__mmask8 __U, __m512i __A, __m512i __B)
__U, __A, __B);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_or_epi32(__m512i __a, __m512i __b)
{
return (__m512i)((__v16su)__a | (__v16su)__b);
@@ -755,7 +755,7 @@ _mm512_maskz_or_epi32(__mmask16 __k, __m512i __a, __m512i __b)
return (__m512i)_mm512_mask_or_epi32(_mm512_setzero_si512(), __k, __a, __b);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_or_epi64(__m512i __a, __m512i __b)
{
return (__m512i)((__v8du)__a | (__v8du)__b);
@@ -775,7 +775,7 @@ _mm512_maskz_or_epi64(__mmask8 __k, __m512i __a, __m512i __b)
return (__m512i)_mm512_mask_or_epi64(_mm512_setzero_si512(), __k, __a, __b);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_xor_epi32(__m512i __a, __m512i __b)
{
return (__m512i)((__v16su)__a ^ (__v16su)__b);
@@ -795,7 +795,7 @@ _mm512_maskz_xor_epi32(__mmask16 __k, __m512i __a, __m512i __b)
return (__m512i)_mm512_mask_xor_epi32(_mm512_setzero_si512(), __k, __a, __b);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_xor_epi64(__m512i __a, __m512i __b)
{
return (__m512i)((__v8du)__a ^ (__v8du)__b);
@@ -815,19 +815,19 @@ _mm512_maskz_xor_epi64(__mmask8 __k, __m512i __a, __m512i __b)
return (__m512i)_mm512_mask_xor_epi64(_mm512_setzero_si512(), __k, __a, __b);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_and_si512(__m512i __a, __m512i __b)
{
return (__m512i)((__v8du)__a & (__v8du)__b);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_or_si512(__m512i __a, __m512i __b)
{
return (__m512i)((__v8du)__a | (__v8du)__b);
}
-static __inline__ __m512i __DEFAULT_FN_ATTRS512
+static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_xor_si512(__m512i __a, __m512i __b)
{
return (__m512i)((__v8du)__a ^ (__v8du)__b);
diff --git a/clang/lib/Headers/avxintrin.h b/clang/lib/Headers/avxintrin.h
index b9ca013c25c7a..5aa09300845bf 100644
--- a/clang/lib/Headers/avxintrin.h
+++ b/clang/lib/Headers/avxintrin.h
@@ -555,7 +555,7 @@ _mm256_rcp_ps(__m256 __a)
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// \returns A 256-bit vector of [4 x double] containing the bitwise AND of the
/// values between both operands.
-static __inline __m256d __DEFAULT_FN_ATTRS
+static __inline __m256d __DEFAULT_FN_ATTRS_CONSTEXPR
_mm256_and_pd(__m256d __a, __m256d __b)
{
return (__m256d)((__v4du)__a & (__v4du)__b);
@@ -573,7 +573,7 @@ _mm256_and_pd(__m256d __a, __m256d __b)
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// \returns A 256-bit vector of [8 x float] containing the bitwise AND of the
/// values between both operands.
-static __inline __m256 __DEFAULT_FN_ATTRS
+static __inline __m256 __DEFAULT_FN_ATTRS_CONSTEXPR
_mm256_and_ps(__m256 __a, __m256 __b)
{
return (__m256)((__v8su)__a & (__v8su)__b);
@@ -594,7 +594,7 @@ _mm256_and_ps(__m256 __a, __m256 __b)
/// \returns A 256-bit vector of [4 x double] containing the bitwise AND of the
/// values of the second operand and the one's complement of the first
/// operand.
-static __inline __m256d __DEFAULT_FN_ATTRS
+static __inline __m256d __DEFAULT_FN_ATTRS_CONSTEXPR
_mm256_andnot_pd(__m256d __a, __m256d __b)
{
return (__m256d)(~(__v4du)__a & (__v4du)__b);
@@ -615,7 +615,7 @@ _mm256_andnot_pd(__m256d __a, __m256d __b)
/// \returns A 256-bit vector of [8 x float] containing the bitwise AND of the
/// values of the second operand and the one's complement of the first
/// operand.
-static __inline __m256 __DEFAULT_FN_ATTRS
+static __inline __m256 __DEFAULT_FN_ATTRS_CONSTEXPR
_mm256_andnot_ps(__m256 __a, __m256 __b)
{
return (__m256)(~(__v8su)__a & (__v8su)__b);
@@ -633,7 +633,7 @@ _mm256_andnot_ps(__m256 __a, __m256 __b)
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// \returns A 256-bit vector of [4 x double] containing the bitwise OR of the
/// values between both operands.
-static __inline __m256d __DEFAULT_FN_ATTRS
+static __inline __m256d __DEFAULT_FN_ATTRS_CONSTEXPR
_mm256_or_pd(__m256d __a, __m256d __b)
{
return (__m256d)((__v4du)__a | (__v4du)__b);
@@ -651,7 +651,7 @@ _mm256_or_pd(__m256d __a, __m256d __b)
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// \returns A 256-bit vector of [8 x float] containing the bitwise OR of the
/// values between both operands.
-static __inline __m256 __DEFAULT_FN_ATTRS
+static __inline __m256 __DEFAULT_FN_ATTRS_CONSTEXPR
_mm256_or_ps(__m256 __a, __m256 __b)
{
return (__m256)((__v8su)__a | (__v8su)__b);
@@ -669,7 +669,7 @@ _mm256_or_ps(__m256 __a, __m256 __b)
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// \returns A 256-bit vector of [4 x double] containing the bitwise XOR of the
/// values between both operands.
-static __inline __m256d __DEFAULT_FN_ATTRS
+static __inline __m256d __DEFAULT_FN_ATTRS_CONSTEXPR
_mm256_xor_pd(__m256d __a, __m256d __b)
{
return (__m256d)((__v4du)__a ^ (__v4du)__b);
@@ -687,7 +687,7 @@ _mm256_xor_pd(__m256d __a, __m256d __b)
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// \returns A 256-bit vector of [8 x float] containing the bitwise XOR of the
/// values between both operands.
-static __inline __m256 __DEFAULT_FN_ATTRS
+static __inline __m256 __DEFAULT_FN_ATTRS_CONSTEXPR
_mm256_xor_ps(__m256 __a, __m256 __b)
{
return (__m256)((__v8su)__a ^ (__v8su)__b);
diff --git a/clang/lib/Headers/emmintrin.h b/clang/lib/Headers/emmintrin.h
index 78e8a422db4c1..9f8f60abdbb35 100644
--- a/clang/lib/Headers/emmintrin.h
+++ b/clang/lib/Headers/emmintrin.h
@@ -2676,8 +2676,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_subs_epu16(__m128i __a,
/// A 128-bit integer vector containing one of the source operands.
/// \returns A 128-bit integer vector containing the bitwise AND of the values
/// in both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_and_si128(__m128i __a,
- __m128i __b) {
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_and_si128(__m128i __a, __m128i __b) {
return (__m128i)((__v2du)__a & (__v2du)__b);
}
@@ -2695,8 +2695,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_and_si128(__m128i __a,
/// A 128-bit vector containing the right source operand.
/// \returns A 128-bit integer vector containing the bitwise AND of the one's
/// complement of the first operand and the values in the second operand.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_andnot_si128(__m128i __a,
- __m128i __b) {
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_andnot_si128(__m128i __a, __m128i __b) {
return (__m128i)(~(__v2du)__a & (__v2du)__b);
}
/// Performs a bitwise OR of two 128-bit integer vectors.
@@ -2711,8 +2711,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_andnot_si128(__m128i __a,
/// A 128-bit integer vector containing one of the source operands.
/// \returns A 128-bit integer vector containing the bitwise OR of the values
/// in both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_or_si128(__m128i __a,
- __m128i __b) {
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_or_si128(__m128i __a, __m128i __b) {
return (__m128i)((__v2du)__a | (__v2du)__b);
}
@@ -2728,8 +2728,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_or_si128(__m128i __a,
/// A 128-bit integer vector containing one of the source operands.
/// \returns A 128-bit integer vector containing the bitwise exclusive OR of the
/// values in both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_xor_si128(__m128i __a,
- __m128i __b) {
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_xor_si128(__m128i __a, __m128i __b) {
return (__m128i)((__v2du)__a ^ (__v2du)__b);
}
diff --git a/clang/test/CodeGen/X86/avx-builtins.c b/clang/test/CodeGen/X86/avx-builtins.c
index a6e70aae420ea..fc389b3a7790c 100644
--- a/clang/test/CodeGen/X86/avx-builtins.c
+++ b/clang/test/CodeGen/X86/avx-builtins.c
@@ -44,12 +44,14 @@ __m256d test_mm256_and_pd(__m256d A, __m256d B) {
// CHECK: and <4 x i64>
return _mm256_and_pd(A, B);
}
+TEST_CONSTEXPR(match_m256d(_mm256_and_pd((__m256d){-4.0, -5.0, +6.0, +7.0}, (__m256d){+0.0, -0.0, -0.0, +7.0}), -0.0, -0.0, +0.0, +7.0));
__m256 test_mm256_and_ps(__m256 A, __m256 B) {
// CHECK-LABEL: test_mm256_and_ps
// CHECK: and <8 x i32>
return _mm256_and_ps(A, B);
}
+TEST_CONSTEXPR(match_m256(_mm256_and_ps((__m256){-4.0f, -5.0f, +6.0f, +7.0f, +7.0f, +6.0f, -5.0f, -4.0f}, (__m256){+0.0f, -0.0f, -0.0f, +7.0f, +7.0f, -0.0f, -0.0f, +0.0f}), -0.0f, -0.0f, +0.0f, +7.0f, +7.0f, +0.0f, -0.0f, -0.0f));
__m256d test_mm256_andnot_pd(__m256d A, __m256d B) {
// CHECK-LABEL: test_mm256_andnot_pd
@@ -57,6 +59,7 @@ __m256d test_mm256_andnot_pd(__m256d A, __m256d B) {
// CHECK: and <4 x i64>
return _mm256_andnot_pd(A, B);
}
+TEST_CONSTEXPR(match_m256d(_mm256_andnot_pd((__m256d){-4.0, -5.0, +6.0, +7.0}, (__m256d){+0.0, -0.0, -0.0, +7.0}), +0.0, +0.0, +0.0, +0.0));
__m256 test_mm256_andnot_ps(__m256 A, __m256 B) {
// CHECK-LABEL: test_mm256_andnot_ps
@@ -64,6 +67,7 @@ __m256 test_mm256_andnot_ps(__m256 A, __m256 B) {
// CHECK: and <8 x i32>
return _mm256_andnot_ps(A, B);
}
+TEST_CONSTEXPR(match_m256(_mm256_andnot_ps((__m256){-4.0f, -5.0f, +6.0f, +7.0f, +7.0f, +6.0f, -5.0f, -4.0f}, (__m256){+0.0f, -0.0f, -0.0f, +7.0f, +7.0f, -0.0f, -0.0f, +0.0f}), +0.0f, +0.0f, +0.0f, +0.0f, +0.0f, +0.0f, +0.0f, +0.0f));
__m256d test_mm256_blend_pd(__m256d A, __m256d B) {
// CHECK-LABEL: test_mm256_blend_pd
@@ -1300,12 +1304,14 @@ __m256d test_mm256_or_pd(__m256d A, __m256d B) {
// CHECK: or <4 x i64>
return _mm256_or_pd(A, B);
}
+TEST_CONSTEXPR(match_m256d(_mm256_or_pd((__m256d){-4.0, -5.0, +6.0, +7.0}, (__m256d){+0.0, -0.0, -0.0, +7.0}), -4.0, -5.0, -6.0, +7.0));
__m256 test_mm256_or_ps(__m256 A, __m256 B) {
// CHECK-LABEL: test_mm256_or_ps
// CHECK: or <8 x i32>
return _mm256_or_ps(A, B);
}
+TEST_CONSTEXPR(match_m256(_mm256_or_ps((__m256){-4.0f, -5.0f, +6.0f, +7.0f, +7.0f, +6.0f, -5.0f, -4.0f}, (__m256){+0.0f, -0.0f, -0.0f, +7.0f, +7.0f, -0.0f, -0.0f, +0.0f}), -4.0f, -5.0f, -6.0f, +7.0f, +7.0f, -6.0f, -5.0f, -4.0f));
__m128d test_mm_permute_pd(__m128d A) {
// CHECK-LABEL: test_mm_permute_pd
@@ -2077,12 +2083,14 @@ __m256d test_mm256_xor_pd(__m256d A, __m256d B) {
// CHECK: xor <4 x i64>
return _mm256_xor_pd(A, B);
}
+TEST_CONSTEXPR(match_m256d(_mm256_xor_pd((__m256d){-4.0, -5.0, +6.0, +7.0}, (__m256d){+0.0, -0.0, -0.0, +7.0}), -4.0, +5.0, -6.0, +0.0));
__m256 test_mm256_xor_ps(__m256 A, __m256 B) {
// CHECK-LABEL: test_mm256_xor_ps
// CHECK: xor <8 x i32>
return _mm256_xor_ps(A, B);
}
+TEST_CONSTEXPR(match_m256(_mm256_xor_ps((__m256){-4.0f, -5.0f, +6.0f, +7.0f, +7.0f, +6.0f, -5.0f, -4.0f}, (__m256){+0.0f, -0.0f, -0.0f, +7.0f, +7.0f, -0.0f, -0.0f, +0.0f}), -4.0f, +5.0f, -6.0f, +0.0f, +0.0f, -6.0f, +5.0f, -4.0f));
void test_mm256_zeroall(void) {
// CHECK-LABEL: test_mm256_zeroall
diff --git a/clang/test/CodeGen/X86/avx2-builtins.c b/clang/test/CodeGen/X86/avx2-builtins.c
index 27da56fb75723..1ed624cf63ab2 100644
--- a/clang/test/CodeGen/X86/avx2-builtins.c
+++ b/clang/test/CodeGen/X86/avx2-builtins.c
@@ -9,6 +9,7 @@
#include <immintrin.h>
+#include "builtin_test_helpers.h"
// NOTE: This should match the tests in llvm/test/CodeGen/X86/avx2-intrinsics-fast-isel.ll
@...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/152305
More information about the cfe-commits
mailing list