llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT--> @llvm/pr-subscribers-clang Author: woruyu (woruyu) <details> <summary>Changes</summary> ### Summary This PR resolves https://github.com/llvm/llvm-project/issues/152315 --- Patch is 36.04 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/152971.diff 8 Files Affected: - (modified) clang/lib/Headers/avx2intrin.h (+24-36) - (modified) clang/lib/Headers/avx512bwintrin.h (+4-6) - (modified) clang/lib/Headers/avx512fintrin.h (+20-30) - (modified) clang/lib/Headers/smmintrin.h (+30-12) - (modified) clang/test/CodeGen/X86/avx2-builtins.c (+24) - (modified) clang/test/CodeGen/X86/avx512bw-builtins.c (+4) - (modified) clang/test/CodeGen/X86/avx512f-builtins.c (+20) - (modified) clang/test/CodeGen/X86/sse41-builtins.c (+25) ``````````diff diff --git a/clang/lib/Headers/avx2intrin.h b/clang/lib/Headers/avx2intrin.h index 55e7102a7d192..f00a8a516ecfe 100644 --- a/clang/lib/Headers/avx2intrin.h +++ b/clang/lib/Headers/avx2intrin.h @@ -1371,9 +1371,8 @@ _mm256_movemask_epi8(__m256i __a) /// A 128-bit integer vector containing the source bytes. /// \returns A 256-bit vector of [16 x i16] containing the sign-extended /// values. -static __inline__ __m256i __DEFAULT_FN_ATTRS256 -_mm256_cvtepi8_epi16(__m128i __V) -{ +static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR +_mm256_cvtepi8_epi16(__m128i __V) { /* This function always performs a signed extension, but __v16qi is a char which may be signed or unsigned, so use __v16qs. */ return (__m256i)__builtin_convertvector((__v16qs)__V, __v16hi); @@ -1399,9 +1398,8 @@ _mm256_cvtepi8_epi16(__m128i __V) /// A 128-bit integer vector containing the source bytes. /// \returns A 256-bit vector of [8 x i32] containing the sign-extended /// values. -static __inline__ __m256i __DEFAULT_FN_ATTRS256 -_mm256_cvtepi8_epi32(__m128i __V) -{ +static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR +_mm256_cvtepi8_epi32(__m128i __V) { /* This function always performs a signed extension, but __v16qi is a char which may be signed or unsigned, so use __v16qs. */ return (__m256i)__builtin_convertvector(__builtin_shufflevector((__v16qs)__V, (__v16qs)__V, 0, 1, 2, 3, 4, 5, 6, 7), __v8si); @@ -1426,9 +1424,8 @@ _mm256_cvtepi8_epi32(__m128i __V) /// A 128-bit integer vector containing the source bytes. /// \returns A 256-bit vector of [4 x i64] containing the sign-extended /// values. -static __inline__ __m256i __DEFAULT_FN_ATTRS256 -_mm256_cvtepi8_epi64(__m128i __V) -{ +static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR +_mm256_cvtepi8_epi64(__m128i __V) { /* This function always performs a signed extension, but __v16qi is a char which may be signed or unsigned, so use __v16qs. */ return (__m256i)__builtin_convertvector(__builtin_shufflevector((__v16qs)__V, (__v16qs)__V, 0, 1, 2, 3), __v4di); @@ -1454,9 +1451,8 @@ _mm256_cvtepi8_epi64(__m128i __V) /// A 128-bit vector of [8 x i16] containing the source values. /// \returns A 256-bit vector of [8 x i32] containing the sign-extended /// values. -static __inline__ __m256i __DEFAULT_FN_ATTRS256 -_mm256_cvtepi16_epi32(__m128i __V) -{ +static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR +_mm256_cvtepi16_epi32(__m128i __V) { return (__m256i)__builtin_convertvector((__v8hi)__V, __v8si); } @@ -1479,9 +1475,8 @@ _mm256_cvtepi16_epi32(__m128i __V) /// A 128-bit vector of [8 x i16] containing the source values. /// \returns A 256-bit vector of [4 x i64] containing the sign-extended /// values. -static __inline__ __m256i __DEFAULT_FN_ATTRS256 -_mm256_cvtepi16_epi64(__m128i __V) -{ +static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR +_mm256_cvtepi16_epi64(__m128i __V) { return (__m256i)__builtin_convertvector(__builtin_shufflevector((__v8hi)__V, (__v8hi)__V, 0, 1, 2, 3), __v4di); } @@ -1504,9 +1499,8 @@ _mm256_cvtepi16_epi64(__m128i __V) /// A 128-bit vector of [4 x i32] containing the source values. /// \returns A 256-bit vector of [4 x i64] containing the sign-extended /// values. -static __inline__ __m256i __DEFAULT_FN_ATTRS256 -_mm256_cvtepi32_epi64(__m128i __V) -{ +static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR +_mm256_cvtepi32_epi64(__m128i __V) { return (__m256i)__builtin_convertvector((__v4si)__V, __v4di); } @@ -1530,9 +1524,8 @@ _mm256_cvtepi32_epi64(__m128i __V) /// A 128-bit integer vector containing the source bytes. /// \returns A 256-bit vector of [16 x i16] containing the zero-extended /// values. -static __inline__ __m256i __DEFAULT_FN_ATTRS256 -_mm256_cvtepu8_epi16(__m128i __V) -{ +static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR +_mm256_cvtepu8_epi16(__m128i __V) { return (__m256i)__builtin_convertvector((__v16qu)__V, __v16hi); } @@ -1556,9 +1549,8 @@ _mm256_cvtepu8_epi16(__m128i __V) /// A 128-bit integer vector containing the source bytes. /// \returns A 256-bit vector of [8 x i32] containing the zero-extended /// values. -static __inline__ __m256i __DEFAULT_FN_ATTRS256 -_mm256_cvtepu8_epi32(__m128i __V) -{ +static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR +_mm256_cvtepu8_epi32(__m128i __V) { return (__m256i)__builtin_convertvector(__builtin_shufflevector((__v16qu)__V, (__v16qu)__V, 0, 1, 2, 3, 4, 5, 6, 7), __v8si); } @@ -1581,9 +1573,8 @@ _mm256_cvtepu8_epi32(__m128i __V) /// A 128-bit integer vector containing the source bytes. /// \returns A 256-bit vector of [4 x i64] containing the zero-extended /// values. -static __inline__ __m256i __DEFAULT_FN_ATTRS256 -_mm256_cvtepu8_epi64(__m128i __V) -{ +static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR +_mm256_cvtepu8_epi64(__m128i __V) { return (__m256i)__builtin_convertvector(__builtin_shufflevector((__v16qu)__V, (__v16qu)__V, 0, 1, 2, 3), __v4di); } @@ -1607,9 +1598,8 @@ _mm256_cvtepu8_epi64(__m128i __V) /// A 128-bit vector of [8 x i16] containing the source values. /// \returns A 256-bit vector of [8 x i32] containing the zero-extended /// values. -static __inline__ __m256i __DEFAULT_FN_ATTRS256 -_mm256_cvtepu16_epi32(__m128i __V) -{ +static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR +_mm256_cvtepu16_epi32(__m128i __V) { return (__m256i)__builtin_convertvector((__v8hu)__V, __v8si); } @@ -1632,9 +1622,8 @@ _mm256_cvtepu16_epi32(__m128i __V) /// A 128-bit vector of [8 x i16] containing the source values. /// \returns A 256-bit vector of [4 x i64] containing the zero-extended /// values. -static __inline__ __m256i __DEFAULT_FN_ATTRS256 -_mm256_cvtepu16_epi64(__m128i __V) -{ +static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR +_mm256_cvtepu16_epi64(__m128i __V) { return (__m256i)__builtin_convertvector(__builtin_shufflevector((__v8hu)__V, (__v8hu)__V, 0, 1, 2, 3), __v4di); } @@ -1657,9 +1646,8 @@ _mm256_cvtepu16_epi64(__m128i __V) /// A 128-bit vector of [4 x i32] containing the source values. /// \returns A 256-bit vector of [4 x i64] containing the zero-extended /// values. -static __inline__ __m256i __DEFAULT_FN_ATTRS256 -_mm256_cvtepu32_epi64(__m128i __V) -{ +static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR +_mm256_cvtepu32_epi64(__m128i __V) { return (__m256i)__builtin_convertvector((__v4su)__V, __v4di); } diff --git a/clang/lib/Headers/avx512bwintrin.h b/clang/lib/Headers/avx512bwintrin.h index 233d4a6574f30..d7377e5d5196c 100644 --- a/clang/lib/Headers/avx512bwintrin.h +++ b/clang/lib/Headers/avx512bwintrin.h @@ -1379,9 +1379,8 @@ _mm512_maskz_unpacklo_epi16(__mmask32 __U, __m512i __A, __m512i __B) { (__v32hi)_mm512_setzero_si512()); } -static __inline__ __m512i __DEFAULT_FN_ATTRS512 -_mm512_cvtepi8_epi16(__m256i __A) -{ +static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR +_mm512_cvtepi8_epi16(__m256i __A) { /* This function always performs a signed extension, but __v32qi is a char which may be signed or unsigned, so use __v32qs. */ return (__m512i)__builtin_convertvector((__v32qs)__A, __v32hi); @@ -1403,9 +1402,8 @@ _mm512_maskz_cvtepi8_epi16(__mmask32 __U, __m256i __A) (__v32hi)_mm512_setzero_si512()); } -static __inline__ __m512i __DEFAULT_FN_ATTRS512 -_mm512_cvtepu8_epi16(__m256i __A) -{ +static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR +_mm512_cvtepu8_epi16(__m256i __A) { return (__m512i)__builtin_convertvector((__v32qu)__A, __v32hi); } diff --git a/clang/lib/Headers/avx512fintrin.h b/clang/lib/Headers/avx512fintrin.h index 95b80cc59bb02..940064675ba45 100644 --- a/clang/lib/Headers/avx512fintrin.h +++ b/clang/lib/Headers/avx512fintrin.h @@ -4714,9 +4714,8 @@ _mm512_knot(__mmask16 __M) #define _mm512_mask_cmpneq_epu64_mask(k, A, B) \ _mm512_mask_cmp_epu64_mask((k), (A), (B), _MM_CMPINT_NE) -static __inline__ __m512i __DEFAULT_FN_ATTRS512 -_mm512_cvtepi8_epi32(__m128i __A) -{ +static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR +_mm512_cvtepi8_epi32(__m128i __A) { /* This function always performs a signed extension, but __v16qi is a char which may be signed or unsigned, so use __v16qs. */ return (__m512i)__builtin_convertvector((__v16qs)__A, __v16si); @@ -4738,9 +4737,8 @@ _mm512_maskz_cvtepi8_epi32(__mmask16 __U, __m128i __A) (__v16si)_mm512_setzero_si512()); } -static __inline__ __m512i __DEFAULT_FN_ATTRS512 -_mm512_cvtepi8_epi64(__m128i __A) -{ +static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR +_mm512_cvtepi8_epi64(__m128i __A) { /* This function always performs a signed extension, but __v16qi is a char which may be signed or unsigned, so use __v16qs. */ return (__m512i)__builtin_convertvector(__builtin_shufflevector((__v16qs)__A, (__v16qs)__A, 0, 1, 2, 3, 4, 5, 6, 7), __v8di); @@ -4762,9 +4760,8 @@ _mm512_maskz_cvtepi8_epi64(__mmask8 __U, __m128i __A) (__v8di)_mm512_setzero_si512 ()); } -static __inline__ __m512i __DEFAULT_FN_ATTRS512 -_mm512_cvtepi32_epi64(__m256i __X) -{ +static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR +_mm512_cvtepi32_epi64(__m256i __X) { return (__m512i)__builtin_convertvector((__v8si)__X, __v8di); } @@ -4784,9 +4781,8 @@ _mm512_maskz_cvtepi32_epi64(__mmask8 __U, __m256i __X) (__v8di)_mm512_setzero_si512()); } -static __inline__ __m512i __DEFAULT_FN_ATTRS512 -_mm512_cvtepi16_epi32(__m256i __A) -{ +static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR +_mm512_cvtepi16_epi32(__m256i __A) { return (__m512i)__builtin_convertvector((__v16hi)__A, __v16si); } @@ -4806,9 +4802,8 @@ _mm512_maskz_cvtepi16_epi32(__mmask16 __U, __m256i __A) (__v16si)_mm512_setzero_si512 ()); } -static __inline__ __m512i __DEFAULT_FN_ATTRS512 -_mm512_cvtepi16_epi64(__m128i __A) -{ +static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR +_mm512_cvtepi16_epi64(__m128i __A) { return (__m512i)__builtin_convertvector((__v8hi)__A, __v8di); } @@ -4828,9 +4823,8 @@ _mm512_maskz_cvtepi16_epi64(__mmask8 __U, __m128i __A) (__v8di)_mm512_setzero_si512()); } -static __inline__ __m512i __DEFAULT_FN_ATTRS512 -_mm512_cvtepu8_epi32(__m128i __A) -{ +static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR +_mm512_cvtepu8_epi32(__m128i __A) { return (__m512i)__builtin_convertvector((__v16qu)__A, __v16si); } @@ -4850,9 +4844,8 @@ _mm512_maskz_cvtepu8_epi32(__mmask16 __U, __m128i __A) (__v16si)_mm512_setzero_si512()); } -static __inline__ __m512i __DEFAULT_FN_ATTRS512 -_mm512_cvtepu8_epi64(__m128i __A) -{ +static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR +_mm512_cvtepu8_epi64(__m128i __A) { return (__m512i)__builtin_convertvector(__builtin_shufflevector((__v16qu)__A, (__v16qu)__A, 0, 1, 2, 3, 4, 5, 6, 7), __v8di); } @@ -4872,9 +4865,8 @@ _mm512_maskz_cvtepu8_epi64(__mmask8 __U, __m128i __A) (__v8di)_mm512_setzero_si512()); } -static __inline__ __m512i __DEFAULT_FN_ATTRS512 -_mm512_cvtepu32_epi64(__m256i __X) -{ +static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR +_mm512_cvtepu32_epi64(__m256i __X) { return (__m512i)__builtin_convertvector((__v8su)__X, __v8di); } @@ -4894,9 +4886,8 @@ _mm512_maskz_cvtepu32_epi64(__mmask8 __U, __m256i __X) (__v8di)_mm512_setzero_si512()); } -static __inline__ __m512i __DEFAULT_FN_ATTRS512 -_mm512_cvtepu16_epi32(__m256i __A) -{ +static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR +_mm512_cvtepu16_epi32(__m256i __A) { return (__m512i)__builtin_convertvector((__v16hu)__A, __v16si); } @@ -4916,9 +4907,8 @@ _mm512_maskz_cvtepu16_epi32(__mmask16 __U, __m256i __A) (__v16si)_mm512_setzero_si512()); } -static __inline__ __m512i __DEFAULT_FN_ATTRS512 -_mm512_cvtepu16_epi64(__m128i __A) -{ +static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR +_mm512_cvtepu16_epi64(__m128i __A) { return (__m512i)__builtin_convertvector((__v8hu)__A, __v8di); } diff --git a/clang/lib/Headers/smmintrin.h b/clang/lib/Headers/smmintrin.h index bc6fe4c801d7e..875ebf08e92fa 100644 --- a/clang/lib/Headers/smmintrin.h +++ b/clang/lib/Headers/smmintrin.h @@ -27,6 +27,12 @@ __min_vector_width__(128))) #endif +#if defined(__cplusplus) && (__cplusplus >= 201103L) +#define __DEFAULT_FN_ATTRS_CONSTEXPR __DEFAULT_FN_ATTRS constexpr +#else +#define __DEFAULT_FN_ATTRS_CONSTEXPR __DEFAULT_FN_ATTRS +#endif + /* SSE4 Rounding macros. */ #define _MM_FROUND_TO_NEAREST_INT 0x00 #define _MM_FROUND_TO_NEG_INF 0x01 @@ -1224,7 +1230,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi64(__m128i __V1, /// A 128-bit vector of [16 x i8]. The lower eight 8-bit elements are /// sign-extended to 16-bit values. /// \returns A 128-bit vector of [8 x i16] containing the sign-extended values. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi8_epi16(__m128i __V) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_cvtepi8_epi16(__m128i __V) { /* This function always performs a signed extension, but __v16qi is a char which may be signed or unsigned, so use __v16qs. */ return (__m128i) __builtin_convertvector( @@ -1246,7 +1253,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi8_epi16(__m128i __V) { /// A 128-bit vector of [16 x i8]. The lower four 8-bit elements are /// sign-extended to 32-bit values. /// \returns A 128-bit vector of [4 x i32] containing the sign-extended values. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi8_epi32(__m128i __V) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_cvtepi8_epi32(__m128i __V) { /* This function always performs a signed extension, but __v16qi is a char which may be signed or unsigned, so use __v16qs. */ return (__m128i) __builtin_convertvector( @@ -1266,7 +1274,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi8_epi32(__m128i __V) { /// A 128-bit vector of [16 x i8]. The lower two 8-bit elements are /// sign-extended to 64-bit values. /// \returns A 128-bit vector of [2 x i64] containing the sign-extended values. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi8_epi64(__m128i __V) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_cvtepi8_epi64(__m128i __V) { /* This function always performs a signed extension, but __v16qi is a char which may be signed or unsigned, so use __v16qs. */ return (__m128i) __builtin_convertvector( @@ -1286,7 +1295,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi8_epi64(__m128i __V) { /// A 128-bit vector of [8 x i16]. The lower four 16-bit elements are /// sign-extended to 32-bit values. /// \returns A 128-bit vector of [4 x i32] containing the sign-extended values. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi16_epi32(__m128i __V) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_cvtepi16_epi32(__m128i __V) { return (__m128i) __builtin_convertvector( __builtin_shufflevector((__v8hi)__V, (__v8hi)__V, 0, 1, 2, 3), __v4si); } @@ -1304,7 +1314,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi16_epi32(__m128i __V) { /// A 128-bit vector of [8 x i16]. The lower two 16-bit elements are /// sign-extended to 64-bit values. /// \returns A 128-bit vector of [2 x i64] containing the sign-extended values. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi16_epi64(__m128i __V) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_cvtepi16_epi64(__m128i __V) { return (__m128i) __builtin_convertvector( __builtin_shufflevector((__v8hi)__V, (__v8hi)__V, 0, 1), __v2di); } @@ -1322,7 +1333,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi16_epi64(__m128i __V) { /// A 128-bit vector of [4 x i32]. The lower two 32-bit elements are /// sign-extended to 64-bit values. /// \returns A 128-bit vector of [2 x i64] containing the sign-extended values. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi32_epi64(__m128i __V) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_cvtepi32_epi64(__m128i __V) { return (__m128i) __builtin_convertvector( __builtin_shufflevector((__v4si)__V, (__v4si)__V, 0, 1), __v2di); } @@ -1341,7 +1353,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepi32_epi64(__m128i __V) { /// A 128-bit vector of [16 x i8]. The lower eight 8-bit elements are /// zero-extended to 16-bit values. /// \returns A 128-bit vector of [8 x i16] containing the zero-extended values. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu8_epi16(__m128i __V) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_cvtepu8_epi16(__m128i __V) { return (__m128i) __builtin_convertvector( __builtin_shufflevector((__v16qu)__V, (__v16qu)__V, 0, 1, 2, 3, 4, 5, 6, 7), @@ -1361,7 +1374,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu8_epi16(__m128i __V) { /// A 128-bit vector of [16 x i8]. The lower four 8-bit elements are /// zero-extended to 32-bit values. /// \returns A 128-bit vector of [4 x i32] containing the zero-extended values. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu8_epi32(__m128i __V) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_cvtepu8_epi32(__m128i __V) { return (__m128i) __builtin_convertvector( __builtin_shufflevector((__v16qu)__V, (__v16qu)__V, 0, 1, 2, 3), __v4si); } @@ -1379,7 +1393,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu8_epi32(__m128i __V) { /// A 128-bit vector of [16 x i8]. The lower two 8-bit elements are /// zero-extended to 64-bit values. /// \returns A 128-bit vector of [2 x i64] containing the zero-extended values. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu8_epi64(__m128i __V) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_cvtepu8_epi64(__m128i __V) { return (__m128i) __builtin_convertvector( __builtin_shufflevector((__v16qu)__V, (__v16qu)__V, 0, 1), __v2di); } @@ -1397,7 +1412,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu8_epi64(__m128i __V) { /// A 128-bit vector of [8 x i16]. The lower four 16-bit elements are /// zero-extended to 32-bit values. /// \returns A 128-bit vector of [4 x i32] containing the zero-extended values. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu16_epi32(__m128i __V) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_cvtepu16_epi32(__m128i __V) { return (__m128i) __builtin_convertvector( __builtin_shufflevector((__v8hu)__V, (__v8hu)__V, 0, 1, 2, 3), __v4si); } @@ -1415,7 +1431,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu16_epi32(__m128i __V) { /// A 128-bit vector of [8 x i16]. The lower two 16-bit elements are /// zero-extended to 64-bit values. /// \returns A 128-bit vector of [2 x i64] containing the zero-extended values. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu16_epi64(__m128i __V) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_cvtepu16_epi64(__m128i __V) { return (__m128i) __builtin_convertvector( __builtin_shufflevector((__v8hu)__V, (__v8hu)__V, 0, 1), __v2di); } @@ -1433,7 +1450,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu16_epi64(__m128i __V) { /// A 128-bit vector of [4 x i32]. The lower two 32-bit elements are /// zero-extended to 64-bit values. /// \returns A 128-bit vector of [2 x i64] containing the zero-extended values. -static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cvtepu32_epi64(__m128i __V) { +static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR +_mm_cvtepu32_epi64(__m128i __V) { return (__m128i) __builtin_convertvector( __builtin_shufflevector((__v4su)__V, (__v4su)__V, 0, 1), __v2di); } diff --git a/clang/test/CodeGen/X86/avx2-builtins.c b/clang/test/CodeGen/X86/avx2-builtins.c index adbb854660a8b..8790485f00a8c 100644 --- a/clang/test/CodeGen/X86/avx2-builtins.c +++ b/clang/test/CodeGen/X86... [truncated] `````````` </details> https://github.com/llvm/llvm-project/pull/152971 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
