[X86] Add MMX/SSE/AVX PHADD/SUB & HADDPS/D intrinsics to be used in constexpr #156822
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
VectorExprEvaluator::VisitCallExpr / InterpretBuiltin - add MMX/SSE/AVX PHADD/SUB & HADDPS/D intrinsics to be used in constexpr llvm#155395
[Clang] VectorExprEvaluator::VisitCallExpr / InterpretBuiltin - add MMX/SSE/AVX PHADD/SUB & HADDPS/D intrinsics to be used in constexpr llvm#155395
|
Thank you for submitting a Pull Request (PR) to the LLVM Project! This PR will be automatically labeled and the relevant teams will be notified. If you wish to, you can add reviewers by using the "Reviewers" section on this page. If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers. If you have further questions, they may be answered by the LLVM GitHub User Guide. You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums. |
llvmbot
added
clang
|
@llvm/pr-subscribers-clang Author: why (whytolearn) Changes[Clang] VectorExprEvaluator::VisitCallExpr / InterpretBuiltin - add MMX/SSE/AVX PHADD/SUB & HADDPS/D intrinsics to be used in constexpr #155395 cover func: _mm_hsub_pi16 _mm_hsub_epi16 _mm256_hsub_epi16 _mm_hadd_pd _mm256_hadd_pd Patch is 32.77 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/156822.diff 8 Files Affected:
diff --git a/clang/lib/Headers/avx2intrin.h b/clang/lib/Headers/avx2intrin.h
index 4d03103ac1e08..54fe41cad6a46 100644
--- a/clang/lib/Headers/avx2intrin.h
+++ b/clang/lib/Headers/avx2intrin.h
@@ -854,10 +854,9 @@ _mm256_cmpgt_epi64(__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_hadd_epi16(__m256i __a, __m256i __b)
-{
- return (__m256i)__builtin_ia32_phaddw256((__v16hi)__a, (__v16hi)__b);
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_hadd_epi16(__m256i __a, __m256i __b) {
+ return (__m256i)__builtin_ia32_phaddw256((__v16hi)__a, (__v16hi)__b);
}
/// Horizontally adds the adjacent pairs of 32-bit integers from two 256-bit
@@ -886,10 +885,9 @@ _mm256_hadd_epi16(__m256i __a, __m256i __b)
/// \param __b
/// A 256-bit vector of [8 x i32] containing one of the source operands.
/// \returns A 256-bit vector of [8 x i32] containing the sums.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_hadd_epi32(__m256i __a, __m256i __b)
-{
- return (__m256i)__builtin_ia32_phaddd256((__v8si)__a, (__v8si)__b);
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_hadd_epi32(__m256i __a, __m256i __b) {
+ return (__m256i)__builtin_ia32_phaddd256((__v8si)__a, (__v8si)__b);
}
/// Horizontally adds the adjacent pairs of 16-bit integers from two 256-bit
@@ -921,10 +919,9 @@ _mm256_hadd_epi32(__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_hadds_epi16(__m256i __a, __m256i __b)
-{
- return (__m256i)__builtin_ia32_phaddsw256((__v16hi)__a, (__v16hi)__b);
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_hadds_epi16(__m256i __a, __m256i __b) {
+ return (__m256i)__builtin_ia32_phaddsw256((__v16hi)__a, (__v16hi)__b);
}
/// Horizontally subtracts adjacent pairs of 16-bit integers from two 256-bit
@@ -957,10 +954,9 @@ _mm256_hadds_epi16(__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 differences.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_hsub_epi16(__m256i __a, __m256i __b)
-{
- return (__m256i)__builtin_ia32_phsubw256((__v16hi)__a, (__v16hi)__b);
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_hsub_epi16(__m256i __a, __m256i __b) {
+ return (__m256i)__builtin_ia32_phsubw256((__v16hi)__a, (__v16hi)__b);
}
/// Horizontally subtracts adjacent pairs of 32-bit integers from two 256-bit
@@ -989,10 +985,9 @@ _mm256_hsub_epi16(__m256i __a, __m256i __b)
/// \param __b
/// A 256-bit vector of [8 x i32] containing one of the source operands.
/// \returns A 256-bit vector of [8 x i32] containing the differences.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_hsub_epi32(__m256i __a, __m256i __b)
-{
- return (__m256i)__builtin_ia32_phsubd256((__v8si)__a, (__v8si)__b);
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_hsub_epi32(__m256i __a, __m256i __b) {
+ return (__m256i)__builtin_ia32_phsubd256((__v8si)__a, (__v8si)__b);
}
/// Horizontally subtracts adjacent pairs of 16-bit integers from two 256-bit
@@ -1025,10 +1020,9 @@ _mm256_hsub_epi32(__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 differences.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_hsubs_epi16(__m256i __a, __m256i __b)
-{
- return (__m256i)__builtin_ia32_phsubsw256((__v16hi)__a, (__v16hi)__b);
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_hsubs_epi16(__m256i __a, __m256i __b) {
+ return (__m256i)__builtin_ia32_phsubsw256((__v16hi)__a, (__v16hi)__b);
}
/// Multiplies each unsigned byte from the 256-bit integer vector in \a __a
diff --git a/clang/lib/Headers/avxintrin.h b/clang/lib/Headers/avxintrin.h
index 03c034f64cbea..fb7110f145c38 100644
--- a/clang/lib/Headers/avxintrin.h
+++ b/clang/lib/Headers/avxintrin.h
@@ -703,9 +703,8 @@ _mm256_xor_ps(__m256 __a, __m256 __b)
/// elements of a vector of [4 x double].
/// \returns A 256-bit vector of [4 x double] containing the horizontal sums of
/// both operands.
-static __inline __m256d __DEFAULT_FN_ATTRS
-_mm256_hadd_pd(__m256d __a, __m256d __b)
-{
+static __inline __m256d __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm256_hadd_pd(__m256d __a, __m256d __b) {
return (__m256d)__builtin_ia32_haddpd256((__v4df)__a, (__v4df)__b);
}
@@ -726,9 +725,8 @@ _mm256_hadd_pd(__m256d __a, __m256d __b)
/// index 2, 3, 6, 7 of a vector of [8 x float].
/// \returns A 256-bit vector of [8 x float] containing the horizontal sums of
/// both operands.
-static __inline __m256 __DEFAULT_FN_ATTRS
-_mm256_hadd_ps(__m256 __a, __m256 __b)
-{
+static __inline __m256 __DEFAULT_FN_ATTRS_CONSTEXPR _mm256_hadd_ps(__m256 __a,
+ __m256 __b) {
return (__m256)__builtin_ia32_haddps256((__v8sf)__a, (__v8sf)__b);
}
@@ -749,9 +747,8 @@ _mm256_hadd_ps(__m256 __a, __m256 __b)
/// odd-indexed elements of a vector of [4 x double].
/// \returns A 256-bit vector of [4 x double] containing the horizontal
/// differences of both operands.
-static __inline __m256d __DEFAULT_FN_ATTRS
-_mm256_hsub_pd(__m256d __a, __m256d __b)
-{
+static __inline __m256d __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm256_hsub_pd(__m256d __a, __m256d __b) {
return (__m256d)__builtin_ia32_hsubpd256((__v4df)__a, (__v4df)__b);
}
@@ -772,9 +769,8 @@ _mm256_hsub_pd(__m256d __a, __m256d __b)
/// elements with index 2, 3, 6, 7 of a vector of [8 x float].
/// \returns A 256-bit vector of [8 x float] containing the horizontal
/// differences of both operands.
-static __inline __m256 __DEFAULT_FN_ATTRS
-_mm256_hsub_ps(__m256 __a, __m256 __b)
-{
+static __inline __m256 __DEFAULT_FN_ATTRS_CONSTEXPR _mm256_hsub_ps(__m256 __a,
+ __m256 __b) {
return (__m256)__builtin_ia32_hsubps256((__v8sf)__a, (__v8sf)__b);
}
diff --git a/clang/lib/Headers/pmmintrin.h b/clang/lib/Headers/pmmintrin.h
index cd605df7fb52d..67f2a7ffd1f56 100644
--- a/clang/lib/Headers/pmmintrin.h
+++ b/clang/lib/Headers/pmmintrin.h
@@ -89,9 +89,8 @@ _mm_addsub_ps(__m128 __a, __m128 __b)
/// destination.
/// \returns A 128-bit vector of [4 x float] containing the horizontal sums of
/// both operands.
-static __inline__ __m128 __DEFAULT_FN_ATTRS
-_mm_hadd_ps(__m128 __a, __m128 __b)
-{
+static __inline__ __m128 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_hadd_ps(__m128 __a,
+ __m128 __b) {
return __builtin_ia32_haddps((__v4sf)__a, (__v4sf)__b);
}
@@ -174,9 +173,8 @@ _mm_moveldup_ps(__m128 __a)
/// A 128-bit vector of [2 x double] containing the right source operand.
/// \returns A 128-bit vector of [2 x double] containing the alternating sums
/// and differences of both operands.
-static __inline__ __m128d __DEFAULT_FN_ATTRS
-_mm_addsub_pd(__m128d __a, __m128d __b)
-{
+static __inline__ __m128d __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_addsub_pd(__m128d __a, __m128d __b) {
return __builtin_ia32_addsubpd((__v2df)__a, (__v2df)__b);
}
@@ -197,9 +195,8 @@ _mm_addsub_pd(__m128d __a, __m128d __b)
/// destination.
/// \returns A 128-bit vector of [2 x double] containing the horizontal sums of
/// both operands.
-static __inline__ __m128d __DEFAULT_FN_ATTRS
-_mm_hadd_pd(__m128d __a, __m128d __b)
-{
+static __inline__ __m128d __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hadd_pd(__m128d __a, __m128d __b) {
return __builtin_ia32_haddpd((__v2df)__a, (__v2df)__b);
}
@@ -220,9 +217,8 @@ _mm_hadd_pd(__m128d __a, __m128d __b)
/// the destination.
/// \returns A 128-bit vector of [2 x double] containing the horizontal
/// differences of both operands.
-static __inline__ __m128d __DEFAULT_FN_ATTRS
-_mm_hsub_pd(__m128d __a, __m128d __b)
-{
+static __inline__ __m128d __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hsub_pd(__m128d __a, __m128d __b) {
return __builtin_ia32_hsubpd((__v2df)__a, (__v2df)__b);
}
diff --git a/clang/lib/Headers/tmmintrin.h b/clang/lib/Headers/tmmintrin.h
index f01c61afa8ea2..b408c6a3404ec 100644
--- a/clang/lib/Headers/tmmintrin.h
+++ b/clang/lib/Headers/tmmintrin.h
@@ -204,10 +204,9 @@ _mm_abs_epi32(__m128i __a) {
/// destination.
/// \returns A 128-bit vector of [8 x i16] containing the horizontal sums of
/// both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_hadd_epi16(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_phaddw128((__v8hi)__a, (__v8hi)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hadd_epi16(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_phaddw128((__v8hi)__a, (__v8hi)__b);
}
/// Horizontally adds the adjacent pairs of values contained in 2 packed
@@ -227,10 +226,9 @@ _mm_hadd_epi16(__m128i __a, __m128i __b)
/// destination.
/// \returns A 128-bit vector of [4 x i32] containing the horizontal sums of
/// both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_hadd_epi32(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_phaddd128((__v4si)__a, (__v4si)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hadd_epi32(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_phaddd128((__v4si)__a, (__v4si)__b);
}
/// Horizontally adds the adjacent pairs of values contained in 2 packed
@@ -250,11 +248,10 @@ _mm_hadd_epi32(__m128i __a, __m128i __b)
/// destination.
/// \returns A 64-bit vector of [4 x i16] containing the horizontal sums of both
/// operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_hadd_pi16(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_phaddw128(
- (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_hadd_pi16(__m64 __a,
+ __m64 __b) {
+ return __trunc64(__builtin_ia32_phaddw128(
+ (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
}
/// Horizontally adds the adjacent pairs of values contained in 2 packed
@@ -274,11 +271,10 @@ _mm_hadd_pi16(__m64 __a, __m64 __b)
/// destination.
/// \returns A 64-bit vector of [2 x i32] containing the horizontal sums of both
/// operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_hadd_pi32(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_phaddd128(
- (__v4si)__builtin_shufflevector(__a, __b, 0, 1), (__v4si){}));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_hadd_pi32(__m64 __a,
+ __m64 __b) {
+ return __trunc64(__builtin_ia32_phaddd128(
+ (__v4si)__builtin_shufflevector(__a, __b, 0, 1), (__v4si){}));
}
/// Horizontally adds, with saturation, the adjacent pairs of values contained
@@ -301,10 +297,9 @@ _mm_hadd_pi32(__m64 __a, __m64 __b)
/// destination.
/// \returns A 128-bit vector of [8 x i16] containing the horizontal saturated
/// sums of both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_hadds_epi16(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_phaddsw128((__v8hi)__a, (__v8hi)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hadds_epi16(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_phaddsw128((__v8hi)__a, (__v8hi)__b);
}
/// Horizontally adds, with saturation, the adjacent pairs of values contained
@@ -327,11 +322,10 @@ _mm_hadds_epi16(__m128i __a, __m128i __b)
/// destination.
/// \returns A 64-bit vector of [4 x i16] containing the horizontal saturated
/// sums of both operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_hadds_pi16(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_phaddsw128(
- (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_hadds_pi16(__m64 __a,
+ __m64 __b) {
+ return __trunc64(__builtin_ia32_phaddsw128(
+ (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
}
/// Horizontally subtracts the adjacent pairs of values contained in 2
@@ -351,10 +345,9 @@ _mm_hadds_pi16(__m64 __a, __m64 __b)
/// the destination.
/// \returns A 128-bit vector of [8 x i16] containing the horizontal differences
/// of both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_hsub_epi16(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_phsubw128((__v8hi)__a, (__v8hi)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hsub_epi16(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_phsubw128((__v8hi)__a, (__v8hi)__b);
}
/// Horizontally subtracts the adjacent pairs of values contained in 2
@@ -374,10 +367,9 @@ _mm_hsub_epi16(__m128i __a, __m128i __b)
/// the destination.
/// \returns A 128-bit vector of [4 x i32] containing the horizontal differences
/// of both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_hsub_epi32(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_phsubd128((__v4si)__a, (__v4si)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hsub_epi32(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_phsubd128((__v4si)__a, (__v4si)__b);
}
/// Horizontally subtracts the adjacent pairs of values contained in 2
@@ -397,11 +389,10 @@ _mm_hsub_epi32(__m128i __a, __m128i __b)
/// the destination.
/// \returns A 64-bit vector of [4 x i16] containing the horizontal differences
/// of both operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_hsub_pi16(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_phsubw128(
- (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_hsub_pi16(__m64 __a,
+ __m64 __b) {
+ return __trunc64(__builtin_ia32_phsubw128(
+ (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
}
/// Horizontally subtracts the adjacent pairs of values contained in 2
@@ -421,11 +412,10 @@ _mm_hsub_pi16(__m64 __a, __m64 __b)
/// the destination.
/// \returns A 64-bit vector of [2 x i32] containing the horizontal differences
/// of both operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_hsub_pi32(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_phsubd128(
- (__v4si)__builtin_shufflevector(__a, __b, 0, 1), (__v4si){}));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_hsub_pi32(__m64 __a,
+ __m64 __b) {
+ return __trunc64(__builtin_ia32_phsubd128(
+ (__v4si)__builtin_shufflevector(__a, __b, 0, 1), (__v4si){}));
}
/// Horizontally subtracts, with saturation, the adjacent pairs of values
@@ -448,10 +438,9 @@ _mm_hsub_pi32(__m64 __a, __m64 __b)
/// the destination.
/// \returns A 128-bit vector of [8 x i16] containing the horizontal saturated
/// differences of both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_hsubs_epi16(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_phsubsw128((__v8hi)__a, (__v8hi)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hsubs_epi16(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_phsubsw128((__v8hi)__a, (__v8hi)__b);
}
/// Horizontally subtracts, with saturation, the adjacent pairs of values
@@ -474,11 +463,10 @@ _mm_hsubs_epi16(__m128i __a, __m128i __b)
/// the destination.
/// \returns A 64-bit vector of [4 x i16] containing the horizontal saturated
/// differences of both operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_hsubs_pi16(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_phsubsw128(
- (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_hsubs_pi16(__m64 __a,
+ __m64 __b) {
+ return __trunc64(__builtin_ia32_phsubsw128(
+ (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
}
/// Multiplies corresponding pairs of packed 8-bit unsigned integer
@@ -509,10 +497,9 @@ _mm_hsubs_pi16(__m64 __a, __m64 __b)
/// \a R5 := (\a __a10 * \a __b10) + (\a __a11 * \a __b11) \n
/// \a R6 := (\a __a12 * \a __b12) + (\a __a13 * \a __b13) \n
/// \a R7 := (\a __a14 * \a __b14) + (\a __a15 * \a __b15)
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_maddubs_epi16(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_pmaddubsw128((__v16qi)__a, (__v16qi)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_maddubs_epi16(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_pmaddubsw128((__v16qi)__a, (__v16qi)__b);
}
/// Multiplies corresponding pairs of packed 8-bit unsigned integer
@@ -539,11 +526,10 @@ _mm_maddubs_epi16(__m128i __a, __m128i __b)
/// \a R1 := (\a __a2 * \a __b2) + (\a __a3 * \a __b3) \n
/// \a R2 := (\a __a4 * \a __b4) + (\a __a5 * \a __b5) \n
/// \a R3 := (\a __a6 * \a __b6) + (\a __a7 * \a __b7)
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_maddubs_pi16(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_pmaddubsw128((__v16qi)__anyext128(__a),
- (__v16qi)__anyext128(__b)));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_maddubs_pi16(__m64 __a, __m64 __b) {
+ return __trunc64(__builtin_ia32_pmaddubsw128((__v16qi)__anyext128(__a),
+ (__v16qi)__anyext128(__b)));
}
/// Multiplies packed 16-bit signed integer values, truncates the 32-bit
@@ -560,10 +546,9 @@ _mm_maddubs_pi16(__m64 __a, __m64 __b)
/// A 128-bit vector of [8 x i16] containing one of the source operands.
/// \returns A 128-bit vector of [8 x i16] containing the rounded and scaled
/// products of both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_mulhrs_epi16(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_pmulhrsw128((__v8hi)__a, (__v8hi)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_mulhrs_epi16(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_pmulhrsw128((__v8hi)__a, (__v8hi)__b);
}
/// Multiplies packed 16-bit signed integer values, truncates the 32-bit
diff --git a/clang/test/CodeGen/X86/avx-builtins.c b/clang/test/CodeGen/X86/avx-builtins.c
index 622ac5d50aaf0..9857b84c94112 100644
--- a/clang/test/CodeGen/X86/avx-builtins.c
+++ b/clang/test/CodeGen/X86/avx-builtins.c
@@ -1083,24 +1083,53 @@ __m256d test_mm256_hadd_pd(__m256d A, __m256d B) {
// CHECK: call {{.*}}<4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double> %{{.*}}, <4 x double> %{{.*}})
return _mm256_hadd_pd(A, B);
}
+constexpr bool test_mm256_hadd_epi32_constexpr() {
+ constexpr __m256d a = _mm256_set_pd(1.0, 2.0, 3.0, 4.0);
+ constexpr __m256d b = _mm256_set_pd(5.0, 6.0, 7.0, 8.0);
+ constexpr __m256d result = _mm256_hadd_pd(a, b);
+ return match_m256d(result,1.0+2.0,3.0+4.0,5.0+6.0,7.0+8.0);
+}
+TEST_CONSTEXPR(test_mm256_hadd_epi32_constexpr())
__m256 test_mm256_hadd_ps(__m256 A, __m256 B) {
// CHECK-LABEL: test_mm256_hadd_ps
// CHECK: call {{.*}}<8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %{{.*}}, <8 x float> %{{.*}})
return _mm256_hadd_ps(A, B);
}
+constexpr bool test_mm256_hadd_ps_constexpr() {
+ constexpr __m256 a = _mm256_set_ps(1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f);
+ constexpr __m256 b = _mm256_set_ps(9.0f, 10.0f, 11.0f, 12.0f, 13.0f, 14.0f, 15.0f, 16.0f);
+ constexpr __m256 result = _mm256_hadd_ps(a, b);
+ return match_m256(result,1.0f+2.0f,3.0f+4.0f,5.0f+6.0f,7.0f+8.0f,
+ 9.0f+10.0f,11.0f+12.0f,13.0f+14.0f,15.0f+16.0f);
+}
+TEST_CONSTEXPR(test_mm256_hadd_ps_constexpr())
...
[truncated]
|
|
@llvm/pr-subscribers-backend-x86 Author: why (whytolearn) Changes[Clang] VectorExprEvaluator::VisitCallExpr / InterpretBuiltin - add MMX/SSE/AVX PHADD/SUB & HADDPS/D intrinsics to be used in constexpr #155395 cover func: _mm_hsub_pi16 _mm_hsub_epi16 _mm256_hsub_epi16 _mm_hadd_pd _mm256_hadd_pd Patch is 32.77 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/156822.diff 8 Files Affected:
diff --git a/clang/lib/Headers/avx2intrin.h b/clang/lib/Headers/avx2intrin.h
index 4d03103ac1e08..54fe41cad6a46 100644
--- a/clang/lib/Headers/avx2intrin.h
+++ b/clang/lib/Headers/avx2intrin.h
@@ -854,10 +854,9 @@ _mm256_cmpgt_epi64(__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_hadd_epi16(__m256i __a, __m256i __b)
-{
- return (__m256i)__builtin_ia32_phaddw256((__v16hi)__a, (__v16hi)__b);
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_hadd_epi16(__m256i __a, __m256i __b) {
+ return (__m256i)__builtin_ia32_phaddw256((__v16hi)__a, (__v16hi)__b);
}
/// Horizontally adds the adjacent pairs of 32-bit integers from two 256-bit
@@ -886,10 +885,9 @@ _mm256_hadd_epi16(__m256i __a, __m256i __b)
/// \param __b
/// A 256-bit vector of [8 x i32] containing one of the source operands.
/// \returns A 256-bit vector of [8 x i32] containing the sums.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_hadd_epi32(__m256i __a, __m256i __b)
-{
- return (__m256i)__builtin_ia32_phaddd256((__v8si)__a, (__v8si)__b);
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_hadd_epi32(__m256i __a, __m256i __b) {
+ return (__m256i)__builtin_ia32_phaddd256((__v8si)__a, (__v8si)__b);
}
/// Horizontally adds the adjacent pairs of 16-bit integers from two 256-bit
@@ -921,10 +919,9 @@ _mm256_hadd_epi32(__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_hadds_epi16(__m256i __a, __m256i __b)
-{
- return (__m256i)__builtin_ia32_phaddsw256((__v16hi)__a, (__v16hi)__b);
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_hadds_epi16(__m256i __a, __m256i __b) {
+ return (__m256i)__builtin_ia32_phaddsw256((__v16hi)__a, (__v16hi)__b);
}
/// Horizontally subtracts adjacent pairs of 16-bit integers from two 256-bit
@@ -957,10 +954,9 @@ _mm256_hadds_epi16(__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 differences.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_hsub_epi16(__m256i __a, __m256i __b)
-{
- return (__m256i)__builtin_ia32_phsubw256((__v16hi)__a, (__v16hi)__b);
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_hsub_epi16(__m256i __a, __m256i __b) {
+ return (__m256i)__builtin_ia32_phsubw256((__v16hi)__a, (__v16hi)__b);
}
/// Horizontally subtracts adjacent pairs of 32-bit integers from two 256-bit
@@ -989,10 +985,9 @@ _mm256_hsub_epi16(__m256i __a, __m256i __b)
/// \param __b
/// A 256-bit vector of [8 x i32] containing one of the source operands.
/// \returns A 256-bit vector of [8 x i32] containing the differences.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_hsub_epi32(__m256i __a, __m256i __b)
-{
- return (__m256i)__builtin_ia32_phsubd256((__v8si)__a, (__v8si)__b);
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_hsub_epi32(__m256i __a, __m256i __b) {
+ return (__m256i)__builtin_ia32_phsubd256((__v8si)__a, (__v8si)__b);
}
/// Horizontally subtracts adjacent pairs of 16-bit integers from two 256-bit
@@ -1025,10 +1020,9 @@ _mm256_hsub_epi32(__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 differences.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
-_mm256_hsubs_epi16(__m256i __a, __m256i __b)
-{
- return (__m256i)__builtin_ia32_phsubsw256((__v16hi)__a, (__v16hi)__b);
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
+_mm256_hsubs_epi16(__m256i __a, __m256i __b) {
+ return (__m256i)__builtin_ia32_phsubsw256((__v16hi)__a, (__v16hi)__b);
}
/// Multiplies each unsigned byte from the 256-bit integer vector in \a __a
diff --git a/clang/lib/Headers/avxintrin.h b/clang/lib/Headers/avxintrin.h
index 03c034f64cbea..fb7110f145c38 100644
--- a/clang/lib/Headers/avxintrin.h
+++ b/clang/lib/Headers/avxintrin.h
@@ -703,9 +703,8 @@ _mm256_xor_ps(__m256 __a, __m256 __b)
/// elements of a vector of [4 x double].
/// \returns A 256-bit vector of [4 x double] containing the horizontal sums of
/// both operands.
-static __inline __m256d __DEFAULT_FN_ATTRS
-_mm256_hadd_pd(__m256d __a, __m256d __b)
-{
+static __inline __m256d __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm256_hadd_pd(__m256d __a, __m256d __b) {
return (__m256d)__builtin_ia32_haddpd256((__v4df)__a, (__v4df)__b);
}
@@ -726,9 +725,8 @@ _mm256_hadd_pd(__m256d __a, __m256d __b)
/// index 2, 3, 6, 7 of a vector of [8 x float].
/// \returns A 256-bit vector of [8 x float] containing the horizontal sums of
/// both operands.
-static __inline __m256 __DEFAULT_FN_ATTRS
-_mm256_hadd_ps(__m256 __a, __m256 __b)
-{
+static __inline __m256 __DEFAULT_FN_ATTRS_CONSTEXPR _mm256_hadd_ps(__m256 __a,
+ __m256 __b) {
return (__m256)__builtin_ia32_haddps256((__v8sf)__a, (__v8sf)__b);
}
@@ -749,9 +747,8 @@ _mm256_hadd_ps(__m256 __a, __m256 __b)
/// odd-indexed elements of a vector of [4 x double].
/// \returns A 256-bit vector of [4 x double] containing the horizontal
/// differences of both operands.
-static __inline __m256d __DEFAULT_FN_ATTRS
-_mm256_hsub_pd(__m256d __a, __m256d __b)
-{
+static __inline __m256d __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm256_hsub_pd(__m256d __a, __m256d __b) {
return (__m256d)__builtin_ia32_hsubpd256((__v4df)__a, (__v4df)__b);
}
@@ -772,9 +769,8 @@ _mm256_hsub_pd(__m256d __a, __m256d __b)
/// elements with index 2, 3, 6, 7 of a vector of [8 x float].
/// \returns A 256-bit vector of [8 x float] containing the horizontal
/// differences of both operands.
-static __inline __m256 __DEFAULT_FN_ATTRS
-_mm256_hsub_ps(__m256 __a, __m256 __b)
-{
+static __inline __m256 __DEFAULT_FN_ATTRS_CONSTEXPR _mm256_hsub_ps(__m256 __a,
+ __m256 __b) {
return (__m256)__builtin_ia32_hsubps256((__v8sf)__a, (__v8sf)__b);
}
diff --git a/clang/lib/Headers/pmmintrin.h b/clang/lib/Headers/pmmintrin.h
index cd605df7fb52d..67f2a7ffd1f56 100644
--- a/clang/lib/Headers/pmmintrin.h
+++ b/clang/lib/Headers/pmmintrin.h
@@ -89,9 +89,8 @@ _mm_addsub_ps(__m128 __a, __m128 __b)
/// destination.
/// \returns A 128-bit vector of [4 x float] containing the horizontal sums of
/// both operands.
-static __inline__ __m128 __DEFAULT_FN_ATTRS
-_mm_hadd_ps(__m128 __a, __m128 __b)
-{
+static __inline__ __m128 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_hadd_ps(__m128 __a,
+ __m128 __b) {
return __builtin_ia32_haddps((__v4sf)__a, (__v4sf)__b);
}
@@ -174,9 +173,8 @@ _mm_moveldup_ps(__m128 __a)
/// A 128-bit vector of [2 x double] containing the right source operand.
/// \returns A 128-bit vector of [2 x double] containing the alternating sums
/// and differences of both operands.
-static __inline__ __m128d __DEFAULT_FN_ATTRS
-_mm_addsub_pd(__m128d __a, __m128d __b)
-{
+static __inline__ __m128d __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_addsub_pd(__m128d __a, __m128d __b) {
return __builtin_ia32_addsubpd((__v2df)__a, (__v2df)__b);
}
@@ -197,9 +195,8 @@ _mm_addsub_pd(__m128d __a, __m128d __b)
/// destination.
/// \returns A 128-bit vector of [2 x double] containing the horizontal sums of
/// both operands.
-static __inline__ __m128d __DEFAULT_FN_ATTRS
-_mm_hadd_pd(__m128d __a, __m128d __b)
-{
+static __inline__ __m128d __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hadd_pd(__m128d __a, __m128d __b) {
return __builtin_ia32_haddpd((__v2df)__a, (__v2df)__b);
}
@@ -220,9 +217,8 @@ _mm_hadd_pd(__m128d __a, __m128d __b)
/// the destination.
/// \returns A 128-bit vector of [2 x double] containing the horizontal
/// differences of both operands.
-static __inline__ __m128d __DEFAULT_FN_ATTRS
-_mm_hsub_pd(__m128d __a, __m128d __b)
-{
+static __inline__ __m128d __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hsub_pd(__m128d __a, __m128d __b) {
return __builtin_ia32_hsubpd((__v2df)__a, (__v2df)__b);
}
diff --git a/clang/lib/Headers/tmmintrin.h b/clang/lib/Headers/tmmintrin.h
index f01c61afa8ea2..b408c6a3404ec 100644
--- a/clang/lib/Headers/tmmintrin.h
+++ b/clang/lib/Headers/tmmintrin.h
@@ -204,10 +204,9 @@ _mm_abs_epi32(__m128i __a) {
/// destination.
/// \returns A 128-bit vector of [8 x i16] containing the horizontal sums of
/// both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_hadd_epi16(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_phaddw128((__v8hi)__a, (__v8hi)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hadd_epi16(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_phaddw128((__v8hi)__a, (__v8hi)__b);
}
/// Horizontally adds the adjacent pairs of values contained in 2 packed
@@ -227,10 +226,9 @@ _mm_hadd_epi16(__m128i __a, __m128i __b)
/// destination.
/// \returns A 128-bit vector of [4 x i32] containing the horizontal sums of
/// both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_hadd_epi32(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_phaddd128((__v4si)__a, (__v4si)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hadd_epi32(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_phaddd128((__v4si)__a, (__v4si)__b);
}
/// Horizontally adds the adjacent pairs of values contained in 2 packed
@@ -250,11 +248,10 @@ _mm_hadd_epi32(__m128i __a, __m128i __b)
/// destination.
/// \returns A 64-bit vector of [4 x i16] containing the horizontal sums of both
/// operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_hadd_pi16(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_phaddw128(
- (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_hadd_pi16(__m64 __a,
+ __m64 __b) {
+ return __trunc64(__builtin_ia32_phaddw128(
+ (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
}
/// Horizontally adds the adjacent pairs of values contained in 2 packed
@@ -274,11 +271,10 @@ _mm_hadd_pi16(__m64 __a, __m64 __b)
/// destination.
/// \returns A 64-bit vector of [2 x i32] containing the horizontal sums of both
/// operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_hadd_pi32(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_phaddd128(
- (__v4si)__builtin_shufflevector(__a, __b, 0, 1), (__v4si){}));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_hadd_pi32(__m64 __a,
+ __m64 __b) {
+ return __trunc64(__builtin_ia32_phaddd128(
+ (__v4si)__builtin_shufflevector(__a, __b, 0, 1), (__v4si){}));
}
/// Horizontally adds, with saturation, the adjacent pairs of values contained
@@ -301,10 +297,9 @@ _mm_hadd_pi32(__m64 __a, __m64 __b)
/// destination.
/// \returns A 128-bit vector of [8 x i16] containing the horizontal saturated
/// sums of both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_hadds_epi16(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_phaddsw128((__v8hi)__a, (__v8hi)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hadds_epi16(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_phaddsw128((__v8hi)__a, (__v8hi)__b);
}
/// Horizontally adds, with saturation, the adjacent pairs of values contained
@@ -327,11 +322,10 @@ _mm_hadds_epi16(__m128i __a, __m128i __b)
/// destination.
/// \returns A 64-bit vector of [4 x i16] containing the horizontal saturated
/// sums of both operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_hadds_pi16(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_phaddsw128(
- (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_hadds_pi16(__m64 __a,
+ __m64 __b) {
+ return __trunc64(__builtin_ia32_phaddsw128(
+ (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
}
/// Horizontally subtracts the adjacent pairs of values contained in 2
@@ -351,10 +345,9 @@ _mm_hadds_pi16(__m64 __a, __m64 __b)
/// the destination.
/// \returns A 128-bit vector of [8 x i16] containing the horizontal differences
/// of both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_hsub_epi16(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_phsubw128((__v8hi)__a, (__v8hi)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hsub_epi16(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_phsubw128((__v8hi)__a, (__v8hi)__b);
}
/// Horizontally subtracts the adjacent pairs of values contained in 2
@@ -374,10 +367,9 @@ _mm_hsub_epi16(__m128i __a, __m128i __b)
/// the destination.
/// \returns A 128-bit vector of [4 x i32] containing the horizontal differences
/// of both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_hsub_epi32(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_phsubd128((__v4si)__a, (__v4si)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hsub_epi32(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_phsubd128((__v4si)__a, (__v4si)__b);
}
/// Horizontally subtracts the adjacent pairs of values contained in 2
@@ -397,11 +389,10 @@ _mm_hsub_epi32(__m128i __a, __m128i __b)
/// the destination.
/// \returns A 64-bit vector of [4 x i16] containing the horizontal differences
/// of both operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_hsub_pi16(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_phsubw128(
- (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_hsub_pi16(__m64 __a,
+ __m64 __b) {
+ return __trunc64(__builtin_ia32_phsubw128(
+ (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
}
/// Horizontally subtracts the adjacent pairs of values contained in 2
@@ -421,11 +412,10 @@ _mm_hsub_pi16(__m64 __a, __m64 __b)
/// the destination.
/// \returns A 64-bit vector of [2 x i32] containing the horizontal differences
/// of both operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_hsub_pi32(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_phsubd128(
- (__v4si)__builtin_shufflevector(__a, __b, 0, 1), (__v4si){}));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_hsub_pi32(__m64 __a,
+ __m64 __b) {
+ return __trunc64(__builtin_ia32_phsubd128(
+ (__v4si)__builtin_shufflevector(__a, __b, 0, 1), (__v4si){}));
}
/// Horizontally subtracts, with saturation, the adjacent pairs of values
@@ -448,10 +438,9 @@ _mm_hsub_pi32(__m64 __a, __m64 __b)
/// the destination.
/// \returns A 128-bit vector of [8 x i16] containing the horizontal saturated
/// differences of both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_hsubs_epi16(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_phsubsw128((__v8hi)__a, (__v8hi)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_hsubs_epi16(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_phsubsw128((__v8hi)__a, (__v8hi)__b);
}
/// Horizontally subtracts, with saturation, the adjacent pairs of values
@@ -474,11 +463,10 @@ _mm_hsubs_epi16(__m128i __a, __m128i __b)
/// the destination.
/// \returns A 64-bit vector of [4 x i16] containing the horizontal saturated
/// differences of both operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_hsubs_pi16(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_phsubsw128(
- (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR _mm_hsubs_pi16(__m64 __a,
+ __m64 __b) {
+ return __trunc64(__builtin_ia32_phsubsw128(
+ (__v8hi)__builtin_shufflevector(__a, __b, 0, 1), (__v8hi){}));
}
/// Multiplies corresponding pairs of packed 8-bit unsigned integer
@@ -509,10 +497,9 @@ _mm_hsubs_pi16(__m64 __a, __m64 __b)
/// \a R5 := (\a __a10 * \a __b10) + (\a __a11 * \a __b11) \n
/// \a R6 := (\a __a12 * \a __b12) + (\a __a13 * \a __b13) \n
/// \a R7 := (\a __a14 * \a __b14) + (\a __a15 * \a __b15)
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_maddubs_epi16(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_pmaddubsw128((__v16qi)__a, (__v16qi)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_maddubs_epi16(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_pmaddubsw128((__v16qi)__a, (__v16qi)__b);
}
/// Multiplies corresponding pairs of packed 8-bit unsigned integer
@@ -539,11 +526,10 @@ _mm_maddubs_epi16(__m128i __a, __m128i __b)
/// \a R1 := (\a __a2 * \a __b2) + (\a __a3 * \a __b3) \n
/// \a R2 := (\a __a4 * \a __b4) + (\a __a5 * \a __b5) \n
/// \a R3 := (\a __a6 * \a __b6) + (\a __a7 * \a __b7)
-static __inline__ __m64 __DEFAULT_FN_ATTRS
-_mm_maddubs_pi16(__m64 __a, __m64 __b)
-{
- return __trunc64(__builtin_ia32_pmaddubsw128((__v16qi)__anyext128(__a),
- (__v16qi)__anyext128(__b)));
+static __inline__ __m64 __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_maddubs_pi16(__m64 __a, __m64 __b) {
+ return __trunc64(__builtin_ia32_pmaddubsw128((__v16qi)__anyext128(__a),
+ (__v16qi)__anyext128(__b)));
}
/// Multiplies packed 16-bit signed integer values, truncates the 32-bit
@@ -560,10 +546,9 @@ _mm_maddubs_pi16(__m64 __a, __m64 __b)
/// A 128-bit vector of [8 x i16] containing one of the source operands.
/// \returns A 128-bit vector of [8 x i16] containing the rounded and scaled
/// products of both operands.
-static __inline__ __m128i __DEFAULT_FN_ATTRS
-_mm_mulhrs_epi16(__m128i __a, __m128i __b)
-{
- return (__m128i)__builtin_ia32_pmulhrsw128((__v8hi)__a, (__v8hi)__b);
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR
+_mm_mulhrs_epi16(__m128i __a, __m128i __b) {
+ return (__m128i)__builtin_ia32_pmulhrsw128((__v8hi)__a, (__v8hi)__b);
}
/// Multiplies packed 16-bit signed integer values, truncates the 32-bit
diff --git a/clang/test/CodeGen/X86/avx-builtins.c b/clang/test/CodeGen/X86/avx-builtins.c
index 622ac5d50aaf0..9857b84c94112 100644
--- a/clang/test/CodeGen/X86/avx-builtins.c
+++ b/clang/test/CodeGen/X86/avx-builtins.c
@@ -1083,24 +1083,53 @@ __m256d test_mm256_hadd_pd(__m256d A, __m256d B) {
// CHECK: call {{.*}}<4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double> %{{.*}}, <4 x double> %{{.*}})
return _mm256_hadd_pd(A, B);
}
+constexpr bool test_mm256_hadd_epi32_constexpr() {
+ constexpr __m256d a = _mm256_set_pd(1.0, 2.0, 3.0, 4.0);
+ constexpr __m256d b = _mm256_set_pd(5.0, 6.0, 7.0, 8.0);
+ constexpr __m256d result = _mm256_hadd_pd(a, b);
+ return match_m256d(result,1.0+2.0,3.0+4.0,5.0+6.0,7.0+8.0);
+}
+TEST_CONSTEXPR(test_mm256_hadd_epi32_constexpr())
__m256 test_mm256_hadd_ps(__m256 A, __m256 B) {
// CHECK-LABEL: test_mm256_hadd_ps
// CHECK: call {{.*}}<8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %{{.*}}, <8 x float> %{{.*}})
return _mm256_hadd_ps(A, B);
}
+constexpr bool test_mm256_hadd_ps_constexpr() {
+ constexpr __m256 a = _mm256_set_ps(1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f);
+ constexpr __m256 b = _mm256_set_ps(9.0f, 10.0f, 11.0f, 12.0f, 13.0f, 14.0f, 15.0f, 16.0f);
+ constexpr __m256 result = _mm256_hadd_ps(a, b);
+ return match_m256(result,1.0f+2.0f,3.0f+4.0f,5.0f+6.0f,7.0f+8.0f,
+ 9.0f+10.0f,11.0f+12.0f,13.0f+14.0f,15.0f+16.0f);
+}
+TEST_CONSTEXPR(test_mm256_hadd_ps_constexpr())
...
[truncated]
|
RKSimon
changed the title
RKSimon
changed the title
|
|
RKSimon
requested changes
Sep 4, 2025
| constexpr __m128i result = _mm_hadd_epi32(a, b); | ||
| return match_v4si(result,1+2,3+4,5+6,7+8); | ||
| } | ||
| TEST_CONSTEXPR(test_mm_hadd_epi32_constexpr()) |
There was a problem hiding this comment.
This needs to be done like:
TEST_CONSTEXPR(match_v4si(_mm_hadd_epi32((__m128i)(__v4si){1,2,3,4}, (__m128i)(__v4si){5,6,7}), 3, 7, 11, 15));
same for all the others - we don't want to have nested intrinsic calls and can't expose constexpr code as the tests run on in a mixture of C/C++ modes
There was a problem hiding this comment.
thank you reply, i will adjust and do rest work
There was a problem hiding this comment.
please help approve the CI workflows again, Thank you!
added 2 commits
September 13, 2025 21:24
llvmbot
added
clang:frontend
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
RKSimon
requested changes
Sep 19, 2025
clang/lib/AST/ExprConstant.cpp
Outdated
| const Expr *ArrayFiller, | ||
| QualType AllocType = QualType()); | ||
| }; | ||
| : |
There was a problem hiding this comment.
(style) don't touch lines not affected by a patch - undo this clang-format change
There was a problem hiding this comment.
i have undo this unintentional formatting of the code. i will more cautious
| TEST_CONSTEXPR(match_v16hi( _mm256_hadds_epi16( | ||
| (__m256i)(__v16hi){32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767}, | ||
| (__m256i)(__v16hi){32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767}, | ||
| 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767, 32767))) |
There was a problem hiding this comment.
don't use uniform values - it makes it very difficult to check the correct element orders have been kept
also - use some non-saturating values as well
| #include "../ExprConstShared.h" | ||
| #include "Boolean.h" | ||
| #include "EvalEmitter.h" | ||
| #include "Floating.h" |
There was a problem hiding this comment.
yes, i have remove this
| #include "llvm/ADT/StringExtras.h" | ||
| #include "llvm/Support/ErrorHandling.h" | ||
| #include "llvm/Support/SipHash.h" | ||
| #include <cassert> |
There was a problem hiding this comment.
yes, have delete
tbaederr
reviewed
Sep 26, 2025
| using T = PrimConv<PT_Float>::T; | ||
| APFloat Elem1 = LHS.elem<T>(I).getAPFloat(); | ||
| APFloat Elem2 = LHS.elem<T>(I + 1).getAPFloat(); | ||
| Dst.elem<T>(DstElem++) = static_cast<T>(APFloat(Fn(Elem1, Elem2, RM))); |
There was a problem hiding this comment.
Why the extra APFloat() here?
There was a problem hiding this comment.
Same with the integer cases above.
There was a problem hiding this comment.
i have do some reference func "interp__builtin_elementwise_int_binop" to deal horizontal add/sub。 So APInt APFloat like same level class。 and when deal flaot, i must consider rounding off mode, So like func "interp__builtin_elementwise_triop_fp" and func "interp__builtin_elementwise_triop", i split it into 2 case
There was a problem hiding this comment.
Fn already returns an APFloat, doesn't it? Why the extra constructor call here?
There was a problem hiding this comment.
thank you remind, i have adjust the code
RKSimon
requested changes
Sep 26, 2025
clang/lib/AST/ExprConstant.cpp
Outdated
| #include "clang/Basic/TargetBuiltins.h" | ||
| #include "clang/Basic/TargetInfo.h" | ||
| #include "llvm/ADT/APFixedPoint.h" | ||
| #include "llvm/ADT/APInt.h" |
| // CHECK: call {{.*}}<4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double> %{{.*}}, <4 x double> %{{.*}}) | ||
| return _mm256_hadd_pd(A, B); | ||
| } | ||
| TEST_CONSTEXPR(match_m256d(_mm256_hadd_pd((__m256d){1.0, 2.0, 3.0, 4.0}, (__m256d){5.0, 6.0, 7.0, 8.0}), 3.0, 7.0, 11.0, 15.0)); |
There was a problem hiding this comment.
The result should be 3.0, 11.0, 7.0, 15.0 - hop are performed at the 128-bit level (so lhs hops then rhs hops per 128) -(interp__builtin_x86_pack has to do something similar):
dst[63:0] := a[127:64] + a[63:0]
dst[127:64] := b[127:64] + b[63:0]
dst[191:128] := a[255:192] + a[191:128]
dst[255:192] := b[255:192] + b[191:128]
dst[MAX:256] := 0
There was a problem hiding this comment.
this is my error, i will more cautious, thank you
|
@whytolearn reverse-ping |
RKSimon
requested changes
Oct 9, 2025
| } | ||
|
|
||
| let Features = "ssse3" in { | ||
| def pmaddubsw128 : X86Builtin<"_Vector<8, short>(_Vector<16, char>, _Vector<16, char>)">; |
There was a problem hiding this comment.
pmaddubsw128 is now defined somewhere else as constexpr - please remove this line
There was a problem hiding this comment.
yes, my merge error
| def mpsadbw256 : X86Builtin<"_Vector<32, char>(_Vector<32, char>, _Vector<32, char>, _Constant char)">; | ||
| def palignr256 : X86Builtin<"_Vector<32, char>(_Vector<32, char>, _Vector<32, char>, _Constant int)">; | ||
| def pmaddubsw256 : X86Builtin<"_Vector<16, short>(_Vector<32, char>, _Vector<32, char>)">; | ||
| def pmaddwd256 : X86Builtin<"_Vector<8, int>(_Vector<16, short>, _Vector<16, short>)">; |
There was a problem hiding this comment.
these needs removing as they're defined as constexpr someplace else
clang/lib/AST/ExprConstant.cpp
Outdated
| return Success(APValue(ResultElements.data(), ResultElements.size()), E); | ||
| } | ||
| case clang::X86::BI__builtin_ia32_haddpd: | ||
| case clang::X86::BI__builtin_ia32_haddps: { |
There was a problem hiding this comment.
all 4 haddps/pd intrinsics should be handled here - using the same NumLanes approach everywhere else
clang/lib/AST/ExprConstant.cpp
Outdated
| return Success(APValue(ResultElements.data(), ResultElements.size()), E); | ||
| } | ||
| case clang::X86::BI__builtin_ia32_hsubpd: | ||
| case clang::X86::BI__builtin_ia32_hsubps: { |
There was a problem hiding this comment.
all 4 hsubps/pd intrinsics should be handled here
RKSimon
requested changes
Oct 13, 2025
Co-authored-by: Simon Pilgrim <[email protected]>
Co-authored-by: Simon Pilgrim <[email protected]>
RKSimon
reviewed
Oct 13, 2025
Co-authored-by: Simon Pilgrim <[email protected]>
RKSimon
requested changes
Oct 13, 2025
clang/lib/Headers/tmmintrin.h
Outdated
| /// \a R6 := (\a __a12 * \a __b12) + (\a __a13 * \a __b13) \n | ||
| /// \a R7 := (\a __a14 * \a __b14) + (\a __a15 * \a __b15) | ||
| static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_maddubs_epi16(__m128i __a, | ||
| __m128i __b) { |
clang/lib/Headers/tmmintrin.h
Outdated
| static __inline__ __m64 __DEFAULT_FN_ATTRS _mm_maddubs_pi16(__m64 __a, | ||
| __m64 __b) { | ||
| return __trunc64(__builtin_ia32_pmaddubsw128((__v16qi)__anyext128(__a), | ||
| (__v16qi)__anyext128(__b))); |
|
@whytolearn Congratulations on having your first Pull Request (PR) merged into the LLVM Project! Your changes will be combined with recent changes from other authors, then tested by our build bots. If there is a problem with a build, you may receive a report in an email or a comment on this PR. Please check whether problems have been caused by your change specifically, as the builds can include changes from many authors. It is not uncommon for your change to be included in a build that fails due to someone else's changes, or infrastructure issues. How to do this, and the rest of the post-merge process, is covered in detail here. If your change does cause a problem, it may be reverted, or you can revert it yourself. This is a normal part of LLVM development. You can fix your changes and open a new PR to merge them again. If you don't get any reports, no action is required from you. Your changes are working as expected, well done! |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/51/builds/25216 Here is the relevant piece of the build log for the reference |
thank you help and code review. cheer !!! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
[Clang] VectorExprEvaluator::VisitCallExpr / InterpretBuiltin - add MMX/SSE/AVX PHADD/SUB & HADDPS/D intrinsics to be used in constexpr
Fixes #155395
cover func:
_mm_hadd_pi16 _mm_hadd_epi16 _mm256_hadd_epi16
_mm_hadd_pi32 _mm_hadd_epi32 _mm256_hadd_epi32
_mm_hadds_pi16 _mm_hadds_epi16 _mm256_hadds_epi16
_mm_hsub_pi16 _mm_hsub_epi16 _mm256_hsub_epi16
_mm_hsub_pi32 _mm_hsub_epi32 _mm256_hsub_epi32
_mm_hsubs_pi16 _mm_hsubs_epi16 _mm256_hsubs_epi16
_mm_hadd_pd _mm256_hadd_pd
_mm_hadd_ps _mm256_hadd_ps
_mm_hsub_pd _mm256_hsub_pd
_mm_hsub_ps _mm256_hsub_ps