[Headers][X86] Allow AVX512 reduction intrinsics to be used in constexpr #152363
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
This PR adds `constexpr` support for the following AVX512 integer reduction intrinsics: - `_mm512_reduce_add_epi32` - `_mm512_reduce_add_epi64` - `_mm512_reduce_mul_epi32` - `_mm512_reduce_mul_epi64` - `_mm512_reduce_and_epi32` - `_mm512_reduce_and_epi64` - `_mm512_reduce_or_epi32` - `_mm512_reduce_or_epi64` - `_mm512_reduce_max_epi32` - `_mm512_reduce_max_epi64` - `_mm512_reduce_min_epi32` - `_mm512_reduce_min_epi64` - `_mm512_reduce_max_epu32` - `_mm512_reduce_max_epu64` - `_mm512_reduce_min_epu32` - `_mm512_reduce_min_epu64`
llvmbot
added
clang
|
@llvm/pr-subscribers-clang Author: Pedro Lobo (pedroclobo) ChangesCloses #152324. This PR adds
Full diff: https://github.com/llvm/llvm-project/pull/152363.diff 1 Files Affected:
diff --git a/clang/lib/Headers/avx512fintrin.h b/clang/lib/Headers/avx512fintrin.h
index 45e7eeb5327d0..2f218150ca867 100644
--- a/clang/lib/Headers/avx512fintrin.h
+++ b/clang/lib/Headers/avx512fintrin.h
@@ -9337,19 +9337,19 @@ _mm512_mask_abs_pd(__m512d __W, __mmask8 __K, __m512d __A)
* This takes log2(n) steps where n is the number of elements in the vector.
*/
-static __inline__ long long __DEFAULT_FN_ATTRS512 _mm512_reduce_add_epi64(__m512i __W) {
+static __inline__ long long __DEFAULT_FN_ATTRS512_CONSTEXPR _mm512_reduce_add_epi64(__m512i __W) {
return __builtin_reduce_add((__v8di)__W);
}
-static __inline__ long long __DEFAULT_FN_ATTRS512 _mm512_reduce_mul_epi64(__m512i __W) {
+static __inline__ long long __DEFAULT_FN_ATTRS512_CONSTEXPR _mm512_reduce_mul_epi64(__m512i __W) {
return __builtin_reduce_mul((__v8di)__W);
}
-static __inline__ long long __DEFAULT_FN_ATTRS512 _mm512_reduce_and_epi64(__m512i __W) {
+static __inline__ long long __DEFAULT_FN_ATTRS512_CONSTEXPR _mm512_reduce_and_epi64(__m512i __W) {
return __builtin_reduce_and((__v8di)__W);
}
-static __inline__ long long __DEFAULT_FN_ATTRS512 _mm512_reduce_or_epi64(__m512i __W) {
+static __inline__ long long __DEFAULT_FN_ATTRS512_CONSTEXPR _mm512_reduce_or_epi64(__m512i __W) {
return __builtin_reduce_or((__v8di)__W);
}
@@ -9400,22 +9400,22 @@ _mm512_mask_reduce_mul_pd(__mmask8 __M, __m512d __W) {
return __builtin_ia32_reduce_fmul_pd512(1.0, __W);
}
-static __inline__ int __DEFAULT_FN_ATTRS512
+static __inline__ int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_add_epi32(__m512i __W) {
return __builtin_reduce_add((__v16si)__W);
}
-static __inline__ int __DEFAULT_FN_ATTRS512
+static __inline__ int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_mul_epi32(__m512i __W) {
return __builtin_reduce_mul((__v16si)__W);
}
-static __inline__ int __DEFAULT_FN_ATTRS512
+static __inline__ int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_and_epi32(__m512i __W) {
return __builtin_reduce_and((__v16si)__W);
}
-static __inline__ int __DEFAULT_FN_ATTRS512
+static __inline__ int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_or_epi32(__m512i __W) {
return __builtin_reduce_or((__v16si)__W);
}
@@ -9466,22 +9466,22 @@ _mm512_mask_reduce_mul_ps(__mmask16 __M, __m512 __W) {
return __builtin_ia32_reduce_fmul_ps512(1.0f, __W);
}
-static __inline__ long long __DEFAULT_FN_ATTRS512
+static __inline__ long long __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_max_epi64(__m512i __V) {
return __builtin_reduce_max((__v8di)__V);
}
-static __inline__ unsigned long long __DEFAULT_FN_ATTRS512
+static __inline__ unsigned long long __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_max_epu64(__m512i __V) {
return __builtin_reduce_max((__v8du)__V);
}
-static __inline__ long long __DEFAULT_FN_ATTRS512
+static __inline__ long long __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_min_epi64(__m512i __V) {
return __builtin_reduce_min((__v8di)__V);
}
-static __inline__ unsigned long long __DEFAULT_FN_ATTRS512
+static __inline__ unsigned long long __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_min_epu64(__m512i __V) {
return __builtin_reduce_min((__v8du)__V);
}
@@ -9509,22 +9509,22 @@ _mm512_mask_reduce_min_epu64(__mmask8 __M, __m512i __V) {
__V = _mm512_mask_mov_epi64(_mm512_set1_epi64(-1LL), __M, __V);
return __builtin_reduce_min((__v8du)__V);
}
-static __inline__ int __DEFAULT_FN_ATTRS512
+static __inline__ int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_max_epi32(__m512i __V) {
return __builtin_reduce_max((__v16si)__V);
}
-static __inline__ unsigned int __DEFAULT_FN_ATTRS512
+static __inline__ unsigned int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_max_epu32(__m512i __V) {
return __builtin_reduce_max((__v16su)__V);
}
-static __inline__ int __DEFAULT_FN_ATTRS512
+static __inline__ int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_min_epi32(__m512i __V) {
return __builtin_reduce_min((__v16si)__V);
}
-static __inline__ unsigned int __DEFAULT_FN_ATTRS512
+static __inline__ unsigned int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_min_epu32(__m512i __V) {
return __builtin_reduce_min((__v16su)__V);
}
|
|
@llvm/pr-subscribers-backend-x86 Author: Pedro Lobo (pedroclobo) ChangesCloses #152324. This PR adds
Full diff: https://github.com/llvm/llvm-project/pull/152363.diff 1 Files Affected:
diff --git a/clang/lib/Headers/avx512fintrin.h b/clang/lib/Headers/avx512fintrin.h
index 45e7eeb5327d0..2f218150ca867 100644
--- a/clang/lib/Headers/avx512fintrin.h
+++ b/clang/lib/Headers/avx512fintrin.h
@@ -9337,19 +9337,19 @@ _mm512_mask_abs_pd(__m512d __W, __mmask8 __K, __m512d __A)
* This takes log2(n) steps where n is the number of elements in the vector.
*/
-static __inline__ long long __DEFAULT_FN_ATTRS512 _mm512_reduce_add_epi64(__m512i __W) {
+static __inline__ long long __DEFAULT_FN_ATTRS512_CONSTEXPR _mm512_reduce_add_epi64(__m512i __W) {
return __builtin_reduce_add((__v8di)__W);
}
-static __inline__ long long __DEFAULT_FN_ATTRS512 _mm512_reduce_mul_epi64(__m512i __W) {
+static __inline__ long long __DEFAULT_FN_ATTRS512_CONSTEXPR _mm512_reduce_mul_epi64(__m512i __W) {
return __builtin_reduce_mul((__v8di)__W);
}
-static __inline__ long long __DEFAULT_FN_ATTRS512 _mm512_reduce_and_epi64(__m512i __W) {
+static __inline__ long long __DEFAULT_FN_ATTRS512_CONSTEXPR _mm512_reduce_and_epi64(__m512i __W) {
return __builtin_reduce_and((__v8di)__W);
}
-static __inline__ long long __DEFAULT_FN_ATTRS512 _mm512_reduce_or_epi64(__m512i __W) {
+static __inline__ long long __DEFAULT_FN_ATTRS512_CONSTEXPR _mm512_reduce_or_epi64(__m512i __W) {
return __builtin_reduce_or((__v8di)__W);
}
@@ -9400,22 +9400,22 @@ _mm512_mask_reduce_mul_pd(__mmask8 __M, __m512d __W) {
return __builtin_ia32_reduce_fmul_pd512(1.0, __W);
}
-static __inline__ int __DEFAULT_FN_ATTRS512
+static __inline__ int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_add_epi32(__m512i __W) {
return __builtin_reduce_add((__v16si)__W);
}
-static __inline__ int __DEFAULT_FN_ATTRS512
+static __inline__ int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_mul_epi32(__m512i __W) {
return __builtin_reduce_mul((__v16si)__W);
}
-static __inline__ int __DEFAULT_FN_ATTRS512
+static __inline__ int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_and_epi32(__m512i __W) {
return __builtin_reduce_and((__v16si)__W);
}
-static __inline__ int __DEFAULT_FN_ATTRS512
+static __inline__ int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_or_epi32(__m512i __W) {
return __builtin_reduce_or((__v16si)__W);
}
@@ -9466,22 +9466,22 @@ _mm512_mask_reduce_mul_ps(__mmask16 __M, __m512 __W) {
return __builtin_ia32_reduce_fmul_ps512(1.0f, __W);
}
-static __inline__ long long __DEFAULT_FN_ATTRS512
+static __inline__ long long __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_max_epi64(__m512i __V) {
return __builtin_reduce_max((__v8di)__V);
}
-static __inline__ unsigned long long __DEFAULT_FN_ATTRS512
+static __inline__ unsigned long long __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_max_epu64(__m512i __V) {
return __builtin_reduce_max((__v8du)__V);
}
-static __inline__ long long __DEFAULT_FN_ATTRS512
+static __inline__ long long __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_min_epi64(__m512i __V) {
return __builtin_reduce_min((__v8di)__V);
}
-static __inline__ unsigned long long __DEFAULT_FN_ATTRS512
+static __inline__ unsigned long long __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_min_epu64(__m512i __V) {
return __builtin_reduce_min((__v8du)__V);
}
@@ -9509,22 +9509,22 @@ _mm512_mask_reduce_min_epu64(__mmask8 __M, __m512i __V) {
__V = _mm512_mask_mov_epi64(_mm512_set1_epi64(-1LL), __M, __V);
return __builtin_reduce_min((__v8du)__V);
}
-static __inline__ int __DEFAULT_FN_ATTRS512
+static __inline__ int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_max_epi32(__m512i __V) {
return __builtin_reduce_max((__v16si)__V);
}
-static __inline__ unsigned int __DEFAULT_FN_ATTRS512
+static __inline__ unsigned int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_max_epu32(__m512i __V) {
return __builtin_reduce_max((__v16su)__V);
}
-static __inline__ int __DEFAULT_FN_ATTRS512
+static __inline__ int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_min_epi32(__m512i __V) {
return __builtin_reduce_min((__v16si)__V);
}
-static __inline__ unsigned int __DEFAULT_FN_ATTRS512
+static __inline__ unsigned int __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_reduce_min_epu32(__m512i __V) {
return __builtin_reduce_min((__v16su)__V);
}
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
RKSimon
reviewed
Aug 6, 2025
I took the liberty of adding the |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Closes #152324.
Part of #30794.
This PR adds
constexprsupport for the following AVX512 integer reduction intrinsics:_mm512_reduce_add_epi32_mm512_reduce_add_epi64_mm512_reduce_mul_epi32_mm512_reduce_mul_epi64_mm512_reduce_and_epi32_mm512_reduce_and_epi64_mm512_reduce_or_epi32_mm512_reduce_or_epi64_mm512_reduce_max_epi32_mm512_reduce_max_epi64_mm512_reduce_min_epi32_mm512_reduce_min_epi64_mm512_reduce_max_epu32_mm512_reduce_max_epu64_mm512_reduce_min_epu32_mm512_reduce_min_epu64