[Headers][X86] Allow FMA intrinsics to be used in constexpr #156385
Merged
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
Fixes llvm#155265 Add constexpr support for the following: _mm512_fmadd_pd _mm512_mask_fmadd_pd _mm512_mask3_fmadd_pd _mm512_maskz_fmadd_pd _mm512_fmadd_ps _mm512_mask_fmadd_ps _mm512_mask3_fmadd_ps _mm512_maskz_fmadd_ps _mm_mask_fmadd_pd _mm_mask3_fmadd_pd _mm_maskz_fmadd_pd _mm_mask_fmadd_ps _mm_mask3_fmadd_ps _mm_maskz_fmadd_ps _mm256_mask_fmadd_pd _mm256_mask3_fmadd_pd _mm256_maskz_fmadd_pd _mm256_mask_fmadd_ps _mm256_mask3_fmadd_ps _mm256_maskz_fmadd_ps _mm512_fmsub_pd _mm512_mask_fmsub_pd _mm512_mask3_fmsub_pd _mm512_maskz_fmsub_pd _mm512_fmsub_ps _mm512_mask_fmsub_ps _mm512_mask3_fmsub_ps _mm512_maskz_fmsub_ps _mm_mask_fmsub_pd _mm_mask3_fmsub_pd _mm_maskz_fmsub_pd _mm_mask_fmsub_ps _mm_mask3_fmsub_ps _mm_maskz_fmsub_ps _mm256_mask_fmsub_pd _mm256_mask3_fmsub_pd _mm256_maskz_fmsub_pd _mm256_mask_fmsub_ps _mm256_mask3_fmsub_ps _mm256_maskz_fmsub_ps _mm512_fnmadd_pd _mm512_mask_fnmadd_pd _mm512_mask3_fnmadd_pd _mm512_maskz_fnmadd_pd _mm512_fnmsub_pd _mm512_mask_fnmsub_pd _mm512_mask3_fnmsub_pd _mm512_maskz_fnmsub_pd _mm_mask_fnmadd_pd _mm_mask3_fnmadd_pd _mm_maskz_fnmadd_pd _mm_mask_fnmadd_ps _mm_mask3_fnmadd_ps _mm_maskz_fnmadd_ps _mm256_mask_fnmadd_pd _mm256_mask3_fnmadd_pd _mm256_maskz_fnmadd_pd _mm256_mask_fnmadd_ps _mm256_mask3_fnmadd_ps _mm256_maskz_fnmadd_ps _mm512_fnmadd_ps _mm512_mask_fnmadd_ps _mm512_mask3_fnmadd_ps _mm512_maskz_fnmadd_ps _mm512_fnmsub_ps _mm512_mask_fnmsub_ps _mm512_mask3_fnmsub_ps _mm512_maskz_fnmsub_ps _mm_mask_fnmsub_pd _mm_mask3_fnmsub_pd _mm_maskz_fnmsub_pd _mm_mask_fnmsub_ps _mm_mask3_fnmsub_ps _mm_maskz_fnmsub_ps _mm256_mask_fnmsub_pd _mm256_mask3_fnmsub_pd _mm256_maskz_fnmsub_pd _mm256_mask_fnmsub_ps _mm256_mask3_fnmsub_ps _mm256_maskz_fnmsub_ps
|
@llvm/pr-subscribers-clang @llvm/pr-subscribers-backend-x86 Author: Chaitanya Koparkar (ckoparkar) ChangesFixes #155265 Add constexpr support for the following: _mm512_fmadd_pd _mm512_mask_fmadd_pd _mm512_mask3_fmadd_pd _mm512_maskz_fmadd_pd _mm512_fmadd_ps _mm512_mask_fmadd_ps _mm512_mask3_fmadd_ps _mm512_maskz_fmadd_ps _mm_mask_fmadd_pd _mm_mask3_fmadd_pd _mm_maskz_fmadd_pd _mm_mask_fmadd_ps _mm_mask3_fmadd_ps _mm_maskz_fmadd_ps _mm256_mask_fmadd_pd _mm256_mask3_fmadd_pd _mm256_maskz_fmadd_pd _mm256_mask_fmadd_ps _mm256_mask3_fmadd_ps _mm256_maskz_fmadd_ps _mm512_fmsub_pd _mm512_mask_fmsub_pd _mm512_mask3_fmsub_pd _mm512_maskz_fmsub_pd _mm512_fmsub_ps _mm512_mask_fmsub_ps _mm512_mask3_fmsub_ps _mm512_maskz_fmsub_ps _mm_mask_fmsub_pd _mm_mask3_fmsub_pd _mm_maskz_fmsub_pd _mm_mask_fmsub_ps _mm_mask3_fmsub_ps _mm_maskz_fmsub_ps _mm256_mask_fmsub_pd _mm256_mask3_fmsub_pd _mm256_maskz_fmsub_pd _mm256_mask_fmsub_ps _mm256_mask3_fmsub_ps _mm256_maskz_fmsub_ps _mm512_fnmadd_pd _mm512_mask_fnmadd_pd _mm512_mask3_fnmadd_pd _mm512_maskz_fnmadd_pd _mm512_fnmsub_pd _mm512_mask_fnmsub_pd _mm512_mask3_fnmsub_pd _mm512_maskz_fnmsub_pd _mm_mask_fnmadd_pd _mm_mask3_fnmadd_pd _mm_maskz_fnmadd_pd _mm_mask_fnmadd_ps _mm_mask3_fnmadd_ps _mm_maskz_fnmadd_ps _mm256_mask_fnmadd_pd _mm256_mask3_fnmadd_pd _mm256_maskz_fnmadd_pd _mm256_mask_fnmadd_ps _mm256_mask3_fnmadd_ps _mm256_maskz_fnmadd_ps _mm512_fnmadd_ps _mm512_mask_fnmadd_ps _mm512_mask3_fnmadd_ps _mm512_maskz_fnmadd_ps _mm512_fnmsub_ps _mm512_mask_fnmsub_ps _mm512_mask3_fnmsub_ps _mm512_maskz_fnmsub_ps _mm_mask_fnmsub_pd _mm_mask3_fnmsub_pd _mm_maskz_fnmsub_pd _mm_mask_fnmsub_ps _mm_mask3_fnmsub_ps _mm_maskz_fnmsub_ps _mm256_mask_fnmsub_pd _mm256_mask3_fnmsub_pd _mm256_maskz_fnmsub_pd _mm256_mask_fnmsub_ps _mm256_mask3_fnmsub_ps _mm256_maskz_fnmsub_ps Patch is 121.35 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/156385.diff 4 Files Affected:
diff --git a/clang/lib/Headers/avx512fintrin.h b/clang/lib/Headers/avx512fintrin.h
index e23b1c0381ab1..741ce26aaa043 100644
--- a/clang/lib/Headers/avx512fintrin.h
+++ b/clang/lib/Headers/avx512fintrin.h
@@ -2502,124 +2502,136 @@ _mm512_maskz_div_ps(__mmask16 __U, __m512 __A, __m512 __B) {
(__mmask8)(U), (int)(R)))
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_fmadd_pd(__m512d __A, __m512d __B, __m512d __C)
{
- return (__m512d) __builtin_ia32_vfmaddpd512_mask ((__v8df) __A,
- (__v8df) __B,
- (__v8df) __C,
- (__mmask8) -1,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512d) __builtin_elementwise_fma((__v8df) __A, (__v8df) __B, (__v8df) __C);
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_fmadd_pd(__m512d __A, __mmask8 __U, __m512d __B, __m512d __C)
{
- return (__m512d) __builtin_ia32_vfmaddpd512_mask ((__v8df) __A,
- (__v8df) __B,
- (__v8df) __C,
- (__mmask8) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512d)__builtin_ia32_selectpd_512(
+ (__mmask8) __U,
+ __builtin_elementwise_fma((__v8df) __A, (__v8df) __B, (__v8df) __C),
+ (__v8df) __A);
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask3_fmadd_pd(__m512d __A, __m512d __B, __m512d __C, __mmask8 __U)
{
- return (__m512d) __builtin_ia32_vfmaddpd512_mask3 ((__v8df) __A,
- (__v8df) __B,
- (__v8df) __C,
- (__mmask8) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512d)__builtin_ia32_selectpd_512(
+ (__mmask8) __U,
+ __builtin_elementwise_fma((__v8df) __A, (__v8df) __B, (__v8df) __C),
+ (__v8df) __C);
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_fmadd_pd(__mmask8 __U, __m512d __A, __m512d __B, __m512d __C)
{
- return (__m512d) __builtin_ia32_vfmaddpd512_maskz ((__v8df) __A,
- (__v8df) __B,
- (__v8df) __C,
- (__mmask8) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512d)__builtin_ia32_selectpd_512(
+ (__mmask8) __U,
+ __builtin_elementwise_fma((__v8df) __A, (__v8df) __B, (__v8df) __C),
+ (__v8df) _mm512_setzero_pd());
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_fmsub_pd(__m512d __A, __m512d __B, __m512d __C)
{
- return (__m512d) __builtin_ia32_vfmaddpd512_mask ((__v8df) __A,
- (__v8df) __B,
- -(__v8df) __C,
- (__mmask8) -1,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512d) __builtin_elementwise_fma((__v8df) __A, (__v8df) __B, -(__v8df) __C);
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_fmsub_pd(__m512d __A, __mmask8 __U, __m512d __B, __m512d __C)
{
- return (__m512d) __builtin_ia32_vfmaddpd512_mask ((__v8df) __A,
- (__v8df) __B,
- -(__v8df) __C,
- (__mmask8) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512d)__builtin_ia32_selectpd_512(
+ (__mmask8) __U,
+ __builtin_elementwise_fma((__v8df) __A, (__v8df) __B, -(__v8df) __C),
+ (__v8df) __A);
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask3_fmsub_pd(__m512d __A, __m512d __B, __m512d __C, __mmask8 __U)
+{
+ return (__m512d)__builtin_ia32_selectpd_512(
+ (__mmask8) __U,
+ __builtin_elementwise_fma((__v8df) __A, (__v8df) __B, -(__v8df) __C),
+ (__v8df) __C);
+}
+
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_fmsub_pd(__mmask8 __U, __m512d __A, __m512d __B, __m512d __C)
{
- return (__m512d) __builtin_ia32_vfmaddpd512_maskz ((__v8df) __A,
- (__v8df) __B,
- -(__v8df) __C,
- (__mmask8) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512d)__builtin_ia32_selectpd_512(
+ (__mmask8) __U,
+ __builtin_elementwise_fma((__v8df) __A, (__v8df) __B, -(__v8df) __C),
+ (__v8df) _mm512_setzero_pd());
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_fnmadd_pd(__m512d __A, __m512d __B, __m512d __C)
{
- return (__m512d) __builtin_ia32_vfmaddpd512_mask ((__v8df) __A,
- -(__v8df) __B,
- (__v8df) __C,
- (__mmask8) -1,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512d)__builtin_elementwise_fma(-(__v8df) __A, (__v8df) __B, (__v8df) __C);
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_fnmadd_pd(__m512d __A, __mmask8 __U, __m512d __B, __m512d __C)
+{
+ return (__m512d)__builtin_ia32_selectpd_512(
+ (__mmask8) __U,
+ __builtin_elementwise_fma(-(__v8df) __A, (__v8df) __B, (__v8df) __C),
+ (__v8df) __A);
+}
+
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask3_fnmadd_pd(__m512d __A, __m512d __B, __m512d __C, __mmask8 __U)
{
- return (__m512d) __builtin_ia32_vfmaddpd512_mask3 (-(__v8df) __A,
- (__v8df) __B,
- (__v8df) __C,
- (__mmask8) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512d)__builtin_ia32_selectpd_512(
+ (__mmask8) __U,
+ __builtin_elementwise_fma(-(__v8df) __A, (__v8df) __B, (__v8df) __C),
+ (__v8df) __C);
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_fnmadd_pd(__mmask8 __U, __m512d __A, __m512d __B, __m512d __C)
{
- return (__m512d) __builtin_ia32_vfmaddpd512_maskz (-(__v8df) __A,
- (__v8df) __B,
- (__v8df) __C,
- (__mmask8) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512d)__builtin_ia32_selectpd_512(
+ (__mmask8) __U,
+ __builtin_elementwise_fma(-(__v8df) __A, (__v8df) __B, (__v8df) __C),
+ (__v8df) _mm512_setzero_pd());
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_fnmsub_pd(__m512d __A, __m512d __B, __m512d __C)
{
- return (__m512d) __builtin_ia32_vfmaddpd512_mask ((__v8df) __A,
- -(__v8df) __B,
- -(__v8df) __C,
- (__mmask8) -1,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512d) __builtin_elementwise_fma(-(__v8df) __A, (__v8df) __B, -(__v8df) __C);
}
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_fnmsub_pd(__m512d __A, __mmask8 __U, __m512d __B, __m512d __C)
+{
+ return (__m512d)__builtin_ia32_selectpd_512(
+ (__mmask8) __U,
+ __builtin_elementwise_fma(-(__v8df) __A, (__v8df) __B, -(__v8df) __C),
+ (__v8df) __A);
+}
+
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask3_fnmsub_pd(__m512d __A, __m512d __B, __m512d __C, __mmask8 __U)
+{
+ return (__m512d)__builtin_ia32_selectpd_512(
+ (__mmask8) __U,
+ __builtin_elementwise_fma(-(__v8df) __A, (__v8df) __B, -(__v8df) __C),
+ (__v8df) __C);
+}
+
+static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_fnmsub_pd(__mmask8 __U, __m512d __A, __m512d __B, __m512d __C)
{
- return (__m512d) __builtin_ia32_vfmaddpd512_maskz (-(__v8df) __A,
- (__v8df) __B,
- -(__v8df) __C,
- (__mmask8) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512d)__builtin_ia32_selectpd_512(
+ (__mmask8) __U,
+ __builtin_elementwise_fma(-(__v8df) __A, (__v8df) __B, -(__v8df) __C),
+ (__v8df) _mm512_setzero_pd());
}
#define _mm512_fmadd_round_ps(A, B, C, R) \
@@ -2706,124 +2718,136 @@ _mm512_maskz_fnmsub_pd(__mmask8 __U, __m512d __A, __m512d __B, __m512d __C)
(__mmask16)(U), (int)(R)))
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_fmadd_ps(__m512 __A, __m512 __B, __m512 __C)
{
- return (__m512) __builtin_ia32_vfmaddps512_mask ((__v16sf) __A,
- (__v16sf) __B,
- (__v16sf) __C,
- (__mmask16) -1,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512 ) __builtin_elementwise_fma((__v16sf)__A, (__v16sf)__B, (__v16sf)__C);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_fmadd_ps(__m512 __A, __mmask16 __U, __m512 __B, __m512 __C)
{
- return (__m512) __builtin_ia32_vfmaddps512_mask ((__v16sf) __A,
- (__v16sf) __B,
- (__v16sf) __C,
- (__mmask16) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512) __builtin_ia32_selectps_512(
+ (__mmask16)__U,
+ __builtin_elementwise_fma((__v16sf)__A, (__v16sf)__B, (__v16sf)__C),
+ (__v16sf)__A);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask3_fmadd_ps(__m512 __A, __m512 __B, __m512 __C, __mmask16 __U)
{
- return (__m512) __builtin_ia32_vfmaddps512_mask3 ((__v16sf) __A,
- (__v16sf) __B,
- (__v16sf) __C,
- (__mmask16) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512) __builtin_ia32_selectps_512(
+ (__mmask16)__U,
+ __builtin_elementwise_fma((__v16sf)__A, (__v16sf)__B, (__v16sf)__C),
+ (__v16sf)__C);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_fmadd_ps(__mmask16 __U, __m512 __A, __m512 __B, __m512 __C)
{
- return (__m512) __builtin_ia32_vfmaddps512_maskz ((__v16sf) __A,
- (__v16sf) __B,
- (__v16sf) __C,
- (__mmask16) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512) __builtin_ia32_selectps_512(
+ (__mmask16)__U,
+ __builtin_elementwise_fma((__v16sf)__A, (__v16sf)__B, (__v16sf)__C),
+ (__v16sf)_mm512_setzero_ps());
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_fmsub_ps(__m512 __A, __m512 __B, __m512 __C)
{
- return (__m512) __builtin_ia32_vfmaddps512_mask ((__v16sf) __A,
- (__v16sf) __B,
- -(__v16sf) __C,
- (__mmask16) -1,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512) __builtin_elementwise_fma((__v16sf)__A, (__v16sf)__B, -(__v16sf)__C);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_fmsub_ps(__m512 __A, __mmask16 __U, __m512 __B, __m512 __C)
{
- return (__m512) __builtin_ia32_vfmaddps512_mask ((__v16sf) __A,
- (__v16sf) __B,
- -(__v16sf) __C,
- (__mmask16) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512) __builtin_ia32_selectps_512(
+ (__mmask16)__U,
+ __builtin_elementwise_fma((__v16sf)__A, (__v16sf)__B, -(__v16sf)__C),
+ (__v16sf)__A);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask3_fmsub_ps(__m512 __A, __m512 __B, __m512 __C, __mmask16 __U)
+{
+ return (__m512) __builtin_ia32_selectps_512(
+ (__mmask16)__U,
+ __builtin_elementwise_fma((__v16sf)__A, (__v16sf)__B, -(__v16sf)__C),
+ (__v16sf)__C);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_fmsub_ps(__mmask16 __U, __m512 __A, __m512 __B, __m512 __C)
{
- return (__m512) __builtin_ia32_vfmaddps512_maskz ((__v16sf) __A,
- (__v16sf) __B,
- -(__v16sf) __C,
- (__mmask16) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512) __builtin_ia32_selectps_512(
+ (__mmask16)__U,
+ __builtin_elementwise_fma((__v16sf)__A, (__v16sf)__B, -(__v16sf)__C),
+ (__v16sf)_mm512_setzero_ps());
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_fnmadd_ps(__m512 __A, __m512 __B, __m512 __C)
{
- return (__m512) __builtin_ia32_vfmaddps512_mask ((__v16sf) __A,
- -(__v16sf) __B,
- (__v16sf) __C,
- (__mmask16) -1,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512) __builtin_elementwise_fma(-(__v16sf)__A, (__v16sf)__B, (__v16sf)__C);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_fnmadd_ps(__m512 __A, __mmask16 __U, __m512 __B, __m512 __C)
+{
+ return (__m512) __builtin_ia32_selectps_512(
+ (__mmask16)__U,
+ __builtin_elementwise_fma(-(__v16sf)__A, (__v16sf)__B, (__v16sf)__C),
+ (__v16sf)__A);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask3_fnmadd_ps(__m512 __A, __m512 __B, __m512 __C, __mmask16 __U)
{
- return (__m512) __builtin_ia32_vfmaddps512_mask3 (-(__v16sf) __A,
- (__v16sf) __B,
- (__v16sf) __C,
- (__mmask16) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512) __builtin_ia32_selectps_512(
+ (__mmask16)__U,
+ __builtin_elementwise_fma(-(__v16sf)__A, (__v16sf)__B, (__v16sf)__C),
+ (__v16sf)__C);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_fnmadd_ps(__mmask16 __U, __m512 __A, __m512 __B, __m512 __C)
{
- return (__m512) __builtin_ia32_vfmaddps512_maskz (-(__v16sf) __A,
- (__v16sf) __B,
- (__v16sf) __C,
- (__mmask16) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512) __builtin_ia32_selectps_512(
+ (__mmask16)__U,
+ __builtin_elementwise_fma(-(__v16sf)__A, (__v16sf)__B, (__v16sf)__C),
+ (__v16sf)_mm512_setzero_ps());
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_fnmsub_ps(__m512 __A, __m512 __B, __m512 __C)
{
- return (__m512) __builtin_ia32_vfmaddps512_mask ((__v16sf) __A,
- -(__v16sf) __B,
- -(__v16sf) __C,
- (__mmask16) -1,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512) __builtin_elementwise_fma(-(__v16sf)__A, (__v16sf)__B, -(__v16sf)__C);
}
-static __inline__ __m512 __DEFAULT_FN_ATTRS512
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask_fnmsub_ps(__m512 __A, __mmask16 __U, __m512 __B, __m512 __C)
+{
+ return (__m512) __builtin_ia32_selectps_512(
+ (__mmask16)__U,
+ __builtin_elementwise_fma(-(__v16sf)__A, (__v16sf)__B, -(__v16sf)__C),
+ (__v16sf)__A);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
+_mm512_mask3_fnmsub_ps(__m512 __A, __m512 __B, __m512 __C, __mmask16 __U)
+{
+ return (__m512) __builtin_ia32_selectps_512(
+ (__mmask16)__U,
+ __builtin_elementwise_fma(-(__v16sf)__A, (__v16sf)__B, -(__v16sf)__C),
+ (__v16sf)__C);
+}
+
+static __inline__ __m512 __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_fnmsub_ps(__mmask16 __U, __m512 __A, __m512 __B, __m512 __C)
{
- return (__m512) __builtin_ia32_vfmaddps512_maskz (-(__v16sf) __A,
- (__v16sf) __B,
- -(__v16sf) __C,
- (__mmask16) __U,
- _MM_FROUND_CUR_DIRECTION);
+ return (__m512) __builtin_ia32_selectps_512(
+ (__mmask16)__U,
+ __builtin_elementwise_fma(-(__v16sf)__A, (__v16sf)__B, -(__v16sf)__C),
+ (__v16sf)_mm512_setzero_ps());
}
#define _mm512_fmaddsub_round_pd(A, B, C, R) \
@@ -3071,15 +3095,6 @@ _mm512_maskz_fmsubadd_ps(__mmask16 __U, __m512 __A, __m512 __B, __m512 __C)
(__mmask8)(U), (int)(R)))
-static __inline__ __m512d __DEFAULT_FN_ATTRS512
-_mm512_mask3_fmsub_pd(__m512d __A, __m512d __B, __m512d __C, __mmask8 __U)
-{
- return (__m512d)__builtin_ia32_vfmsubpd512_mask3 ((__v8df) __A,
- (__v8df) __B,
- (__v8df) __C,
- (__mmask8) __U,
- _MM_FROUND_CUR_DIRECTION);
-}
#define _mm512_mask3_fmsub_round_ps(A, B, C, U, R) \
((__m512)__builtin_ia32_vfmsubps512_mask3((__v16sf)(__m512)(A), \
@@ -3087,16 +3102,6 @@ _mm512_mask3_fmsub_pd(__m512d __A, __m512d __B, __m512d __C, __mmask8 __U)
...
[truncated]
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
e546bec to
b9bc673
Compare
RKSimon
requested changes
Sep 2, 2025
RKSimon
reviewed
Sep 3, 2025
RKSimon
approved these changes
Sep 3, 2025
|
Indeed. I considered rearranging them, and the tests too, but chose to leave it for another time to not make this patch even longer. |
|
Thanks for the review and merge! |
RKSimon
pushed a commit
that referenced
this pull request
Sep 4, 2025
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.
Fixes #155265
Add constexpr support for the following:
_mm512_fmadd_pd _mm512_mask_fmadd_pd _mm512_mask3_fmadd_pd _mm512_maskz_fmadd_pd _mm512_fmadd_ps _mm512_mask_fmadd_ps _mm512_mask3_fmadd_ps _mm512_maskz_fmadd_ps _mm_mask_fmadd_pd _mm_mask3_fmadd_pd _mm_maskz_fmadd_pd _mm_mask_fmadd_ps _mm_mask3_fmadd_ps _mm_maskz_fmadd_ps _mm256_mask_fmadd_pd _mm256_mask3_fmadd_pd _mm256_maskz_fmadd_pd _mm256_mask_fmadd_ps _mm256_mask3_fmadd_ps _mm256_maskz_fmadd_ps
_mm512_fmsub_pd _mm512_mask_fmsub_pd _mm512_mask3_fmsub_pd _mm512_maskz_fmsub_pd _mm512_fmsub_ps _mm512_mask_fmsub_ps _mm512_mask3_fmsub_ps _mm512_maskz_fmsub_ps _mm_mask_fmsub_pd _mm_mask3_fmsub_pd _mm_maskz_fmsub_pd _mm_mask_fmsub_ps _mm_mask3_fmsub_ps _mm_maskz_fmsub_ps _mm256_mask_fmsub_pd _mm256_mask3_fmsub_pd _mm256_maskz_fmsub_pd _mm256_mask_fmsub_ps _mm256_mask3_fmsub_ps _mm256_maskz_fmsub_ps
_mm512_fnmadd_pd _mm512_mask_fnmadd_pd _mm512_mask3_fnmadd_pd _mm512_maskz_fnmadd_pd _mm512_fnmsub_pd _mm512_mask_fnmsub_pd _mm512_mask3_fnmsub_pd _mm512_maskz_fnmsub_pd _mm_mask_fnmadd_pd _mm_mask3_fnmadd_pd _mm_maskz_fnmadd_pd _mm_mask_fnmadd_ps _mm_mask3_fnmadd_ps _mm_maskz_fnmadd_ps _mm256_mask_fnmadd_pd _mm256_mask3_fnmadd_pd _mm256_maskz_fnmadd_pd _mm256_mask_fnmadd_ps _mm256_mask3_fnmadd_ps _mm256_maskz_fnmadd_ps
_mm512_fnmadd_ps _mm512_mask_fnmadd_ps _mm512_mask3_fnmadd_ps _mm512_maskz_fnmadd_ps _mm512_fnmsub_ps _mm512_mask_fnmsub_ps _mm512_mask3_fnmsub_ps _mm512_maskz_fnmsub_ps _mm_mask_fnmsub_pd _mm_mask3_fnmsub_pd _mm_maskz_fnmsub_pd _mm_mask_fnmsub_ps _mm_mask3_fnmsub_ps _mm_maskz_fnmsub_ps _mm256_mask_fnmsub_pd _mm256_mask3_fnmsub_pd _mm256_maskz_fnmsub_pd _mm256_mask_fnmsub_ps _mm256_mask3_fnmsub_ps _mm256_maskz_fnmsub_ps