i386: Enable AVX512 memory broadcast for FNMADD

Many AVX512 vector operations can broadcast from a scalar memory source.
This patch enables memory broadcast for FNMADD operations.  In order to
support AVX512 memory broadcast for FNMADD, FNMADD builtin functions are
also added, instead of passing the negated value to FMA builtin functions.

gcc/

	PR target/72782
	* config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
	__builtin_ia32_vfnmaddpd512_mask.
	(_mm512_mask_fnmadd_round_pd): Likewise.
	(_mm512_fnmadd_pd): Likewise.
	(_mm512_mask_fnmadd_pd): Likewise.
	(_mm512_maskz_fnmadd_round_pd): Use
	__builtin_ia32_vfnmaddpd512_maskz.
	(_mm512_maskz_fnmadd_pd): Likewise.
	(_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
	(_mm512_mask_fnmadd_round_ps): Likewise.
	(_mm512_fnmadd_ps): Likewise.
	(_mm512_mask_fnmadd_ps): Likewise.
	(_mm512_maskz_fnmadd_round_ps): Use
	__builtin_ia32_vfnmaddps512_maskz.
	(_mm512_maskz_fnmadd_ps): Likewise.
	* config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
	__builtin_ia32_vfnmaddpd256_mask.
	(_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
	(_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
	(_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
	(_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
	(_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
	(_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
	(_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
	(_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
	* config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
	__builtin_ia32_vfnmaddpd.
	(_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
	(_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
	(_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
	(_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
	(_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
	* config/i386/i386-builtin.def: Add
	__builtin_ia32_vfnmaddpd256_mask,
	__builtin_ia32_vfnmaddpd256_maskz,
	__builtin_ia32_vfnmaddpd128_mask,
	__builtin_ia32_vfnmaddpd128_maskz,
	__builtin_ia32_vfnmaddps256_mask,
	__builtin_ia32_vfnmaddps256_maskz,
	__builtin_ia32_vfnmaddps128_mask,
	__builtin_ia32_vfnmaddps128_maskz,
	__builtin_ia32_vfnmaddpd512_mask,
	__builtin_ia32_vfnmaddpd512_maskz,
	__builtin_ia32_vfnmaddps512_mask,
	__builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
	__builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
	__builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
	__builtin_ia32_vfnmaddpd256.
	* config/i386/sse.md (fma4i_fnmadd_<mode>): New.
	(<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
	(*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
	Likewise.
	(*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
	Likewise.
	(*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
	Likewise.
	(fmai_vmfnmadd_<mode><round_name>): Likewise.

gcc/testsuite/

	PR target/72782
	* gcc.target/i386/avx512f-fnmadd-df-zmm-1.c: New test.
	* gcc.target/i386/avx512f-fnmadd-sf-zmm-1.c: Likewise.
	* gcc.target/i386/avx512f-fnmadd-sf-zmm-2.c: Likewise.
	* gcc.target/i386/avx512f-fnmadd-sf-zmm-3.c: Likewise.
	* gcc.target/i386/avx512f-fnmadd-sf-zmm-4.c: Likewise.
	* gcc.target/i386/avx512f-fnmadd-sf-zmm-5.c: Likewise.
	* gcc.target/i386/avx512f-fnmadd-sf-zmm-6.c: Likewise.
	* gcc.target/i386/avx512f-fnmadd-sf-zmm-7.c: Likewise.
	* gcc.target/i386/avx512f-fnmadd-sf-zmm-8.c: Likewise.
	* gcc.target/i386/avx512vl-fnmadd-sf-xmm-1.c: Likewise.
	* gcc.target/i386/avx512vl-fnmadd-sf-ymm-1.c: Likewise.

From-SVN: r265357

5 files changed, 203 insertions, 106 deletions

diff --git a/gcc/config/i386/avx512fintrin.h b/gcc/config/i386/avx512fintrin.h
index c0c8fa1..1445e9e 100644
--- a/gcc/config/i386/avx512fintrin.h
+++ b/gcc/config/i386/avx512fintrin.h
@@ -3613,10 +3613,10 @@ extern __inline __m512d
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm512_fnmadd_round_pd (__m512d __A, __m512d __B, __m512d __C, const int __R)
 {
-  return (__m512d) __builtin_ia32_vfmaddpd512_mask (-(__v8df) __A,
-						    (__v8df) __B,
-						    (__v8df) __C,
-						    (__mmask8) -1, __R);
+  return (__m512d) __builtin_ia32_vfnmaddpd512_mask ((__v8df) __A,
+						     (__v8df) __B,
+						     (__v8df) __C,
+						     (__mmask8) -1, __R);
 }
 
 extern __inline __m512d
@@ -3635,10 +3635,10 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm512_mask3_fnmadd_round_pd (__m512d __A, __m512d __B, __m512d __C,
 			      __mmask8 __U, const int __R)
 {
-  return (__m512d) __builtin_ia32_vfmaddpd512_mask3 (-(__v8df) __A,
-						     (__v8df) __B,
-						     (__v8df) __C,
-						     (__mmask8) __U, __R);
+  return (__m512d) __builtin_ia32_vfnmaddpd512_mask3 ((__v8df) __A,
+						      (__v8df) __B,
+						      (__v8df) __C,
+						      (__mmask8) __U, __R);
 }
 
 extern __inline __m512d
@@ -3646,20 +3646,20 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm512_maskz_fnmadd_round_pd (__mmask8 __U, __m512d __A, __m512d __B,
 			      __m512d __C, const int __R)
 {
-  return (__m512d) __builtin_ia32_vfmaddpd512_maskz (-(__v8df) __A,
-						     (__v8df) __B,
-						     (__v8df) __C,
-						     (__mmask8) __U, __R);
+  return (__m512d) __builtin_ia32_vfnmaddpd512_maskz ((__v8df) __A,
+						      (__v8df) __B,
+						      (__v8df) __C,
+						      (__mmask8) __U, __R);
 }
 
 extern __inline __m512
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm512_fnmadd_round_ps (__m512 __A, __m512 __B, __m512 __C, const int __R)
 {
-  return (__m512) __builtin_ia32_vfmaddps512_mask (-(__v16sf) __A,
-						   (__v16sf) __B,
-						   (__v16sf) __C,
-						   (__mmask16) -1, __R);
+  return (__m512) __builtin_ia32_vfnmaddps512_mask ((__v16sf) __A,
+						    (__v16sf) __B,
+						    (__v16sf) __C,
+						    (__mmask16) -1, __R);
 }
 
 extern __inline __m512
@@ -3678,10 +3678,10 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm512_mask3_fnmadd_round_ps (__m512 __A, __m512 __B, __m512 __C,
 			      __mmask16 __U, const int __R)
 {
-  return (__m512) __builtin_ia32_vfmaddps512_mask3 (-(__v16sf) __A,
-						    (__v16sf) __B,
-						    (__v16sf) __C,
-						    (__mmask16) __U, __R);
+  return (__m512) __builtin_ia32_vfnmaddps512_mask3 ((__v16sf) __A,
+						     (__v16sf) __B,
+						     (__v16sf) __C,
+						     (__mmask16) __U, __R);
 }
 
 extern __inline __m512
@@ -3689,10 +3689,10 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm512_maskz_fnmadd_round_ps (__mmask16 __U, __m512 __A, __m512 __B,
 			      __m512 __C, const int __R)
 {
-  return (__m512) __builtin_ia32_vfmaddps512_maskz (-(__v16sf) __A,
-						    (__v16sf) __B,
-						    (__v16sf) __C,
-						    (__mmask16) __U, __R);
+  return (__m512) __builtin_ia32_vfnmaddps512_maskz ((__v16sf) __A,
+						     (__v16sf) __B,
+						     (__v16sf) __C,
+						     (__mmask16) __U, __R);
 }
 
 extern __inline __m512d
@@ -3878,28 +3878,28 @@ _mm512_maskz_fnmsub_round_ps (__mmask16 __U, __m512 __A, __m512 __B,
     (__m512)__builtin_ia32_vfmaddsubps512_maskz(A, B, -(C), U, R)
 
 #define _mm512_fnmadd_round_pd(A, B, C, R)            \
-    (__m512d)__builtin_ia32_vfmaddpd512_mask(-(A), B, C, -1, R)
+    (__m512d)__builtin_ia32_vfnmaddpd512_mask(A, B, C, -1, R)
 
 #define _mm512_mask_fnmadd_round_pd(A, U, B, C, R)    \
-    (__m512d)__builtin_ia32_vfnmaddpd512_mask(-(A), B, C, U, R)
+    (__m512d)__builtin_ia32_vfnmaddpd512_mask(A, B, C, U, R)
 
 #define _mm512_mask3_fnmadd_round_pd(A, B, C, U, R)   \
-    (__m512d)__builtin_ia32_vfmaddpd512_mask3(-(A), B, C, U, R)
+    (__m512d)__builtin_ia32_vfnmaddpd512_mask3(A, B, C, U, R)
 
 #define _mm512_maskz_fnmadd_round_pd(U, A, B, C, R)   \
-    (__m512d)__builtin_ia32_vfmaddpd512_maskz(-(A), B, C, U, R)
+    (__m512d)__builtin_ia32_vfnmaddpd512_maskz(A, B, C, U, R)
 
 #define _mm512_fnmadd_round_ps(A, B, C, R)            \
-    (__m512)__builtin_ia32_vfmaddps512_mask(-(A), B, C, -1, R)
+    (__m512)__builtin_ia32_vfnmaddps512_mask(A, B, C, -1, R)
 
 #define _mm512_mask_fnmadd_round_ps(A, U, B, C, R)    \
-    (__m512)__builtin_ia32_vfnmaddps512_mask(-(A), B, C, U, R)
+    (__m512)__builtin_ia32_vfnmaddps512_mask(A, B, C, U, R)
 
 #define _mm512_mask3_fnmadd_round_ps(A, B, C, U, R)   \
-    (__m512)__builtin_ia32_vfmaddps512_mask3(-(A), B, C, U, R)
+    (__m512)__builtin_ia32_vfnmaddps512_mask3(A, B, C, U, R)
 
 #define _mm512_maskz_fnmadd_round_ps(U, A, B, C, R)   \
-    (__m512)__builtin_ia32_vfmaddps512_maskz(-(A), B, C, U, R)
+    (__m512)__builtin_ia32_vfnmaddps512_maskz(A, B, C, U, R)
 
 #define _mm512_fnmsub_round_pd(A, B, C, R)            \
     (__m512d)__builtin_ia32_vfmaddpd512_mask(-(A), B, -(C), -1, R)
@@ -12680,11 +12680,11 @@ extern __inline __m512d
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _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_ia32_vfnmaddpd512_mask ((__v8df) __A,
+						     (__v8df) __B,
+						     (__v8df) __C,
+						     (__mmask8) -1,
+						     _MM_FROUND_CUR_DIRECTION);
 }
 
 extern __inline __m512d
@@ -12702,33 +12702,33 @@ extern __inline __m512d
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _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_vfnmaddpd512_mask3 ((__v8df) __A,
+						      (__v8df) __B,
+						      (__v8df) __C,
+						      (__mmask8) __U,
+						      _MM_FROUND_CUR_DIRECTION);
 }
 
 extern __inline __m512d
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _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_vfnmaddpd512_maskz ((__v8df) __A,
+						      (__v8df) __B,
+						      (__v8df) __C,
+						      (__mmask8) __U,
+						      _MM_FROUND_CUR_DIRECTION);
 }
 
 extern __inline __m512
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _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_ia32_vfnmaddps512_mask ((__v16sf) __A,
+						    (__v16sf) __B,
+						    (__v16sf) __C,
+						    (__mmask16) -1,
+						    _MM_FROUND_CUR_DIRECTION);
 }
 
 extern __inline __m512
@@ -12746,22 +12746,22 @@ extern __inline __m512
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _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_vfnmaddps512_mask3 ((__v16sf) __A,
+						     (__v16sf) __B,
+						     (__v16sf) __C,
+						     (__mmask16) __U,
+						     _MM_FROUND_CUR_DIRECTION);
 }
 
 extern __inline __m512
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _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_vfnmaddps512_maskz ((__v16sf) __A,
+						     (__v16sf) __B,
+						     (__v16sf) __C,
+						     (__mmask16) __U,
+						     _MM_FROUND_CUR_DIRECTION);
 }
 
 extern __inline __m512d
diff --git a/gcc/config/i386/avx512vlintrin.h b/gcc/config/i386/avx512vlintrin.h
index fcc35c3..b46c38e 100644
--- a/gcc/config/i386/avx512vlintrin.h
+++ b/gcc/config/i386/avx512vlintrin.h
@@ -4525,10 +4525,10 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm256_mask3_fnmadd_pd (__m256d __A, __m256d __B, __m256d __C,
 			__mmask8 __U)
 {
-  return (__m256d) __builtin_ia32_vfmaddpd256_mask3 (-(__v4df) __A,
-						     (__v4df) __B,
-						     (__v4df) __C,
-						     (__mmask8) __U);
+  return (__m256d) __builtin_ia32_vfnmaddpd256_mask3 ((__v4df) __A,
+						      (__v4df) __B,
+						      (__v4df) __C,
+						      (__mmask8) __U);
 }
 
 extern __inline __m256d
@@ -4536,10 +4536,10 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm256_maskz_fnmadd_pd (__mmask8 __U, __m256d __A, __m256d __B,
 			__m256d __C)
 {
-  return (__m256d) __builtin_ia32_vfmaddpd256_maskz (-(__v4df) __A,
-						     (__v4df) __B,
-						     (__v4df) __C,
-						     (__mmask8) __U);
+  return (__m256d) __builtin_ia32_vfnmaddpd256_maskz ((__v4df) __A,
+						      (__v4df) __B,
+						      (__v4df) __C,
+						      (__mmask8) __U);
 }
 
 extern __inline __m128d
@@ -4558,10 +4558,10 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_mask3_fnmadd_pd (__m128d __A, __m128d __B, __m128d __C,
 		     __mmask8 __U)
 {
-  return (__m128d) __builtin_ia32_vfmaddpd128_mask3 (-(__v2df) __A,
-						     (__v2df) __B,
-						     (__v2df) __C,
-						     (__mmask8) __U);
+  return (__m128d) __builtin_ia32_vfnmaddpd128_mask3 ((__v2df) __A,
+						      (__v2df) __B,
+						      (__v2df) __C,
+						      (__mmask8) __U);
 }
 
 extern __inline __m128d
@@ -4569,10 +4569,10 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_maskz_fnmadd_pd (__mmask8 __U, __m128d __A, __m128d __B,
 		     __m128d __C)
 {
-  return (__m128d) __builtin_ia32_vfmaddpd128_maskz (-(__v2df) __A,
-						     (__v2df) __B,
-						     (__v2df) __C,
-						     (__mmask8) __U);
+  return (__m128d) __builtin_ia32_vfnmaddpd128_maskz ((__v2df) __A,
+						      (__v2df) __B,
+						      (__v2df) __C,
+						      (__mmask8) __U);
 }
 
 extern __inline __m256
@@ -4591,10 +4591,10 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm256_mask3_fnmadd_ps (__m256 __A, __m256 __B, __m256 __C,
 			__mmask8 __U)
 {
-  return (__m256) __builtin_ia32_vfmaddps256_mask3 (-(__v8sf) __A,
-						    (__v8sf) __B,
-						    (__v8sf) __C,
-						    (__mmask8) __U);
+  return (__m256) __builtin_ia32_vfnmaddps256_mask3 ((__v8sf) __A,
+						     (__v8sf) __B,
+						     (__v8sf) __C,
+						     (__mmask8) __U);
 }
 
 extern __inline __m256
@@ -4602,10 +4602,10 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm256_maskz_fnmadd_ps (__mmask8 __U, __m256 __A, __m256 __B,
 			__m256 __C)
 {
-  return (__m256) __builtin_ia32_vfmaddps256_maskz (-(__v8sf) __A,
-						    (__v8sf) __B,
-						    (__v8sf) __C,
-						    (__mmask8) __U);
+  return (__m256) __builtin_ia32_vfnmaddps256_maskz ((__v8sf) __A,
+						     (__v8sf) __B,
+						     (__v8sf) __C,
+						     (__mmask8) __U);
 }
 
 extern __inline __m128
@@ -4622,20 +4622,20 @@ extern __inline __m128
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_mask3_fnmadd_ps (__m128 __A, __m128 __B, __m128 __C, __mmask8 __U)
 {
-  return (__m128) __builtin_ia32_vfmaddps128_mask3 (-(__v4sf) __A,
-						    (__v4sf) __B,
-						    (__v4sf) __C,
-						    (__mmask8) __U);
+  return (__m128) __builtin_ia32_vfnmaddps128_mask3 ((__v4sf) __A,
+						     (__v4sf) __B,
+						     (__v4sf) __C,
+						     (__mmask8) __U);
 }
 
 extern __inline __m128
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_maskz_fnmadd_ps (__mmask8 __U, __m128 __A, __m128 __B, __m128 __C)
 {
-  return (__m128) __builtin_ia32_vfmaddps128_maskz (-(__v4sf) __A,
-						    (__v4sf) __B,
-						    (__v4sf) __C,
-						    (__mmask8) __U);
+  return (__m128) __builtin_ia32_vfnmaddps128_maskz ((__v4sf) __A,
+						     (__v4sf) __B,
+						     (__v4sf) __C,
+						     (__mmask8) __U);
 }
 
 extern __inline __m256d
diff --git a/gcc/config/i386/fmaintrin.h b/gcc/config/i386/fmaintrin.h
index 2eddd89..0a2f4a7 100644
--- a/gcc/config/i386/fmaintrin.h
+++ b/gcc/config/i386/fmaintrin.h
@@ -134,48 +134,48 @@ extern __inline __m128d
 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _mm_fnmadd_pd (__m128d __A, __m128d __B, __m128d __C)
 {
-  return (__m128d)__builtin_ia32_vfmaddpd (-(__v2df)__A, (__v2df)__B,
-                                           (__v2df)__C);
+  return (__m128d)__builtin_ia32_vfnmaddpd ((__v2df)__A, (__v2df)__B,
+					    (__v2df)__C);
 }
 
 extern __inline __m256d
 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _mm256_fnmadd_pd (__m256d __A, __m256d __B, __m256d __C)
 {
-  return (__m256d)__builtin_ia32_vfmaddpd256 (-(__v4df)__A, (__v4df)__B,
-                                              (__v4df)__C);
+  return (__m256d)__builtin_ia32_vfnmaddpd256 ((__v4df)__A, (__v4df)__B,
+					       (__v4df)__C);
 }
 
 extern __inline __m128
 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _mm_fnmadd_ps (__m128 __A, __m128 __B, __m128 __C)
 {
-  return (__m128)__builtin_ia32_vfmaddps (-(__v4sf)__A, (__v4sf)__B,
-                                          (__v4sf)__C);
+  return (__m128)__builtin_ia32_vfnmaddps ((__v4sf)__A, (__v4sf)__B,
+					   (__v4sf)__C);
 }
 
 extern __inline __m256
 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _mm256_fnmadd_ps (__m256 __A, __m256 __B, __m256 __C)
 {
-  return (__m256)__builtin_ia32_vfmaddps256 (-(__v8sf)__A, (__v8sf)__B,
-                                             (__v8sf)__C);
+  return (__m256)__builtin_ia32_vfnmaddps256 ((__v8sf)__A, (__v8sf)__B,
+					      (__v8sf)__C);
 }
 
 extern __inline __m128d
 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _mm_fnmadd_sd (__m128d __A, __m128d __B, __m128d __C)
 {
-  return (__m128d)__builtin_ia32_vfmaddsd3 ((__v2df)__A, -(__v2df)__B,
-                                            (__v2df)__C);
+  return (__m128d)__builtin_ia32_vfnmaddsd3 ((__v2df)__A, (__v2df)__B,
+					     (__v2df)__C);
 }
 
 extern __inline __m128
 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
 _mm_fnmadd_ss (__m128 __A, __m128 __B, __m128 __C)
 {
-  return (__m128)__builtin_ia32_vfmaddss3 ((__v4sf)__A, -(__v4sf)__B,
-                                           (__v4sf)__C);
+  return (__m128)__builtin_ia32_vfnmaddss3 ((__v4sf)__A, (__v4sf)__B,
+					    (__v4sf)__C);
 }
 
 extern __inline __m128d
diff --git a/gcc/config/i386/i386-builtin.def b/gcc/config/i386/i386-builtin.def
index f5b5e56..74343db 100644
--- a/gcc/config/i386/i386-builtin.def
+++ b/gcc/config/i386/i386-builtin.def
@@ -1916,9 +1916,19 @@ BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fmsub_v4sf_mask, "__builtin_i
 BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fmsub_v4sf_mask3, "__builtin_ia32_vfmsubps128_mask3", IX86_BUILTIN_VFMSUBPS128_MASK3, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF_V4SF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fmsub_v4sf_maskz, "__builtin_ia32_vfmsubps128_maskz", IX86_BUILTIN_VFMSUBPS128_MASKZ, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF_V4SF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmadd_v4df_mask, "__builtin_ia32_vfnmaddpd256_mask", IX86_BUILTIN_VFNMADDPD256_MASK, UNKNOWN, (int) V4DF_FTYPE_V4DF_V4DF_V4DF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmadd_v4df_mask3, "__builtin_ia32_vfnmaddpd256_mask3", IX86_BUILTIN_VFNMADDPD256_MASK3, UNKNOWN, (int) V4DF_FTYPE_V4DF_V4DF_V4DF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmadd_v4df_maskz,
+"__builtin_ia32_vfnmaddpd256_maskz", IX86_BUILTIN_VFNMADDPD256_MASKZ, UNKNOWN, (int) V4DF_FTYPE_V4DF_V4DF_V4DF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmadd_v2df_mask, "__builtin_ia32_vfnmaddpd128_mask", IX86_BUILTIN_VFNMADDPD128_MASK, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF_V2DF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmadd_v2df_mask3, "__builtin_ia32_vfnmaddpd128_mask3", IX86_BUILTIN_VFNMADDPD128_MASK3, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF_V2DF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmadd_v2df_maskz,
+"__builtin_ia32_vfnmaddpd128_maskz", IX86_BUILTIN_VFNMADDPD128_MASKZ, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF_V2DF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmadd_v8sf_mask, "__builtin_ia32_vfnmaddps256_mask", IX86_BUILTIN_VFNMADDPS256_MASK, UNKNOWN, (int) V8SF_FTYPE_V8SF_V8SF_V8SF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmadd_v8sf_mask3, "__builtin_ia32_vfnmaddps256_mask3", IX86_BUILTIN_VFNMADDPS256_MASK3, UNKNOWN, (int) V8SF_FTYPE_V8SF_V8SF_V8SF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmadd_v8sf_maskz, "__builtin_ia32_vfnmaddps256_maskz", IX86_BUILTIN_VFNMADDPS256_MASKZ, UNKNOWN, (int) V8SF_FTYPE_V8SF_V8SF_V8SF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmadd_v4sf_mask, "__builtin_ia32_vfnmaddps128_mask", IX86_BUILTIN_VFNMADDPS128_MASK, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF_V4SF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmadd_v4sf_mask3, "__builtin_ia32_vfnmaddps128_mask3", IX86_BUILTIN_VFNMADDPS128_MASK3, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF_V4SF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmadd_v4sf_maskz, "__builtin_ia32_vfnmaddps128_maskz", IX86_BUILTIN_VFNMADDPS128_MASKZ, UNKNOWN, (int) V4SF_FTYPE_V4SF_V4SF_V4SF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmsub_v4df_mask, "__builtin_ia32_vfnmsubpd256_mask", IX86_BUILTIN_VFNMSUBPD256_MASK, UNKNOWN, (int) V4DF_FTYPE_V4DF_V4DF_V4DF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmsub_v4df_mask3, "__builtin_ia32_vfnmsubpd256_mask3", IX86_BUILTIN_VFNMSUBPD256_MASK3, UNKNOWN, (int) V4DF_FTYPE_V4DF_V4DF_V4DF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, CODE_FOR_avx512vl_fnmsub_v2df_mask, "__builtin_ia32_vfnmsubpd128_mask", IX86_BUILTIN_VFNMSUBPD128_MASK, UNKNOWN, (int) V2DF_FTYPE_V2DF_V2DF_V2DF_UQI)
@@ -2783,7 +2793,11 @@ BDESC (OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fmsub_v16sf_mask_round, "__buil
 BDESC (OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fmsub_v16sf_mask3_round, "__builtin_ia32_vfmsubps512_mask3", IX86_BUILTIN_VFMSUBPS512_MASK3, UNKNOWN, (int) V16SF_FTYPE_V16SF_V16SF_V16SF_HI_INT)
 BDESC (OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fmsub_v16sf_maskz_round, "__builtin_ia32_vfmsubps512_maskz", IX86_BUILTIN_VFMSUBPS512_MASKZ, UNKNOWN, (int) V16SF_FTYPE_V16SF_V16SF_V16SF_HI_INT)
 BDESC (OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fnmadd_v8df_mask_round, "__builtin_ia32_vfnmaddpd512_mask", IX86_BUILTIN_VFNMADDPD512_MASK, UNKNOWN, (int) V8DF_FTYPE_V8DF_V8DF_V8DF_UQI_INT)
+BDESC (OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fnmadd_v8df_mask3_round, "__builtin_ia32_vfnmaddpd512_mask3", IX86_BUILTIN_VFNMADDPD512_MASK3, UNKNOWN, (int) V8DF_FTYPE_V8DF_V8DF_V8DF_UQI_INT)
+BDESC (OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fnmadd_v8df_maskz_round, "__builtin_ia32_vfnmaddpd512_maskz", IX86_BUILTIN_VFNMADDPD512_MASKZ, UNKNOWN, (int) V8DF_FTYPE_V8DF_V8DF_V8DF_UQI_INT)
 BDESC (OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fnmadd_v16sf_mask_round, "__builtin_ia32_vfnmaddps512_mask", IX86_BUILTIN_VFNMADDPS512_MASK, UNKNOWN, (int) V16SF_FTYPE_V16SF_V16SF_V16SF_HI_INT)
+BDESC (OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fnmadd_v16sf_mask3_round, "__builtin_ia32_vfnmaddps512_mask3", IX86_BUILTIN_VFNMADDPS512_MASK3, UNKNOWN, (int) V16SF_FTYPE_V16SF_V16SF_V16SF_HI_INT)
+BDESC (OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fnmadd_v16sf_maskz_round, "__builtin_ia32_vfnmaddps512_maskz", IX86_BUILTIN_VFNMADDPS512_MASKZ, UNKNOWN, (int) V16SF_FTYPE_V16SF_V16SF_V16SF_HI_INT)
 BDESC (OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fnmsub_v8df_mask_round, "__builtin_ia32_vfnmsubpd512_mask", IX86_BUILTIN_VFNMSUBPD512_MASK, UNKNOWN, (int) V8DF_FTYPE_V8DF_V8DF_V8DF_UQI_INT)
 BDESC (OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fnmsub_v8df_mask3_round, "__builtin_ia32_vfnmsubpd512_mask3", IX86_BUILTIN_VFNMSUBPD512_MASK3, UNKNOWN, (int) V8DF_FTYPE_V8DF_V8DF_V8DF_UQI_INT)
 BDESC (OPTION_MASK_ISA_AVX512F, CODE_FOR_avx512f_fnmsub_v16sf_mask_round, "__builtin_ia32_vfnmsubps512_mask", IX86_BUILTIN_VFNMSUBPS512_MASK, UNKNOWN, (int) V16SF_FTYPE_V16SF_V16SF_V16SF_HI_INT)
@@ -2869,6 +2883,8 @@ BDESC (OPTION_MASK_ISA_FMA, CODE_FOR_fmai_vmfmadd_v4sf, "__builtin_ia32_vfmaddss
 BDESC (OPTION_MASK_ISA_FMA, CODE_FOR_fmai_vmfmadd_v2df, "__builtin_ia32_vfmaddsd3", IX86_BUILTIN_VFMADDSD3, UNKNOWN, (int)MULTI_ARG_3_DF)
 BDESC (OPTION_MASK_ISA_FMA, CODE_FOR_fmai_vmfmsub_v4sf, "__builtin_ia32_vfmsubss3", IX86_BUILTIN_VFMSUBSS3, UNKNOWN, (int)MULTI_ARG_3_SF)
 BDESC (OPTION_MASK_ISA_FMA, CODE_FOR_fmai_vmfmsub_v2df, "__builtin_ia32_vfmsubsd3", IX86_BUILTIN_VFMSUBSD3, UNKNOWN, (int)MULTI_ARG_3_DF)
+BDESC (OPTION_MASK_ISA_FMA, CODE_FOR_fmai_vmfnmadd_v4sf, "__builtin_ia32_vfnmaddss3", IX86_BUILTIN_VFNMADDSS3, UNKNOWN, (int)MULTI_ARG_3_SF)
+BDESC (OPTION_MASK_ISA_FMA, CODE_FOR_fmai_vmfnmadd_v2df, "__builtin_ia32_vfnmaddsd3", IX86_BUILTIN_VFNMADDSD3, UNKNOWN, (int)MULTI_ARG_3_DF)
 
 BDESC (OPTION_MASK_ISA_FMA | OPTION_MASK_ISA_FMA4, CODE_FOR_fma4i_fmadd_v4sf, "__builtin_ia32_vfmaddps", IX86_BUILTIN_VFMADDPS, UNKNOWN, (int)MULTI_ARG_3_SF)
 BDESC (OPTION_MASK_ISA_FMA | OPTION_MASK_ISA_FMA4, CODE_FOR_fma4i_fmadd_v2df, "__builtin_ia32_vfmaddpd", IX86_BUILTIN_VFMADDPD, UNKNOWN, (int)MULTI_ARG_3_DF)
@@ -2878,6 +2894,10 @@ BDESC (OPTION_MASK_ISA_FMA | OPTION_MASK_ISA_FMA4, CODE_FOR_fma4i_fmsub_v4sf, "_
 BDESC (OPTION_MASK_ISA_FMA | OPTION_MASK_ISA_FMA4, CODE_FOR_fma4i_fmsub_v2df, "__builtin_ia32_vfmsubpd", IX86_BUILTIN_VFMSUBPD, UNKNOWN, (int)MULTI_ARG_3_DF)
 BDESC (OPTION_MASK_ISA_FMA | OPTION_MASK_ISA_FMA4, CODE_FOR_fma4i_fmsub_v8sf, "__builtin_ia32_vfmsubps256", IX86_BUILTIN_VFMSUBPS256, UNKNOWN, (int)MULTI_ARG_3_SF2)
 BDESC (OPTION_MASK_ISA_FMA | OPTION_MASK_ISA_FMA4, CODE_FOR_fma4i_fmsub_v4df, "__builtin_ia32_vfmsubpd256", IX86_BUILTIN_VFMSUBPD256, UNKNOWN, (int)MULTI_ARG_3_DF2)
+BDESC (OPTION_MASK_ISA_FMA | OPTION_MASK_ISA_FMA4, CODE_FOR_fma4i_fnmadd_v4sf, "__builtin_ia32_vfnmaddps", IX86_BUILTIN_VFNMADDPS, UNKNOWN, (int)MULTI_ARG_3_SF)
+BDESC (OPTION_MASK_ISA_FMA | OPTION_MASK_ISA_FMA4, CODE_FOR_fma4i_fnmadd_v2df, "__builtin_ia32_vfnmaddpd", IX86_BUILTIN_VFNMADDPD, UNKNOWN, (int)MULTI_ARG_3_DF)
+BDESC (OPTION_MASK_ISA_FMA | OPTION_MASK_ISA_FMA4, CODE_FOR_fma4i_fnmadd_v8sf, "__builtin_ia32_vfnmaddps256", IX86_BUILTIN_VFNMADDPS256, UNKNOWN, (int)MULTI_ARG_3_SF2)
+BDESC (OPTION_MASK_ISA_FMA | OPTION_MASK_ISA_FMA4, CODE_FOR_fma4i_fnmadd_v4df, "__builtin_ia32_vfnmaddpd256", IX86_BUILTIN_VFNMADDPD256, UNKNOWN, (int)MULTI_ARG_3_DF2)
 
 BDESC (OPTION_MASK_ISA_FMA | OPTION_MASK_ISA_FMA4, CODE_FOR_fmaddsub_v4sf, "__builtin_ia32_vfmaddsubps", IX86_BUILTIN_VFMADDSUBPS, UNKNOWN, (int)MULTI_ARG_3_SF)
 BDESC (OPTION_MASK_ISA_FMA | OPTION_MASK_ISA_FMA4, CODE_FOR_fmaddsub_v2df, "__builtin_ia32_vfmaddsubpd", IX86_BUILTIN_VFMADDSUBPD, UNKNOWN, (int)MULTI_ARG_3_DF)
diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md
index 0fdaaed..0426da4 100644
--- a/gcc/config/i386/sse.md
+++ b/gcc/config/i386/sse.md
@@ -3756,6 +3756,14 @@
 	  (neg:FMAMODE_AVX512
 	    (match_operand:FMAMODE_AVX512 3 "nonimmediate_operand"))))])
 
+(define_expand "fma4i_fnmadd_<mode>"
+  [(set (match_operand:FMAMODE_AVX512 0 "register_operand")
+	(fma:FMAMODE_AVX512
+	  (neg:FMAMODE_AVX512
+	    (match_operand:FMAMODE_AVX512 1 "nonimmediate_operand"))
+	  (match_operand:FMAMODE_AVX512 2 "nonimmediate_operand")
+	  (match_operand:FMAMODE_AVX512 3 "nonimmediate_operand")))])
+
 (define_expand "<avx512>_fmadd_<mode>_maskz<round_expand_name>"
   [(match_operand:VF_AVX512VL 0 "register_operand")
    (match_operand:VF_AVX512VL 1 "<round_expand_nimm_predicate>")
@@ -4016,6 +4024,20 @@
    (set_attr "type" "ssemuladd")
    (set_attr "mode" "<MODE>")])
 
+(define_expand "<avx512>_fnmadd_<mode>_maskz<round_expand_name>"
+  [(match_operand:VF_AVX512VL 0 "register_operand")
+   (match_operand:VF_AVX512VL 1 "<round_expand_nimm_predicate>")
+   (match_operand:VF_AVX512VL 2 "<round_expand_nimm_predicate>")
+   (match_operand:VF_AVX512VL 3 "<round_expand_nimm_predicate>")
+   (match_operand:<avx512fmaskmode> 4 "register_operand")]
+  "TARGET_AVX512F && <round_mode512bit_condition>"
+{
+  emit_insn (gen_fma_fnmadd_<mode>_maskz_1<round_expand_name> (
+    operands[0], operands[1], operands[2], operands[3],
+    CONST0_RTX (<MODE>mode), operands[4]<round_expand_operand>));
+  DONE;
+})
+
 (define_insn "<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>"
   [(set (match_operand:VF_SF_AVX512VL 0 "register_operand" "=v,v,v")
 	(fma:VF_SF_AVX512VL
@@ -4031,6 +4053,49 @@
   [(set_attr "type" "ssemuladd")
    (set_attr "mode" "<MODE>")])
 
+(define_insn "*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1"
+  [(set (match_operand:VF_AVX512 0 "register_operand" "=v,v")
+	(fma:VF_AVX512
+	  (neg:VF_AVX512
+	    (match_operand:VF_AVX512 1 "register_operand" "0,v"))
+	  (match_operand:VF_AVX512 2 "register_operand" "v,0")
+	  (vec_duplicate:VF_AVX512
+	    (match_operand:<ssescalarmode> 3 "memory_operand" "m,m"))))]
+  "TARGET_AVX512F && <sd_mask_mode512bit_condition>"
+  "vfnmadd213<ssemodesuffix>\t{%3<avx512bcst>, %2, %0<sd_mask_op4>|%0<sd_mask_op4>, %2, %3<avx512bcst>}"
+  [(set_attr "type" "ssemuladd")
+   (set_attr "mode" "<MODE>")])
+
+(define_insn "*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2"
+  [(set (match_operand:VF_AVX512 0 "register_operand" "=v,v")
+	(fma:VF_AVX512
+	  (neg:VF_AVX512
+	    (vec_duplicate:VF_AVX512
+	      (match_operand:<ssescalarmode> 1 "memory_operand" "m,m")))
+	  (match_operand:VF_AVX512 2 "register_operand" "0,v")
+	  (match_operand:VF_AVX512 3 "register_operand" "v,0")))]
+  "TARGET_AVX512F && <sd_mask_mode512bit_condition>"
+  "@
+   vfnmadd132<ssemodesuffix>\t{%1<avx512bcst>, %3, %0<sd_mask_op4>|%0<sd_mask_op4>, %3, %1<avx512bcst>}
+   vfnmadd231<ssemodesuffix>\t{%1<avx512bcst>, %2, %0<sd_mask_op4>|%0<sd_mask_op4>, %2, %1<avx512bcst>}"
+  [(set_attr "type" "ssemuladd")
+   (set_attr "mode" "<MODE>")])
+
+(define_insn "*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3"
+  [(set (match_operand:VF_AVX512 0 "register_operand" "=v,v")
+	(fma:VF_AVX512
+	  (neg:VF_AVX512
+	    (match_operand:VF_AVX512 1 "register_operand" "0,v"))
+	  (vec_duplicate:VF_AVX512
+	    (match_operand:<ssescalarmode> 2 "memory_operand" "m,m"))
+	  (match_operand:VF_AVX512 3 "register_operand" "v,0")))]
+  "TARGET_AVX512F && <sd_mask_mode512bit_condition>"
+  "@
+   vfnmadd132<ssemodesuffix>\t{%2<avx512bcst>, %3, %0<sd_mask_op4>|%0<sd_mask_op4>, %3, %2<avx512bcst>}
+   vfnmadd231<ssemodesuffix>\t{%2<avx512bcst>, %1, %0<sd_mask_op4>|%0<sd_mask_op4>, %1, %2<avx512bcst>}"
+  [(set_attr "type" "ssemuladd")
+   (set_attr "mode" "<MODE>")])
+
 (define_insn "<avx512>_fnmadd_<mode>_mask<round_name>"
   [(set (match_operand:VF_AVX512VL 0 "register_operand" "=v,v")
 	(vec_merge:VF_AVX512VL
@@ -4326,6 +4391,18 @@
 	  (const_int 1)))]
   "TARGET_FMA")
 
+(define_expand "fmai_vmfnmadd_<mode><round_name>"
+  [(set (match_operand:VF_128 0 "register_operand")
+	(vec_merge:VF_128
+	  (fma:VF_128
+	    (neg:VF_128
+	      (match_operand:VF_128 2 "<round_nimm_predicate>"))
+	    (match_operand:VF_128 1 "<round_nimm_predicate>")
+	    (match_operand:VF_128 3 "<round_nimm_predicate>"))
+	  (match_dup 1)
+	  (const_int 1)))]
+  "TARGET_FMA")
+
 (define_insn "*fmai_fmadd_<mode>"
   [(set (match_operand:VF_128 0 "register_operand" "=v,v")
         (vec_merge:VF_128