diff mbox series

AVX512FP16: Adjust builtin for mask complex fma

Message ID 20211013090659.43060-1-hongyu.wang@intel.com
State New
Headers show
Series AVX512FP16: Adjust builtin for mask complex fma | expand

Commit Message

Hongyu Wang Oct. 13, 2021, 9:06 a.m. UTC
Hi,

Current mask/mask3 implementation for complex fma contains
duplicated parameter in macro, which may cause error at -O0.
Refactor macro implementation to builtins to avoid potential
error.

For round intrinsic with NO_ROUND as input, ix86_erase_embedded_rounding
erases embedded_rounding upspec but could break other emit_insn in
expanders. Skip those expanders with multiple emit_insn for this
function and check rounding in expander with subst.

Bootstrapped/regtested on x86_64-pc-linux-gnu{-m32,} and sde{-m32,}.
OK for master?

gcc/ChangeLog:

	* config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
	Adjust builtin call.
	(_mm512_mask3_fcmadd_pch): Likewise.
	(_mm512_mask_fmadd_pch): Likewise
	(_mm512_mask3_fmadd_pch): Likewise
	(_mm512_mask_fcmadd_round_pch): Likewise
	(_mm512_mask3_fcmadd_round_pch): Likewise
	(_mm512_mask_fmadd_round_pch): Likewise
	(_mm512_mask3_fmadd_round_pch): Likewise
	(_mm_mask_fcmadd_sch): Likewise
	(_mm_mask3_fcmadd_sch): Likewise
	(_mm_mask_fmadd_sch): Likewise
	(_mm_mask3_fmadd_sch): Likewise
	(_mm_mask_fcmadd_round_sch): Likewise
	(_mm_mask3_fcmadd_round_sch): Likewise
	(_mm_mask_fmadd_round_sch): Likewise
	(_mm_mask3_fmadd_round_sch): Likewise
	(_mm_fcmadd_round_sch): Likewise
	* config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
	Adjust builtin call.
	(_mm_mask3_fmadd_pch): Likewise
	(_mm256_mask_fmadd_pch): Likewise
	(_mm256_mask3_fmadd_pch): Likewise
	(_mm_mask_fcmadd_pch): Likewise
	(_mm_mask3_fcmadd_pch): Likewise
	(_mm256_mask_fcmadd_pch): Likewise
	(_mm256_mask3_fcmadd_pch): Likewise
	* config/i386/i386-builtin.def: Add mask3 builtin for complex
	fma, and adjust mask_builtin to corresponding expander.
	* config/i386/i386-expand.c (ix86_expand_round_builtin):
	Skip eraseing embedded rounding for expanders that emits
	multiple insns.
	* config/i386/sse.md (complexmove): New mode_attr.
	(<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
	(<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
	(avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
	(avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
	(avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
	(avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
	* config/i386/subst.md (round_embedded_complex): New subst.

gcc/testsuite/ChangeLog:

	* gcc.target/i386/avx-1.c: Add new mask3 builtins.
	* gcc.target/i386/sse-13.c: Ditto.
	* gcc.target/i386/sse-23.c: Ditto.
	* gcc.target/i386/avx512fp16-vfcmaddcsh-1a.c: Add scanning for
	mask/mask3 intrinsic.
	* gcc.target/i386/avx512fp16-vfmaddcsh-1a.c: Ditto.
	* gcc.target/i386/avx512fp16-vfcmaddcsh-1c.c: New test for
	-mavx512vl.
	* gcc.target/i386/avx512fp16-vfmaddcsh-1c.c: Ditto.
---
 gcc/config/i386/avx512fp16intrin.h            | 261 ++++++------------
 gcc/config/i386/avx512fp16vlintrin.h          |  56 ++--
 gcc/config/i386/i386-builtin.def              |  24 +-
 gcc/config/i386/i386-expand.c                 |  22 +-
 gcc/config/i386/sse.md                        | 183 ++++++++++++
 gcc/config/i386/subst.md                      |   3 +
 gcc/testsuite/gcc.target/i386/avx-1.c         |   4 +
 .../i386/avx512fp16-vfcmaddcsh-1a.c           |   2 +
 .../i386/avx512fp16-vfcmaddcsh-1c.c           |  13 +
 .../gcc.target/i386/avx512fp16-vfmaddcsh-1a.c |   2 +
 .../gcc.target/i386/avx512fp16-vfmaddcsh-1c.c |  13 +
 gcc/testsuite/gcc.target/i386/sse-13.c        |   4 +
 gcc/testsuite/gcc.target/i386/sse-23.c        |   4 +
 13 files changed, 375 insertions(+), 216 deletions(-)
 create mode 100644 gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1c.c
 create mode 100644 gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1c.c

Comments

Hongtao Liu Oct. 14, 2021, 1:36 a.m. UTC | #1
On Wed, Oct 13, 2021 at 5:07 PM Hongyu Wang via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> Hi,
>
> Current mask/mask3 implementation for complex fma contains
> duplicated parameter in macro, which may cause error at -O0.
> Refactor macro implementation to builtins to avoid potential
> error.
>
> For round intrinsic with NO_ROUND as input, ix86_erase_embedded_rounding
> erases embedded_rounding upspec but could break other emit_insn in
> expanders. Skip those expanders with multiple emit_insn for this
> function and check rounding in expander with subst.
>
> Bootstrapped/regtested on x86_64-pc-linux-gnu{-m32,} and sde{-m32,}.
> OK for master?
Ok.
>
> gcc/ChangeLog:
>
>         * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
>         Adjust builtin call.
>         (_mm512_mask3_fcmadd_pch): Likewise.
>         (_mm512_mask_fmadd_pch): Likewise
>         (_mm512_mask3_fmadd_pch): Likewise
>         (_mm512_mask_fcmadd_round_pch): Likewise
>         (_mm512_mask3_fcmadd_round_pch): Likewise
>         (_mm512_mask_fmadd_round_pch): Likewise
>         (_mm512_mask3_fmadd_round_pch): Likewise
>         (_mm_mask_fcmadd_sch): Likewise
>         (_mm_mask3_fcmadd_sch): Likewise
>         (_mm_mask_fmadd_sch): Likewise
>         (_mm_mask3_fmadd_sch): Likewise
>         (_mm_mask_fcmadd_round_sch): Likewise
>         (_mm_mask3_fcmadd_round_sch): Likewise
>         (_mm_mask_fmadd_round_sch): Likewise
>         (_mm_mask3_fmadd_round_sch): Likewise
>         (_mm_fcmadd_round_sch): Likewise
>         * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
>         Adjust builtin call.
>         (_mm_mask3_fmadd_pch): Likewise
>         (_mm256_mask_fmadd_pch): Likewise
>         (_mm256_mask3_fmadd_pch): Likewise
>         (_mm_mask_fcmadd_pch): Likewise
>         (_mm_mask3_fcmadd_pch): Likewise
>         (_mm256_mask_fcmadd_pch): Likewise
>         (_mm256_mask3_fcmadd_pch): Likewise
>         * config/i386/i386-builtin.def: Add mask3 builtin for complex
>         fma, and adjust mask_builtin to corresponding expander.
>         * config/i386/i386-expand.c (ix86_expand_round_builtin):
>         Skip eraseing embedded rounding for expanders that emits
>         multiple insns.
>         * config/i386/sse.md (complexmove): New mode_attr.
>         (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
>         (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
>         (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
>         (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
>         (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
>         (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
>         * config/i386/subst.md (round_embedded_complex): New subst.
>
> gcc/testsuite/ChangeLog:
>
>         * gcc.target/i386/avx-1.c: Add new mask3 builtins.
>         * gcc.target/i386/sse-13.c: Ditto.
>         * gcc.target/i386/sse-23.c: Ditto.
>         * gcc.target/i386/avx512fp16-vfcmaddcsh-1a.c: Add scanning for
>         mask/mask3 intrinsic.
>         * gcc.target/i386/avx512fp16-vfmaddcsh-1a.c: Ditto.
>         * gcc.target/i386/avx512fp16-vfcmaddcsh-1c.c: New test for
>         -mavx512vl.
>         * gcc.target/i386/avx512fp16-vfmaddcsh-1c.c: Ditto.
> ---
>  gcc/config/i386/avx512fp16intrin.h            | 261 ++++++------------
>  gcc/config/i386/avx512fp16vlintrin.h          |  56 ++--
>  gcc/config/i386/i386-builtin.def              |  24 +-
>  gcc/config/i386/i386-expand.c                 |  22 +-
>  gcc/config/i386/sse.md                        | 183 ++++++++++++
>  gcc/config/i386/subst.md                      |   3 +
>  gcc/testsuite/gcc.target/i386/avx-1.c         |   4 +
>  .../i386/avx512fp16-vfcmaddcsh-1a.c           |   2 +
>  .../i386/avx512fp16-vfcmaddcsh-1c.c           |  13 +
>  .../gcc.target/i386/avx512fp16-vfmaddcsh-1a.c |   2 +
>  .../gcc.target/i386/avx512fp16-vfmaddcsh-1c.c |  13 +
>  gcc/testsuite/gcc.target/i386/sse-13.c        |   4 +
>  gcc/testsuite/gcc.target/i386/sse-23.c        |   4 +
>  13 files changed, 375 insertions(+), 216 deletions(-)
>  create mode 100644 gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1c.c
>  create mode 100644 gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1c.c
>
> diff --git a/gcc/config/i386/avx512fp16intrin.h b/gcc/config/i386/avx512fp16intrin.h
> index 29cf6792335..5e49447a020 100644
> --- a/gcc/config/i386/avx512fp16intrin.h
> +++ b/gcc/config/i386/avx512fp16intrin.h
> @@ -6258,13 +6258,11 @@ extern __inline __m512h
>  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm512_mask_fcmadd_pch (__m512h __A, __mmask16 __B, __m512h __C, __m512h __D)
>  {
> -  return (__m512h) __builtin_ia32_movaps512_mask
> -    ((__v16sf)
> -     __builtin_ia32_vfcmaddcph512_mask_round ((__v32hf) __A,
> -                                             (__v32hf) __C,
> -                                             (__v32hf) __D, __B,
> -                                             _MM_FROUND_CUR_DIRECTION),
> -     (__v16sf) __A, __B);
> +  return (__m512h)
> +    __builtin_ia32_vfcmaddcph512_mask_round ((__v32hf) __A,
> +                                            (__v32hf) __C,
> +                                            (__v32hf) __D, __B,
> +                                            _MM_FROUND_CUR_DIRECTION);
>  }
>
>  extern __inline __m512h
> @@ -6272,10 +6270,10 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm512_mask3_fcmadd_pch (__m512h __A, __m512h __B, __m512h __C, __mmask16 __D)
>  {
>    return (__m512h)
> -    __builtin_ia32_vfcmaddcph512_mask_round ((__v32hf) __A,
> -                                            (__v32hf) __B,
> -                                            (__v32hf) __C,
> -                                            __D, _MM_FROUND_CUR_DIRECTION);
> +    __builtin_ia32_vfcmaddcph512_mask3_round ((__v32hf) __A,
> +                                             (__v32hf) __B,
> +                                             (__v32hf) __C,
> +                                             __D, _MM_FROUND_CUR_DIRECTION);
>  }
>
>  extern __inline __m512h
> @@ -6304,13 +6302,11 @@ extern __inline __m512h
>  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm512_mask_fmadd_pch (__m512h __A, __mmask16 __B, __m512h __C, __m512h __D)
>  {
> -  return (__m512h) __builtin_ia32_movaps512_mask
> -    ((__v16sf)
> -     __builtin_ia32_vfmaddcph512_mask_round ((__v32hf) __A,
> -                                            (__v32hf) __C,
> -                                            (__v32hf) __D, __B,
> -                                            _MM_FROUND_CUR_DIRECTION),
> -     (__v16sf) __A, __B);
> +  return (__m512h)
> +    __builtin_ia32_vfmaddcph512_mask_round ((__v32hf) __A,
> +                                           (__v32hf) __C,
> +                                           (__v32hf) __D, __B,
> +                                           _MM_FROUND_CUR_DIRECTION);
>  }
>
>  extern __inline __m512h
> @@ -6318,10 +6314,10 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm512_mask3_fmadd_pch (__m512h __A, __m512h __B, __m512h __C, __mmask16 __D)
>  {
>    return (__m512h)
> -    __builtin_ia32_vfmaddcph512_mask_round ((__v32hf) __A,
> -                                           (__v32hf) __B,
> -                                           (__v32hf) __C,
> -                                           __D, _MM_FROUND_CUR_DIRECTION);
> +    __builtin_ia32_vfmaddcph512_mask3_round ((__v32hf) __A,
> +                                            (__v32hf) __B,
> +                                            (__v32hf) __C,
> +                                            __D, _MM_FROUND_CUR_DIRECTION);
>  }
>
>  extern __inline __m512h
> @@ -6352,13 +6348,11 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm512_mask_fcmadd_round_pch (__m512h __A, __mmask16 __B, __m512h __C,
>                               __m512h __D, const int __E)
>  {
> -  return (__m512h) __builtin_ia32_movaps512_mask
> -    ((__v16sf)
> -     __builtin_ia32_vfcmaddcph512_mask_round ((__v32hf) __A,
> -                                             (__v32hf) __C,
> -                                             (__v32hf) __D, __B,
> -                                             __E),
> -     (__v16sf) __A, __B);
> +  return (__m512h)
> +    __builtin_ia32_vfcmaddcph512_mask_round ((__v32hf) __A,
> +                                            (__v32hf) __C,
> +                                            (__v32hf) __D, __B,
> +                                            __E);
>  }
>
>  extern __inline __m512h
> @@ -6367,10 +6361,10 @@ _mm512_mask3_fcmadd_round_pch (__m512h __A, __m512h __B, __m512h __C,
>                                __mmask16 __D, const int __E)
>  {
>    return (__m512h)
> -    __builtin_ia32_vfcmaddcph512_mask_round ((__v32hf) __A,
> -                                            (__v32hf) __B,
> -                                            (__v32hf) __C,
> -                                            __D, __E);
> +    __builtin_ia32_vfcmaddcph512_mask3_round ((__v32hf) __A,
> +                                             (__v32hf) __B,
> +                                             (__v32hf) __C,
> +                                             __D, __E);
>  }
>
>  extern __inline __m512h
> @@ -6401,13 +6395,11 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm512_mask_fmadd_round_pch (__m512h __A, __mmask16 __B, __m512h __C,
>                              __m512h __D, const int __E)
>  {
> -  return (__m512h) __builtin_ia32_movaps512_mask
> -    ((__v16sf)
> -     __builtin_ia32_vfmaddcph512_mask_round ((__v32hf) __A,
> -                                            (__v32hf) __C,
> -                                            (__v32hf) __D, __B,
> -                                            __E),
> -     (__v16sf) __A, __B);
> +  return (__m512h)
> +    __builtin_ia32_vfmaddcph512_mask_round ((__v32hf) __A,
> +                                           (__v32hf) __C,
> +                                           (__v32hf) __D, __B,
> +                                           __E);
>  }
>
>  extern __inline __m512h
> @@ -6416,10 +6408,10 @@ _mm512_mask3_fmadd_round_pch (__m512h __A, __m512h __B, __m512h __C,
>                               __mmask16 __D, const int __E)
>  {
>    return (__m512h)
> -    __builtin_ia32_vfmaddcph512_mask_round ((__v32hf) __A,
> -                                           (__v32hf) __B,
> -                                           (__v32hf) __C,
> -                                           __D, __E);
> +    __builtin_ia32_vfmaddcph512_mask3_round ((__v32hf) __A,
> +                                            (__v32hf) __B,
> +                                            (__v32hf) __C,
> +                                            __D, __E);
>  }
>
>  extern __inline __m512h
> @@ -6439,18 +6431,16 @@ _mm512_maskz_fmadd_round_pch (__mmask16 __A, __m512h __B, __m512h __C,
>    (__m512h) __builtin_ia32_vfcmaddcph512_round ((A), (B), (C), (D))
>
>  #define _mm512_mask_fcmadd_round_pch(A, B, C, D, E)                    \
> -  ((__m512h) __builtin_ia32_movaps512_mask (                           \
> -   (__v16sf)                                                           \
> +  ((__m512h)                                                           \
>      __builtin_ia32_vfcmaddcph512_mask_round ((__v32hf) (A),            \
>                                              (__v32hf) (C),             \
>                                              (__v32hf) (D),             \
> -                                            (B), (E)),                 \
> -                                            (__v16sf) (A), (B)));
> +                                            (B), (E)))
>
>
>  #define _mm512_mask3_fcmadd_round_pch(A, B, C, D, E)                   \
>    ((__m512h)                                                           \
> -   __builtin_ia32_vfcmaddcph512_mask_round ((A), (B), (C), (D), (E)))
> +   __builtin_ia32_vfcmaddcph512_mask3_round ((A), (B), (C), (D), (E)))
>
>  #define _mm512_maskz_fcmadd_round_pch(A, B, C, D, E)                   \
>    (__m512h)                                                            \
> @@ -6460,17 +6450,15 @@ _mm512_maskz_fmadd_round_pch (__mmask16 __A, __m512h __B, __m512h __C,
>    (__m512h) __builtin_ia32_vfmaddcph512_round ((A), (B), (C), (D))
>
>  #define _mm512_mask_fmadd_round_pch(A, B, C, D, E)                     \
> -  ((__m512h) __builtin_ia32_movaps512_mask (                           \
> -   (__v16sf)                                                           \
> +  ((__m512h)                                                           \
>      __builtin_ia32_vfmaddcph512_mask_round ((__v32hf) (A),             \
>                                             (__v32hf) (C),              \
>                                             (__v32hf) (D),              \
> -                                           (B), (E)),                  \
> -                                           (__v16sf) (A), (B)));
> +                                           (B), (E)))
>
>  #define _mm512_mask3_fmadd_round_pch(A, B, C, D, E)                    \
>    (__m512h)                                                            \
> -   __builtin_ia32_vfmaddcph512_mask_round ((A), (B), (C), (D), (E))
> +   __builtin_ia32_vfmaddcph512_mask3_round ((A), (B), (C), (D), (E))
>
>  #define _mm512_maskz_fmadd_round_pch(A, B, C, D, E)                    \
>    (__m512h)                                                            \
> @@ -6643,35 +6631,22 @@ extern __inline __m128h
>  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm_mask_fcmadd_sch (__m128h __A, __mmask8 __B, __m128h __C, __m128h __D)
>  {
> -#ifdef __AVX512VL__
> -  return (__m128h) __builtin_ia32_movaps128_mask (
> -    (__v4sf)
> -    __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) __A,
> -                                         (__v8hf) __C,
> -                                         (__v8hf) __D, __B,
> -                                         _MM_FROUND_CUR_DIRECTION),
> -    (__v4sf) __A, __B);
> -#else
> -  return (__m128h) __builtin_ia32_blendvps ((__v4sf) __A,
> -    (__v4sf)
> +  return (__m128h)
>      __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) __A,
>                                           (__v8hf) __C,
>                                           (__v8hf) __D, __B,
> -                                         _MM_FROUND_CUR_DIRECTION),
> -    (__v4sf) _mm_set_ss ((float) ((int) __B << 31)));
> -#endif
> +                                         _MM_FROUND_CUR_DIRECTION);
>  }
>
>  extern __inline __m128h
>  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm_mask3_fcmadd_sch (__m128h __A, __m128h __B, __m128h __C, __mmask8 __D)
>  {
> -  return (__m128h) _mm_move_ss ((__m128) __C,
> -    (__m128)
> -    __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) __A,
> -                                         (__v8hf) __B,
> -                                         (__v8hf) __C, __D,
> -                                         _MM_FROUND_CUR_DIRECTION));
> +  return (__m128h)
> +    __builtin_ia32_vfcmaddcsh_mask3_round ((__v8hf) __A,
> +                                          (__v8hf) __B,
> +                                          (__v8hf) __C, __D,
> +                                          _MM_FROUND_CUR_DIRECTION);
>  }
>
>  extern __inline __m128h
> @@ -6700,35 +6675,22 @@ extern __inline __m128h
>  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm_mask_fmadd_sch (__m128h __A, __mmask8 __B, __m128h __C, __m128h __D)
>  {
> -#ifdef __AVX512VL__
> -  return (__m128h) __builtin_ia32_movaps128_mask (
> -    (__v4sf)
> -    __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) __A,
> -                                        (__v8hf) __C,
> -                                        (__v8hf) __D, __B,
> -                                        _MM_FROUND_CUR_DIRECTION),
> -    (__v4sf) __A, __B);
> -#else
> -  return (__m128h) __builtin_ia32_blendvps ((__v4sf) __A,
> -    (__v4sf)
> +  return (__m128h)
>      __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) __A,
>                                          (__v8hf) __C,
>                                          (__v8hf) __D, __B,
> -                                        _MM_FROUND_CUR_DIRECTION),
> -    (__v4sf) _mm_set_ss ((float) ((int) __B << 31)));
> -#endif
> +                                        _MM_FROUND_CUR_DIRECTION);
>  }
>
>  extern __inline __m128h
>  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm_mask3_fmadd_sch (__m128h __A, __m128h __B, __m128h __C, __mmask8 __D)
>  {
> -  return (__m128h) _mm_move_ss ((__m128) __C,
> -    (__m128)
> -    __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) __A,
> -                                        (__v8hf) __B,
> -                                        (__v8hf) __C, __D,
> -                                        _MM_FROUND_CUR_DIRECTION));
> +  return (__m128h)
> +    __builtin_ia32_vfmaddcsh_mask3_round ((__v8hf) __A,
> +                                         (__v8hf) __B,
> +                                         (__v8hf) __C, __D,
> +                                         _MM_FROUND_CUR_DIRECTION);
>  }
>
>  extern __inline __m128h
> @@ -6759,23 +6721,11 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm_mask_fcmadd_round_sch (__m128h __A, __mmask8 __B, __m128h __C,
>                            __m128h __D, const int __E)
>  {
> -#ifdef __AVX512VL__
> -  return (__m128h) __builtin_ia32_movaps128_mask (
> -    (__v4sf)
> -    __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) __A,
> -                                         (__v8hf) __C,
> -                                         (__v8hf) __D,
> -                                         __B, __E),
> -    (__v4sf) __A, __B);
> -#else
> -  return (__m128h) __builtin_ia32_blendvps ((__v4sf) __A,
> -    (__v4sf)
> +  return (__m128h)
>      __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) __A,
>                                           (__v8hf) __C,
>                                           (__v8hf) __D,
> -                                         __B, __E),
> -    (__v4sf) _mm_set_ss ((float) ((int) __B << 31)));
> -#endif
> +                                         __B, __E);
>  }
>
>  extern __inline __m128h
> @@ -6783,12 +6733,11 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm_mask3_fcmadd_round_sch (__m128h __A, __m128h __B, __m128h __C,
>                             __mmask8 __D, const int __E)
>  {
> -  return (__m128h) _mm_move_ss ((__m128) __C,
> -    (__m128)
> -    __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) __A,
> -                                         (__v8hf) __B,
> -                                         (__v8hf) __C,
> -                                         __D, __E));
> +  return (__m128h)
> +    __builtin_ia32_vfcmaddcsh_mask3_round ((__v8hf) __A,
> +                                          (__v8hf) __B,
> +                                          (__v8hf) __C,
> +                                          __D, __E);
>  }
>
>  extern __inline __m128h
> @@ -6819,23 +6768,11 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm_mask_fmadd_round_sch (__m128h __A, __mmask8 __B, __m128h __C,
>                           __m128h __D, const int __E)
>  {
> -#ifdef __AVX512VL__
> -  return (__m128h) __builtin_ia32_movaps128_mask (
> -    (__v4sf)
> -    __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) __A,
> -                                        (__v8hf) __C,
> -                                        (__v8hf) __D,
> -                                        __B, __E),
> -    (__v4sf) __A, __B);
> -#else
> -  return (__m128h) __builtin_ia32_blendvps ((__v4sf) __A,
> -    (__v4sf)
> +  return (__m128h)
>      __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) __A,
>                                          (__v8hf) __C,
>                                          (__v8hf) __D,
> -                                        __B, __E),
> -    (__v4sf) _mm_set_ss ((float) ((int) __B << 31)));
> -#endif
> +                                        __B, __E);
>  }
>
>  extern __inline __m128h
> @@ -6843,12 +6780,11 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm_mask3_fmadd_round_sch (__m128h __A, __m128h __B, __m128h __C,
>                            __mmask8 __D, const int __E)
>  {
> -  return (__m128h) _mm_move_ss ((__m128) __C,
> -    (__m128)
> -    __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) __A,
> -                                        (__v8hf) __B,
> -                                        (__v8hf) __C,
> -                                        __D, __E));
> +  return (__m128h)
> +    __builtin_ia32_vfmaddcsh_mask3_round ((__v8hf) __A,
> +                                         (__v8hf) __B,
> +                                         (__v8hf) __C,
> +                                         __D, __E);
>  }
>
>  extern __inline __m128h
> @@ -6874,34 +6810,20 @@ _mm_fmadd_round_sch (__m128h __A, __m128h __B, __m128h __C, const int __D)
>                                     __D);
>  }
>  #else
> -#ifdef __AVX512VL__
>  #define _mm_mask_fcmadd_round_sch(A, B, C, D, E)                       \
> -    ((__m128h) __builtin_ia32_movaps128_mask (                         \
> -     (__v4sf)                                                          \
> +    ((__m128h)                                                         \
>       __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) (A),               \
>                                            (__v8hf) (C),                \
>                                            (__v8hf) (D),                \
> -                                          (B), (E)),                   \
> -                                          (__v4sf) (A), (B)))
> +                                          (B), (E)))
>
> -#else
> -#define _mm_mask_fcmadd_round_sch(A, B, C, D, E)                       \
> -  ((__m128h) __builtin_ia32_blendvps ((__v4sf) (A),                    \
> -   (__v4sf)                                                            \
> -   __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) (A),         \
> -                                        (__v8hf) (C),          \
> -                                        (__v8hf) (D),          \
> -                                        (B), (E)),             \
> -    (__v4sf) _mm_set_ss ((float) ((int) (B) << 31))))
> -#endif
>
>  #define _mm_mask3_fcmadd_round_sch(A, B, C, D, E)                      \
> -  ((__m128h) _mm_move_ss ((__m128) (C),                                        \
> -   (__m128)                                                            \
> -   __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) (A),         \
> -                                        (__v8hf) (B),          \
> -                                        (__v8hf) (C),          \
> -                                        (D), (E))))
> +  ((__m128h)                                                           \
> +   __builtin_ia32_vfcmaddcsh_mask3_round ((__v8hf) (A),                \
> +                                         (__v8hf) (B),         \
> +                                         (__v8hf) (C),         \
> +                                         (D), (E)))
>
>  #define _mm_maskz_fcmadd_round_sch(A, B, C, D, E)              \
>    __builtin_ia32_vfcmaddcsh_maskz_round ((B), (C), (D), (A), (E))
> @@ -6909,34 +6831,19 @@ _mm_fmadd_round_sch (__m128h __A, __m128h __B, __m128h __C, const int __D)
>  #define _mm_fcmadd_round_sch(A, B, C, D)               \
>    __builtin_ia32_vfcmaddcsh_round ((A), (B), (C), (D))
>
> -#ifdef __AVX512VL__
>  #define _mm_mask_fmadd_round_sch(A, B, C, D, E)                                \
> -    ((__m128h) __builtin_ia32_movaps128_mask (                         \
> -     (__v4sf)                                                          \
> +    ((__m128h)                                                         \
>       __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) (A),                \
>                                           (__v8hf) (C),         \
>                                           (__v8hf) (D),         \
> -                                         (B), (E)),            \
> -                                         (__v4sf) (A), (B)))
> -
> -#else
> -#define _mm_mask_fmadd_round_sch(A, B, C, D, E)                                \
> -  ((__m128h) __builtin_ia32_blendvps ((__v4sf) (A),                    \
> -   (__v4sf)                                                            \
> -   __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) (A),          \
> -                                       (__v8hf) (C),           \
> -                                       (__v8hf) (D),           \
> -                                       (B), (E)),              \
> -    (__v4sf) _mm_set_ss ((float) ((int) (B) << 31))))
> -#endif
> +                                         (B), (E)))
>
>  #define _mm_mask3_fmadd_round_sch(A, B, C, D, E)                       \
> -  ((__m128h) _mm_move_ss ((__m128) (C),                                        \
> -   (__m128)                                                            \
> -   __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) (A),          \
> -                                       (__v8hf) (B),           \
> -                                       (__v8hf) (C),           \
> -                                       (D), (E))))
> +  ((__m128h)                                                           \
> +   __builtin_ia32_vfmaddcsh_mask3_round ((__v8hf) (A),         \
> +                                        (__v8hf) (B),          \
> +                                        (__v8hf) (C),          \
> +                                        (D), (E)))
>
>  #define _mm_maskz_fmadd_round_sch(A, B, C, D, E)               \
>    __builtin_ia32_vfmaddcsh_maskz_round ((B), (C), (D), (A), (E))
> diff --git a/gcc/config/i386/avx512fp16vlintrin.h b/gcc/config/i386/avx512fp16vlintrin.h
> index 3d3de964224..0b1f1cbe6ad 100644
> --- a/gcc/config/i386/avx512fp16vlintrin.h
> +++ b/gcc/config/i386/avx512fp16vlintrin.h
> @@ -2898,21 +2898,20 @@ extern __inline __m128h
>  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm_mask_fmadd_pch (__m128h __A, __mmask8 __B, __m128h __C, __m128h __D)
>  {
> -  return (__m128h) __builtin_ia32_movaps128_mask
> -    ((__v4sf)
> -     __builtin_ia32_vfmaddcph128_mask ((__v8hf) __A,
> -                                      (__v8hf) __C,
> -                                      (__v8hf) __D, __B),
> -     (__v4sf) __A, __B);
> +  return (__m128h)
> +    __builtin_ia32_vfmaddcph128_mask ((__v8hf) __A,
> +                                     (__v8hf) __C,
> +                                     (__v8hf) __D, __B);
>  }
>
>  extern __inline __m128h
>  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm_mask3_fmadd_pch (__m128h __A, __m128h __B, __m128h __C,  __mmask8 __D)
>  {
> -  return (__m128h) __builtin_ia32_vfmaddcph128_mask ((__v8hf) __A,
> -                                                    (__v8hf) __B,
> -                                                    (__v8hf) __C, __D);
> +  return (__m128h)
> +    __builtin_ia32_vfmaddcph128_mask3 ((__v8hf) __A,
> +                                      (__v8hf) __B,
> +                                      (__v8hf) __C, __D);
>  }
>
>  extern __inline __m128h
> @@ -2937,21 +2936,20 @@ extern __inline __m256h
>  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm256_mask_fmadd_pch (__m256h __A, __mmask8 __B, __m256h __C, __m256h __D)
>  {
> -  return (__m256h) __builtin_ia32_movaps256_mask
> -    ((__v8sf)
> +  return (__m256h)
>       __builtin_ia32_vfmaddcph256_mask ((__v16hf) __A,
>                                        (__v16hf) __C,
> -                                      (__v16hf) __D, __B),
> -     (__v8sf) __A, __B);
> +                                      (__v16hf) __D, __B);
>  }
>
>  extern __inline __m256h
>  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm256_mask3_fmadd_pch (__m256h __A, __m256h __B, __m256h __C,  __mmask8 __D)
>  {
> -  return (__m256h) __builtin_ia32_vfmaddcph256_mask ((__v16hf) __A,
> -                                                    (__v16hf) __B,
> -                                                    (__v16hf) __C, __D);
> +  return (__m256h)
> +    __builtin_ia32_vfmaddcph256_mask3 ((__v16hf) __A,
> +                                      (__v16hf) __B,
> +                                      (__v16hf) __C, __D);
>  }
>
>  extern __inline __m256h
> @@ -2976,21 +2974,20 @@ extern __inline __m128h
>  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm_mask_fcmadd_pch (__m128h __A, __mmask8 __B, __m128h __C, __m128h __D)
>  {
> -  return (__m128h) __builtin_ia32_movaps128_mask
> -    ((__v4sf)
> +  return (__m128h)
>       __builtin_ia32_vfcmaddcph128_mask ((__v8hf) __A,
>                                         (__v8hf) __C,
> -                                       (__v8hf) __D, __B),
> -     (__v4sf) __A, __B);
> +                                       (__v8hf) __D, __B);
>  }
>
>  extern __inline __m128h
>  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm_mask3_fcmadd_pch (__m128h __A, __m128h __B, __m128h __C,  __mmask8 __D)
>  {
> -  return (__m128h) __builtin_ia32_vfcmaddcph128_mask ((__v8hf) __A,
> -                                                     (__v8hf) __B,
> -                                                     (__v8hf) __C, __D);
> +  return (__m128h)
> +    __builtin_ia32_vfcmaddcph128_mask3 ((__v8hf) __A,
> +                                       (__v8hf) __B,
> +                                       (__v8hf) __C, __D);
>  }
>
>  extern __inline __m128h
> @@ -3015,21 +3012,20 @@ extern __inline __m256h
>  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm256_mask_fcmadd_pch (__m256h __A, __mmask8 __B, __m256h __C, __m256h __D)
>  {
> -  return (__m256h) __builtin_ia32_movaps256_mask
> -    ((__v8sf)
> +  return (__m256h)
>       __builtin_ia32_vfcmaddcph256_mask ((__v16hf) __A,
>                                         (__v16hf) __C,
> -                                       (__v16hf) __D, __B),
> -     (__v8sf) __A, __B);
> +                                       (__v16hf) __D, __B);
>  }
>
>  extern __inline __m256h
>  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
>  _mm256_mask3_fcmadd_pch (__m256h __A, __m256h __B, __m256h __C,  __mmask8 __D)
>  {
> -  return (__m256h) __builtin_ia32_vfcmaddcph256_mask ((__v16hf) __A,
> -                                                     (__v16hf) __B,
> -                                                     (__v16hf) __C, __D);
> +  return (__m256h)
> +    __builtin_ia32_vfcmaddcph256_mask3 ((__v16hf) __A,
> +                                       (__v16hf) __B,
> +                                       (__v16hf) __C, __D);
>  }
>
>  extern __inline __m256h
> diff --git a/gcc/config/i386/i386-builtin.def b/gcc/config/i386/i386-builtin.def
> index 302e1bc6502..99217d08d37 100644
> --- a/gcc/config/i386/i386-builtin.def
> +++ b/gcc/config/i386/i386-builtin.def
> @@ -2912,16 +2912,20 @@ BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp1
>  BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fnmsub_v8hf_mask3, "__builtin_ia32_vfnmsubph128_mask3", IX86_BUILTIN_VFNMSUBPH128_MASK3, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
>  BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fnmsub_v8hf_maskz, "__builtin_ia32_vfnmsubph128_maskz", IX86_BUILTIN_VFNMSUBPH128_MASKZ, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
>  BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_fma_fmaddc_v8hf, "__builtin_ia32_vfmaddcph128", IX86_BUILTIN_VFMADDCPH_V8HF, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF)
> -BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddc_v8hf_mask, "__builtin_ia32_vfmaddcph128_mask", IX86_BUILTIN_VFMADDCPH128_MASK, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
> +BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddc_v8hf_mask1, "__builtin_ia32_vfmaddcph128_mask", IX86_BUILTIN_VFMADDCPH128_MASK, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
> +BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddc_v8hf_mask, "__builtin_ia32_vfmaddcph128_mask3", IX86_BUILTIN_VFMADDCPH128_MASK3, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
>  BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddc_v8hf_maskz, "__builtin_ia32_vfmaddcph128_maskz", IX86_BUILTIN_VFMADDCPH128_MASKZ, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
>  BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_fma_fmaddc_v16hf, "__builtin_ia32_vfmaddcph256", IX86_BUILTIN_VFMADDCPH_V16HF, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF)
> -BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fmaddc_v16hf_mask, "__builtin_ia32_vfmaddcph256_mask", IX86_BUILTIN_VFMADDCPH256_MASK, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
> +BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fmaddc_v16hf_mask1, "__builtin_ia32_vfmaddcph256_mask", IX86_BUILTIN_VFMADDCPH256_MASK, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
> +BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fmaddc_v16hf_mask, "__builtin_ia32_vfmaddcph256_mask3", IX86_BUILTIN_VFMADDCPH256_MASK3, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
>  BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fmaddc_v16hf_maskz, "__builtin_ia32_vfmaddcph256_maskz", IX86_BUILTIN_VFMADDCPH256_MASKZ, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
>  BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_fma_fcmaddc_v8hf, "__builtin_ia32_vfcmaddcph128", IX86_BUILTIN_VFCMADDCPH_V8HF, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF)
> -BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddc_v8hf_mask, "__builtin_ia32_vfcmaddcph128_mask", IX86_BUILTIN_VFCMADDCPH128_MASK, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
> +BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddc_v8hf_mask1, "__builtin_ia32_vfcmaddcph128_mask", IX86_BUILTIN_VFCMADDCPH128_MASK, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
> +BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddc_v8hf_mask, "__builtin_ia32_vfcmaddcph128_mask3", IX86_BUILTIN_VFCMADDCPH128_MASK3, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
>  BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddc_v8hf_maskz, "__builtin_ia32_vfcmaddcph128_maskz", IX86_BUILTIN_VFCMADDCPH128_MASKZ, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
>  BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_fma_fcmaddc_v16hf, "__builtin_ia32_vfcmaddcph256", IX86_BUILTIN_VFCMADDCPH_V16HF, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF)
> -BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fcmaddc_v16hf_mask, "__builtin_ia32_vfcmaddcph256_mask", IX86_BUILTIN_VFCMADDCPH256_MASK, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
> +BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fcmaddc_v16hf_mask1, "__builtin_ia32_vfcmaddcph256_mask", IX86_BUILTIN_VFCMADDCPH256_MASK, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
> +BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fcmaddc_v16hf_mask, "__builtin_ia32_vfcmaddcph256_mask3", IX86_BUILTIN_VFCMADDCPH256_MASK3, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
>  BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fcmaddc_v16hf_maskz, "__builtin_ia32_vfcmaddcph256_maskz", IX86_BUILTIN_VFCMADDCPH256_MASKZ, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
>  BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmulc_v8hf, "__builtin_ia32_vfcmulcph128", IX86_BUILTIN_VFCMULCPH_V8HF, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF)
>  BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmulc_v8hf_mask, "__builtin_ia32_vfcmulcph128_mask", IX86_BUILTIN_VFCMULCPH128_MASK, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
> @@ -3222,20 +3226,24 @@ BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512f_vmfnmadd_v8hf_mask3_roun
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512f_vmfnmadd_v8hf_maskz_round, "__builtin_ia32_vfnmaddsh3_maskz", IX86_BUILTIN_VFNMADDSH3_MASKZ, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512f_vmfmsub_v8hf_mask3_round, "__builtin_ia32_vfmsubsh3_mask3", IX86_BUILTIN_VFMSUBSH3_MASK3, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_fma_fmaddc_v32hf_round, "__builtin_ia32_vfmaddcph512_round", IX86_BUILTIN_VFMADDCPH512_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_INT)
> -BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fmaddc_v32hf_mask_round, "__builtin_ia32_vfmaddcph512_mask_round", IX86_BUILTIN_VFMADDCPH512_MASK_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
> +BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fmaddc_v32hf_mask1_round, "__builtin_ia32_vfmaddcph512_mask_round", IX86_BUILTIN_VFMADDCPH512_MASK_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
> +BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fmaddc_v32hf_mask_round, "__builtin_ia32_vfmaddcph512_mask3_round", IX86_BUILTIN_VFMADDCPH512_MASK3_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fmaddc_v32hf_maskz_round, "__builtin_ia32_vfmaddcph512_maskz_round", IX86_BUILTIN_VFMADDCPH512_MASKZ_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_fma_fcmaddc_v32hf_round, "__builtin_ia32_vfcmaddcph512_round", IX86_BUILTIN_VFCMADDCPH512_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_INT)
> -BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fcmaddc_v32hf_mask_round, "__builtin_ia32_vfcmaddcph512_mask_round", IX86_BUILTIN_VFCMADDCPH512_MASK_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
> +BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fcmaddc_v32hf_mask1_round, "__builtin_ia32_vfcmaddcph512_mask_round", IX86_BUILTIN_VFCMADDCPH512_MASK_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
> +BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fcmaddc_v32hf_mask_round, "__builtin_ia32_vfcmaddcph512_mask3_round", IX86_BUILTIN_VFCMADDCPH512_MASK3_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fcmaddc_v32hf_maskz_round, "__builtin_ia32_vfcmaddcph512_maskz_round", IX86_BUILTIN_VFCMADDCPH512_MASKZ_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fcmulc_v32hf_round, "__builtin_ia32_vfcmulcph512_round", IX86_BUILTIN_VFCMULCPH512_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fcmulc_v32hf_mask_round, "__builtin_ia32_vfcmulcph512_mask_round", IX86_BUILTIN_VFCMULCPH512_MASK_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fmulc_v32hf_round, "__builtin_ia32_vfmulcph512_round", IX86_BUILTIN_VFMULCPH512_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fmulc_v32hf_mask_round, "__builtin_ia32_vfmulcph512_mask_round", IX86_BUILTIN_VFMULCPH512_MASK_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fma_fcmaddcsh_v8hf_round, "__builtin_ia32_vfcmaddcsh_round", IX86_BUILTIN_VFCMADDCSH_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_INT)
> -BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddcsh_v8hf_mask_round, "__builtin_ia32_vfcmaddcsh_mask_round", IX86_BUILTIN_VFCMADDCSH_MASK_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
> +BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddcsh_v8hf_mask1_round, "__builtin_ia32_vfcmaddcsh_mask_round", IX86_BUILTIN_VFCMADDCSH_MASK_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
> +BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddcsh_v8hf_mask3_round, "__builtin_ia32_vfcmaddcsh_mask3_round", IX86_BUILTIN_VFCMADDCSH_MASK3_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddcsh_v8hf_maskz_round, "__builtin_ia32_vfcmaddcsh_maskz_round", IX86_BUILTIN_VFCMADDCSH_MASKZ_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fma_fmaddcsh_v8hf_round, "__builtin_ia32_vfmaddcsh_round", IX86_BUILTIN_VFMADDCSH_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_INT)
> -BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddcsh_v8hf_mask_round, "__builtin_ia32_vfmaddcsh_mask_round", IX86_BUILTIN_VFMADDCSH_MASK_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
> +BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddcsh_v8hf_mask1_round, "__builtin_ia32_vfmaddcsh_mask_round", IX86_BUILTIN_VFMADDCSH_MASK_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
> +BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddcsh_v8hf_mask3_round, "__builtin_ia32_vfmaddcsh_mask3_round", IX86_BUILTIN_VFMADDCSH_MASK3_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddcsh_v8hf_maskz_round, "__builtin_ia32_vfmaddcsh_maskz_round", IX86_BUILTIN_VFMADDCSH_MASKZ_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmulcsh_v8hf_round, "__builtin_ia32_vfcmulcsh_round", IX86_BUILTIN_VFCMULCSH_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_INT)
>  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmulcsh_v8hf_mask_round, "__builtin_ia32_vfcmulcsh_mask_round", IX86_BUILTIN_VFCMULCSH_MASK_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
> diff --git a/gcc/config/i386/i386-expand.c b/gcc/config/i386/i386-expand.c
> index c0924a59efb..de4fe9ce147 100644
> --- a/gcc/config/i386/i386-expand.c
> +++ b/gcc/config/i386/i386-expand.c
> @@ -10877,7 +10877,27 @@ ix86_expand_round_builtin (const struct builtin_description *d,
>
>           /* If there is no rounding use normal version of the pattern.  */
>           if (INTVAL (op) == NO_ROUND)
> -           redundant_embed_rnd = 1;
> +           {
> +             /* Skip erasing embedded rounding for below expanders who
> +                generates multiple insns.  In ix86_erase_embedded_rounding
> +                the pattern will be transformed to a single set, and emit_insn
> +                appends the set insead of insert it to chain.  So the insns
> +                emitted inside define_expander would be ignored.  */
> +             switch (icode)
> +               {
> +               case CODE_FOR_avx512bw_fmaddc_v32hf_mask1_round:
> +               case CODE_FOR_avx512bw_fcmaddc_v32hf_mask1_round:
> +               case CODE_FOR_avx512fp16_fmaddcsh_v8hf_mask1_round:
> +               case CODE_FOR_avx512fp16_fcmaddcsh_v8hf_mask1_round:
> +               case CODE_FOR_avx512fp16_fmaddcsh_v8hf_mask3_round:
> +               case CODE_FOR_avx512fp16_fcmaddcsh_v8hf_mask3_round:
> +                 redundant_embed_rnd = 0;
> +                 break;
> +               default:
> +                 redundant_embed_rnd = 1;
> +                 break;
> +               }
> +           }
>         }
>        else
>         {
> diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md
> index a3c4a3f1e62..bdd34f379dd 100644
> --- a/gcc/config/i386/sse.md
> +++ b/gcc/config/i386/sse.md
> @@ -5864,6 +5864,34 @@
>          (UNSPEC_COMPLEX_FMUL "fmulc")
>          (UNSPEC_COMPLEX_FCMUL "fcmulc")])
>
> +(define_mode_attr complexmove
> +  [(V32HF "avx512f_loadv16sf")
> +   (V16HF "avx512vl_loadv8sf")
> +   (V8HF "avx512vl_loadv4sf")])
> +
> +(define_expand "<avx512>_fmaddc_<mode>_mask1<round_expand_name>"
> +  [(match_operand:VF_AVX512FP16VL 0 "register_operand")
> +   (match_operand:VF_AVX512FP16VL 1 "<round_expand_nimm_predicate>")
> +   (match_operand:VF_AVX512FP16VL 2 "<round_expand_nimm_predicate>")
> +   (match_operand:VF_AVX512FP16VL 3 "<round_expand_nimm_predicate>")
> +   (match_operand:<avx512fmaskcmode> 4 "register_operand")]
> +  "TARGET_AVX512FP16 && <round_mode512bit_condition>"
> +{
> +  rtx op0, op1;
> +  if (<round_embedded_complex>)
> +    emit_insn (gen_<avx512>_fmaddc_<mode>_mask<round_expand_name> (
> +      operands[0], operands[1], operands[2], operands[3],
> +      operands[4]<round_expand_operand>));
> +  else
> +    emit_insn (gen_<avx512>_fmaddc_<mode>_mask (operands[0],
> +      operands[1], operands[2], operands[3], operands[4]));
> +
> +  op0 = lowpart_subreg (<ssePSmode>mode, operands[0], <MODE>mode);
> +  op1 = lowpart_subreg (<ssePSmode>mode, operands[1], <MODE>mode);
> +  emit_insn (gen_<complexmove>_mask (op0, op0, op1, operands[4]));
> +  DONE;
> +})
> +
>  (define_expand "<avx512>_fmaddc_<mode>_maskz<round_expand_name>"
>    [(match_operand:VF_AVX512FP16VL 0 "register_operand")
>     (match_operand:VF_AVX512FP16VL 1 "<round_expand_nimm_predicate>")
> @@ -5878,6 +5906,31 @@
>    DONE;
>  })
>
> +(define_expand "<avx512>_fcmaddc_<mode>_mask1<round_expand_name>"
> +  [(match_operand:VF_AVX512FP16VL 0 "register_operand")
> +   (match_operand:VF_AVX512FP16VL 1 "<round_expand_nimm_predicate>")
> +   (match_operand:VF_AVX512FP16VL 2 "<round_expand_nimm_predicate>")
> +   (match_operand:VF_AVX512FP16VL 3 "<round_expand_nimm_predicate>")
> +   (match_operand:<avx512fmaskcmode> 4 "register_operand")]
> +  "TARGET_AVX512FP16 && <round_mode512bit_condition>"
> +{
> +  rtx op0, op1;
> +  if (<round_embedded_complex>)
> +    emit_insn (gen_<avx512>_fcmaddc_<mode>_mask<round_expand_name> (
> +      operands[0], operands[1], operands[2], operands[3],
> +      operands[4]<round_expand_operand>));
> +  else
> +  {
> +    emit_insn (gen_<avx512>_fcmaddc_<mode>_mask (operands[0],
> +      operands[1], operands[2], operands[3], operands[4]));
> +  }
> +
> +  op0 = lowpart_subreg (<ssePSmode>mode, operands[0], <MODE>mode);
> +  op1 = lowpart_subreg (<ssePSmode>mode, operands[1], <MODE>mode);
> +  emit_insn (gen_<complexmove>_mask (op0, op0, op1, operands[4]));
> +  DONE;
> +})
> +
>  (define_expand "<avx512>_fcmaddc_<mode>_maskz<round_expand_name>"
>    [(match_operand:VF_AVX512FP16VL 0 "register_operand")
>     (match_operand:VF_AVX512FP16VL 1 "<round_expand_nimm_predicate>")
> @@ -5946,6 +5999,47 @@
>    DONE;
>  })
>
> +(define_expand "avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>"
> +  [(match_operand:V8HF 0 "register_operand")
> +   (match_operand:V8HF 1 "<round_expand_nimm_predicate>")
> +   (match_operand:V8HF 2 "<round_expand_nimm_predicate>")
> +   (match_operand:V8HF 3 "<round_expand_nimm_predicate>")
> +   (match_operand:QI 4 "register_operand")]
> +  "TARGET_AVX512FP16 && <round_mode512bit_condition>"
> +{
> +  rtx op0, op1;
> +
> +  if (<round_embedded_complex>)
> +    emit_insn (gen_avx512fp16_fmaddcsh_v8hf_mask<round_expand_name> (
> +      operands[0], operands[1], operands[2], operands[3],
> +      operands[4]<round_expand_operand>));
> +  else
> +    emit_insn (gen_avx512fp16_fmaddcsh_v8hf_mask (operands[0],
> +      operands[1], operands[2], operands[3], operands[4]));
> +
> +  if (TARGET_AVX512VL)
> +  {
> +    op0 = lowpart_subreg (V4SFmode, operands[0], V8HFmode);
> +    op1 = lowpart_subreg (V4SFmode, operands[1], V8HFmode);
> +    emit_insn (gen_avx512vl_loadv4sf_mask (op0, op0, op1, operands[4]));
> +  }
> +  else
> +  {
> +    rtx mask, tmp, vec_mask;
> +    mask = lowpart_subreg (SImode, operands[4], QImode),
> +    tmp = gen_reg_rtx (SImode);
> +    emit_insn (gen_ashlsi3 (tmp, mask, GEN_INT (31)));
> +    vec_mask = gen_reg_rtx (V4SImode);
> +    emit_insn (gen_rtx_SET (vec_mask, CONST0_RTX (V4SImode)));
> +    emit_insn (gen_vec_setv4si_0 (vec_mask, vec_mask, tmp));
> +    vec_mask = lowpart_subreg (V4SFmode, vec_mask, V4SImode);
> +    op0 = lowpart_subreg (V4SFmode, operands[0], V8HFmode);
> +    op1 = lowpart_subreg (V4SFmode, operands[1], V8HFmode);
> +    emit_insn (gen_sse4_1_blendvps (op0, op1, op0, vec_mask));
> +  }
> +  DONE;
> +})
> +
>  (define_expand "avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>"
>    [(match_operand:V8HF 0 "register_operand")
>     (match_operand:V8HF 1 "<round_expand_nimm_predicate>")
> @@ -5960,6 +6054,95 @@
>    DONE;
>  })
>
> +(define_expand "avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>"
> +  [(match_operand:V8HF 0 "register_operand")
> +   (match_operand:V8HF 1 "<round_expand_nimm_predicate>")
> +   (match_operand:V8HF 2 "<round_expand_nimm_predicate>")
> +   (match_operand:V8HF 3 "<round_expand_nimm_predicate>")
> +   (match_operand:QI 4 "register_operand")]
> +  "TARGET_AVX512FP16 && <round_mode512bit_condition>"
> +{
> +  rtx op0, op1;
> +
> +  if (<round_embedded_complex>)
> +    emit_insn (gen_avx512fp16_fcmaddcsh_v8hf_mask<round_expand_name> (
> +      operands[0], operands[1], operands[2], operands[3],
> +      operands[4]<round_expand_operand>));
> +  else
> +    emit_insn (gen_avx512fp16_fcmaddcsh_v8hf_mask (operands[0],
> +      operands[1], operands[2], operands[3], operands[4]));
> +
> +  if (TARGET_AVX512VL)
> +  {
> +    op0 = lowpart_subreg (V4SFmode, operands[0], V8HFmode);
> +    op1 = lowpart_subreg (V4SFmode, operands[1], V8HFmode);
> +    emit_insn (gen_avx512vl_loadv4sf_mask (op0, op0, op1, operands[4]));
> +  }
> +  else
> +  {
> +    rtx mask, tmp, vec_mask;
> +    mask = lowpart_subreg (SImode, operands[4], QImode),
> +    tmp = gen_reg_rtx (SImode);
> +    emit_insn (gen_ashlsi3 (tmp, mask, GEN_INT (31)));
> +    vec_mask = gen_reg_rtx (V4SImode);
> +    emit_insn (gen_rtx_SET (vec_mask, CONST0_RTX (V4SImode)));
> +    emit_insn (gen_vec_setv4si_0 (vec_mask, vec_mask, tmp));
> +    vec_mask = lowpart_subreg (V4SFmode, vec_mask, V4SImode);
> +    op0 = lowpart_subreg (V4SFmode, operands[0], V8HFmode);
> +    op1 = lowpart_subreg (V4SFmode, operands[1], V8HFmode);
> +    emit_insn (gen_sse4_1_blendvps (op0, op1, op0, vec_mask));
> +  }
> +  DONE;
> +})
> +
> +(define_expand "avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>"
> +  [(match_operand:V8HF 0 "register_operand")
> +   (match_operand:V8HF 1 "<round_expand_nimm_predicate>")
> +   (match_operand:V8HF 2 "<round_expand_nimm_predicate>")
> +   (match_operand:V8HF 3 "<round_expand_nimm_predicate>")
> +   (match_operand:QI 4 "register_operand")]
> +  "TARGET_AVX512FP16 && <round_mode512bit_condition>"
> +{
> +  rtx op0, op1;
> +
> +  if (<round_embedded_complex>)
> +    emit_insn (gen_avx512fp16_fcmaddcsh_v8hf_mask<round_expand_name> (
> +      operands[0], operands[1], operands[2], operands[3],
> +      operands[4]<round_expand_operand>));
> +  else
> +    emit_insn (gen_avx512fp16_fcmaddcsh_v8hf_mask (operands[0],
> +      operands[1], operands[2], operands[3], operands[4]));
> +
> +  op0 = lowpart_subreg (V4SFmode, operands[0], V8HFmode);
> +  op1 = lowpart_subreg (V4SFmode, operands[3], V8HFmode);
> +  emit_insn (gen_sse_movss (op0, op1, op0));
> +  DONE;
> +})
> +
> +(define_expand "avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>"
> +  [(match_operand:V8HF 0 "register_operand")
> +   (match_operand:V8HF 1 "<round_expand_nimm_predicate>")
> +   (match_operand:V8HF 2 "<round_expand_nimm_predicate>")
> +   (match_operand:V8HF 3 "<round_expand_nimm_predicate>")
> +   (match_operand:QI 4 "register_operand")]
> +  "TARGET_AVX512FP16 && <round_mode512bit_condition>"
> +{
> +  rtx op0, op1;
> +
> +  if (<round_embedded_complex>)
> +    emit_insn (gen_avx512fp16_fmaddcsh_v8hf_mask<round_expand_name> (
> +      operands[0], operands[1], operands[2], operands[3],
> +      operands[4]<round_expand_operand>));
> +  else
> +    emit_insn (gen_avx512fp16_fmaddcsh_v8hf_mask (operands[0],
> +      operands[1], operands[2], operands[3], operands[4]));
> +
> +  op0 = lowpart_subreg (V4SFmode, operands[0], V8HFmode);
> +  op1 = lowpart_subreg (V4SFmode, operands[3], V8HFmode);
> +  emit_insn (gen_sse_movss (op0, op1, op0));
> +  DONE;
> +})
> +
>  (define_insn "avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>"
>    [(set (match_operand:V8HF 0 "register_operand" "=&v")
>         (vec_merge:V8HF
> diff --git a/gcc/config/i386/subst.md b/gcc/config/i386/subst.md
> index 11e62c67fcc..510e82c84f9 100644
> --- a/gcc/config/i386/subst.md
> +++ b/gcc/config/i386/subst.md
> @@ -276,6 +276,9 @@
>  (define_subst_attr "round_expand_name" "round_expand" "" "_round")
>  (define_subst_attr "round_expand_nimm_predicate" "round_expand" "nonimmediate_operand" "register_operand")
>  (define_subst_attr "round_expand_operand" "round_expand" "" ", operands[5]")
> +(define_subst_attr "round_embedded_complex" "round_expand" "0" "!(CONST_INT_P (operands[5])
> +                                                                 && (INTVAL (operands[5])
> +                                                                     == NO_ROUND))")
>
>  (define_subst "round_expand"
>   [(match_operand:SUBST_V 0)
> diff --git a/gcc/testsuite/gcc.target/i386/avx-1.c b/gcc/testsuite/gcc.target/i386/avx-1.c
> index 8744aa7df55..154e7b3b107 100644
> --- a/gcc/testsuite/gcc.target/i386/avx-1.c
> +++ b/gcc/testsuite/gcc.target/i386/avx-1.c
> @@ -789,9 +789,11 @@
>  #define __builtin_ia32_vfnmsubsh3_maskz(A, B, C, D, E) __builtin_ia32_vfnmsubsh3_maskz(A, B, C, D, 8)
>  #define __builtin_ia32_vfcmaddcph512_round(A, B, C, D) __builtin_ia32_vfcmaddcph512_round(A, B, C, 8)
>  #define __builtin_ia32_vfcmaddcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfcmaddcph512_mask_round(A, C, D, B, 8)
> +#define __builtin_ia32_vfcmaddcph512_mask3_round(A, C, D, B, E) __builtin_ia32_vfcmaddcph512_mask3_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfcmaddcph512_maskz_round(B, C, D, A, E) __builtin_ia32_vfcmaddcph512_maskz_round(B, C, D, A, 8)
>  #define __builtin_ia32_vfmaddcph512_round(A, B, C, D) __builtin_ia32_vfmaddcph512_round(A, B, C, 8)
>  #define __builtin_ia32_vfmaddcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfmaddcph512_mask_round(A, C, D, B, 8)
> +#define __builtin_ia32_vfmaddcph512_mask3_round(A, C, D, B, E) __builtin_ia32_vfmaddcph512_mask3_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfmaddcph512_maskz_round(B, C, D, A, E) __builtin_ia32_vfmaddcph512_maskz_round(B, C, D, A, 8)
>  #define __builtin_ia32_vfmulcph512_round(A, B, C) __builtin_ia32_vfmulcph512_round(A, B, 8)
>  #define __builtin_ia32_vfmulcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfmulcph512_mask_round(A, C, D, B, 8)
> @@ -799,9 +801,11 @@
>  #define __builtin_ia32_vfcmulcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfcmulcph512_mask_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfmaddcsh_round(A, B, C, D) __builtin_ia32_vfmaddcsh_round(A, B, C, 8)
>  #define __builtin_ia32_vfmaddcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfmaddcsh_mask_round(A, C, D, B, 8)
> +#define __builtin_ia32_vfmaddcsh_mask3_round(A, C, D, B, E) __builtin_ia32_vfmaddcsh_mask3_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfmaddcsh_maskz_round(B, C, D, A, E) __builtin_ia32_vfmaddcsh_maskz_round(B, C, D, A, 8)
>  #define __builtin_ia32_vfcmaddcsh_round(A, B, C, D) __builtin_ia32_vfcmaddcsh_round(A, B, C, 8)
>  #define __builtin_ia32_vfcmaddcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfcmaddcsh_mask_round(A, C, D, B, 8)
> +#define __builtin_ia32_vfcmaddcsh_mask3_round(A, C, D, B, E) __builtin_ia32_vfcmaddcsh_mask3_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfcmaddcsh_maskz_round(B, C, D, A, E) __builtin_ia32_vfcmaddcsh_maskz_round(B, C, D, A, 8)
>  #define __builtin_ia32_vfmulcsh_round(A, B, C) __builtin_ia32_vfmulcsh_round(A, B, 8)
>  #define __builtin_ia32_vfmulcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfmulcsh_mask_round(A, C, D, B, 8)
> diff --git a/gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1a.c b/gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1a.c
> index 8bd8eebd8df..8ff2092c325 100644
> --- a/gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1a.c
> +++ b/gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1a.c
> @@ -6,6 +6,8 @@
>  /* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
>  /* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
>  /* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\{rz-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\{z\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 1 } } */
> +/* { dg-final { scan-assembler-times "vblendvps\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */
> +/* { dg-final { scan-assembler-times "vmovss\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */
>
>  #include <immintrin.h>
>
> diff --git a/gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1c.c b/gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1c.c
> new file mode 100644
> index 00000000000..79a295f722c
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1c.c
> @@ -0,0 +1,13 @@
> +/* { dg-do compile } */
> +/* { dg-options "-mavx512fp16 -mavx512vl -O2" } */
> +/* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
> +/* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\{\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
> +/* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\{z\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 1 } } */
> +/* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
> +/* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
> +/* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\{rz-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\{z\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 1 } } */
> +/* { dg-final { scan-assembler-times "vblendmps\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\{\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
> +/* { dg-final { scan-assembler-times "vmovss\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */
> +
> +#include "avx512fp16-vfcmaddcsh-1a.c"
> +
> diff --git a/gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1a.c b/gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1a.c
> index 1e376b4a2bb..2ebe1f8ddd7 100644
> --- a/gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1a.c
> +++ b/gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1a.c
> @@ -6,6 +6,8 @@
>  /* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
>  /* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
>  /* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\{rz-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\{z\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 1 } } */
> +/* { dg-final { scan-assembler-times "vblendvps\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */
> +/* { dg-final { scan-assembler-times "vmovss\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */
>
>  #include <immintrin.h>
>
> diff --git a/gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1c.c b/gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1c.c
> new file mode 100644
> index 00000000000..7863f8f9af9
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1c.c
> @@ -0,0 +1,13 @@
> +/* { dg-do compile } */
> +/* { dg-options "-mavx512fp16 -mavx512vl -O2" } */
> +/* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
> +/* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\{\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
> +/* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\{z\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 1 } } */
> +/* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
> +/* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
> +/* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\{rz-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\{z\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 1 } } */
> +/* { dg-final { scan-assembler-times "vblendmps\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\{\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
> +/* { dg-final { scan-assembler-times "vmovss\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */
> +
> +#include "avx512fp16-vfmaddcsh-1a.c"
> +
> diff --git a/gcc/testsuite/gcc.target/i386/sse-13.c b/gcc/testsuite/gcc.target/i386/sse-13.c
> index f6d54e3a815..e285c307d00 100644
> --- a/gcc/testsuite/gcc.target/i386/sse-13.c
> +++ b/gcc/testsuite/gcc.target/i386/sse-13.c
> @@ -806,9 +806,11 @@
>  #define __builtin_ia32_vfnmsubsh3_maskz(A, B, C, D, E) __builtin_ia32_vfnmsubsh3_maskz(A, B, C, D, 8)
>  #define __builtin_ia32_vfcmaddcph512_round(A, B, C, D) __builtin_ia32_vfcmaddcph512_round(A, B, C, 8)
>  #define __builtin_ia32_vfcmaddcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfcmaddcph512_mask_round(A, C, D, B, 8)
> +#define __builtin_ia32_vfcmaddcph512_mask3_round(A, C, D, B, E) __builtin_ia32_vfcmaddcph512_mask3_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfcmaddcph512_maskz_round(B, C, D, A, E) __builtin_ia32_vfcmaddcph512_maskz_round(B, C, D, A, 8)
>  #define __builtin_ia32_vfmaddcph512_round(A, B, C, D) __builtin_ia32_vfmaddcph512_round(A, B, C, 8)
>  #define __builtin_ia32_vfmaddcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfmaddcph512_mask_round(A, C, D, B, 8)
> +#define __builtin_ia32_vfmaddcph512_mask3_round(A, C, D, B, E) __builtin_ia32_vfmaddcph512_mask3_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfmaddcph512_maskz_round(B, C, D, A, E) __builtin_ia32_vfmaddcph512_maskz_round(B, C, D, A, 8)
>  #define __builtin_ia32_vfmulcph512_round(A, B, C) __builtin_ia32_vfmulcph512_round(A, B, 8)
>  #define __builtin_ia32_vfmulcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfmulcph512_mask_round(A, C, D, B, 8)
> @@ -816,9 +818,11 @@
>  #define __builtin_ia32_vfcmulcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfcmulcph512_mask_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfmaddcsh_round(A, B, C, D) __builtin_ia32_vfmaddcsh_round(A, B, C, 8)
>  #define __builtin_ia32_vfmaddcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfmaddcsh_mask_round(A, C, D, B, 8)
> +#define __builtin_ia32_vfmaddcsh_mask3_round(A, C, D, B, E) __builtin_ia32_vfmaddcsh_mask3_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfmaddcsh_maskz_round(B, C, D, A, E) __builtin_ia32_vfmaddcsh_maskz_round(B, C, D, A, 8)
>  #define __builtin_ia32_vfcmaddcsh_round(A, B, C, D) __builtin_ia32_vfcmaddcsh_round(A, B, C, 8)
>  #define __builtin_ia32_vfcmaddcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfcmaddcsh_mask_round(A, C, D, B, 8)
> +#define __builtin_ia32_vfcmaddcsh_mask3_round(A, C, D, B, E) __builtin_ia32_vfcmaddcsh_mask3_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfcmaddcsh_maskz_round(B, C, D, A, E) __builtin_ia32_vfcmaddcsh_maskz_round(B, C, D, A, 8)
>  #define __builtin_ia32_vfmulcsh_round(A, B, C) __builtin_ia32_vfmulcsh_round(A, B, 8)
>  #define __builtin_ia32_vfmulcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfmulcsh_mask_round(A, C, D, B, 8)
> diff --git a/gcc/testsuite/gcc.target/i386/sse-23.c b/gcc/testsuite/gcc.target/i386/sse-23.c
> index 98251269a64..f71a7b29157 100644
> --- a/gcc/testsuite/gcc.target/i386/sse-23.c
> +++ b/gcc/testsuite/gcc.target/i386/sse-23.c
> @@ -807,9 +807,11 @@
>  #define __builtin_ia32_vfnmsubsh3_maskz(A, B, C, D, E) __builtin_ia32_vfnmsubsh3_maskz(A, B, C, D, 8)
>  #define __builtin_ia32_vfcmaddcph512_round(A, B, C, D) __builtin_ia32_vfcmaddcph512_round(A, B, C, 8)
>  #define __builtin_ia32_vfcmaddcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfcmaddcph512_mask_round(A, C, D, B, 8)
> +#define __builtin_ia32_vfcmaddcph512_mask3_round(A, C, D, B, E) __builtin_ia32_vfcmaddcph512_mask3_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfcmaddcph512_maskz_round(B, C, D, A, E) __builtin_ia32_vfcmaddcph512_maskz_round(B, C, D, A, 8)
>  #define __builtin_ia32_vfmaddcph512_round(A, B, C, D) __builtin_ia32_vfmaddcph512_round(A, B, C, 8)
>  #define __builtin_ia32_vfmaddcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfmaddcph512_mask_round(A, C, D, B, 8)
> +#define __builtin_ia32_vfmaddcph512_mask3_round(A, C, D, B, E) __builtin_ia32_vfmaddcph512_mask3_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfmaddcph512_maskz_round(B, C, D, A, E) __builtin_ia32_vfmaddcph512_maskz_round(B, C, D, A, 8)
>  #define __builtin_ia32_vfmulcph512_round(A, B, C) __builtin_ia32_vfmulcph512_round(A, B, 8)
>  #define __builtin_ia32_vfmulcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfmulcph512_mask_round(A, C, D, B, 8)
> @@ -817,9 +819,11 @@
>  #define __builtin_ia32_vfcmulcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfcmulcph512_mask_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfmaddcsh_round(A, B, C, D) __builtin_ia32_vfmaddcsh_round(A, B, C, 8)
>  #define __builtin_ia32_vfmaddcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfmaddcsh_mask_round(A, C, D, B, 8)
> +#define __builtin_ia32_vfmaddcsh_mask3_round(A, C, D, B, E) __builtin_ia32_vfmaddcsh_mask3_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfmaddcsh_maskz_round(B, C, D, A, E) __builtin_ia32_vfmaddcsh_maskz_round(B, C, D, A, 8)
>  #define __builtin_ia32_vfcmaddcsh_round(A, B, C, D) __builtin_ia32_vfcmaddcsh_round(A, B, C, 8)
>  #define __builtin_ia32_vfcmaddcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfcmaddcsh_mask_round(A, C, D, B, 8)
> +#define __builtin_ia32_vfcmaddcsh_mask3_round(A, C, D, B, E) __builtin_ia32_vfcmaddcsh_mask3_round(A, C, D, B, 8)
>  #define __builtin_ia32_vfcmaddcsh_maskz_round(B, C, D, A, E) __builtin_ia32_vfcmaddcsh_maskz_round(B, C, D, A, 8)
>  #define __builtin_ia32_vfmulcsh_round(A, B, C) __builtin_ia32_vfmulcsh_round(A, B, 8)
>  #define __builtin_ia32_vfmulcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfmulcsh_mask_round(A, C, D, B, 8)
> --
> 2.18.1
>
diff mbox series

Patch

diff --git a/gcc/config/i386/avx512fp16intrin.h b/gcc/config/i386/avx512fp16intrin.h
index 29cf6792335..5e49447a020 100644
--- a/gcc/config/i386/avx512fp16intrin.h
+++ b/gcc/config/i386/avx512fp16intrin.h
@@ -6258,13 +6258,11 @@  extern __inline __m512h
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm512_mask_fcmadd_pch (__m512h __A, __mmask16 __B, __m512h __C, __m512h __D)
 {
-  return (__m512h) __builtin_ia32_movaps512_mask
-    ((__v16sf)
-     __builtin_ia32_vfcmaddcph512_mask_round ((__v32hf) __A,
-					      (__v32hf) __C,
-					      (__v32hf) __D, __B,
-					      _MM_FROUND_CUR_DIRECTION),
-     (__v16sf) __A, __B);
+  return (__m512h)
+    __builtin_ia32_vfcmaddcph512_mask_round ((__v32hf) __A,
+					     (__v32hf) __C,
+					     (__v32hf) __D, __B,
+					     _MM_FROUND_CUR_DIRECTION);
 }
 
 extern __inline __m512h
@@ -6272,10 +6270,10 @@  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm512_mask3_fcmadd_pch (__m512h __A, __m512h __B, __m512h __C, __mmask16 __D)
 {
   return (__m512h)
-    __builtin_ia32_vfcmaddcph512_mask_round ((__v32hf) __A,
-					     (__v32hf) __B,
-					     (__v32hf) __C,
-					     __D, _MM_FROUND_CUR_DIRECTION);
+    __builtin_ia32_vfcmaddcph512_mask3_round ((__v32hf) __A,
+					      (__v32hf) __B,
+					      (__v32hf) __C,
+					      __D, _MM_FROUND_CUR_DIRECTION);
 }
 
 extern __inline __m512h
@@ -6304,13 +6302,11 @@  extern __inline __m512h
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm512_mask_fmadd_pch (__m512h __A, __mmask16 __B, __m512h __C, __m512h __D)
 {
-  return (__m512h) __builtin_ia32_movaps512_mask
-    ((__v16sf)
-     __builtin_ia32_vfmaddcph512_mask_round ((__v32hf) __A,
-					     (__v32hf) __C,
-					     (__v32hf) __D, __B,
-					     _MM_FROUND_CUR_DIRECTION),
-     (__v16sf) __A, __B);
+  return (__m512h)
+    __builtin_ia32_vfmaddcph512_mask_round ((__v32hf) __A,
+					    (__v32hf) __C,
+					    (__v32hf) __D, __B,
+					    _MM_FROUND_CUR_DIRECTION);
 }
 
 extern __inline __m512h
@@ -6318,10 +6314,10 @@  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm512_mask3_fmadd_pch (__m512h __A, __m512h __B, __m512h __C, __mmask16 __D)
 {
   return (__m512h)
-    __builtin_ia32_vfmaddcph512_mask_round ((__v32hf) __A,
-					    (__v32hf) __B,
-					    (__v32hf) __C,
-					    __D, _MM_FROUND_CUR_DIRECTION);
+    __builtin_ia32_vfmaddcph512_mask3_round ((__v32hf) __A,
+					     (__v32hf) __B,
+					     (__v32hf) __C,
+					     __D, _MM_FROUND_CUR_DIRECTION);
 }
 
 extern __inline __m512h
@@ -6352,13 +6348,11 @@  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm512_mask_fcmadd_round_pch (__m512h __A, __mmask16 __B, __m512h __C,
 			      __m512h __D, const int __E)
 {
-  return (__m512h) __builtin_ia32_movaps512_mask
-    ((__v16sf)
-     __builtin_ia32_vfcmaddcph512_mask_round ((__v32hf) __A,
-					      (__v32hf) __C,
-					      (__v32hf) __D, __B,
-					      __E),
-     (__v16sf) __A, __B);
+  return (__m512h)
+    __builtin_ia32_vfcmaddcph512_mask_round ((__v32hf) __A,
+					     (__v32hf) __C,
+					     (__v32hf) __D, __B,
+					     __E);
 }
 
 extern __inline __m512h
@@ -6367,10 +6361,10 @@  _mm512_mask3_fcmadd_round_pch (__m512h __A, __m512h __B, __m512h __C,
 			       __mmask16 __D, const int __E)
 {
   return (__m512h)
-    __builtin_ia32_vfcmaddcph512_mask_round ((__v32hf) __A,
-					     (__v32hf) __B,
-					     (__v32hf) __C,
-					     __D, __E);
+    __builtin_ia32_vfcmaddcph512_mask3_round ((__v32hf) __A,
+					      (__v32hf) __B,
+					      (__v32hf) __C,
+					      __D, __E);
 }
 
 extern __inline __m512h
@@ -6401,13 +6395,11 @@  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm512_mask_fmadd_round_pch (__m512h __A, __mmask16 __B, __m512h __C,
 			     __m512h __D, const int __E)
 {
-  return (__m512h) __builtin_ia32_movaps512_mask
-    ((__v16sf)
-     __builtin_ia32_vfmaddcph512_mask_round ((__v32hf) __A,
-					     (__v32hf) __C,
-					     (__v32hf) __D, __B,
-					     __E),
-     (__v16sf) __A, __B);
+  return (__m512h)
+    __builtin_ia32_vfmaddcph512_mask_round ((__v32hf) __A,
+					    (__v32hf) __C,
+					    (__v32hf) __D, __B,
+					    __E);
 }
 
 extern __inline __m512h
@@ -6416,10 +6408,10 @@  _mm512_mask3_fmadd_round_pch (__m512h __A, __m512h __B, __m512h __C,
 			      __mmask16 __D, const int __E)
 {
   return (__m512h)
-    __builtin_ia32_vfmaddcph512_mask_round ((__v32hf) __A,
-					    (__v32hf) __B,
-					    (__v32hf) __C,
-					    __D, __E);
+    __builtin_ia32_vfmaddcph512_mask3_round ((__v32hf) __A,
+					     (__v32hf) __B,
+					     (__v32hf) __C,
+					     __D, __E);
 }
 
 extern __inline __m512h
@@ -6439,18 +6431,16 @@  _mm512_maskz_fmadd_round_pch (__mmask16 __A, __m512h __B, __m512h __C,
   (__m512h) __builtin_ia32_vfcmaddcph512_round ((A), (B), (C), (D))
 
 #define _mm512_mask_fcmadd_round_pch(A, B, C, D, E)			\
-  ((__m512h) __builtin_ia32_movaps512_mask (				\
-   (__v16sf)								\
+  ((__m512h) 								\
     __builtin_ia32_vfcmaddcph512_mask_round ((__v32hf) (A),		\
 					     (__v32hf) (C),		\
 					     (__v32hf) (D),		\
-					     (B), (E)),			\
-					     (__v16sf) (A), (B)));
+					     (B), (E)))
 
 
 #define _mm512_mask3_fcmadd_round_pch(A, B, C, D, E)			\
   ((__m512h)								\
-   __builtin_ia32_vfcmaddcph512_mask_round ((A), (B), (C), (D), (E)))
+   __builtin_ia32_vfcmaddcph512_mask3_round ((A), (B), (C), (D), (E)))
 
 #define _mm512_maskz_fcmadd_round_pch(A, B, C, D, E)			\
   (__m512h)								\
@@ -6460,17 +6450,15 @@  _mm512_maskz_fmadd_round_pch (__mmask16 __A, __m512h __B, __m512h __C,
   (__m512h) __builtin_ia32_vfmaddcph512_round ((A), (B), (C), (D))
 
 #define _mm512_mask_fmadd_round_pch(A, B, C, D, E)			\
-  ((__m512h) __builtin_ia32_movaps512_mask (				\
-   (__v16sf)								\
+  ((__m512h)								\
     __builtin_ia32_vfmaddcph512_mask_round ((__v32hf) (A),		\
 					    (__v32hf) (C),		\
 					    (__v32hf) (D),		\
-					    (B), (E)),			\
-					    (__v16sf) (A), (B)));
+					    (B), (E)))
 
 #define _mm512_mask3_fmadd_round_pch(A, B, C, D, E)			\
   (__m512h)								\
-   __builtin_ia32_vfmaddcph512_mask_round ((A), (B), (C), (D), (E))
+   __builtin_ia32_vfmaddcph512_mask3_round ((A), (B), (C), (D), (E))
 
 #define _mm512_maskz_fmadd_round_pch(A, B, C, D, E)			\
   (__m512h)								\
@@ -6643,35 +6631,22 @@  extern __inline __m128h
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_mask_fcmadd_sch (__m128h __A, __mmask8 __B, __m128h __C, __m128h __D)
 {
-#ifdef __AVX512VL__
-  return (__m128h) __builtin_ia32_movaps128_mask (
-    (__v4sf)
-    __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) __A,
-					  (__v8hf) __C,
-					  (__v8hf) __D, __B,
-					  _MM_FROUND_CUR_DIRECTION),
-    (__v4sf) __A, __B);
-#else
-  return (__m128h) __builtin_ia32_blendvps ((__v4sf) __A,
-    (__v4sf)
+  return (__m128h)
     __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) __A,
 					  (__v8hf) __C,
 					  (__v8hf) __D, __B,
-					  _MM_FROUND_CUR_DIRECTION),
-    (__v4sf) _mm_set_ss ((float) ((int) __B << 31)));
-#endif
+					  _MM_FROUND_CUR_DIRECTION);
 }
 
 extern __inline __m128h
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_mask3_fcmadd_sch (__m128h __A, __m128h __B, __m128h __C, __mmask8 __D)
 {
-  return (__m128h) _mm_move_ss ((__m128) __C,
-    (__m128)
-    __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) __A,
-					  (__v8hf) __B,
-					  (__v8hf) __C, __D,
-					  _MM_FROUND_CUR_DIRECTION));
+  return (__m128h)
+    __builtin_ia32_vfcmaddcsh_mask3_round ((__v8hf) __A,
+					   (__v8hf) __B,
+					   (__v8hf) __C, __D,
+					   _MM_FROUND_CUR_DIRECTION);
 }
 
 extern __inline __m128h
@@ -6700,35 +6675,22 @@  extern __inline __m128h
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_mask_fmadd_sch (__m128h __A, __mmask8 __B, __m128h __C, __m128h __D)
 {
-#ifdef __AVX512VL__
-  return (__m128h) __builtin_ia32_movaps128_mask (
-    (__v4sf)
-    __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) __A,
-					 (__v8hf) __C,
-					 (__v8hf) __D, __B,
-					 _MM_FROUND_CUR_DIRECTION),
-    (__v4sf) __A, __B);
-#else
-  return (__m128h) __builtin_ia32_blendvps ((__v4sf) __A,
-    (__v4sf)
+  return (__m128h)
     __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) __A,
 					 (__v8hf) __C,
 					 (__v8hf) __D, __B,
-					 _MM_FROUND_CUR_DIRECTION),
-    (__v4sf) _mm_set_ss ((float) ((int) __B << 31)));
-#endif
+					 _MM_FROUND_CUR_DIRECTION);
 }
 
 extern __inline __m128h
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_mask3_fmadd_sch (__m128h __A, __m128h __B, __m128h __C, __mmask8 __D)
 {
-  return (__m128h) _mm_move_ss ((__m128) __C,
-    (__m128)
-    __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) __A,
-					 (__v8hf) __B,
-					 (__v8hf) __C, __D,
-					 _MM_FROUND_CUR_DIRECTION));
+  return (__m128h)
+    __builtin_ia32_vfmaddcsh_mask3_round ((__v8hf) __A,
+					  (__v8hf) __B,
+					  (__v8hf) __C, __D,
+					  _MM_FROUND_CUR_DIRECTION);
 }
 
 extern __inline __m128h
@@ -6759,23 +6721,11 @@  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_mask_fcmadd_round_sch (__m128h __A, __mmask8 __B, __m128h __C,
 			   __m128h __D, const int __E)
 {
-#ifdef __AVX512VL__
-  return (__m128h) __builtin_ia32_movaps128_mask (
-    (__v4sf)
-    __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) __A,
-					  (__v8hf) __C,
-					  (__v8hf) __D,
-					  __B, __E),
-    (__v4sf) __A, __B);
-#else
-  return (__m128h) __builtin_ia32_blendvps ((__v4sf) __A,
-    (__v4sf)
+  return (__m128h)
     __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) __A,
 					  (__v8hf) __C,
 					  (__v8hf) __D,
-					  __B, __E),
-    (__v4sf) _mm_set_ss ((float) ((int) __B << 31)));
-#endif
+					  __B, __E);
 }
 
 extern __inline __m128h
@@ -6783,12 +6733,11 @@  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_mask3_fcmadd_round_sch (__m128h __A, __m128h __B, __m128h __C,
 			    __mmask8 __D, const int __E)
 {
-  return (__m128h) _mm_move_ss ((__m128) __C,
-    (__m128)
-    __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) __A,
-					  (__v8hf) __B,
-					  (__v8hf) __C,
-					  __D, __E));
+  return (__m128h)
+    __builtin_ia32_vfcmaddcsh_mask3_round ((__v8hf) __A,
+					   (__v8hf) __B,
+					   (__v8hf) __C,
+					   __D, __E);
 }
 
 extern __inline __m128h
@@ -6819,23 +6768,11 @@  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_mask_fmadd_round_sch (__m128h __A, __mmask8 __B, __m128h __C,
 			  __m128h __D, const int __E)
 {
-#ifdef __AVX512VL__
-  return (__m128h) __builtin_ia32_movaps128_mask (
-    (__v4sf)
-    __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) __A,
-					 (__v8hf) __C,
-					 (__v8hf) __D,
-					 __B, __E),
-    (__v4sf) __A, __B);
-#else
-  return (__m128h) __builtin_ia32_blendvps ((__v4sf) __A,
-    (__v4sf)
+  return (__m128h)
     __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) __A,
 					 (__v8hf) __C,
 					 (__v8hf) __D,
-					 __B, __E),
-    (__v4sf) _mm_set_ss ((float) ((int) __B << 31)));
-#endif
+					 __B, __E);
 }
 
 extern __inline __m128h
@@ -6843,12 +6780,11 @@  __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_mask3_fmadd_round_sch (__m128h __A, __m128h __B, __m128h __C,
 			   __mmask8 __D, const int __E)
 {
-  return (__m128h) _mm_move_ss ((__m128) __C,
-    (__m128)
-    __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) __A,
-					 (__v8hf) __B,
-					 (__v8hf) __C,
-					 __D, __E));
+  return (__m128h)
+    __builtin_ia32_vfmaddcsh_mask3_round ((__v8hf) __A,
+					  (__v8hf) __B,
+					  (__v8hf) __C,
+					  __D, __E);
 }
 
 extern __inline __m128h
@@ -6874,34 +6810,20 @@  _mm_fmadd_round_sch (__m128h __A, __m128h __B, __m128h __C, const int __D)
 				    __D);
 }
 #else
-#ifdef __AVX512VL__
 #define _mm_mask_fcmadd_round_sch(A, B, C, D, E)			\
-    ((__m128h) __builtin_ia32_movaps128_mask (				\
-     (__v4sf)								\
+    ((__m128h)								\
      __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) (A),		\
 					   (__v8hf) (C),		\
 					   (__v8hf) (D),		\
-					   (B), (E)),			\
-					   (__v4sf) (A), (B)))
+					   (B), (E)))
 
-#else
-#define _mm_mask_fcmadd_round_sch(A, B, C, D, E)			\
-  ((__m128h) __builtin_ia32_blendvps ((__v4sf) (A),			\
-   (__v4sf)								\
-   __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) (A),		\
-					 (__v8hf) (C),		\
-					 (__v8hf) (D),		\
-					 (B), (E)),		\
-    (__v4sf) _mm_set_ss ((float) ((int) (B) << 31))))
-#endif
 
 #define _mm_mask3_fcmadd_round_sch(A, B, C, D, E)			\
-  ((__m128h) _mm_move_ss ((__m128) (C),					\
-   (__m128)								\
-   __builtin_ia32_vfcmaddcsh_mask_round ((__v8hf) (A),		\
-					 (__v8hf) (B),		\
-					 (__v8hf) (C),		\
-					 (D), (E))))
+  ((__m128h)								\
+   __builtin_ia32_vfcmaddcsh_mask3_round ((__v8hf) (A),		\
+					  (__v8hf) (B),		\
+					  (__v8hf) (C),		\
+					  (D), (E)))
 
 #define _mm_maskz_fcmadd_round_sch(A, B, C, D, E)		\
   __builtin_ia32_vfcmaddcsh_maskz_round ((B), (C), (D), (A), (E))
@@ -6909,34 +6831,19 @@  _mm_fmadd_round_sch (__m128h __A, __m128h __B, __m128h __C, const int __D)
 #define _mm_fcmadd_round_sch(A, B, C, D)		\
   __builtin_ia32_vfcmaddcsh_round ((A), (B), (C), (D))
 
-#ifdef __AVX512VL__
 #define _mm_mask_fmadd_round_sch(A, B, C, D, E)				\
-    ((__m128h) __builtin_ia32_movaps128_mask (				\
-     (__v4sf)								\
+    ((__m128h)								\
      __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) (A),		\
 					  (__v8hf) (C),		\
 					  (__v8hf) (D),		\
-					  (B), (E)),		\
-					  (__v4sf) (A), (B)))
-
-#else
-#define _mm_mask_fmadd_round_sch(A, B, C, D, E)				\
-  ((__m128h) __builtin_ia32_blendvps ((__v4sf) (A),			\
-   (__v4sf)								\
-   __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) (A),		\
-					(__v8hf) (C),		\
-					(__v8hf) (D),		\
-					(B), (E)),		\
-    (__v4sf) _mm_set_ss ((float) ((int) (B) << 31))))
-#endif
+					  (B), (E)))
 
 #define _mm_mask3_fmadd_round_sch(A, B, C, D, E)			\
-  ((__m128h) _mm_move_ss ((__m128) (C),					\
-   (__m128)								\
-   __builtin_ia32_vfmaddcsh_mask_round ((__v8hf) (A),		\
-					(__v8hf) (B),		\
-					(__v8hf) (C),		\
-					(D), (E))))
+  ((__m128h)								\
+   __builtin_ia32_vfmaddcsh_mask3_round ((__v8hf) (A),		\
+					 (__v8hf) (B),		\
+					 (__v8hf) (C),		\
+					 (D), (E)))
 
 #define _mm_maskz_fmadd_round_sch(A, B, C, D, E)		\
   __builtin_ia32_vfmaddcsh_maskz_round ((B), (C), (D), (A), (E))
diff --git a/gcc/config/i386/avx512fp16vlintrin.h b/gcc/config/i386/avx512fp16vlintrin.h
index 3d3de964224..0b1f1cbe6ad 100644
--- a/gcc/config/i386/avx512fp16vlintrin.h
+++ b/gcc/config/i386/avx512fp16vlintrin.h
@@ -2898,21 +2898,20 @@  extern __inline __m128h
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_mask_fmadd_pch (__m128h __A, __mmask8 __B, __m128h __C, __m128h __D)
 {
-  return (__m128h) __builtin_ia32_movaps128_mask
-    ((__v4sf)
-     __builtin_ia32_vfmaddcph128_mask ((__v8hf) __A,
-				       (__v8hf) __C,
-				       (__v8hf) __D, __B),
-     (__v4sf) __A, __B);
+  return (__m128h)
+    __builtin_ia32_vfmaddcph128_mask ((__v8hf) __A,
+				      (__v8hf) __C,
+				      (__v8hf) __D, __B);
 }
 
 extern __inline __m128h
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_mask3_fmadd_pch (__m128h __A, __m128h __B, __m128h __C,  __mmask8 __D)
 {
-  return (__m128h) __builtin_ia32_vfmaddcph128_mask ((__v8hf) __A,
-						     (__v8hf) __B,
-						     (__v8hf) __C, __D);
+  return (__m128h)
+    __builtin_ia32_vfmaddcph128_mask3 ((__v8hf) __A,
+				       (__v8hf) __B,
+				       (__v8hf) __C, __D);
 }
 
 extern __inline __m128h
@@ -2937,21 +2936,20 @@  extern __inline __m256h
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm256_mask_fmadd_pch (__m256h __A, __mmask8 __B, __m256h __C, __m256h __D)
 {
-  return (__m256h) __builtin_ia32_movaps256_mask
-    ((__v8sf)
+  return (__m256h)
      __builtin_ia32_vfmaddcph256_mask ((__v16hf) __A,
 				       (__v16hf) __C,
-				       (__v16hf) __D, __B),
-     (__v8sf) __A, __B);
+				       (__v16hf) __D, __B);
 }
 
 extern __inline __m256h
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm256_mask3_fmadd_pch (__m256h __A, __m256h __B, __m256h __C,  __mmask8 __D)
 {
-  return (__m256h) __builtin_ia32_vfmaddcph256_mask ((__v16hf) __A,
-						     (__v16hf) __B,
-						     (__v16hf) __C, __D);
+  return (__m256h)
+    __builtin_ia32_vfmaddcph256_mask3 ((__v16hf) __A,
+				       (__v16hf) __B,
+				       (__v16hf) __C, __D);
 }
 
 extern __inline __m256h
@@ -2976,21 +2974,20 @@  extern __inline __m128h
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_mask_fcmadd_pch (__m128h __A, __mmask8 __B, __m128h __C, __m128h __D)
 {
-  return (__m128h) __builtin_ia32_movaps128_mask
-    ((__v4sf)
+  return (__m128h)
      __builtin_ia32_vfcmaddcph128_mask ((__v8hf) __A,
 					(__v8hf) __C,
-					(__v8hf) __D, __B),
-     (__v4sf) __A, __B);
+					(__v8hf) __D, __B);
 }
 
 extern __inline __m128h
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm_mask3_fcmadd_pch (__m128h __A, __m128h __B, __m128h __C,  __mmask8 __D)
 {
-  return (__m128h) __builtin_ia32_vfcmaddcph128_mask ((__v8hf) __A,
-						      (__v8hf) __B,
-						      (__v8hf) __C, __D);
+  return (__m128h)
+    __builtin_ia32_vfcmaddcph128_mask3 ((__v8hf) __A,
+					(__v8hf) __B,
+					(__v8hf) __C, __D);
 }
 
 extern __inline __m128h
@@ -3015,21 +3012,20 @@  extern __inline __m256h
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm256_mask_fcmadd_pch (__m256h __A, __mmask8 __B, __m256h __C, __m256h __D)
 {
-  return (__m256h) __builtin_ia32_movaps256_mask
-    ((__v8sf)
+  return (__m256h)
      __builtin_ia32_vfcmaddcph256_mask ((__v16hf) __A,
 					(__v16hf) __C,
-					(__v16hf) __D, __B),
-     (__v8sf) __A, __B);
+					(__v16hf) __D, __B);
 }
 
 extern __inline __m256h
 __attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
 _mm256_mask3_fcmadd_pch (__m256h __A, __m256h __B, __m256h __C,  __mmask8 __D)
 {
-  return (__m256h) __builtin_ia32_vfcmaddcph256_mask ((__v16hf) __A,
-						      (__v16hf) __B,
-						      (__v16hf) __C, __D);
+  return (__m256h)
+    __builtin_ia32_vfcmaddcph256_mask3 ((__v16hf) __A,
+					(__v16hf) __B,
+					(__v16hf) __C, __D);
 }
 
 extern __inline __m256h
diff --git a/gcc/config/i386/i386-builtin.def b/gcc/config/i386/i386-builtin.def
index 302e1bc6502..99217d08d37 100644
--- a/gcc/config/i386/i386-builtin.def
+++ b/gcc/config/i386/i386-builtin.def
@@ -2912,16 +2912,20 @@  BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp1
 BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fnmsub_v8hf_mask3, "__builtin_ia32_vfnmsubph128_mask3", IX86_BUILTIN_VFNMSUBPH128_MASK3, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fnmsub_v8hf_maskz, "__builtin_ia32_vfnmsubph128_maskz", IX86_BUILTIN_VFNMSUBPH128_MASKZ, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_fma_fmaddc_v8hf, "__builtin_ia32_vfmaddcph128", IX86_BUILTIN_VFMADDCPH_V8HF, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF)
-BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddc_v8hf_mask, "__builtin_ia32_vfmaddcph128_mask", IX86_BUILTIN_VFMADDCPH128_MASK, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddc_v8hf_mask1, "__builtin_ia32_vfmaddcph128_mask", IX86_BUILTIN_VFMADDCPH128_MASK, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddc_v8hf_mask, "__builtin_ia32_vfmaddcph128_mask3", IX86_BUILTIN_VFMADDCPH128_MASK3, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddc_v8hf_maskz, "__builtin_ia32_vfmaddcph128_maskz", IX86_BUILTIN_VFMADDCPH128_MASKZ, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_fma_fmaddc_v16hf, "__builtin_ia32_vfmaddcph256", IX86_BUILTIN_VFMADDCPH_V16HF, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF)
-BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fmaddc_v16hf_mask, "__builtin_ia32_vfmaddcph256_mask", IX86_BUILTIN_VFMADDCPH256_MASK, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fmaddc_v16hf_mask1, "__builtin_ia32_vfmaddcph256_mask", IX86_BUILTIN_VFMADDCPH256_MASK, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fmaddc_v16hf_mask, "__builtin_ia32_vfmaddcph256_mask3", IX86_BUILTIN_VFMADDCPH256_MASK3, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fmaddc_v16hf_maskz, "__builtin_ia32_vfmaddcph256_maskz", IX86_BUILTIN_VFMADDCPH256_MASKZ, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_fma_fcmaddc_v8hf, "__builtin_ia32_vfcmaddcph128", IX86_BUILTIN_VFCMADDCPH_V8HF, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF)
-BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddc_v8hf_mask, "__builtin_ia32_vfcmaddcph128_mask", IX86_BUILTIN_VFCMADDCPH128_MASK, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddc_v8hf_mask1, "__builtin_ia32_vfcmaddcph128_mask", IX86_BUILTIN_VFCMADDCPH128_MASK, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddc_v8hf_mask, "__builtin_ia32_vfcmaddcph128_mask3", IX86_BUILTIN_VFCMADDCPH128_MASK3, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddc_v8hf_maskz, "__builtin_ia32_vfcmaddcph128_maskz", IX86_BUILTIN_VFCMADDCPH128_MASKZ, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_fma_fcmaddc_v16hf, "__builtin_ia32_vfcmaddcph256", IX86_BUILTIN_VFCMADDCPH_V16HF, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF)
-BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fcmaddc_v16hf_mask, "__builtin_ia32_vfcmaddcph256_mask", IX86_BUILTIN_VFCMADDCPH256_MASK, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fcmaddc_v16hf_mask1, "__builtin_ia32_vfcmaddcph256_mask", IX86_BUILTIN_VFCMADDCPH256_MASK, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
+BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fcmaddc_v16hf_mask, "__builtin_ia32_vfcmaddcph256_mask3", IX86_BUILTIN_VFCMADDCPH256_MASK3, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512vl_fcmaddc_v16hf_maskz, "__builtin_ia32_vfcmaddcph256_maskz", IX86_BUILTIN_VFCMADDCPH256_MASKZ, UNKNOWN, (int) V16HF_FTYPE_V16HF_V16HF_V16HF_UQI)
 BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmulc_v8hf, "__builtin_ia32_vfcmulcph128", IX86_BUILTIN_VFCMULCPH_V8HF, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF)
 BDESC (OPTION_MASK_ISA_AVX512VL, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmulc_v8hf_mask, "__builtin_ia32_vfcmulcph128_mask", IX86_BUILTIN_VFCMULCPH128_MASK, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI)
@@ -3222,20 +3226,24 @@  BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512f_vmfnmadd_v8hf_mask3_roun
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512f_vmfnmadd_v8hf_maskz_round, "__builtin_ia32_vfnmaddsh3_maskz", IX86_BUILTIN_VFNMADDSH3_MASKZ, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512f_vmfmsub_v8hf_mask3_round, "__builtin_ia32_vfmsubsh3_mask3", IX86_BUILTIN_VFMSUBSH3_MASK3, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_fma_fmaddc_v32hf_round, "__builtin_ia32_vfmaddcph512_round", IX86_BUILTIN_VFMADDCPH512_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_INT)
-BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fmaddc_v32hf_mask_round, "__builtin_ia32_vfmaddcph512_mask_round", IX86_BUILTIN_VFMADDCPH512_MASK_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fmaddc_v32hf_mask1_round, "__builtin_ia32_vfmaddcph512_mask_round", IX86_BUILTIN_VFMADDCPH512_MASK_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fmaddc_v32hf_mask_round, "__builtin_ia32_vfmaddcph512_mask3_round", IX86_BUILTIN_VFMADDCPH512_MASK3_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fmaddc_v32hf_maskz_round, "__builtin_ia32_vfmaddcph512_maskz_round", IX86_BUILTIN_VFMADDCPH512_MASKZ_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_fma_fcmaddc_v32hf_round, "__builtin_ia32_vfcmaddcph512_round", IX86_BUILTIN_VFCMADDCPH512_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_INT)
-BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fcmaddc_v32hf_mask_round, "__builtin_ia32_vfcmaddcph512_mask_round", IX86_BUILTIN_VFCMADDCPH512_MASK_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fcmaddc_v32hf_mask1_round, "__builtin_ia32_vfcmaddcph512_mask_round", IX86_BUILTIN_VFCMADDCPH512_MASK_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fcmaddc_v32hf_mask_round, "__builtin_ia32_vfcmaddcph512_mask3_round", IX86_BUILTIN_VFCMADDCPH512_MASK3_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fcmaddc_v32hf_maskz_round, "__builtin_ia32_vfcmaddcph512_maskz_round", IX86_BUILTIN_VFCMADDCPH512_MASKZ_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fcmulc_v32hf_round, "__builtin_ia32_vfcmulcph512_round", IX86_BUILTIN_VFCMULCPH512_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fcmulc_v32hf_mask_round, "__builtin_ia32_vfcmulcph512_mask_round", IX86_BUILTIN_VFCMULCPH512_MASK_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fmulc_v32hf_round, "__builtin_ia32_vfmulcph512_round", IX86_BUILTIN_VFMULCPH512_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512bw_fmulc_v32hf_mask_round, "__builtin_ia32_vfmulcph512_mask_round", IX86_BUILTIN_VFMULCPH512_MASK_ROUND, UNKNOWN, (int) V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fma_fcmaddcsh_v8hf_round, "__builtin_ia32_vfcmaddcsh_round", IX86_BUILTIN_VFCMADDCSH_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_INT)
-BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddcsh_v8hf_mask_round, "__builtin_ia32_vfcmaddcsh_mask_round", IX86_BUILTIN_VFCMADDCSH_MASK_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddcsh_v8hf_mask1_round, "__builtin_ia32_vfcmaddcsh_mask_round", IX86_BUILTIN_VFCMADDCSH_MASK_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddcsh_v8hf_mask3_round, "__builtin_ia32_vfcmaddcsh_mask3_round", IX86_BUILTIN_VFCMADDCSH_MASK3_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmaddcsh_v8hf_maskz_round, "__builtin_ia32_vfcmaddcsh_maskz_round", IX86_BUILTIN_VFCMADDCSH_MASKZ_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fma_fmaddcsh_v8hf_round, "__builtin_ia32_vfmaddcsh_round", IX86_BUILTIN_VFMADDCSH_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_INT)
-BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddcsh_v8hf_mask_round, "__builtin_ia32_vfmaddcsh_mask_round", IX86_BUILTIN_VFMADDCSH_MASK_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddcsh_v8hf_mask1_round, "__builtin_ia32_vfmaddcsh_mask_round", IX86_BUILTIN_VFMADDCSH_MASK_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
+BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddcsh_v8hf_mask3_round, "__builtin_ia32_vfmaddcsh_mask3_round", IX86_BUILTIN_VFMADDCSH_MASK3_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fmaddcsh_v8hf_maskz_round, "__builtin_ia32_vfmaddcsh_maskz_round", IX86_BUILTIN_VFMADDCSH_MASKZ_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmulcsh_v8hf_round, "__builtin_ia32_vfcmulcsh_round", IX86_BUILTIN_VFCMULCSH_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_INT)
 BDESC (0, OPTION_MASK_ISA2_AVX512FP16, CODE_FOR_avx512fp16_fcmulcsh_v8hf_mask_round, "__builtin_ia32_vfcmulcsh_mask_round", IX86_BUILTIN_VFCMULCSH_MASK_ROUND, UNKNOWN, (int) V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT)
diff --git a/gcc/config/i386/i386-expand.c b/gcc/config/i386/i386-expand.c
index c0924a59efb..de4fe9ce147 100644
--- a/gcc/config/i386/i386-expand.c
+++ b/gcc/config/i386/i386-expand.c
@@ -10877,7 +10877,27 @@  ix86_expand_round_builtin (const struct builtin_description *d,
 
 	  /* If there is no rounding use normal version of the pattern.  */
 	  if (INTVAL (op) == NO_ROUND)
-	    redundant_embed_rnd = 1;
+	    {
+	      /* Skip erasing embedded rounding for below expanders who
+		 generates multiple insns.  In ix86_erase_embedded_rounding
+		 the pattern will be transformed to a single set, and emit_insn
+		 appends the set insead of insert it to chain.  So the insns
+		 emitted inside define_expander would be ignored.  */
+	      switch (icode)
+		{
+		case CODE_FOR_avx512bw_fmaddc_v32hf_mask1_round:
+		case CODE_FOR_avx512bw_fcmaddc_v32hf_mask1_round:
+		case CODE_FOR_avx512fp16_fmaddcsh_v8hf_mask1_round:
+		case CODE_FOR_avx512fp16_fcmaddcsh_v8hf_mask1_round:
+		case CODE_FOR_avx512fp16_fmaddcsh_v8hf_mask3_round:
+		case CODE_FOR_avx512fp16_fcmaddcsh_v8hf_mask3_round:
+		  redundant_embed_rnd = 0;
+		  break;
+		default:
+		  redundant_embed_rnd = 1;
+		  break;
+		}
+	    }
 	}
       else
 	{
diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md
index a3c4a3f1e62..bdd34f379dd 100644
--- a/gcc/config/i386/sse.md
+++ b/gcc/config/i386/sse.md
@@ -5864,6 +5864,34 @@ 
 	 (UNSPEC_COMPLEX_FMUL "fmulc")
 	 (UNSPEC_COMPLEX_FCMUL "fcmulc")])
 
+(define_mode_attr complexmove
+  [(V32HF "avx512f_loadv16sf")
+   (V16HF "avx512vl_loadv8sf")
+   (V8HF "avx512vl_loadv4sf")])
+
+(define_expand "<avx512>_fmaddc_<mode>_mask1<round_expand_name>"
+  [(match_operand:VF_AVX512FP16VL 0 "register_operand")
+   (match_operand:VF_AVX512FP16VL 1 "<round_expand_nimm_predicate>")
+   (match_operand:VF_AVX512FP16VL 2 "<round_expand_nimm_predicate>")
+   (match_operand:VF_AVX512FP16VL 3 "<round_expand_nimm_predicate>")
+   (match_operand:<avx512fmaskcmode> 4 "register_operand")]
+  "TARGET_AVX512FP16 && <round_mode512bit_condition>"
+{
+  rtx op0, op1;
+  if (<round_embedded_complex>)
+    emit_insn (gen_<avx512>_fmaddc_<mode>_mask<round_expand_name> (
+      operands[0], operands[1], operands[2], operands[3],
+      operands[4]<round_expand_operand>));
+  else
+    emit_insn (gen_<avx512>_fmaddc_<mode>_mask (operands[0],
+      operands[1], operands[2], operands[3], operands[4]));
+
+  op0 = lowpart_subreg (<ssePSmode>mode, operands[0], <MODE>mode);
+  op1 = lowpart_subreg (<ssePSmode>mode, operands[1], <MODE>mode);
+  emit_insn (gen_<complexmove>_mask (op0, op0, op1, operands[4]));
+  DONE;
+})
+
 (define_expand "<avx512>_fmaddc_<mode>_maskz<round_expand_name>"
   [(match_operand:VF_AVX512FP16VL 0 "register_operand")
    (match_operand:VF_AVX512FP16VL 1 "<round_expand_nimm_predicate>")
@@ -5878,6 +5906,31 @@ 
   DONE;
 })
 
+(define_expand "<avx512>_fcmaddc_<mode>_mask1<round_expand_name>"
+  [(match_operand:VF_AVX512FP16VL 0 "register_operand")
+   (match_operand:VF_AVX512FP16VL 1 "<round_expand_nimm_predicate>")
+   (match_operand:VF_AVX512FP16VL 2 "<round_expand_nimm_predicate>")
+   (match_operand:VF_AVX512FP16VL 3 "<round_expand_nimm_predicate>")
+   (match_operand:<avx512fmaskcmode> 4 "register_operand")]
+  "TARGET_AVX512FP16 && <round_mode512bit_condition>"
+{
+  rtx op0, op1;
+  if (<round_embedded_complex>)
+    emit_insn (gen_<avx512>_fcmaddc_<mode>_mask<round_expand_name> (
+      operands[0], operands[1], operands[2], operands[3],
+      operands[4]<round_expand_operand>));
+  else
+  {
+    emit_insn (gen_<avx512>_fcmaddc_<mode>_mask (operands[0],
+      operands[1], operands[2], operands[3], operands[4]));
+  }
+
+  op0 = lowpart_subreg (<ssePSmode>mode, operands[0], <MODE>mode);
+  op1 = lowpart_subreg (<ssePSmode>mode, operands[1], <MODE>mode);
+  emit_insn (gen_<complexmove>_mask (op0, op0, op1, operands[4]));
+  DONE;
+})
+
 (define_expand "<avx512>_fcmaddc_<mode>_maskz<round_expand_name>"
   [(match_operand:VF_AVX512FP16VL 0 "register_operand")
    (match_operand:VF_AVX512FP16VL 1 "<round_expand_nimm_predicate>")
@@ -5946,6 +5999,47 @@ 
   DONE;
 })
 
+(define_expand "avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>"
+  [(match_operand:V8HF 0 "register_operand")
+   (match_operand:V8HF 1 "<round_expand_nimm_predicate>")
+   (match_operand:V8HF 2 "<round_expand_nimm_predicate>")
+   (match_operand:V8HF 3 "<round_expand_nimm_predicate>")
+   (match_operand:QI 4 "register_operand")]
+  "TARGET_AVX512FP16 && <round_mode512bit_condition>"
+{
+  rtx op0, op1;
+
+  if (<round_embedded_complex>)
+    emit_insn (gen_avx512fp16_fmaddcsh_v8hf_mask<round_expand_name> (
+      operands[0], operands[1], operands[2], operands[3],
+      operands[4]<round_expand_operand>));
+  else
+    emit_insn (gen_avx512fp16_fmaddcsh_v8hf_mask (operands[0],
+      operands[1], operands[2], operands[3], operands[4]));
+
+  if (TARGET_AVX512VL)
+  {
+    op0 = lowpart_subreg (V4SFmode, operands[0], V8HFmode);
+    op1 = lowpart_subreg (V4SFmode, operands[1], V8HFmode);
+    emit_insn (gen_avx512vl_loadv4sf_mask (op0, op0, op1, operands[4]));
+  }
+  else
+  {
+    rtx mask, tmp, vec_mask;
+    mask = lowpart_subreg (SImode, operands[4], QImode),
+    tmp = gen_reg_rtx (SImode);
+    emit_insn (gen_ashlsi3 (tmp, mask, GEN_INT (31)));
+    vec_mask = gen_reg_rtx (V4SImode);
+    emit_insn (gen_rtx_SET (vec_mask, CONST0_RTX (V4SImode)));
+    emit_insn (gen_vec_setv4si_0 (vec_mask, vec_mask, tmp));
+    vec_mask = lowpart_subreg (V4SFmode, vec_mask, V4SImode);
+    op0 = lowpart_subreg (V4SFmode, operands[0], V8HFmode);
+    op1 = lowpart_subreg (V4SFmode, operands[1], V8HFmode);
+    emit_insn (gen_sse4_1_blendvps (op0, op1, op0, vec_mask));
+  }
+  DONE;
+})
+
 (define_expand "avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>"
   [(match_operand:V8HF 0 "register_operand")
    (match_operand:V8HF 1 "<round_expand_nimm_predicate>")
@@ -5960,6 +6054,95 @@ 
   DONE;
 })
 
+(define_expand "avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>"
+  [(match_operand:V8HF 0 "register_operand")
+   (match_operand:V8HF 1 "<round_expand_nimm_predicate>")
+   (match_operand:V8HF 2 "<round_expand_nimm_predicate>")
+   (match_operand:V8HF 3 "<round_expand_nimm_predicate>")
+   (match_operand:QI 4 "register_operand")]
+  "TARGET_AVX512FP16 && <round_mode512bit_condition>"
+{
+  rtx op0, op1;
+
+  if (<round_embedded_complex>)
+    emit_insn (gen_avx512fp16_fcmaddcsh_v8hf_mask<round_expand_name> (
+      operands[0], operands[1], operands[2], operands[3],
+      operands[4]<round_expand_operand>));
+  else
+    emit_insn (gen_avx512fp16_fcmaddcsh_v8hf_mask (operands[0],
+      operands[1], operands[2], operands[3], operands[4]));
+
+  if (TARGET_AVX512VL)
+  {
+    op0 = lowpart_subreg (V4SFmode, operands[0], V8HFmode);
+    op1 = lowpart_subreg (V4SFmode, operands[1], V8HFmode);
+    emit_insn (gen_avx512vl_loadv4sf_mask (op0, op0, op1, operands[4]));
+  }
+  else
+  {
+    rtx mask, tmp, vec_mask;
+    mask = lowpart_subreg (SImode, operands[4], QImode),
+    tmp = gen_reg_rtx (SImode);
+    emit_insn (gen_ashlsi3 (tmp, mask, GEN_INT (31)));
+    vec_mask = gen_reg_rtx (V4SImode);
+    emit_insn (gen_rtx_SET (vec_mask, CONST0_RTX (V4SImode)));
+    emit_insn (gen_vec_setv4si_0 (vec_mask, vec_mask, tmp));
+    vec_mask = lowpart_subreg (V4SFmode, vec_mask, V4SImode);
+    op0 = lowpart_subreg (V4SFmode, operands[0], V8HFmode);
+    op1 = lowpart_subreg (V4SFmode, operands[1], V8HFmode);
+    emit_insn (gen_sse4_1_blendvps (op0, op1, op0, vec_mask));
+  }
+  DONE;
+})
+
+(define_expand "avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>"
+  [(match_operand:V8HF 0 "register_operand")
+   (match_operand:V8HF 1 "<round_expand_nimm_predicate>")
+   (match_operand:V8HF 2 "<round_expand_nimm_predicate>")
+   (match_operand:V8HF 3 "<round_expand_nimm_predicate>")
+   (match_operand:QI 4 "register_operand")]
+  "TARGET_AVX512FP16 && <round_mode512bit_condition>"
+{
+  rtx op0, op1;
+
+  if (<round_embedded_complex>)
+    emit_insn (gen_avx512fp16_fcmaddcsh_v8hf_mask<round_expand_name> (
+      operands[0], operands[1], operands[2], operands[3],
+      operands[4]<round_expand_operand>));
+  else
+    emit_insn (gen_avx512fp16_fcmaddcsh_v8hf_mask (operands[0],
+      operands[1], operands[2], operands[3], operands[4]));
+
+  op0 = lowpart_subreg (V4SFmode, operands[0], V8HFmode);
+  op1 = lowpart_subreg (V4SFmode, operands[3], V8HFmode);
+  emit_insn (gen_sse_movss (op0, op1, op0));
+  DONE;
+})
+
+(define_expand "avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>"
+  [(match_operand:V8HF 0 "register_operand")
+   (match_operand:V8HF 1 "<round_expand_nimm_predicate>")
+   (match_operand:V8HF 2 "<round_expand_nimm_predicate>")
+   (match_operand:V8HF 3 "<round_expand_nimm_predicate>")
+   (match_operand:QI 4 "register_operand")]
+  "TARGET_AVX512FP16 && <round_mode512bit_condition>"
+{
+  rtx op0, op1;
+
+  if (<round_embedded_complex>)
+    emit_insn (gen_avx512fp16_fmaddcsh_v8hf_mask<round_expand_name> (
+      operands[0], operands[1], operands[2], operands[3],
+      operands[4]<round_expand_operand>));
+  else
+    emit_insn (gen_avx512fp16_fmaddcsh_v8hf_mask (operands[0],
+      operands[1], operands[2], operands[3], operands[4]));
+
+  op0 = lowpart_subreg (V4SFmode, operands[0], V8HFmode);
+  op1 = lowpart_subreg (V4SFmode, operands[3], V8HFmode);
+  emit_insn (gen_sse_movss (op0, op1, op0));
+  DONE;
+})
+
 (define_insn "avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>"
   [(set (match_operand:V8HF 0 "register_operand" "=&v")
 	(vec_merge:V8HF
diff --git a/gcc/config/i386/subst.md b/gcc/config/i386/subst.md
index 11e62c67fcc..510e82c84f9 100644
--- a/gcc/config/i386/subst.md
+++ b/gcc/config/i386/subst.md
@@ -276,6 +276,9 @@ 
 (define_subst_attr "round_expand_name" "round_expand" "" "_round")
 (define_subst_attr "round_expand_nimm_predicate" "round_expand" "nonimmediate_operand" "register_operand")
 (define_subst_attr "round_expand_operand" "round_expand" "" ", operands[5]")
+(define_subst_attr "round_embedded_complex" "round_expand" "0" "!(CONST_INT_P (operands[5])
+								  && (INTVAL (operands[5])
+								      == NO_ROUND))")
 
 (define_subst "round_expand"
  [(match_operand:SUBST_V 0)
diff --git a/gcc/testsuite/gcc.target/i386/avx-1.c b/gcc/testsuite/gcc.target/i386/avx-1.c
index 8744aa7df55..154e7b3b107 100644
--- a/gcc/testsuite/gcc.target/i386/avx-1.c
+++ b/gcc/testsuite/gcc.target/i386/avx-1.c
@@ -789,9 +789,11 @@ 
 #define __builtin_ia32_vfnmsubsh3_maskz(A, B, C, D, E) __builtin_ia32_vfnmsubsh3_maskz(A, B, C, D, 8)
 #define __builtin_ia32_vfcmaddcph512_round(A, B, C, D) __builtin_ia32_vfcmaddcph512_round(A, B, C, 8)
 #define __builtin_ia32_vfcmaddcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfcmaddcph512_mask_round(A, C, D, B, 8)
+#define __builtin_ia32_vfcmaddcph512_mask3_round(A, C, D, B, E) __builtin_ia32_vfcmaddcph512_mask3_round(A, C, D, B, 8)
 #define __builtin_ia32_vfcmaddcph512_maskz_round(B, C, D, A, E) __builtin_ia32_vfcmaddcph512_maskz_round(B, C, D, A, 8)
 #define __builtin_ia32_vfmaddcph512_round(A, B, C, D) __builtin_ia32_vfmaddcph512_round(A, B, C, 8)
 #define __builtin_ia32_vfmaddcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfmaddcph512_mask_round(A, C, D, B, 8)
+#define __builtin_ia32_vfmaddcph512_mask3_round(A, C, D, B, E) __builtin_ia32_vfmaddcph512_mask3_round(A, C, D, B, 8)
 #define __builtin_ia32_vfmaddcph512_maskz_round(B, C, D, A, E) __builtin_ia32_vfmaddcph512_maskz_round(B, C, D, A, 8)
 #define __builtin_ia32_vfmulcph512_round(A, B, C) __builtin_ia32_vfmulcph512_round(A, B, 8)
 #define __builtin_ia32_vfmulcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfmulcph512_mask_round(A, C, D, B, 8)
@@ -799,9 +801,11 @@ 
 #define __builtin_ia32_vfcmulcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfcmulcph512_mask_round(A, C, D, B, 8)
 #define __builtin_ia32_vfmaddcsh_round(A, B, C, D) __builtin_ia32_vfmaddcsh_round(A, B, C, 8)
 #define __builtin_ia32_vfmaddcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfmaddcsh_mask_round(A, C, D, B, 8)
+#define __builtin_ia32_vfmaddcsh_mask3_round(A, C, D, B, E) __builtin_ia32_vfmaddcsh_mask3_round(A, C, D, B, 8)
 #define __builtin_ia32_vfmaddcsh_maskz_round(B, C, D, A, E) __builtin_ia32_vfmaddcsh_maskz_round(B, C, D, A, 8)
 #define __builtin_ia32_vfcmaddcsh_round(A, B, C, D) __builtin_ia32_vfcmaddcsh_round(A, B, C, 8)
 #define __builtin_ia32_vfcmaddcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfcmaddcsh_mask_round(A, C, D, B, 8)
+#define __builtin_ia32_vfcmaddcsh_mask3_round(A, C, D, B, E) __builtin_ia32_vfcmaddcsh_mask3_round(A, C, D, B, 8)
 #define __builtin_ia32_vfcmaddcsh_maskz_round(B, C, D, A, E) __builtin_ia32_vfcmaddcsh_maskz_round(B, C, D, A, 8)
 #define __builtin_ia32_vfmulcsh_round(A, B, C) __builtin_ia32_vfmulcsh_round(A, B, 8)
 #define __builtin_ia32_vfmulcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfmulcsh_mask_round(A, C, D, B, 8)
diff --git a/gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1a.c b/gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1a.c
index 8bd8eebd8df..8ff2092c325 100644
--- a/gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1a.c
+++ b/gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1a.c
@@ -6,6 +6,8 @@ 
 /* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
 /* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
 /* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\{rz-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\{z\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vblendvps\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */
+/* { dg-final { scan-assembler-times "vmovss\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */
 
 #include <immintrin.h>
 
diff --git a/gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1c.c b/gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1c.c
new file mode 100644
index 00000000000..79a295f722c
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/avx512fp16-vfcmaddcsh-1c.c
@@ -0,0 +1,13 @@ 
+/* { dg-do compile } */
+/* { dg-options "-mavx512fp16 -mavx512vl -O2" } */
+/* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\{\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
+/* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\{z\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
+/* { dg-final { scan-assembler-times "vfcmaddcsh\[ \\t\]+\{rz-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\{z\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vblendmps\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\{\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
+/* { dg-final { scan-assembler-times "vmovss\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */
+
+#include "avx512fp16-vfcmaddcsh-1a.c"
+
diff --git a/gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1a.c b/gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1a.c
index 1e376b4a2bb..2ebe1f8ddd7 100644
--- a/gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1a.c
+++ b/gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1a.c
@@ -6,6 +6,8 @@ 
 /* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
 /* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
 /* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\{rz-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\{z\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vblendvps\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */
+/* { dg-final { scan-assembler-times "vmovss\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */
 
 #include <immintrin.h>
 
diff --git a/gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1c.c b/gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1c.c
new file mode 100644
index 00000000000..7863f8f9af9
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/avx512fp16-vfmaddcsh-1c.c
@@ -0,0 +1,13 @@ 
+/* { dg-do compile } */
+/* { dg-options "-mavx512fp16 -mavx512vl -O2" } */
+/* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\{\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
+/* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\{z\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\{rn-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
+/* { dg-final { scan-assembler-times "vfmaddcsh\[ \\t\]+\{rz-sae\}\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\{z\}\[^\n\r]*(?:\n|\[ \\t\]+#)" 1 } } */
+/* { dg-final { scan-assembler-times "vblendmps\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\{%k\[0-9\]\}\[^\{\n\r]*(?:\n|\[ \\t\]+#)" 2 } } */
+/* { dg-final { scan-assembler-times "vmovss\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+\[^\n\r]*%xmm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */
+
+#include "avx512fp16-vfmaddcsh-1a.c"
+
diff --git a/gcc/testsuite/gcc.target/i386/sse-13.c b/gcc/testsuite/gcc.target/i386/sse-13.c
index f6d54e3a815..e285c307d00 100644
--- a/gcc/testsuite/gcc.target/i386/sse-13.c
+++ b/gcc/testsuite/gcc.target/i386/sse-13.c
@@ -806,9 +806,11 @@ 
 #define __builtin_ia32_vfnmsubsh3_maskz(A, B, C, D, E) __builtin_ia32_vfnmsubsh3_maskz(A, B, C, D, 8)
 #define __builtin_ia32_vfcmaddcph512_round(A, B, C, D) __builtin_ia32_vfcmaddcph512_round(A, B, C, 8)
 #define __builtin_ia32_vfcmaddcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfcmaddcph512_mask_round(A, C, D, B, 8)
+#define __builtin_ia32_vfcmaddcph512_mask3_round(A, C, D, B, E) __builtin_ia32_vfcmaddcph512_mask3_round(A, C, D, B, 8)
 #define __builtin_ia32_vfcmaddcph512_maskz_round(B, C, D, A, E) __builtin_ia32_vfcmaddcph512_maskz_round(B, C, D, A, 8)
 #define __builtin_ia32_vfmaddcph512_round(A, B, C, D) __builtin_ia32_vfmaddcph512_round(A, B, C, 8)
 #define __builtin_ia32_vfmaddcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfmaddcph512_mask_round(A, C, D, B, 8)
+#define __builtin_ia32_vfmaddcph512_mask3_round(A, C, D, B, E) __builtin_ia32_vfmaddcph512_mask3_round(A, C, D, B, 8)
 #define __builtin_ia32_vfmaddcph512_maskz_round(B, C, D, A, E) __builtin_ia32_vfmaddcph512_maskz_round(B, C, D, A, 8)
 #define __builtin_ia32_vfmulcph512_round(A, B, C) __builtin_ia32_vfmulcph512_round(A, B, 8)
 #define __builtin_ia32_vfmulcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfmulcph512_mask_round(A, C, D, B, 8)
@@ -816,9 +818,11 @@ 
 #define __builtin_ia32_vfcmulcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfcmulcph512_mask_round(A, C, D, B, 8)
 #define __builtin_ia32_vfmaddcsh_round(A, B, C, D) __builtin_ia32_vfmaddcsh_round(A, B, C, 8)
 #define __builtin_ia32_vfmaddcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfmaddcsh_mask_round(A, C, D, B, 8)
+#define __builtin_ia32_vfmaddcsh_mask3_round(A, C, D, B, E) __builtin_ia32_vfmaddcsh_mask3_round(A, C, D, B, 8)
 #define __builtin_ia32_vfmaddcsh_maskz_round(B, C, D, A, E) __builtin_ia32_vfmaddcsh_maskz_round(B, C, D, A, 8)
 #define __builtin_ia32_vfcmaddcsh_round(A, B, C, D) __builtin_ia32_vfcmaddcsh_round(A, B, C, 8)
 #define __builtin_ia32_vfcmaddcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfcmaddcsh_mask_round(A, C, D, B, 8)
+#define __builtin_ia32_vfcmaddcsh_mask3_round(A, C, D, B, E) __builtin_ia32_vfcmaddcsh_mask3_round(A, C, D, B, 8)
 #define __builtin_ia32_vfcmaddcsh_maskz_round(B, C, D, A, E) __builtin_ia32_vfcmaddcsh_maskz_round(B, C, D, A, 8)
 #define __builtin_ia32_vfmulcsh_round(A, B, C) __builtin_ia32_vfmulcsh_round(A, B, 8)
 #define __builtin_ia32_vfmulcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfmulcsh_mask_round(A, C, D, B, 8)
diff --git a/gcc/testsuite/gcc.target/i386/sse-23.c b/gcc/testsuite/gcc.target/i386/sse-23.c
index 98251269a64..f71a7b29157 100644
--- a/gcc/testsuite/gcc.target/i386/sse-23.c
+++ b/gcc/testsuite/gcc.target/i386/sse-23.c
@@ -807,9 +807,11 @@ 
 #define __builtin_ia32_vfnmsubsh3_maskz(A, B, C, D, E) __builtin_ia32_vfnmsubsh3_maskz(A, B, C, D, 8)
 #define __builtin_ia32_vfcmaddcph512_round(A, B, C, D) __builtin_ia32_vfcmaddcph512_round(A, B, C, 8)
 #define __builtin_ia32_vfcmaddcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfcmaddcph512_mask_round(A, C, D, B, 8)
+#define __builtin_ia32_vfcmaddcph512_mask3_round(A, C, D, B, E) __builtin_ia32_vfcmaddcph512_mask3_round(A, C, D, B, 8)
 #define __builtin_ia32_vfcmaddcph512_maskz_round(B, C, D, A, E) __builtin_ia32_vfcmaddcph512_maskz_round(B, C, D, A, 8)
 #define __builtin_ia32_vfmaddcph512_round(A, B, C, D) __builtin_ia32_vfmaddcph512_round(A, B, C, 8)
 #define __builtin_ia32_vfmaddcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfmaddcph512_mask_round(A, C, D, B, 8)
+#define __builtin_ia32_vfmaddcph512_mask3_round(A, C, D, B, E) __builtin_ia32_vfmaddcph512_mask3_round(A, C, D, B, 8)
 #define __builtin_ia32_vfmaddcph512_maskz_round(B, C, D, A, E) __builtin_ia32_vfmaddcph512_maskz_round(B, C, D, A, 8)
 #define __builtin_ia32_vfmulcph512_round(A, B, C) __builtin_ia32_vfmulcph512_round(A, B, 8)
 #define __builtin_ia32_vfmulcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfmulcph512_mask_round(A, C, D, B, 8)
@@ -817,9 +819,11 @@ 
 #define __builtin_ia32_vfcmulcph512_mask_round(A, C, D, B, E) __builtin_ia32_vfcmulcph512_mask_round(A, C, D, B, 8)
 #define __builtin_ia32_vfmaddcsh_round(A, B, C, D) __builtin_ia32_vfmaddcsh_round(A, B, C, 8)
 #define __builtin_ia32_vfmaddcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfmaddcsh_mask_round(A, C, D, B, 8)
+#define __builtin_ia32_vfmaddcsh_mask3_round(A, C, D, B, E) __builtin_ia32_vfmaddcsh_mask3_round(A, C, D, B, 8)
 #define __builtin_ia32_vfmaddcsh_maskz_round(B, C, D, A, E) __builtin_ia32_vfmaddcsh_maskz_round(B, C, D, A, 8)
 #define __builtin_ia32_vfcmaddcsh_round(A, B, C, D) __builtin_ia32_vfcmaddcsh_round(A, B, C, 8)
 #define __builtin_ia32_vfcmaddcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfcmaddcsh_mask_round(A, C, D, B, 8)
+#define __builtin_ia32_vfcmaddcsh_mask3_round(A, C, D, B, E) __builtin_ia32_vfcmaddcsh_mask3_round(A, C, D, B, 8)
 #define __builtin_ia32_vfcmaddcsh_maskz_round(B, C, D, A, E) __builtin_ia32_vfcmaddcsh_maskz_round(B, C, D, A, 8)
 #define __builtin_ia32_vfmulcsh_round(A, B, C) __builtin_ia32_vfmulcsh_round(A, B, 8)
 #define __builtin_ia32_vfmulcsh_mask_round(A, C, D, B, E) __builtin_ia32_vfmulcsh_mask_round(A, C, D, B, 8)