On Thu, Jul 1, 2021 at 2:18 PM liuhongt <hongtao....@intel.com> wrote:
>
> gcc/ChangeLog:
>
> * config/i386/i386-features.c (i386-features.c): Handle
> E_HFmode.
> * config/i386/i386.md (sqrthf2): New expander.
> (*sqrt<mode>2_sse): Extend to MODEFH.
> * config/i386/sse.md
> (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
> Extend to VFH_128.
>
> gcc/testsuite/ChangeLog:
>
> * gcc.target/i386/avx512fp16-builtin-sqrt-1.c: New test.
> * gcc.target/i386/avx512fp16vl-builtin-sqrt-1.c: New test.
> ---
> gcc/config/i386/i386-features.c | 15 +++++++++++----
> gcc/config/i386/i386.md | 12 +++++++++---
> gcc/config/i386/sse.md | 8 ++++----
> .../i386/avx512fp16-builtin-sqrt-1.c | 18 ++++++++++++++++++
> .../i386/avx512fp16vl-builtin-sqrt-1.c | 19 +++++++++++++++++++
> 5 files changed, 61 insertions(+), 11 deletions(-)
> create mode 100644 gcc/testsuite/gcc.target/i386/avx512fp16-builtin-sqrt-1.c
> create mode 100644
> gcc/testsuite/gcc.target/i386/avx512fp16vl-builtin-sqrt-1.c
>
> diff --git a/gcc/config/i386/i386-features.c b/gcc/config/i386/i386-features.c
> index a25769ae478..0b5a1a3af53 100644
> --- a/gcc/config/i386/i386-features.c
> +++ b/gcc/config/i386/i386-features.c
> @@ -2238,15 +2238,22 @@ remove_partial_avx_dependency (void)
>
> rtx zero;
> machine_mode dest_vecmode;
> - if (dest_mode == E_SFmode)
> + switch (dest_mode)
> {
> + case E_HFmode:
> + dest_vecmode = V8HFmode;
> + zero = gen_rtx_SUBREG (V8HFmode, v4sf_const0, 0);
> + break;
> + case E_SFmode:
> dest_vecmode = V4SFmode;
> zero = v4sf_const0;
> - }
> - else
> - {
> + break;
> + case E_DFmode:
> dest_vecmode = V2DFmode;
> zero = gen_rtx_SUBREG (V2DFmode, v4sf_const0, 0);
> + break;
> + default:
> + gcc_unreachable ();
> }
>
> /* Change source to vector mode. */
> diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
> index a85c23d74f1..81c893c60de 100644
> --- a/gcc/config/i386/i386.md
> +++ b/gcc/config/i386/i386.md
> @@ -16561,9 +16561,9 @@ (define_expand "rsqrtsf2"
> })
>
> (define_insn "*sqrt<mode>2_sse"
> - [(set (match_operand:MODEF 0 "register_operand" "=v,v,v")
> - (sqrt:MODEF
> - (match_operand:MODEF 1 "nonimmediate_operand" "0,v,m")))]
> + [(set (match_operand:MODEFH 0 "register_operand" "=v,v,v")
> + (sqrt:MODEFH
> + (match_operand:MODEFH 1 "nonimmediate_operand" "0,v,m")))]
> "SSE_FLOAT_MODE_P (<MODE>mode) && TARGET_SSE_MATH"
> "@
> %vsqrt<ssemodesuffix>\t{%d1, %0|%0, %d1}
> @@ -16583,6 +16583,12 @@ (define_insn "*sqrt<mode>2_sse"
> ]
> (symbol_ref "true")))])
>
As mentioned by uros, l think this also better has a separate pattern for hf.
> +(define_expand "sqrthf2"
> + [(set (match_operand:HF 0 "register_operand")
> + (sqrt:HF
> + (match_operand:HF 1 "nonimmediate_operand")))]
> + "TARGET_AVX512FP16")
> +
> (define_expand "sqrt<mode>2"
> [(set (match_operand:MODEF 0 "register_operand")
> (sqrt:MODEF
> diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md
> index 2c3dba5bdb0..b47e7f0b82a 100644
> --- a/gcc/config/i386/sse.md
> +++ b/gcc/config/i386/sse.md
> @@ -2389,12 +2389,12 @@ (define_insn
> "<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>"
> (set_attr "mode" "<ssescalarmode>")])
>
> (define_insn "*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>"
> - [(set (match_operand:VF_128 0 "register_operand" "=x,v")
> - (vec_merge:VF_128
> - (vec_duplicate:VF_128
> + [(set (match_operand:VFH_128 0 "register_operand" "=x,v")
> + (vec_merge:VFH_128
> + (vec_duplicate:VFH_128
> (sqrt:<ssescalarmode>
> (match_operand:<ssescalarmode> 1 "nonimmediate_operand"
> "xm,<round_scalar_constraint>")))
> - (match_operand:VF_128 2 "register_operand" "0,v")
> + (match_operand:VFH_128 2 "register_operand" "0,v")
> (const_int 1)))]
> "TARGET_SSE"
> "@
> diff --git a/gcc/testsuite/gcc.target/i386/avx512fp16-builtin-sqrt-1.c
> b/gcc/testsuite/gcc.target/i386/avx512fp16-builtin-sqrt-1.c
> new file mode 100644
> index 00000000000..38cdf23fef7
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/avx512fp16-builtin-sqrt-1.c
> @@ -0,0 +1,18 @@
> +/* { dg-do compile } */
> +/* { dg-options "-Ofast -mavx512fp16" } */
> +
> +_Float16
> +f1 (_Float16 x)
> +{
> + return __builtin_sqrtf16 (x);
> +}
> +
> +void
> +f2 (_Float16* __restrict psrc, _Float16* __restrict pdst)
> +{
> + for (int i = 0; i != 32; i++)
> + pdst[i] = __builtin_sqrtf16 (psrc[i]);
> +}
> +
> +/* { dg-final { scan-assembler-times "vsqrtsh\[^\n\r\]*xmm\[0-9\]" 1 } } */
> +/* { dg-final { scan-assembler-times "vsqrtph\[^\n\r\]*zmm\[0-9\]" 1 } } */
> diff --git a/gcc/testsuite/gcc.target/i386/avx512fp16vl-builtin-sqrt-1.c
> b/gcc/testsuite/gcc.target/i386/avx512fp16vl-builtin-sqrt-1.c
> new file mode 100644
> index 00000000000..08deb3ea470
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/avx512fp16vl-builtin-sqrt-1.c
> @@ -0,0 +1,19 @@
> +/* { dg-do compile } */
> +/* { dg-options "-Ofast -mavx512fp16 -mavx512vl" } */
> +
> +void
> +f1 (_Float16* __restrict psrc, _Float16* __restrict pdst)
> +{
> + for (int i = 0; i != 8; i++)
> + pdst[i] = __builtin_sqrtf16 (psrc[i]);
> +}
> +
> +void
> +f2 (_Float16* __restrict psrc, _Float16* __restrict pdst)
> +{
> + for (int i = 0; i != 16; i++)
> + pdst[i] = __builtin_sqrtf16 (psrc[i]);
> +}
> +
> +/* { dg-final { scan-assembler-times "vsqrtph\[^\n\r\]*xmm\[0-9\]" 1 } } */
> +/* { dg-final { scan-assembler-times "vsqrtph\[^\n\r\]*ymm\[0-9\]" 1 } } */
> --
> 2.18.1
>
--
BR,
Hongtao
