> -----Original Message-----
> From: Kewen Lin <[email protected]>
> Sent: Thursday, June 4, 2026 10:00 AM
> To: Hongtao Liu <[email protected]>
> Cc: [email protected]; Qingkuan Lai <[email protected]>;
> Xiaomei Xue <[email protected]>; Uros Bizjak <[email protected]>;
> [email protected]; Liu, Hongtao <[email protected]>; Feng Xue
> <[email protected]>
> Subject: [PATCH v2] i386: Extend MOVNTDQ support to cover various packed
> integer types
> 
> Hi Hongtao,
> 
> 在 2026/6/2 10:15, Hongtao Liu 写道:
> > On Tue, Jun 2, 2026 at 9:16 AM Kewen Lin <[email protected]> wrote:
> >>
> >> CC maintainers for this.
> >>
> >> 在 2025/12/23 13:44, xuexiaomei 写道:
> >>> From: "[email protected]" <[email protected]>
> >>>
> >>> gcc/ChangeLog:
> >>>
> >>>         * config/i386/sse.md
> >>>         (define_mode_attr sse2):
> >>>         Add V4SI/V8SI/V16SI/V8HI/V16HI/V32HI.
> >>>         (<sse2>_movnt<mode>): Change operand mode from VI8 to
> >>>         VI_AVX_AVX512F.
> >>>         (define_mode_iterator STORENT_MODE):
> >>>         Add V4SI/V8SI/V16SI/V8HI/V16HI/V32HI/V16QI/V32QI/V64QI.
> >>>
> >>
> >> Though patch 1/2 needs rework as Richi's review comments, IMHO this
> >> patch 2/2 is independent of patch 1/2.
> > Ok.
> >
> 
> Thanks for the review!  I tweaked the patch a bit as below, the main change is
> to guard mode V4SI/V8HI/V16QI in STORENT_MODE with TARGET_SSE2 to
Thanks for catching this.

> align with <sse2>_movnt<mode> insn condition, along with some trivial
> formating adjustments.  It's tested well as before.
> 
> Is this v2 ok for trunk?
Ok.

> 
> BR,
> Kewen
> -----
> [PATCH] i386: Extend MOVNTDQ support to cover various packed integer
> types
> 
> As the associated mode iterator VI8 shows, MOVNTDQ is only used for *QI
> and *DI packed integer modes, this patch extends it to cover *HI and *SI
> packed integer modes.
> 
> gcc/ChangeLog:
> 
>       * config/i386/sse.md (define_mode_attr sse2): Add V4SI/V8SI/V16SI/
>       V8HI/V16HI/V32HI.
>       (<sse2>_movnt<mode>): Change mode iterator from VI8 to
> VI_AVX_AVX512F.
>       (define_mode_iterator STORENT_MODE): Add
> V4SI/V8SI/V16SI/V8HI/V16HI/
>       V32HI/V16QI/V32QI/V64QI.
> 
> Signed-off-by: Xiaomei Xue <[email protected]>
> ---
>  gcc/config/i386/sse.md | 16 +++++++++++-----
>  1 file changed, 11 insertions(+), 5 deletions(-)
> 
> diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md index
> 21773384973..a632e256e91 100644
> --- a/gcc/config/i386/sse.md
> +++ b/gcc/config/i386/sse.md
> @@ -1024,6 +1024,8 @@ (define_mode_attr sse
> 
>  (define_mode_attr sse2
>    [(V16QI "sse2") (V32QI "avx") (V64QI "avx512f")
> +   (V8HI "sse2") (V16HI "avx") (V32HI "avx512f")
> +   (V4SI "sse2") (V8SI "avx") (V16SI "avx512f")
>     (V2DI "sse2") (V4DI "avx") (V8DI "avx512f")])
> 
>  (define_mode_attr sse3
> @@ -2057,9 +2059,10 @@ (define_insn "<sse>_movnt<mode>"
>     (set_attr "mode" "<MODE>")])
> 
>  (define_insn "<sse2>_movnt<mode>"
> -  [(set (match_operand:VI8 0 "memory_operand" "=m")
> -     (unspec:VI8 [(match_operand:VI8 1 "register_operand" "v")]
> -                 UNSPEC_MOVNT))]
> +  [(set (match_operand:VI_AVX_AVX512F 0 "memory_operand" "=m")
> +     (unspec:VI_AVX_AVX512F
> +       [(match_operand:VI_AVX_AVX512F 1 "register_operand" "v")]
> +       UNSPEC_MOVNT))]
>    "TARGET_SSE2"
>    "%vmovntdq\t{%1, %0|%0, %1}"
>    [(set_attr "type" "ssecvt")
> @@ -2080,9 +2083,12 @@ (define_insn "<sse2>_movnt<mode>"
>  (define_mode_iterator STORENT_MODE
>    [(DI "TARGET_SSE2 && TARGET_64BIT") (SI "TARGET_SSE2")
>     (SF "TARGET_SSE4A") (DF "TARGET_SSE4A")
> -   (V8DI "TARGET_AVX512F") (V4DI "TARGET_AVX") (V2DI "TARGET_SSE2")
>     (V16SF "TARGET_AVX512F") (V8SF "TARGET_AVX") V4SF
> -   (V8DF "TARGET_AVX512F") (V4DF "TARGET_AVX") (V2DF "TARGET_SSE2")])
> +   (V8DF "TARGET_AVX512F") (V4DF "TARGET_AVX") (V2DF "TARGET_SSE2")
> +   (V8DI "TARGET_AVX512F") (V4DI "TARGET_AVX") (V2DI "TARGET_SSE2")
> +   (V16SI "TARGET_AVX512F") (V8SI "TARGET_AVX") (V4SI "TARGET_SSE2")
> +   (V32HI "TARGET_AVX512F") (V16HI "TARGET_AVX") (V8HI "TARGET_SSE2")
> +   (V64QI "TARGET_AVX512F") (V32QI "TARGET_AVX") (V16QI
> + "TARGET_SSE2")])
> 
>  (define_expand "storent<mode>"
>    [(set (match_operand:STORENT_MODE 0 "memory_operand")
> --
> 2.53.0

Reply via email to