This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH i386 AVX512] [19-1/n] Fix shuf*64x2 assembler operand.
- From: Uros Bizjak <ubizjak at gmail dot com>
- To: Kirill Yukhin <kirill dot yukhin at gmail dot com>
- Cc: Jakub Jelinek <jakub at redhat dot com>, Richard Henderson <rth at redhat dot com>, GCC Patches <gcc-patches at gcc dot gnu dot org>
- Date: Wed, 27 Aug 2014 12:31:36 +0200
- Subject: Re: [PATCH i386 AVX512] [19-1/n] Fix shuf*64x2 assembler operand.
- Authentication-results: sourceware.org; auth=none
- References: <20140815115159 dot GE35144 at msticlxl57 dot ims dot intel dot com> <CAFULd4YGOu7kjGzdgyY0DkKcD+f5N+RWbO52WV9Ntjrx+Of9vw at mail dot gmail dot com> <20140820125150 dot GB37280 at msticlxl57 dot ims dot intel dot com> <CAFULd4axbApstDDF7yjPzK3=pkC0StEoemvrK6HyUU0FYkTHsQ at mail dot gmail dot com> <20140827091154 dot GB33178 at msticlxl57 dot ims dot intel dot com>
On Wed, Aug 27, 2014 at 11:11 AM, Kirill Yukhin <kirill.yukhin@gmail.com> wrote:
>> OK with this change.
>
> I've discovered a problem with assembler operand of
> vshuf*64x2 insn.
>
> Patch below bootstrapped and avx512-regtested.
>
> Is it ok for trunk?
OK with a couple of nits.
Thanks,
Uros.
> gcc/
> * config/i386/sse.md (define_mode_attr concat_tg_mode):
> Move up.
No need for the above entry.
> (define_insn "<mask_codefor>avx512dq_broadcast<mode><mask_name>_1"):
> Use `concat_tg_mode' attribute to determine asm register size.
>
> --
> Thanks, K
>
> diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md
> index fa23330..e1d42f8 100644
> --- a/gcc/config/i386/sse.md
> +++ b/gcc/config/i386/sse.md
> @@ -718,6 +718,11 @@
> (V16SF "ss") (V8SF "ss") (V4SF "ss")
> (V8DF "sd") (V4DF "sd") (V2DF "sd")])
>
> +;; Tight mode of assembler operand to mode iterator
Tie mode ...
> +(define_mode_attr concat_tg_mode
> + [(V32QI "t") (V16HI "t") (V8SI "t") (V4DI "t") (V8SF "t") (V4DF "t")
> + (V64QI "g") (V32HI "g") (V16SI "g") (V8DI "g") (V16SF "g") (V8DF "g")])
> +