From: "H.J. Lu" <hjl.tools@gmail.com>
To: liuhongt <hongtao.liu@intel.com>
Cc: gcc-patches@gcc.gnu.org, crazylht@gmail.com, ubizjak@gmail.com
Subject: Re: [PATCH] Canonicalize vec_perm index to make the first index come from the first vector.
Date: Tue, 18 Oct 2022 16:49:13 -0700 [thread overview]
Message-ID: <CAMe9rOoJyJhOjmbLqMwM0_QmbQcmA9X1nm2Z_=6zrAgh7BVN9g@mail.gmail.com> (raw)
In-Reply-To: <20221018232301.264776-1-hongtao.liu@intel.com>
On Tue, Oct 18, 2022 at 4:25 PM liuhongt <hongtao.liu@intel.com> wrote:
>
> Fix unexpected non-canon form from gimple vector selector.
>
> Bootstrapped and regtested on x86_64-pc-linux-gnu{-m32,}.
> Ok for trunk?
>
> gcc/ChangeLog:
>
> PR target/107271
> * config/i386/i386-expand.cc (ix86_vec_perm_index_canon): New.
> (expand_vec_perm_shufps_shufps): Call
> ix86_vec_perm_index_canon
>
> gcc/testsuite/ChangeLog:
>
> * gcc.target/i386/pr107271.c: New test.
> ---
> gcc/config/i386/i386-expand.cc | 17 +++++++++++++++++
> gcc/testsuite/gcc.target/i386/pr107271.c | 16 ++++++++++++++++
> 2 files changed, 33 insertions(+)
> create mode 100644 gcc/testsuite/gcc.target/i386/pr107271.c
>
> diff --git a/gcc/config/i386/i386-expand.cc b/gcc/config/i386/i386-expand.cc
> index 6baff6d0e61..4f121516091 100644
> --- a/gcc/config/i386/i386-expand.cc
> +++ b/gcc/config/i386/i386-expand.cc
> @@ -19604,6 +19604,22 @@ expand_vec_perm_1 (struct expand_vec_perm_d *d)
> return false;
> }
>
> +/* Canonicalize vec_perm index to make the first index
> + always comes from the first index. */
vector?
> +static void
> +ix86_vec_perm_index_canon (struct expand_vec_perm_d *d)
> +{
> + unsigned nelt = d->nelt;
> + if (d->perm[0] < nelt)
> + return;
> +
> + for (unsigned i = 0; i != nelt; i++)
> + d->perm[i] = (d->perm[i] + nelt) % (2 * nelt);
> +
> + std::swap (d->op0, d->op1);
> + return;
> +}
> +
> /* A subroutine of ix86_expand_vec_perm_const_1. Try to implement D
> in terms of a pair of shufps+ shufps/pshufd instructions. */
> static bool
> @@ -19621,6 +19637,7 @@ expand_vec_perm_shufps_shufps (struct expand_vec_perm_d *d)
> if (d->testing_p)
> return true;
>
> + ix86_vec_perm_index_canon (d);
> for (i = 0; i < 4; ++i)
> count += d->perm[i] > 3 ? 1 : 0;
>
> diff --git a/gcc/testsuite/gcc.target/i386/pr107271.c b/gcc/testsuite/gcc.target/i386/pr107271.c
> new file mode 100644
> index 00000000000..fe89c9a5bef
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/pr107271.c
> @@ -0,0 +1,16 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O0" } */
> +
> +typedef int __attribute__((__vector_size__ (16))) V;
> +
> +static inline __attribute__((__always_inline__)) V
> +bar (V v128u32_0)
> +{
> + return __builtin_shuffle ((V){}, v128u32_0, v128u32_0);
> +}
> +
> +V
> +foo (void)
> +{
> + return bar ((V){7, 4, 4});
> +}
> --
> 2.27.0
>
--
H.J.
next prev parent reply other threads:[~2022-10-18 23:49 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-10-18 23:23 liuhongt
2022-10-18 23:49 ` H.J. Lu [this message]
2022-10-18 23:51 ` Hongtao Liu
2022-10-19 5:58 ` Uros Bizjak
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAMe9rOoJyJhOjmbLqMwM0_QmbQcmA9X1nm2Z_=6zrAgh7BVN9g@mail.gmail.com' \
--to=hjl.tools@gmail.com \
--cc=crazylht@gmail.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=hongtao.liu@intel.com \
--cc=ubizjak@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).