public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Richard Biener <richard.guenther@gmail.com>
To: Andrew Pinski <apinski@marvell.com>
Cc: GCC Patches <gcc-patches@gcc.gnu.org>
Subject: Re: [PATCH] Fix PR 10153: tail recusion for vector types.
Date: Wed, 21 Jul 2021 16:12:58 +0200	[thread overview]
Message-ID: <CAFiYyc2XQyz+9MPGkN8dr100WvD5ZuNrZAB-owAfeN2AL+1BbA@mail.gmail.com> (raw)
In-Reply-To: <1626818523-4848-1-git-send-email-apinski@marvell.com>

On Wed, Jul 21, 2021 at 12:03 AM apinski--- via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> From: Andrew Pinski <apinski@marvell.com>
>
> The problem here is we try to an initialized value
> from a scalar constant. For vectors we need to do
> a vect_dup instead.  This fixes that issue and we
> get the correct code even and it does not crash.
>
> OK? Bootstrapped and tested on aarch64-linux-gnu with no regressions.

I think the create_tailcall_accumulators should simply
use build_{one,zero}_cst (type) instead of passing down
integer_{zero,one}_node.  Then we can even elide the fold_convert ...

> gcc/ChangeLog:
>
>         PR tree-optimize/10153
>         * tree-tailcall.c (create_tailcall_accumulator):
>         For vector types create a duplicated VECTOR_CST
>         before calling fold_convert.
>
> gcc/testsuite/ChangeLog:
>
>         PR tree-optimize/10153
>         * gcc.c-torture/compile/pr10153-1.c: New test.
>         * gcc.c-torture/compile/pr10153-2.c: New test.
> ---
>  gcc/testsuite/gcc.c-torture/compile/pr10153-1.c | 7 +++++++
>  gcc/testsuite/gcc.c-torture/compile/pr10153-2.c | 9 +++++++++
>  gcc/tree-tailcall.c                             | 3 +++
>  3 files changed, 19 insertions(+)
>  create mode 100644 gcc/testsuite/gcc.c-torture/compile/pr10153-1.c
>  create mode 100644 gcc/testsuite/gcc.c-torture/compile/pr10153-2.c
>
> diff --git a/gcc/testsuite/gcc.c-torture/compile/pr10153-1.c b/gcc/testsuite/gcc.c-torture/compile/pr10153-1.c
> new file mode 100644
> index 00000000000..3f2040f32a1
> --- /dev/null
> +++ b/gcc/testsuite/gcc.c-torture/compile/pr10153-1.c
> @@ -0,0 +1,7 @@
> +typedef int V __attribute__ ((vector_size (2 * sizeof (int))));
> +V
> +foo (void)
> +{
> +  V v = { };
> +  return v - foo();
> +}
> diff --git a/gcc/testsuite/gcc.c-torture/compile/pr10153-2.c b/gcc/testsuite/gcc.c-torture/compile/pr10153-2.c
> new file mode 100644
> index 00000000000..1af4c8e2a36
> --- /dev/null
> +++ b/gcc/testsuite/gcc.c-torture/compile/pr10153-2.c
> @@ -0,0 +1,9 @@
> +typedef int V __attribute__ ((vector_size (2 * sizeof (int))));
> +V
> +foo (int t)
> +{
> +  if (t < 10)
> +    return (V){1, 1};
> +  V v = { };
> +  return v - foo(t - 1);
> +}
> diff --git a/gcc/tree-tailcall.c b/gcc/tree-tailcall.c
> index a4d31c90c49..9d1a98b1cfd 100644
> --- a/gcc/tree-tailcall.c
> +++ b/gcc/tree-tailcall.c
> @@ -1080,6 +1080,9 @@ create_tailcall_accumulator (const char *label, basic_block bb, tree init)
>
>    phi = create_phi_node (tmp, bb);
>    /* RET_TYPE can be a float when -ffast-maths is enabled.  */
> +  /* For vectors create a dup. */
> +  if (VECTOR_TYPE_P (ret_type))
> +    init = build_vector_from_val (ret_type, fold_convert (TREE_TYPE (ret_type), init));
>    add_phi_arg (phi, fold_convert (ret_type, init), single_pred_edge (bb),
>                UNKNOWN_LOCATION);
>    return PHI_RESULT (phi);
> --
> 2.27.0
>

  reply	other threads:[~2021-07-21 14:13 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-07-20 22:02 apinski
2021-07-21 14:12 ` Richard Biener [this message]
2021-07-21 19:26 apinski
2021-07-22 11:53 ` Richard Biener

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=CAFiYyc2XQyz+9MPGkN8dr100WvD5ZuNrZAB-owAfeN2AL+1BbA@mail.gmail.com \
    --to=richard.guenther@gmail.com \
    --cc=apinski@marvell.com \
    --cc=gcc-patches@gcc.gnu.org \
    /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).