public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
From: "goon.pri.low at gmail dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug tree-optimization/112657] New: missed optimization: cmove not used with multiple returns
Date: Tue, 21 Nov 2023 21:42:59 +0000	[thread overview]
Message-ID: <bug-112657-4@http.gcc.gnu.org/bugzilla/> (raw)

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112657

            Bug ID: 112657
           Summary: missed optimization: cmove not used with multiple
                    returns
           Product: gcc
           Version: 13.2.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: goon.pri.low at gmail dot com
  Target Milestone: ---

This function

int unopt(int c) {
    if (c == 14)
        return -9;
    else
        return c;
}

unopt:
        mov     eax, edi
        cmp     edi, 14
        je      .L4
        ret
.L4:
        mov     eax, -9
        ret

Should probably be optimized to:

int opt(int c) {
    if (c == 14)
        c = -9;

    return c;
}

opt:
        cmp     edi, 14
        mov     eax, -9
        cmovne  eax, edi
        ret

This seems to only really happen when negative numbers are used,

int positive(int c) {
    if (c == 14)
        return 9;
    else
        return c;
}

positive:
        mov     eax, edi
        cmp     edi, 14
        mov     edx, 9
        cmove   eax, edx
        ret

Though use of positive values still isn't completely optimized (possibly same
as https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97968)

Also it seems like if the order is reversed:

int reverse(int c) {
    if (c != 14)
        c = 9120;

    return c;
}

reverse:
        cmp     edi, 14
        mov     edx, 14
        mov     eax, 9120
        cmove   eax, edx
        ret

We could use %edi in the cmove and eliminate the 2nd instruction.

             reply	other threads:[~2023-11-21 21:42 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-11-21 21:42 goon.pri.low at gmail dot com [this message]
2023-11-21 21:53 ` [Bug target/112657] " pinskia at gcc dot gnu.org
2023-11-21 21:53 ` pinskia at gcc dot gnu.org
2023-11-21 21:59 ` [Bug rtl-optimization/112657] [13/14 Regression] " pinskia at gcc dot gnu.org
2023-11-22  9:16 ` ubizjak at gmail dot com
2023-11-22  9:28 ` ubizjak at gmail dot com
2023-11-22  9:38 ` ubizjak at gmail dot com
2023-11-22  9:46 ` ubizjak at gmail dot com
2023-11-22 10:48 ` rguenth at gcc dot gnu.org
2023-11-22 13:35 ` hubicka at gcc dot gnu.org
2024-03-07 21:01 ` law at gcc dot gnu.org
2024-05-21  9:18 ` [Bug rtl-optimization/112657] [13/14/15 " jakub at gcc dot gnu.org

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=bug-112657-4@http.gcc.gnu.org/bugzilla/ \
    --to=gcc-bugzilla@gcc.gnu.org \
    --cc=gcc-bugs@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).