public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
From: "pinskia at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug tree-optimization/115202] [11/12/13/14/15 Regression] Missed optimization: std::min(f ?  (unsigned short)m : a, ~0)
Date: Thu, 23 May 2024 12:18:41 +0000	[thread overview]
Message-ID: <bug-115202-4-kzgA3PVnoH@http.gcc.gnu.org/bugzilla/> (raw)
In-Reply-To: <bug-115202-4@http.gcc.gnu.org/bugzilla/>

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115202

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|14.2                        |11.5
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2024-05-23
            Summary|[14/15 Regression] Missed   |[11/12/13/14/15 Regression]
                   |optimization: std::min(f ?  |Missed optimization:
                   |(unsigned short)m : a, ~0)  |std::min(f ?  (unsigned
                   |                            |short)m : a, ~0)
     Ever confirmed|0                           |1

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.
Here is a reduced testcase:
```
unsigned short m;
template<typename T>
const T min(const T a, const T b)
{
        if (a < b)
          return a;
        return b;
}
unsigned short func(int a, int f) {
    return min(f ? m : a, ~0);
}
```

The above is a regression from GCC 8 where in GCC 9 adds an early phiopt.


And here is a version which is not a regression:
```
unsigned short m;
template<typename T>
inline const T min(const T a, const T b)
{
  return a < b ? a : b;
}
unsigned short func(int a, int f) {
    return min(f ? m : a, ~0);
}
```


The real issue is we don't optimize MIN<PHI<>, -1>.
From the IR:
```
  # RANGE [irange] int [0, 65535] MASK 0xffff VALUE 0x0
  iftmp.0_6 = (intD.9) m.1_1;
;;    succ:       4 [always]  count:536870912 (estimated locally, freq 0.5000)
(FALLTHRU,EXECUTABLE)

;;   basic block 4, loop depth 0, count 1073741824 (estimated locally, freq
1.0000), maybe hot
;;    prev block 3, next block 1, flags: (NEW, REACHABLE, VISITED)
;;    pred:       3 [always]  count:536870912 (estimated locally, freq 0.5000)
(FALLTHRU,EXECUTABLE)
;;                2 [50.0% (guessed)]  count:536870912 (estimated locally, freq
0.5000) (FALSE_VALUE,EXECUTABLE)
  # iftmp.0_2 = PHI <iftmp.0_6(3), a_4(D)(2)>
  # RANGE [irange] int [-INF, -1]
  _7 = MIN_EXPR <iftmp.0_2, -1>;
```

Obviously here `MIN_EXPR <X, -1>` for the iftmp.0_6 part of the PHI, it will be
-1.

      parent reply	other threads:[~2024-05-23 12:18 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-05-23  7:44 [Bug tree-optimization/115202] New: [14/15 " 652023330028 at smail dot nju.edu.cn
2024-05-23  9:02 ` [Bug tree-optimization/115202] " rguenth at gcc dot gnu.org
2024-05-23 12:18 ` pinskia at gcc dot gnu.org [this message]

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-115202-4-kzgA3PVnoH@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).