public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug tree-optimization/113380] New: `(a CMP CST) |  (a*a CMP CST1)` is not optimized as decent as with `||`
@ 2024-01-13 17:01 pinskia at gcc dot gnu.org
  2024-01-15  8:37 ` [Bug tree-optimization/113380] " rguenth at gcc dot gnu.org
  2024-01-15  9:21 ` pinskia at gcc dot gnu.org
  0 siblings, 2 replies; 3+ messages in thread
From: pinskia at gcc dot gnu.org @ 2024-01-13 17:01 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113380

            Bug ID: 113380
           Summary: `(a CMP CST) |  (a*a CMP CST1)` is not optimized as
                    decent as with `||`
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: enhancement
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pinskia at gcc dot gnu.org
  Target Milestone: ---

Take:
```
int foo(int a)
{
   return (a < 4) | (a * a > 7);
}

int bar(int a)
{
   return (a < 4) | (a * a < 7);
}

int baz(int a)
{
   return (a > -4) | (a * a > 7);
}

int qux(int a)
{
   return (a > -4) | (a * a < 7);
}
```

This is not as optimized as:
```
int foo(int a)
{
   return (a < 4) || (a * a > 7);
}

int bar(int a)
{
   return (a < 4) || (a * a < 7);
}

int baz(int a)
{
   return (a > -4) || (a * a > 7);
}

int qux(int a)
{
   return (a > -4) || (a * a < 7);
}
```

^ permalink raw reply	[flat|nested] 3+ messages in thread

* [Bug tree-optimization/113380] `(a CMP CST) |  (a*a CMP CST1)` is not optimized as decent as with `||`
  2024-01-13 17:01 [Bug tree-optimization/113380] New: `(a CMP CST) | (a*a CMP CST1)` is not optimized as decent as with `||` pinskia at gcc dot gnu.org
@ 2024-01-15  8:37 ` rguenth at gcc dot gnu.org
  2024-01-15  9:21 ` pinskia at gcc dot gnu.org
  1 sibling, 0 replies; 3+ messages in thread
From: rguenth at gcc dot gnu.org @ 2024-01-15  8:37 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113380

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Interesting - smells like fold-const.cc stuff not in match.pd.

^ permalink raw reply	[flat|nested] 3+ messages in thread

* [Bug tree-optimization/113380] `(a CMP CST) |  (a*a CMP CST1)` is not optimized as decent as with `||`
  2024-01-13 17:01 [Bug tree-optimization/113380] New: `(a CMP CST) | (a*a CMP CST1)` is not optimized as decent as with `||` pinskia at gcc dot gnu.org
  2024-01-15  8:37 ` [Bug tree-optimization/113380] " rguenth at gcc dot gnu.org
@ 2024-01-15  9:21 ` pinskia at gcc dot gnu.org
  1 sibling, 0 replies; 3+ messages in thread
From: pinskia at gcc dot gnu.org @ 2024-01-15  9:21 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113380

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #1)
> Interesting - smells like fold-const.cc stuff not in match.pd.

It is not, In the `||` case, the ethread (ranger) is able to opimize away the
second part to either 0 or 1 due to having basic blocks. While in the `|` case
there is no basic block for ranger to anchor to.

I am not 100% sure how to solve this but an idea is if you have a `bool |
bool`, always create during gimplification `if (a) goto c; else if (b) goto c;
else goto d; ` Likewise for `bool & bool`.

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2024-01-15  9:21 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-01-13 17:01 [Bug tree-optimization/113380] New: `(a CMP CST) | (a*a CMP CST1)` is not optimized as decent as with `||` pinskia at gcc dot gnu.org
2024-01-15  8:37 ` [Bug tree-optimization/113380] " rguenth at gcc dot gnu.org
2024-01-15  9:21 ` pinskia at gcc dot gnu.org

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).