public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug tree-optimization/110335] New: int converted to double should be considered non-nan and non-nan*non-nan is still non-nan
@ 2023-06-21 7:35 pinskia at gcc dot gnu.org
2023-06-22 7:39 ` [Bug tree-optimization/110335] " pinskia at gcc dot gnu.org
0 siblings, 1 reply; 2+ messages in thread
From: pinskia at gcc dot gnu.org @ 2023-06-21 7:35 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110335
Bug ID: 110335
Summary: int converted to double should be considered non-nan
and non-nan*non-nan is still non-nan
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:
```
void
h (int n, int from, int to)
{
double dto = to;
double damt = n * dto;
if (damt != damt)
__builtin_trap();
}
void
h1 (double dto, double n)
{
if (dto != dto) __builtin_unreachable();
if (n != n) __builtin_unreachable();
double damt = n * dto;
if (damt != damt)
__builtin_trap();
}
```
I would assume this two functions be optimized to just `return;`
LLVM is able to optimize h but not h1.
^ permalink raw reply [flat|nested] 2+ messages in thread
* [Bug tree-optimization/110335] int converted to double should be considered non-nan and non-nan*non-nan is still non-nan
2023-06-21 7:35 [Bug tree-optimization/110335] New: int converted to double should be considered non-nan and non-nan*non-nan is still non-nan pinskia at gcc dot gnu.org
@ 2023-06-22 7:39 ` pinskia at gcc dot gnu.org
0 siblings, 0 replies; 2+ messages in thread
From: pinskia at gcc dot gnu.org @ 2023-06-22 7:39 UTC (permalink / raw)
To: gcc-bugs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110335
--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Another testcase:
```
void
h3(int to)
{
double dto = to;
double damt = dto +2;
if (damt != damt)
__builtin_trap();
}
```
Note we do optimize:
```
void
done(int to)
{
double dto = to;
double damt = -dto;
if (damt != damt)
__builtin_trap();
}
```
But via match and simplify.
Note there is tree_expr_maybe_nan_p and tree_expr_finite_p in fold-const.cc
Which can be used as a model for some frange related things too.
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2023-06-22 7:39 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-06-21 7:35 [Bug tree-optimization/110335] New: int converted to double should be considered non-nan and non-nan*non-nan is still non-nan pinskia at gcc dot gnu.org
2023-06-22 7:39 ` [Bug tree-optimization/110335] " 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).