https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93970
Martin Sebor <msebor at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |RESOLVED
Blocks|49774 |
Resolution|--- |INVALID
--- Comment #1 from Martin Sebor <msebor at gcc dot gnu.org> ---
The test case in comment #0 cannot be optimized as suggested because *q can be
equal to NaN. Replacing the (unfortunately chosen) double with some other type
like long shows that the expected optimization is done as expected.
$ cat pr93970.c && gcc -O2 -S -Wall -fdump-tree-optimized=/dev/stdout pr93970.c
void f (int *__restrict p, int *__restrict q)
{
int t = *q;
*p = 0;
if (t != *q) // folded to false
__builtin_abort ();
}
void g (int *__restrict p, long *__restrict q)
{
long t = *q;
*p = 0;
if (t != *q) // also folded as expected
__builtin_abort ();
}
;; Function f (f, funcdef_no=0, decl_uid=1931, cgraph_uid=1, symbol_order=0)
f (int * restrict p, int * restrict q)
{
<bb 2> [local count: 1073741824]:
*p_3(D) = 0;
return;
}
;; Function g (g, funcdef_no=3, decl_uid=1936, cgraph_uid=2, symbol_order=1)
g (int * restrict p, long int * restrict q)
{
<bb 2> [local count: 1073741824]:
*p_2(D) = 0;
return;
}
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=49774
[Bug 49774] [meta-bug] restrict qualification aliasing issues
