https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101250
Bug ID: 101250
Summary: adjust_iv_update_pos update the iv statement
unexpectedly cause memory address offset mismatch
Product: gcc
Version: 12.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
Assignee: unassigned at gcc dot gnu.org
Reporter: luoxhu at gcc dot gnu.org
Target Milestone: ---
Test case:
unsigned int foo (unsigned char *ip, unsigned char *ref, unsigned int maxlen)
{
unsigned int len = 2;
do {
len++;
}while(len < maxlen && ip[len] == ref[len]);
return len;
}
ivopts:
<bb 4> [local count: 1014686026]:
_3 = MEM[(unsigned char *)ip_10(D) + ivtmp.16_15 * 1];
ivtmp.16_16 = ivtmp.16_15 + 1;
_19 = ref_12(D) + 18446744073709551615;
_6 = MEM[(unsigned char *)_19 + ivtmp.16_16 * 1];
if (_3 == _6)
goto <bb 6>; [94.50%]
else
goto <bb 10>; [5.50%]
Disable adjust_iv_update_pos will produce:
<bb 4> [local count: 1014686026]:
_3 = MEM[(unsigned char *)ip_10(D) + ivtmp.16_15 * 1];
_6 = MEM[(unsigned char *)ref_12(D) + ivtmp.16_15 * 1];
ivtmp.16_16 = ivtmp.16_15 + 1;
if (_3 == _6)
goto <bb 6>; [94.50%]
else
goto <bb 10>; [5.50%]
discussions:
https://gcc.gnu.org/pipermail/gcc-patches/2021-June/573709.html