https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127838
Bug ID: 127838
Summary: Incorrect vectorization with -march=x86-64-v2
Product: gcc
Version: 17.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: kristerw at gcc dot gnu.org
Blocks: 118443
Target Milestone: ---
The function f below is miscompiled on x86_64 with -O3 -march=x86-64-v2.
The overlap check inserted by the vectorizer overflows and therefore fails to
detect that writing to res may modify the element accessed in the next
iteration.
#include <sys/mman.h>
#include <unistd.h>
#include <stdint.h>
__attribute__((noipa))
void f (const uint32_t *s, char *res, int length)
{
unsigned long i;
for (i = 0; i < length; i++)
{
if (s[i] > 255)
__builtin_exit (0);
res[i] = (char)s[i];
}
}
int main()
{
long pgsz = sysconf (_SC_PAGESIZE);
void *p = mmap (NULL, pgsz * 2, PROT_READ|PROT_WRITE,
MAP_ANONYMOUS|MAP_PRIVATE, 0, 0);
if (p == MAP_FAILED)
return 0;
mprotect (p + pgsz, pgsz, PROT_NONE);
uint32_t *s = p + pgsz - 64;
s[0] = 1;
for (int i = 1; i < 16; i++)
s[i] = 0;
char *res = p + pgsz - 64 + 5;
f(s, res, 0xfffffff0);
return 0;
}
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118443
[Bug 118443] [Meta bug] Bugs triggered by and blocking more smtgcc testing
