https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102838
--- Comment #15 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Ah, bet Solaris aligned_alloc relies on:
"the value of size shall be an integral multiple of alignment"
(glibc aligned_alloc doesn't).
Does memalign or posix_memalign rely on that too, or just aligned_alloc?
If just aligned_alloc, we could do incrementally:
2021-11-17 Jakub Jelinek <ja...@redhat.com>
PR libgomp/102838
* alloc.c (gomp_aligned_alloc): Prefer _aligned_alloc over
memalign over posix_memalign over aligned_alloc over fallback
with malloc instead of aligned_alloc over _aligned_alloc over
posix_memalign over memalign over fallback with malloc. For
aligned_alloc, round up size up to multiple of al.
--- libgomp/alloc.c.jj 2021-01-04 10:25:56.157037659 +0100
+++ libgomp/alloc.c 2021-11-17 13:32:25.246271672 +0100
@@ -65,18 +65,24 @@ gomp_aligned_alloc (size_t al, size_t si
void *ret;
if (al < sizeof (void *))
al = sizeof (void *);
-#ifdef HAVE_ALIGNED_ALLOC
- ret = aligned_alloc (al, size);
-#elif defined(HAVE__ALIGNED_MALLOC)
+#ifdef HAVE__ALIGNED_MALLOC
ret = _aligned_malloc (size, al);
-#elif defined(HAVE_POSIX_MEMALIGN)
- if (posix_memalign (&ret, al, size) != 0)
- ret = NULL;
#elif defined(HAVE_MEMALIGN)
{
extern void *memalign (size_t, size_t);
ret = memalign (al, size);
}
+#elif defined(HAVE_POSIX_MEMALIGN)
+ if (posix_memalign (&ret, al, size) != 0)
+ ret = NULL;
+#lif defined(HAVE_ALIGNED_ALLOC)
+ {
+ size_t sz = (size + al - 1) & ~(al - 1);
+ if (__builtin_expect (sz >= size, 1))
+ ret = aligned_alloc (al, sz);
+ else
+ ret = NULL;
+ }
#else
ret = NULL;
if ((al & (al - 1)) == 0 && size)
