https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120086
--- Comment #20 from John David Anglin <danglin at gcc dot gnu.org> ---
(In reply to John David Anglin from comment #19)
> It's not clear why gcov-30 passes on Linux. I searched the entire tree and
> couldn't find how TARGET_HAVE_LIBATOMIC gets set on Linux. Also, always
> setting TARGET_HAVE_LIBATOMIC to true doesn't seem 100% correct because I
> believe it's possible to disable building libatomic.
It's not set on Linux.
I see issue. We have in tree-profile.cc:
/* Verify whether we can utilize atomic update operations. */
bool can_support_atomic = targetm.have_libatomic;
unsigned HOST_WIDE_INT gcov_type_size
= tree_to_uhwi (TYPE_SIZE_UNIT (get_gcov_type ()));
bool have_atomic_4
= HAVE_sync_compare_and_swapsi || HAVE_atomic_compare_and_swapsi;
bool have_atomic_8
= HAVE_sync_compare_and_swapdi || HAVE_atomic_compare_and_swapdi;
bool needs_split = gcov_type_size == 8 && !have_atomic_8 && have_atomic_4;
if (!can_support_atomic)
{
if (gcov_type_size == 4)
can_support_atomic = have_atomic_4;
else if (gcov_type_size == 8)
can_support_atomic = have_atomic_8;
}
HAVE_sync_compare_and_swapdi isn't defined because this operation is
implemented as a libcall. This is also true on Linux but we have
in pa-linux.h:
/* The SYNC operations are implemented as library functions, not
INSN patterns. As a result, the HAVE defines for the patterns are
not defined. We need to define them to generate the corresponding
__GCC_HAVE_SYNC_COMPARE_AND_SWAP_* and __GCC_ATOMIC_*_LOCK_FREE
defines. */
#define HAVE_sync_compare_and_swapqi 1
#define HAVE_sync_compare_and_swaphi 1
#define HAVE_sync_compare_and_swapsi 1
#define HAVE_sync_compare_and_swapdi 1
