void
tabort6 (int *mem)
{
/* Here global + 255 gets reloaded into a reg. Better would be to
just reload global or *mem and get the +255 for free as address
arithmetic. */
__builtin_tabort (*mem + 255);
}
foo (v4si a)
{
a += (v4si){ 1, 1, 1, 1 };
if (__builtin_tbegin (0) == 0)
{
a += (v4si){ 1, 1, 1, 1 };
__builtin_tabort (256);
__builtin_tend ();
}
else
a -= (v4si){ 1, 1, 1, 1 };
return a;
}
int main (void) {
#ifdef HAS_ISA_2_07
/* Just get the compiler to generate each of the TM instructions
* so we can verify that valgrind recognizes them.
* For now, only the tbegin instruction does anything in valgrind.
* The rest are just treated as NOPS.
*/
__builtin_tabort (0);
__builtin_tabortdc (0,0,0);
__builtin_tabortdci (0,0,0);
__builtin_tabortwc (0,0,0);
__builtin_tabortwci (0,0,0);
__builtin_tbegin (0);
__builtin_tend (0);
// __builtin_tcheck (0); tcheck not recognized by compiler
__builtin_trechkpt (); // not recognized by early HW
__builtin_treclaim (0); // not recognized by early HW
__builtin_tsr (0);
#endif
return 0;
}
int
__lll_trylock_elision (int *futex, short *adapt_count)
{
__asm__ volatile (".machinemode \"zarch_nohighgprs\"\n\t"
".machine \"all\""
: : : "memory");
/* Implement POSIX semantics by forbiding nesting elided trylocks.
Sorry. After the abort the code is re-executed
non transactional and if the lock was already locked
return an error. */
if (__builtin_tx_nesting_depth () > 0)
{
/* Note that this abort may terminate an outermost transaction that
was created outside glibc.
This persistently aborts the current transactions to force
them to use the default lock instead of retrying transactions
until their try_tbegin is zero.
*/
__builtin_tabort (_HTM_FIRST_USER_ABORT_CODE | 1);
}
/* Only try a transaction if it's worth it. */
if (*adapt_count <= 0)
{
unsigned status;
if (__builtin_expect
((status = __builtin_tbegin ((void *)0)) == _HTM_TBEGIN_STARTED, 1))
{
if (*futex == 0)
return 0;
/* Lock was busy. Fall back to normal locking. */
/* Since we are in a non-nested transaction there is no need to abort,
which is expensive. */
__builtin_tend ();
/* Note: Changing the adapt_count here might abort a transaction on a
different cpu, but that could happen anyway when the futex is
acquired, so there's no need to check the nesting depth here. */
if (aconf.skip_lock_busy > 0)
*adapt_count = aconf.skip_lock_busy;
}
else
{
if (status != _HTM_TBEGIN_TRANSIENT)
{
/* A persistent abort (cc 1 or 3) indicates that a retry is
probably futile. Use the normal locking now and for the
next couple of calls.
Be careful to avoid writing to the lock. */
if (aconf.skip_trylock_internal_abort > 0)
*adapt_count = aconf.skip_trylock_internal_abort;
}
}
/* Could do some retries here. */
}
else
{
/* Lost updates are possible, but harmless. Due to races this might lead
to *adapt_count becoming less than zero. */
(*adapt_count)--;
}
return lll_trylock (*futex);
}
int
foo (void)
{
return __builtin_tabort (10);
}
void
tabort8 ()
{
__builtin_tabort (-1);
}
void
tabort7 ()
{
__builtin_tabort (global + 255);
}
void
tabort5 ()
{
__builtin_tabort (global);
}
void
tabort4 (int *mem)
{
__builtin_tabort (*mem);
}
void
tabort3 (int reg)
{
/* { dg-final { scan-assembler-times "tabort\t255" 1 } } */
__builtin_tabort (reg + 255);
}
void
tabort2 (int reg)
{
__builtin_tabort (reg);
}
void
tabort1 ()
{
__builtin_tabort (256);
}
