On 12/18/2014 09:02 PM, Mike Holmes wrote:
Maxim is this scheduled for 0.7 ? I think the tests from
Yan,Barry,Mario require it
If it's ready I think we don't need to wait and include it. Working on
validation / merge...
Maxim.
On 10 December 2014 at 08:29, Bill Fischofer
<[email protected] <mailto:[email protected]>> wrote:
On Wed, Dec 10, 2014 at 4:55 AM, Ola Liljedahl
<[email protected] <mailto:[email protected]>> wrote:
Signed-off-by: Ola Liljedahl <[email protected]
<mailto:[email protected]>>
Reviewed-by: Bill Fischofer <[email protected]
<mailto:[email protected]>>
---
(This document/code contribution attached is provided under
the terms of
agreement LES-LTM-21309)
Implemented the missing odp_ticketlock_trylock().
platform/linux-generic/odp_ticketlock.c | 30
++++++++++++++++++++++++++++++
1 file changed, 30 insertions(+)
diff --git a/platform/linux-generic/odp_ticketlock.c
b/platform/linux-generic/odp_ticketlock.c
index 682b01b..6525786 100644
--- a/platform/linux-generic/odp_ticketlock.c
+++ b/platform/linux-generic/odp_ticketlock.c
@@ -34,6 +34,36 @@ void odp_ticketlock_lock(odp_ticketlock_t
*ticketlock)
odp_spin();
}
+int odp_ticketlock_trylock(odp_ticketlock_t *tklock)
+{
+ /* We read 'next_ticket' and 'cur_ticket'
non-atomically which should
+ * not be a problem as they are not independent of
each other.
+ * 'cur_ticket' is always <= to 'next_ticket' and if
we see an
+ * older value of 'cur_ticket', this only means the
lock will
+ * look busy and trylock will fail. */
+ uint32_t next = odp_atomic_load_u32(&tklock->next_ticket);
+ uint32_t cur = odp_atomic_load_u32(&tklock->cur_ticket);
+ /* First check that lock is available and possible to
take without
+ * spinning. */
+ if (next == cur) {
+ /* Then try to take the lock by incrementing
'next_ticket'
+ * but only if it still has the original value
which is
+ * equal to 'cur_ticket'.
+ * We don't have to include 'cur_ticket' in
the comparison
+ * because it cannot be larger than
'next_ticket' (only
+ * smaller if the lock is busy).
+ * If CAS fails, it means some other thread
intercepted and
+ * took a ticket which means the lock is not
available
+ * anymore */
+ if
(_odp_atomic_u32_cmp_xchg_strong_mm(&tklock->next_ticket,
+ &next,
+ next + 1,
+ _ODP_MEMMODEL_ACQ,
+ _ODP_MEMMODEL_RLX))
+ return 1;
+ }
+ return 0;
+}
void odp_ticketlock_unlock(odp_ticketlock_t *ticketlock)
{
--
1.9.1
_______________________________________________
lng-odp mailing list
[email protected] <mailto:[email protected]>
http://lists.linaro.org/mailman/listinfo/lng-odp
_______________________________________________
lng-odp mailing list
[email protected] <mailto:[email protected]>
http://lists.linaro.org/mailman/listinfo/lng-odp
--
*Mike Holmes*
Linaro Sr Technical Manager
LNG - ODP
_______________________________________________
lng-odp mailing list
[email protected]
http://lists.linaro.org/mailman/listinfo/lng-odp
_______________________________________________
lng-odp mailing list
[email protected]
http://lists.linaro.org/mailman/listinfo/lng-odp
