[Ipopt-tickets] [Ipopt] #167: Any plan to make Ipopt thread-safe
Ipopt
coin-trac at coin-or.org
Sat Apr 5 14:31:50 EDT 2014
#167: Any plan to make Ipopt thread-safe
--------------------------+---------------------------
Reporter: FabF1 | Owner: ipopt-team
Type: enhancement | Status: closed
Priority: normal | Component: Ipopt
Version: 3.8 | Severity: normal
Resolution: fixed | Keywords: thread-safety
--------------------------+---------------------------
Comment (by stefan):
The method of using a combination of memory address and object-local tag
counter has reported to be buggy, as an object can obtain the same memory
address as a previously destructed object from the same class, so the tag
was not unique.
With r2424, we tried to make this problem happen less likely by
initializing the tag counter to the CPU times (in microseconds).
However, as this is more a workaround than a solution, r2472 restored the
original unique_tag methodology (by reverting r2424, r2269, r2267, r2250,
r2187), i.e., ensure unique tags for a single-threaded run.
Furthermore, with r2473, the unique_tag variable is now declared to be
[[http://en.wikipedia.org/wiki/Thread-local_storage|thread-local]], so
thread-safety should be restored.
--
Ticket URL: <https://projects.coin-or.org/Ipopt/ticket/167#comment:6>
Ipopt <http://projects.coin-or.org/Ipopt>
Interior-point optimizer for nonlinear programs.
More information about the Ipopt-tickets
mailing list