AndrewGoodrich
commented
2 years ago This change seems reasonable, but I would have implemented it a bit differently. I am not a big fan of having references to internal fields in sc_simcontext by methods in other classes, and I certainly would prefer those methods did not change sc_simcontext fields. My preference would be methods in sc_simcontext that return the events, creating them if necessary. An additional question is whether you want the methods in sc_event and sc_process_handle if you can just call the sc_simcontext methods. Also, as far as I can tell a single event for both none and non_event would be fine.
markfoodyburton
nmeum
The changes proposed here remove all(?) global
sc_events (sc_event::noneandsc_process_handle::non_event) by converting them to functions returning a reference to ansc_eventin the current simulation context. This eases reset of the simulation context, as discussed in #8, as thesesc_events no longer contain any reference to the initial simulation context.A useful side effect of these changes: Previously the simulation context was created on program startup by the
sc_eventconstructor (through the invocation ofsc_get_curr_simcontext()) by the aforementioned globalsc_events. With these changes applied it is created on the first invocation ofsc_get_curr_simcontext().It is currently unclear to me if it would be possible to return the same reference for
sc_event::none()andsc_process_handle::non_event(), currently two different class members in thesc_simcontextare used.To the best of my knowledge neither
sc_event::nonenorsc_process_handle::non_eventare mandated by IEEE Std 1666-2011. For this reason, the proposed changes should not affect existing standard-conforming SystemC programs.