Daniele,
I can't comment on the DNUCA issues, as I know very little about it
myself. But as far as triggerEvents() vs. triggerAllEvents(), changing
between them will definately change Ruby's functionality. Specifically,
triggerAllEvents() would essentially cause all outstanding accesses to
complete... at which time Ruby would have no remaining events.
triggerEvents(current+1000) *may* do the same thing, but, depending on
latencies, contention, etc., may leave some requests only partially
completed.
Regards,
Dan
Daniele Bordes wrote:
Hi all. I'm using tester.exec to perform trace driven simulations, and
I have a question: after all references in the trace file have been
read and a void line is found in the file by the tracer, are all the
events generated by the last reference processed or does the
simulation end without processing some of the events?
I have this doubts because I don't understand some ruby output
statistics about DNUCA_L2CACHE hits, misses, allocations and
promotions.
Another question:
Do I introduce a bug or problems if in function
"Tracer::playbackTrace" I substitute the function call:
g_eventQueue_ptr->triggerEvents(g_eventQueue_ptr->getTime() + 1000);
// FIXME - should be smarter
with:
g_eventQueue_ptr->triggerAllEvents()
?
Sorry if I am a nuisance.
Thank you all very much.
_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users
Use Google to search the GEMS Users mailing list by adding "site:https://lists.cs.wisc.edu/archive/gems-users/" to your search.
|