> Hence in this one specific
> case you have to synthesize a no-op just to have something to mark as
> faulting.
This sounds like it was a FS problem that was solved but then kind of
trickled down to SE right? I think I remember this happening as well
but formerly in SE mode whenever you got a fault that was not a page
fault to increase the stack then we died so thats why I think it is
something that must of got permeated through.

> Korey, do you recall what smtNumFetchingThreads does with respect to
> the fetch stage?
Yes, that *should* be how many threads fetch from a cycle at once.
That number should be 1 by default.

> If you're saying that it wasn't designed to do that, but in
> reality it is doing that in the default configuration, then that
> sounds bad.
If that's what's happening, that definitely is bad as we made the
design decision that since the majority of SMT implementations would
want to fetch from 1 thread at a time. I'm not sure how that got set
to anything other than 1 other than it was edited somehow when we were
translating to the new type of params. In the *old* days, the params
the O3 configuration file set this to 1 I'm pretty sure so we never
saw this problem.

Setting that default to 1 should solve the initial problem.

For the case where we really do want to fetch from multiple threads,
we added the initial framework/functionality for it in anticipation of
future work but we dont have any workloads that would stress that
situation as it is uncommon.

-- 
----------
Korey L Sewell
Graduate Student - PhD Candidate
Computer Science & Engineering
University of Michigan
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users

Reply via email to