On Tue, Nov 16, 2010 at 1:13 PM, Nilay Vaish <[email protected]> wrote:


> All in all, we would ideally like to both speed up individual calls and
>> reduce the number of calls. IIRC, gprof indicated that findTagInSet() was
>> called 4-5X more frequently than there were cache accesses, which makes no
>> sense to me; it seems like a typical cache hit should only require a
>> single
>> tag lookup.
>>
>
> Should this not be true in case a multiprocessor system is being simulated?
> I am not aware the configuration that ruby_fs.py makes use of.


I don't quite follow your question, and I don't recall the stats I looked at
enough to remember exactly what the "cache access" statistic I was looking
at referred to.  However, the bottom line is that an L1 hit should not
require more than one tag lookup, and most accesses should be L1 hits.  It's
true that if you have a miss and you have to send out multiple probes then
you will have several tag lookups, but that should be relatively rare
compared to the total number of memory accesses.

Steve
_______________________________________________
m5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/m5-dev

Reply via email to