Historically there have been a couple of problems with this:
- it's hard to guarantee that the arrays are sized properly (e.g.,
originally a cache didn't know how many CPUs/threads it was connected to)
- some packets (like writebacks) don't have an associated ID

Lisa did a bunch of work to clean these issues up, see for example:
http://repo.m5sim.org/m5/rev/ab05e20dc4a7
but it looks like she only fixed up a few of the stats.

Note that you really want contextId() and not threadId() (the existing
commented-out references to threadId() are very old).

Maybe Lisa can speak up if there's anything else to add.

Steve

On Wed, Mar 2, 2011 at 10:25 AM, Anthony Gutierrez <[email protected]>wrote:

> I am looking at the mshr_hits, mshr_misses, etc. statistics in
> cache_impl.hh and I see that they are indexed by pkt->cmdToIndex() for each
> command, and then that vector is indexed by a hard coded 0. However, I see
> that there is a commented out index pkt->req->threadId(). I want to keep
> track of these stats on a per thread basis so I uncommented this out. It
> compiles and works fine. I was just wondering what was the reason for
> commenting this out? Does it not work the way it should or is there anything
> I should be aware of when doing this?
>
> -Tony
> _______________________________________________
> m5-users mailing list
> [email protected]
> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users

Reply via email to