I'm thinking you should nuke it. What I do for things like that is
just make a patch with mq. That way you always have access to it, but
it doesn't otherwise clutter things up for you or anybody else.
Gabe
Quoting Korey Sewell <[email protected]>:
Good points, Sorry about that. As the comment says, that's just a stat that
I was using a while back for debugging but cluttered things up when I was
looking at the "real" processor stats.
maybe a "#if 0" will do or I can just get rid of them altogether.
I'm leaning toward the former though as they are useful for "deep" debugging
of the model.
Does anybody care if I just "#if 0" that?
If not,
I'll knock that out and commit soon.
On Fri, Jun 25, 2010 at 5:14 PM, Steve Reinhardt <[email protected]> wrote:
Hi Korey,
You really threw me for a loop with this one... I was testing out a
simple change, which I often do using m5.debug since it compiles
faster, and the inorder tests kept failing. I could see that it was
due to some changes you made recently, but I had pulled both the code
changes and the stats updates, and the regressions on zizzer were
passing, so I was stumped. Then I looked more closely at your
changes... the bottom line is that you can't have stats that depend on
how the binary is compiled, as that messes up our regression framework
which expects all versions of m5 to get the same output for the test
runs.
If you want, you can use some other flag to ifdef them, or maybe just
get rid of them entirely knowing that you can resurrect them out of
the hg archive if you need to.
Thanks,
Steve
On Wed, Jun 23, 2010 at 3:26 PM, Korey Sewell <[email protected]> wrote:
> changeset 128cce9f92bc in /z/repo/m5
> details: http://repo.m5sim.org/m5?cmd=changeset;node=128cce9f92bc
> description:
> inorder-stats: add instruction type stats
> also, remove inst-req stats as default.good for debugging
> but in terms of pure processor stats they aren't useful
>
[...]
> diff -r e3311623d9f0 -r 128cce9f92bc src/cpu/inorder/resource.cc
> --- a/src/cpu/inorder/resource.cc Wed Jun 23 18:15:23 2010 -0400
> +++ b/src/cpu/inorder/resource.cc Wed Jun 23 18:18:20 2010 -0400
> @@ -79,11 +79,13 @@
> void
> Resource::regStats()
> {
> +#ifdef DEBUG
> instReqsProcessed
> .name(name() + ".instReqsProcessed")
> .desc("Number of Instructions Requests that completed in "
> "this resource.")
> .prereq(instReqsProcessed);
> +#endif
> }
>
> int
> @@ -474,8 +476,10 @@
>
> // change slot # to -1, since we check slotNum to see if request is
still valid
> slotNum = -1;
> -
> +
> +#ifdef DEBUG
> res->instReqsProcessed++;
> +#endif
> }
>
> ResourceEvent::ResourceEvent()
> diff -r e3311623d9f0 -r 128cce9f92bc src/cpu/inorder/resource.hh
> --- a/src/cpu/inorder/resource.hh Wed Jun 23 18:15:23 2010 -0400
> +++ b/src/cpu/inorder/resource.hh Wed Jun 23 18:18:20 2010 -0400
> @@ -253,8 +253,10 @@
> // DEFAULT RESOURCE STATISTICS
> //
> /////////////////////////////////////////////////////////////////
> +#ifdef DEBUG
> /** Number of Instruction Requests the Resource Processes */
> Stats::Scalar instReqsProcessed;
> +#endif
> };
>
> class ResourceEvent : public Event
_______________________________________________
m5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/m5-dev
--
- Korey
_______________________________________________
m5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/m5-dev
