I'd second that point.
And the optimal way to do it would probaby just find the stats that
you want and simout (?) which stats you want.
The way I showed is quite unsophisticated in that you print out ALL
stats every so often.
I went through and changed a bunch of stats (esp. in the cache) to not
show up if they were a zero value (you can set a flag to do that -
check M5 docs).
If I were reallllly smart, I would maybe create some kind of function
called "printMyStats" that only printed the stats I cared about every
so often...
Once you get to the problem of too large of stat files you probably
want to consider Ali's option...
On 5/8/08, Ali Saidi <[EMAIL PROTECTED]> wrote:
> I again want to point out that if you're interested in a couple of
> statistics every cycle, this is not the optimal way to do it. Not only will
> you end up with an extremely large text file, you'll also slow down the
> simulation maybe 10x or more to dump stats in this manner. What you probably
> want to do is just open up a file (with simout.create()) and write the data
> that interests you every cycle.
>
> Ali
>
>
> On May 8, 2008, at 11:37 AM, jerry lrui wrote:
>
> > Hi, Korey
> >
> > Thank you so much for your help. I really appreciate it.
> >
> > Jerry
> >
> > On 5/8/08, Korey Sewell <[EMAIL PROTECTED]> wrote: OK,
> > so it's a few simple things you can do to dump stats at a certain
> > interval. I adopted this from the checkpoint (?) dumping code and
> > implemented it in the SimpleCPU.
> >
> > (1) src/simple/base.cc: In the constructor add these lines:
> > Tick when = ticks(p->start_stats);
> > Tick repeat = ticks(p->sample_period);
> >
> > printf("Stats starting @ %i w/period of %i.\n", (int) when, (int)
> repeat);
> >
> > Stats::StatEvent(true, true, when, repeat);
> >
> > You need to substitute the right values for the when/repeat variables.
> >
> > *NOTE: DEPENDING ON YOUR CONFIG 1 cycle might equal 500 ticks. There
> > is a email about a few weeks back that describes how to make a 1:1
> > correlation between cycles and ticks.
> >
> > Make sure these header files available:
> > #include "sim/stats.hh"
> > #include "sim/system.hh"
> > #include "sim/stat_control.hh"
> >
> >
> > (2) I print out headers so that I know that there stats for each
> > sample. In src/base/stats/text.cc add lines to the Text::output()
> > function to mark what stat sample you are on. My function looks like
> > this:
> > Text::output()
> > {
> > using namespace Database;
> >
> > static int sample = 0;
> >
> > ccprintf(*stream, "\n---------- Begin Simulation Statistics :
> > Sample %i ----------\n", sample);
> > stat_list_t::const_iterator i, end = stats().end();
> > for (i = stats().begin(); i != end; ++i)
> > (*i)->visit(*this);
> > ccprintf(*stream, "\n---------- End Simulation Statistics : Sample
> > %i ----------\n", sample++);
> > stream->flush();
> > }
> >
> >
> >
> > Anyway, I hope that helps what you are doing.
> >
> > On Tue, May 6, 2008 at 12:53 AM, jerry lrui <[EMAIL PROTECTED]> wrote:
> > > Thanks. you are right. it's impractical to dump all these stats every
> cycle.
> > > According to your suggestion, I'll further look through the codes to
> find
> > > correct way to handle these stats i needed. Thanks again.
> > >
> > >
> > > Jerry
> > >
> > > On 5/6/08, Korey Sewell <[EMAIL PROTECTED]> wrote:
> > >>
> > >> I agree with Lisa that you dont want 50 million files...
> > >>
> > >> I'll try to churn out a patch for you tomorrow... The patch I will
> > >> generate keeps dumping to the same file so what's going to happen if
> > >> you go per-cycle is you have a REALLY BIG stat file.
> > >>
> > >> I used the patch to dump I think a max of like10,000 samples and that
> > >> was pretty ridiculous given the # of cache stats that are turned out.
> > >> I had to go through the cache and change the majority of stats to not
> > >> even show if they were a value of ZERO in order to help condense the
> > >> file size.
> > >>
> > >> Because I did so much hacking, it makes a tougher patch (this was
> > >> pre-mercurial queue days for me)... sorry about that!
> > >>
> > >> Actually, it might be easy to just copy/paste code to you.... i think
> > >> it's just a few lines...
> > >>
> > >> On Mon, May 5, 2008 at 8:47 PM, Lisa Hsu <[EMAIL PROTECTED]> wrote:
> > >> > i don't think you want that. dumping stats, as geoff and i have
> > >> > mentioned,
> > >> > will dump EVERY stat. if your simulation is even 50,000 cycles,
> you'll
> > >> > have
> > >> > 50,000 files to sift through, most of which it sounds like you don't
> > >> > care
> > >> > about. and i'm sure your sims are way longer than that.
> > >> >
> > >> > if i were you, i'd look through the tree to find out exactly how a
> Stat
> > >> > type
> > >> > gets evaluated and then sent to the m5 stats file, and then do that
> > >> > yourself
> > >> > on every tick to send to your own personal output file. if i recall
> > >> > correctly, a Stat type isn't evaluated until it is dumped, so you
> have
> > >> > to
> > >> > find out how to get it to evaluate for your own purposes, set up your
> > >> > own
> > >> > output file, and have it dump into there. i don't remember where it
> is
> > >> > or
> > >> > anything, but i imagine if you follow the chain of calls from
> dumpstats
> > >> > you'll find your answer.
> > >> >
> > >> > lisa
> > >> >
> > >> >
> > >> >
> > >> > On Mon, May 5, 2008 at 9:09 PM, jerry lrui <[EMAIL PROTECTED]>
> wrote:
> > >> > >
> > >> > > Thanks! I'm using BaseCPU, too.
> > >> > >
> > >> > >
> > >> > >
> > >> > >
> > >> > > On 5/6/08, Korey Sewell <[EMAIL PROTECTED]> wrote:
> > >> > > > I have something in my tree that lets me dump stats every X
> cycles.
> > >> > > > That's probably what you are looking for. It's similar to dumping
> > >> > > > checkpoints at a certain interval, but instead I dump the
> stats...
> > >> > > >
> > >> > > > I think I implemented it in BaseCPU... what CPU are you using
> this
> > >> > > > for?
> > >> > > >
> > >> > > > I'll look into what I got later at the next opportunity...
> > >> > > >
> > >> > > > On Mon, May 5, 2008 at 9:09 AM, jerry lrui <[EMAIL PROTECTED]>
> > >> > > > wrote:
> > >> > > > > Is there a better way to get these stats after very cycle, if I
> > >> > > > > can
> > >> > not dump
> > >> > > > > all these stats? I only need few stats such as cache hits to
> > >> > > > > analyze
> > >> > the
> > >> > > > > runtime characteristic. Thanks.
> > >> > > > >
> > >> > > > > Jerry
> > >> > > > >
> > >> > > > >
> > >> > > > >
> > >> > > > > On 5/5/08, Geoffrey Blake <[EMAIL PROTECTED]> wrote:
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > > If you just need aggregate stats at the end of the simulation
> > >> > > > > > run,
> > >> > such as
> > >> > > > > number of dcache.ReadReq_hits, M5 already does this. When the
> > >> > simulation
> > >> > > > > finishes, it dumps the file: m5stats.txt to your current
> working
> > >> > directory
> > >> > > > > with everything added up for you.
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > > Dumping stats every cycle can easily lead to multi gigabyte
> > >> > > > > > files
> > >> > that
> > >> > > > > will be very hard to work with and store.
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > > Geoff
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > > From: [EMAIL PROTECTED]
> > >> > > > > > [mailto:[EMAIL PROTECTED]
> > >> > On
> > >> > > > > Behalf Of jerry lrui
> > >> > > > > > Sent: Monday, May 05, 2008 11:17 AM
> > >> > > > > > To: M5 users mailing list
> > >> > > > > > Subject: Re: [m5-users] How to dump m5stats per cycle?
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > > Hi, Lisa
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > > I only need few stats to analyze the runtime characteristics
> of
> > >> > > > > > an
> > >> > > > > application. For example the number of dcache.ReadReq_hits,
> > >> > > > > icache.ReadReq_hits, etc. I added these codes
> > >> > > > > "Stats::StatEvent(true,false);" in function simulate(Tick
> > >> > > > > num_cycles)
> > >> > > > > located simulate.cc. But I didn't get the result I wanted.
> Could
> > >> > > > > you
> > >> > tell me
> > >> > > > > what is the correct way to get thest stats? Thank you very
> much!
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > > Jerry
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > > On 5/5/08, Lisa Hsu <[EMAIL PROTECTED]> wrote:
> > >> > > > > >
> > >> > > > > > doing this will result in millions of large files full of
> stat
> > >> > > > > > text
> > >> > - what
> > >> > > > > is it that you want this for?
> > >> > > > > >
> > >> > > > > > lisa
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > > On Sat, May 3, 2008 at 11:39 PM, jerry lrui
> > >> > > > > > <[EMAIL PROTECTED]>
> > >> > wrote:
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > > Hi,
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > > I want to dump m5stats after per cycle. I've looked up the
> code.
> > >> > Maybe I
> > >> > > > > can add some codes in function simulate(Tick num_cycles)
> located
> > >> > > > > simulate.cc. Fox example, add a StatEvent like
> > >> > > > > "Stats::StatEvent(true,false);" in this function? But I'm not
> sure
> > >> > > > > how
> > >> > to do
> > >> > > > > it correctly. I'm using m5sim 2.0b5 in SE model. What is the
> > >> > > > > correct
> > >> > way to
> > >> > > > > get this information? Thanks!
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > > Jerry
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> _______________________________________________
> > >> > > > > > 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
> > >> > > > > >
> > >> > > > > >
> > >> > > > > >
> > >> > > > > > No virus found in this incoming message.
> > >> > > > > > Checked by AVG.
> > >> > > > > > Version: 7.5.524 / Virus Database: 269.23.8/1413 - Release
> Date:
> > >> > 5/3/2008
> > >> > > > > 11:22 AM
> > >> > > > > >
> > >> > > > > >
> > >> > > > > > No virus found in this outgoing message.
> > >> > > > > > Checked by AVG.
> > >> > > > > > Version: 7.5.524 / Virus Database: 269.23.8/1413 - Release
> Date:
> > >> > 5/3/2008
> > >> > > > > 11:22 AM
> > >> > > > > >
> > >> > > > > >
> _______________________________________________
> > >> > > > > > 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
> > >> > > > >
> > >> > > >
> > >> > > >
> > >> > > >
> > >> > > > --
> > >> > > > ----------
> > >> > > > 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
> > >> > > >
> > >> > >
> > >> > >
> > >> > > _______________________________________________
> > >> > > 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
> > >> >
> > >>
> > >>
> > >>
> > >> --
> > >> ----------
> > >> 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
> > >
> > >
> > > _______________________________________________
> > > m5-users mailing list
> > > [email protected]
> > > http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
> > >
> >
> >
> >
> > --
> > ----------
> > 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
> >
> > _______________________________________________
> > 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
>
--
----------
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
