Hi Andi,
You are not correctly using the timer. You only print the cumulative
CPU time. For real examples of TStopwatch, see
$ROOTSYS/test/bench.cxx or $ROOTSYS/test/MainEvent.cxx.
Rene Brun
On
Tue, 22 Jul 2003, andi klein wrote:
> Hi Andrei
> I might be doing something wrong again, but I noticed that with
> increasing number of events, the cpu time / event goes up.
> What you see following is the start of the eventloop. I am using 100000
> events.
>
>
> Int_t ihelp=-1;
> clock_total->Start(kTRUE);
> for(Int_t icount=0;icount<n_track+1;icount++)
> {
> clock->Start(kTRUE); // start stop watch and reset it every time
> ihelp++;
> if(ihelp % 10000 ==0){
> cout<<" have calculated "<<ihelp<<" events"<<endl;
> cout<<"Only "<<n_track-ihelp<< " left to go"<<endl;
> clock->Stop();
> cout<<" for 10K events used : "<<clock->CpuTime()<<" sec
> cputime"<<endl<<endl;
> clock->ResetCpuTime(cptime);
>
> }
>
> The output from my program looks as follows, where I print out the
> cputime every 10K events. At the end I print out the time from
> a stopwatch which never gets reset.
> andi
>
> ********************output*****************
>
> Top volume is exp. Master volume is exp
> det added to browser
> List of materials :
> Fixing runtime shapes...
> Counting nodes...
> Voxelizing...
> Building caches for nodes and matrices...
> ### 19 nodes/ 19 volume UID's in Sputter experiment
> ----------------modeler ready----------------
> --- number of nodes on screen : 18
> Total size of x3d primitives:
> gSize3D.numPoints= 1342
> gSize3D.numSegs = 2672
> gSize3D.numPolys = 1336
> --- number of nodes on screen : 18
> have calculated 0 events
> Only 100000 left to go
> for 10K events used : 2.44217e-17 sec cputime
>
> have calculated 10000 events
> Only 90000 left to go
> for 10K events used : 5.84 sec cputime
>
> have calculated 20000 events
> Only 80000 left to go
> for 10K events used : 16.17 sec cputime
>
> have calculated 30000 events
> Only 70000 left to go
> for 10K events used : 25.66 sec cputime
>
> have calculated 40000 events
> Only 60000 left to go
> for 10K events used : 34.85 sec cputime
>
> have calculated 50000 events
> Only 50000 left to go
> for 10K events used : 44.41 sec cputime
>
> have calculated 60000 events
> Only 40000 left to go
> for 10K events used : 53.8 sec cputime
>
> have calculated 70000 events
> Only 30000 left to go
> for 10K events used : 63.01 sec cputime
>
> have calculated 80000 events
> Only 20000 left to go
> for 10K events used : 72.37 sec cputime
>
> have calculated 90000 events
> Only 10000 left to go
> for 10K events used : 83.58 sec cputime
>
> have calculated 100000 events
> Only 0 left to go
> for 10K events used : 92.18 sec cputime
>
> total cputime used 491.93
> Average time per event 0.0049193 sec/event
> ****************************************************
>
> number of sucessful foil tracks 80798
> number of sucessful iris tracks 56113
> number of sucessful coll tracks 510
> number of sucessful tracks 460
>
>
>
This archive was generated by hypermail 2b29 : Thu Jan 01 2004 - 17:50:13 MET