Hi Edward,
Maybe the network is an issue here. I'm also chaining a lot of files
located on my afs account, and afs seems to cache them the first time they
are chained. The second time, it's much faster, because the files are
local.
Also, you are using castor. Maybe some of your files are not staged when
you stage them, so castor first has to retreive them from tape.
cheers,
Colin
On Fri, 12 Sep 2003, Edward Moyse wrote:
> The files are about 4 Mb each.
> Per file it seems (subjectively) to be going quicker now ... about 2-4s per
> file. This is just chaining them though ... so it's going to take 2 hours
> before it even starts running.
>
> I have a (crude) loop like below:
>
> for (Int_t i =0 ; i<numberOfFiles ; ++i){
> TString num="";
> num+=i;
> TString numPrefix="";
> for (Int_t j = 0 ; j < (sizeOfNum-num.Length() ) ; ++j) numPrefix+="0";
> chain->Add(pre+numPrefix+num+post);
> }
>
> and
> pre="rfio:/castor/cern.ch/atlas/transfer/emoyse/dc1.002000.lumi02.recon.010._";
> post =".hlt.pythia_jet_17.root";
> sizeOfNum=5;
>
> Even if this loop is sub-optimal (which I can quite believe) I don't see that
> it can make that much difference.
>
> Oh, and I'm using the latest version 3.05.07.
>
> Ed
>
> On Friday 12 September 2003 3:23 pm, Rene Brun wrote:
> > Ed,
> >
> > How big are your files? 1 Mbyte, 100 MBytes, 1 GByte ?
> > How long does it take to process one single file?
> > TChain should not add any overhead.
> > Which version are you using?
> >
> > Rene Brun
>
>
>
This archive was generated by hypermail 2b29 : Thu Jan 01 2004 - 17:50:15 MET