[Rivet] RE : RE : installation problems (It works!)

[Judith Katzy]

Hi Andy,

just for your information: Atlas uses the .2 versions with common block
size 10000.

cheers,
Judith

On Tue, 29 Jul 2008, Andy Buckley wrote:

> Aaargh! That's the first time anyone has ever mentioned that! Is it
> documented anywhere? Wouldn't it be more normal to have the '.2'
> versions as the exception rather than the norm?
>
> Maybe it should also be recommended to the collaborations that they use
> this trick to dynamically work out the common block size rather than
> hard-coding it... although I'm not sure if this propagates to Fortran
> gens like TAUOLA, PHOTOS, etc. who might like to dig into the HEPEVT
> common block and can't do this dynamic size trick.
>
> I thought that the '.2' versions were Genser patches (hence our requests
> that they be hidden from users, who shouldn't be exposed to details of
> the build system patches). That makes more sense now: AGILe requests the
> 4000-size generators, expecting that name to be a symlink to the latest
> patch version. I'll update the AGILe loader to prefer the '.2' Pythia
> and Herwig libs, and these fixes will make us more robust against using
> the "wrong" version in future.
>
> Thanks for the help and clarification!
> Andy
>
>
> James Monk wrote:
>> So in that case AGILe should probably load the xx.2 version of the
>> generators
>>
>> On 29 Jul 2008, at 18:12, Mikhail Kirsanov wrote:
>>
>>> Hello,
>>> In GENSER, .2 in the version always means HEPEVT size 10000. So,
>>> pythia 418 has a size
>>> 4000, and 418.2 has a size 10000.
>>> For some time we thoght that all experiments moved to 10000, but then
>>> it turned out that
>>> CMS still uses the size 4000. For this reason, for some pythia
>>> versions there is no 4000 subversion.
>>>                                             Mikhail
>>>
>>>
>>> -----Original Message-----
>>> From: James Monk [mailto:jmonk at hep.ucl.ac.uk]
>>> Sent: Tue 7/29/2008 6:59 PM
>>> To: Andy Buckley
>>> Cc: Mikhail Kirsanov; Holger Schulz; genser-developers;
>>> rivet at projects.hepforge.org
>>> Subject: Re: RE : RE : [Rivet] installation problems (It works!)
>>>
>>> It seems like the max number of records in the event is actually set
>>> at 4000 (found in pyhepc.F as the parameter NMXHEP), which is what
>>> caused the problem.  The fix Andy submitted to AGILe works for me
>>> because NMXHEP gets copied into MSTU(8).
>>>
>>> I have a question tho'.  It looks to me that the AGILe number of
>>> entries has been 10,000 for quite some time (at least as far back as
>>> the original AGILe 1.0 release), yet this is the first time it
>>> breaks.  Did NMXHEP get decreased at some point in the Genser release
>>> from 10,000->4000?  Is 4000 a long enough record at the LHC (I thought
>>> the option of 10,000 was added specifically for the LHC!).
>>>
>>> In fact, I just checked ATLAS' Pythia interface and I see this:
>>>
>>> HepMC::HEPEVT_Wrapper::set_max_number_entries(10000);
>>>
>>> This is not so good, right?
>>>
>>> -James
>>>
>>> On 29 Jul 2008, at 17:01, Andy Buckley wrote:
>>>
>>>> Mikhail Kirsanov wrote:
>>>>> I would recommend to set it automatically, example is here:
>>>>> $MCGENERATORS/pythia6/418/share/examples/main78.cc
>>>>
>>>> Thanks Mikhail, this seems to work for me. I've committed it to SVN -
>>>> can you try, James?
>>>>
>>>>> There is the corresponding possibility also in Herwig installed
>>>>> in GENSER, if you want I can give you example.
>>>>
>>>> I'd be very grateful if you could point me at it... it will be good to
>>>> have this fragility eliminated at last!
>>>>
>>>> Andy
>>>>
>>>
>>>
>>
>
>