|Re: GEANE on 64bit Ubuntu 9.04 [message #8451 is a reply to message #8403]
||Wed, 06 May 2009 19:16
Registered: June 2007
the warning come from the routine TRPRFN, we do not use
TRPROP because we have the flag E (Exact tracking) in the
interface. ERPROP uses the same message both for TRPROP
and TRPFN. In this routine eqs. A1-A20 from Strandlie and
Wittek, NIM A 566(2006)687 are coded and calculated.
These formulae refer to the propagation of the error matrix from
one trsnverse (SC) system to another during tracking.
Some of these formulae become singular when cos lambda (dip angle) is equal to zero, because it appears on the denominator.
This happen when the particle is exactly
along the z-axis of the Master system, because in this case the
phi angle is undefined. Since the equations are meaningful also for very small cos lambda values, in GENAE only the conditions cos lambda=0 EXACTLY is flagged and no actions are triggered,
apart from the printed message.
In this case perhaps one could set to zero all the derivatives
involving phi in the jacobian and skip the message.
A lot of messages of this typre could happen when you send the
particles exactly along the z-axis. In all the other cases this
should happen very rarely! Even more strange is the dependence
of the error on the 32-64 bit configuration.
Therefore, to understand better what's going on, I would
1) do you send particles along the Z- axis in your generator?
2) what is the pervcentage of messages you have on the whole
in a typical production?
3) In a 32 bit configuration you have few messages. Are the
result reliable with no artifacts?
Thank you in advance