From: Francesco Cerutti (Francesco.Cerutti@cern.ch)
Date: Wed Jan 24 2007 - 10:43:46 CET
Hi Martin,
1. the Comptn_2 messages indicate a bad cosine normalization
(abs(1-u**2-v**2-w**2)>10^-12) which is immediately corrected.
As for Geofar and GEOMETRY SEARCH ARRAY FULL, just quoting the manual
(Error messages):
2. "The following message indicates a real problem if repeated more than a
few times:
GEOFAR, TXYZ: 1. -0.0721463599 -0.409276348 -0.909553612
Nfrom, Nreg, X, Y, Z 1001 3 -0.108787724 -1.26878781 8.78769155
Geofar: Particle in region 3 (cell # 0) in position 1.000000000E+00
0.000000000E+00 1.000000000E+00
is now causing trouble, requesting a step of 6.258867675E-07 cm
to direction -2.285059979E-01 -9.412338141E-01 2.487245789E-01, error
count: 0
[...skipped...]
Particle index 3 total energy 5.189748600E-04 GeV Nsurf 0
We succeeded in saving the particle: current region is n. 2 (cell #
0)
As it can be seen, the program has some difficulty to track a particle in
a certain direction, and it tries to fix the problem by "nudging" the
particle by a small amount, in case the problem is due to a rounding
error near a boundary. If the message appears often, it is recommended to
run the geometry debugger centreing around the position reported in order
to find if there is an error in the geometry description".
3. "Another geometry error message is the following:
****************************************
GEOMETRY SEARCH ARRAY FULL
****************************************
This message indicates that insufficient memory has been allocated for
the "contiguity list" (list of zones contiguous to each zone). This is
not an actual error, but it is suggested that the user could optimise
computer time by increasing the values of the NAZ variable in the
geometry region specifications".
About that, a more extensive explanation by Alberto Fasso' follows:
"The message means the following.
To save time, for each region, FLUKA sets up a list of "neighbor
regions" to be tested first when a particle leaves any of the
bodies making up that region.
If the new region is not found in the list, the other regions
are searched. When found, the region is added to the list of
neighbors, provided there is enough space left in the array.
The space allocated is by default 5 neighbors per region, which is
normally more than sufficient. The value can be modified by the user
as explained in the manual (description of region data)
=> the integer in columns 6-10 is the number of regions which can be
entered by a particle leaving any of the bodies defined for the
region being described (leave blank in continuation cards). This
number is used to allocate memory, and it is not essential that it
be exact (if left blank, it is set to 5). Any number is accepted,
but if it is close to the actual value it may slightly improve
tracking speed.
Note that memory allocation is done globally, not region by region,
I suspect that you have set those numbers to a very small value
(1 or 2), because I always work with the default and I have never
got that message.
Anyway, in general, experience shows that CPU time is not
strongly affected by this setting, unless you have an enormous number
of regions through which FLUKA loops every time a particle exits a
region".
Cheers
Francesco
**************************************************
Francesco Cerutti
CERN-AB
CH-1211 Geneva 23
Switzerland
tel. ++41 22 7678962
fax ++41 22 7668854
This archive was generated by hypermail 2.1.6 : Wed Jan 24 2007 - 14:08:31 CET