RE: Re: R: RE: Program received signal SIGFPE: Floating-point exception - erroneous arithmetic operation.

From: Vasilis Vlachoudis <Vasilis.Vlachoudis_at_cern.ch>
Date: Wed, 15 May 2013 16:43:02 +0000

Dear Francesca,

I could reproduce the error, which appeared to be a floating point precision problem. Now is in the hands of level 3 support (Alfredo Ferrari) see below his reply:

... it is precision mismatch apparently peculiar of gfortran.
It comes down from the enionf routine which sets the momentum (PLA)
to 0 and the energy (EKIN) as initial kinetic - (total - mass) which
apart rounding should be zero as well. The code has some tolerance on the
rounding but apparently in some cases with gfortran the rounding is larger
than expected, EKIN larger than 'epsilon" and PLA still zero...

I changed the logic in enionf so that to get *exactly* zero for EKIN as
well, I'll make a respin later

Best Regards
Vasilis

________________________________________
From: owner-fluka-discuss_at_mi.infn.it [owner-fluka-discuss_at_mi.infn.it] on behalf of francesca.belloni_at_libero.it [francesca.belloni_at_libero.it]
Sent: 15 May 2013 00:26
To: fasso_at_slac.stanford.edu; fluka-discuss_at_fluka.org
Subject: R: Re: R: RE: Program received signal SIGFPE: Floating-point exception - erroneous arithmetic operation.

Dear Alberto,
           thank you really a lot for your answer, because it corrects mistakes
that I usually do.
I modified the file source.f as you told, but the same mistake still shows up.
In attachment is the new source.f and the *.err, *.log and *.out files.

Thank you,
best regards,
Francesca
Received on Wed May 15 2013 - 23:03:46 CEST