Re: use of CERNLIB with Fluka (source)

Dear Noemi,

i)
> I used the lflukac script in
> order to link Fluka with CERN Library but at the end of the procedure I
> obtained the following error message:
> "source_prot_min_new.f:130: undefined reference to `dgauss_'
> collect2: ld returned 1 exit status". Can you find out where is the error?

I guess that you forgot the option -C when you launched the lflukac script

(when the CERNPATH variable in lflukac is set wrong, you get a different
error message)

ii)
> 2) In source.f the sum
> (TXFLK(NPFLKA))**2+(TYFLK(NPFLKA))**2+(TXFLK(NPFLKA))**2 must be exactly
> equal 1 (double precision). In my source_prot_min_new.f I set
> TXFLK(NPFLKA)=costx and TYFLK(NPFLKA)=costy (where costx and costy were read
> from an external data file) and I obtained TZFLK(NPFLKA) from the algorithm
> (which was commented on the original source.f in usermvax) reported at line
> 228 TZFLK(NPFLKA) = SQRT(ONEONE - TXFLK(NPFLKA)**2- TYFLK(NPFLKA)**2 ) but
> sometime it happen that the argument of SQRT is negative , moreover by
> using this algorithm TZFLK(NPFLKA) is always positive while I need also
> negative values in order to simulate an isotropic particle flux.

if the argument of SQRT is negative, it means that the input costx and
costy values are wrong (the sum of two squared direction cosines cannot be
greater than 1). Very likely it is a problem of accuracy: the number of
digits that you provide is not enough, i.e. the approximation on them is
too rough. You might solve both the mentioned problems by writing

TXFLK (NPFLKA) = costx / SQRT(costx**2+costy**2+costz**2)
TYFLK (NPFLKA) = costy / SQRT(costx**2+costy**2+costz**2)
TZFLK (NPFLKA) = costz / SQRT(costx**2+costy**2+costz**2)

But you can implement an isotropic sampling directly in your source
routine, according to the Sebastien's suggestion or, e.g., sampling
uniformly between ZERZER and TWOPIP the azimuthal angle and between
-ONEONE and ONEONE the cosine of the polar angle.

Note that in your source routine all real numbers are in single precision
(D0 missing), you redefine - in single precision! - something already
available (your PI is supposed to be PIPIPI in (DBLPRC)), and you write

IF(costz.ne.0.)

(bad because 0. is not ZERZER or 0d0 - as pointed out before - and
because to check the equality of two real numbers is quite dangerous and
unrecommended) instead of

IF(ABS(costz).LE.AZRZRZ)

Ciao

Francesco

**************************************************
Francesco Cerutti
CERN-AB
CH-1211 Geneva 23
Switzerland
tel. ++41 22 7678962
fax ++41 22 7668854
Received on Thu Mar 27 18:45:07 2008