[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [ccp4bb]: REFMAC5 - error (72): floating overflow



***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***

> Limits of asymmetric unit      : 1.00 1.00 0.50
>  Grid spacing to be used        :   288  144  250
>  Maximuum H,K,L                 :    99   49   90
>  Minimum acceptable grid spacing:   214  107  193
> forrtl: error (72): floating overflow            <<<<<

We have seen this with other Alpha Fortran programs.  This error usually 
occurs when the precision of a real variable is insufficient to contain 
the value of a calculation.  Following a cursory look at the code in the 
refmac5_ directory, primarily in hkon_occup.f, it seems that all 
floating point types are REAL.  Sometimes, a cheap, sleazy way around 
insufficient-precision REALs on the Alpha is to recompile with the flag 
"-real_size 64".  Of course, this presumes that there isn't somehow 
something wrong with the input values (e.g., a value much, much closer 
to zero than it should be)...and completely dodges any underlying deeper 
issues with the code.

Alternatively, one would probably need to put lots of print statements 
into the source in refmac5_, starting with hkon_occup.f and 
solvent_contr_mask.f, to see if the location of the relevant calculation 
can be narrowed down.  Keep the herbs handy.

Regards,
Michael

-- 
Michael Strickler, Ph.D.
Research Specialist
Center for Structural Biology
Howard Hughes Medical Institute
Yale University