 |
 |
> Date: Thu, 24 Aug 2000 20:08:39 +0010 > From: [log in to unmask] > Harvey Richardson wrote: > >>> The Fortran 90 standard says > >>> > >>> "In a storage association context > >>> ... > >>> A nonpointer scalar object of type double precision real or > >>> default complex occupies two contiguous numeric storage units." > >>> > >>> so you could use COMMON or EQUIVALENCE to have a real arrays > >>> mapped to the complex array. Only if the COMPLEX variable is of default precision. > >>In which order are the real and imaginary parts of each complex number stored? > > > >I don't know but I'll answer the question "which of the numeric > >storage units contain the real and imaginary parts?"... > > > >For Fortran 77 the first is real and the second imaginary. > > > >When I wrote the above I did not realise that the Fortran 90 standard > >does not contain the text that states this in the Fortran 77 standard > >(4.6). At least I (and others) couldn't find it. > > I think one of the others who expressed a similar view to yours was Richard > Maine. > > If this is true about the "oversight" in the F90/F95(?) standard, then > potentially a lot of code could be broken by a new "compiler vendor kid on the > block" and maintaining compliance to the standard. > > Be they good practices or not, many engineering programs have been written with > equivalence to the real and imaginary parts of complex numbers. I can remember > instances from my youth, when good programming practices were those that used > every devious trick I could invent. > > If this is an oversight (and before BG decides to go into the new "standard" BG > Fortran compiler), is there a mechanism to correct this before potential harm? > > Regards, Paddy > > Paddy O'Brien, %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%