Don Simons wrote:
>
> Dirk Laurie <[EMAIL PROTECTED]> at INTERNET-MAIL wrote
>
> > [Christof] <[EMAIL PROTECTED]> wrote:
> >>
> >> as known, pmx can handle only a limited number of voices.
>
> >Of all PMX restrictions, this is the easiest one to lift. You simply
> >replace all occurrences of nm=7 in the Fortran
> >source by nm=15 or whatever. For some values there may be common
> >block alignment problems on some computers, but stepping up by some
> >multiple of 8 seems to work most of the time.
>
> I thought the common block alignment warnings were due to the order of
> variable names in common lists, not the dimensions of the variables.
> In any event, it would be useful if poeple who have attempted to
> compile my sources, regardless of the value of nm, would send me
> copies of the warnings, so maybe I can rearrange some of the blocks
> and eliminate the warnings.
>
You are right, but certain dimensions can make things work out lucky
even when the order is not yet perfect.
What happens is this: the blocks get rearranged to get rid of the
warnings, but this does not yet imply that the ideal of descending
length (which would be immune to changes in nm) has been attained.
There may be places left where something like
REAL, CHAR, CHAR(NM), REAL
occurs. This does not give an error message for nm=7 or nm=15,
but may do so for nm=11 when REAL has length 8, and usually does
for other nm between 7 and 15.
Ideally those of us who have finicky Fortran compilers should
try compiling at two consecutive values of nm so we can send you
a longer list of alignment errors :-/
Dirk
