🚀 go-pugleaf

MS-DOS FORTRAN 3.2 and above, Lahey F77, and probably IBM
FORTRAN 2.0, and IBM Professional FORTRAN handle larger
models for source and data

i understand that by being clever one can use about 128K with
hardly any degrading of performance

in regards to performance, some data derived from many
runs of a discrete simulation package suggests the 
following rules of thumb:

AT performance is 3x of PC
math co-processor increases performance by a factor of 5
using memory larger than 64k for source or data decreases performance
   by half


will be glad to hear from anyone else whose opinions, rules of
thumb, or experience is different