FORTRAN floating point accuracy problems

[HP67]

Thanks again, Dieter. Modern Fortran has many more functions including ceil, floor, and nearest integer but I'm using FORTRAN IV which has only truncation. I want to look over the hardware doc before I decide what to do. I don't understand the floating point instructions sine I never used them. I do remember there is some kind of hardware rounding function. It might be better to write whatever primitives I need in assembler and call them from FORTRAN. I'll post again after I have a better idea what hardware support there is.

What I should have mentioned: Modern Fortran has IEEE support and depending on the compiler that should include all IEEE rounding modes, etc. But, a long time ago (F90, possibly F77) Fortran added a feature called internal reads where you can "read" and "write" into variables without any I/O. Since FORTRAN always rounds on input/output this was a way to get a standard rounding mode for whatever precision you wanted. However it's not available as far back as I am working, so...