Floating point math in FORTRAN IV on PDP-8

Kyle Owen kylevowen at gmail.com
Sun Sep 23 20:59:12 CDT 2018


On Sun, Sep 23, 2018 at 8:28 PM Rick Murphy via cctalk <
cctalk at classiccmp.org> wrote:

>
> That seems like you have a mixed-up F4 compiler/runtime.
> On my system (SIMH) I get the same results, FPP enabled or disabled.
> Don't know how to enable/disable EAE on SIMH, unfortunately.
>
> What you're getting means that the compiler, runtime, and libraries are
> not from the same release.
>      -Rick
>

Hmm...I was using the exact same RK05 image in both SIMH and the real
machine.

I've been informed that "set cpu noeae" will disable the EAE.

Just tried it with that...and there's the problem. No FPP or EAE seems like
a bad combination for running FORTRAN IV code that does any bit of math
operations.

>From SIMH (with EAE and FPP disabled):

.R F4
*FLOAT/G$
    1.000002    1.000002    0.000000
    2.000002    1.000031    0.693147
    3.000002    1.224747    1.098614
    4.000002    2.000002    1.386296
    5.000002    5.798176    1.609439
    6.000002    4.015202    1.791761
    7.000002    3.522211    1.837092
    8.000001    2.000061   -1.920560
    9.000002    2.121345   -1.802777
   10.000008    2.236078   -1.697417
   11.000007    2.345213   -2.130667
   12.000012    2.449492   -1.515095
   13.000011    2.549511   -1.141335
   14.000010    3.331171   -1.469765
   15.000009    3.338015   -1.037061
   16.000008    4.000002    2.772591
   17.000007    4.123107    2.833214
   18.000012    4.242642    2.890373
   19.000011    4.358900    2.967808
   20.000010   11.596357    2.995733
   21.000009   10.150188    3.122650
   22.000008    9.188697    2.562484
   23.000013    8.516458    2.604795
   24.000012    8.030402    3.178056
   25.000011    7.671024    3.206363
   26.000010    7.401572    3.551816
   27.000010    7.198119    3.612231
   28.000015    7.044420    3.223386
   29.000014    6.929063    1.967658
   30.000013    6.843800    3.656090
   31.000012    6.782515    3.713244
   32.000011    4.000121    3.465738
[snip]

Maybe someday I'll delve into why it doesn't play nicely with such math
without EAE and FPP.

So, in summary, either EAE or FPP (or both) works fine. Without at least
one, don't count on any floating point math to be right (based on my
limited assessment).

Surely this must've been known about 40+ years ago...right?

Kyle


More information about the cctalk mailing list