28 Jul
2005
28 Jul
'05
3:43 p.m.
FORTRAN is simply not structured to easily allow
multiple operations on succeeding values...
The Pi program I mentioned was in Fortran, probably
even in F66. This didn't stop me. Easy /vs/ hard in this
case just meant slower code, no great pains in
programming at all. If I were to make a library, I'd just put
the exponent in the array with the Integer part(s).
Since I do NOT find that to be a problem, I am not
very concerned.
Great
But in case there was already a library around, I
thought I would ask.
Looks like you have two Solutions, Gnu-MP and hand
coding to your needs. Both seem like the same amount
of work, one to build up, the other to simplify down.
{add and subtract too?}
My basic approach was to store numbers in <= half the bits
of the whole integer (like 8b out of I16), perform the operation,
whatever it was piecewise across the array, then "normalize",
cleaning up, by sweeping across and carrying the excess bits
to the next array element(s). Two pass and slowish, but getting
"Integer" perfect results on real sized numbers was the thrill of it.
Rolling in a normalizer into the four operations individually would
increase code size 2 ~ 3X but may be what you want.
John A.