Re: A82: OP1 stuff, I guess


[Prev][Next][Index][Thread]

Re: A82: OP1 stuff, I guess




>Extended digits: These digits are only used in intermittent results are
part
>of some complex calculations.  However, these digits are not stored in
>variables so they really don't make much of a difference...

The extra digits are used in most calculations, and they do make a
difference. The use of extra digits makes the calculations more precise,
by avoiding some of the errors you would get if you rounded of the numbers
all the time. The rounding error which occurs can be really anoiying, if
you have a look at some programs which uses FP numbers you will see that
this causes a lot of problems.

>Mantissa: Binary coded decimal - the hex digits represent decimal digits.
>In the op data above, the mantissa is 1.23456789000.  The decimal point
isthan -99 with this system, and the system will work
>correctly with these numbers.  However, most calculations will error (IE

If the routine you are using starts out by adjusting the exponent loading
the exponent with an invalid number wont work. The routines i have seen
whicoint numbers on the 82 (but not other calcs), this
>number is either $00 or $80.  $00 = positive mantissa, $80 = negative
>mantissa.  (Special notaph has to be updated. I have not looked in to
this, but my guess is that they are used to control wether GRAPH_MEM
contain valid data. The three bits correspond to the 3 graphmodes
function,parametric and polar. If the bit is set one of the functions used
to display graphs depends on this value. This means that by monetoring
these bit the system can see if it can use the GRAPH_MEM for drawing the
graphs.
>
>For real vars stored in memory, the format is the same, except the extra
>digit bytes are not there.


There is one exception to this, the values on the FP stacks (there seem to
be two FP stacks) might not have the type byte set to the correct value.

There are some functions which uses the OPs for temp storage, and they
might therefore destroy values which you need in your calculations. 

Dines


_______________________________________

Dines Justesen
Email: dines@post1.com or
       c958362@student.dtu.dk
WWW  : http://www.gbar.dtu.dk/~c958362/
_______________________________________