Reputation: 73
For example, A polynomial is defined as follows:
f := (x, y) -> 333.75y^6 + x^2(11x^2y^2 - y^6 - 12y^4 - 2) + 5.5y^8 + 1/2*x/y
In maple, I look to evaluate this to 5 significant figures like so:
evalf[5](f(77617,33096))
And obtain a value that is: 1*10^32.
Why is this not to 5 sig fig? Why is this not close to a value of 7.878 * 10^29 as you increase the number of sig fig required?
Thanks!
Upvotes: 1
Views: 203
Reputation: 7271
Don't reduce the working precision that low, especially if you are trying to compute an accurate answer (and then round it for convenience).
More importantly, for compound expressions the floating-point working precision (Digits
, or the index of an evalf
call) is just that: a specification of working precision and not an accuracy request.
By lowering the working precision so much you are seeing greater roundoff error in the floating-point computation.
restart;
f := (x, y) -> 333.75*y^6
+ x^2*(11*x^2*y^2 - y^6 - 12*y^4 - 2)
+ 5.5*y^8 + 1/2*x/y:
for d from 5 to 15 do
evalf[5](evalf[d](f(77617,33096)));
end do;
32
1 10
31
-3 10
30
1 10
29
8 10
29
7.9 10
29
7.88 10
29
7.878 10
29
7.8784 10
29
7.8785 10
29
7.8785 10
29
7.8785 10
Upvotes: 2