Pete
for high precision sqr approximation using the Newton-Raphson method you can start with low precision and double the precision on each iteration
some pseudo code
you can apply the logic for n-root approximation
the first approximation will be more complicated if you want to go beyond the exponential range of double, you could try _Float for the first approximation but unless you use the CRT library it won't work because there's very limited function support for _Float
for high precision sqr approximation using the Newton-Raphson method you can start with low precision and double the precision on each iteration
some pseudo code
Code: (Select All)
l=log(NUM_DIGITS*0.0625)*1.5 'appromimation to l=log(NUM_DIGITS/16)/log(2) ''+ a little extra
'get the first approximation using double arithmetic
'then
limit&&=16
for k=1 to l+1
limit&&=2*limit&&
tmp = sm_div(n, r, limit&&)
r2 = sm_add(r, tmp, limit&&)
r = sm_mul(r2, half, limit&&)
next
return rthe first approximation will be more complicated if you want to go beyond the exponential range of double, you could try _Float for the first approximation but unless you use the CRT library it won't work because there's very limited function support for _Float