On Thursday, October 31, 2013 8:01:53 PM UTC-7, Ben Bacarisse wrote: > email@example.com writes: > > <snip> > > > Actually Ben i have a similar that may be easier for you to follow, > > > any square can be divided into 4 sub squares. And if we have a number > > > we can find the 10^x above it and 10^x-1 below it. > > > > > > So 10^x is 1 now we can chose if we want 0 at real zero or zero at > > > square 10^x-1 If we choose the later we close in faster. Now the area > > > between the lesser and bigger square or if we use zero, can be > > > described as a percentage ratio of the height. > > > > Sorry, I can't make head nor tail of this. > > > > <snip> > > > The perfect square we find is subtracted from our number and now we > > > work same approach for this smaller square. This is repeated until the > > > full number is encoded to a series of squares + a small integer less > > > then 4. > > > > Yes, this bit I've understood, but why? What's the point of doing this? > >
I'd like a SQRT(X) algorithm that goes to 10,000 digits.
useful for 1 time cypher pads in this age of eavesdropping!
Binary is fine too!
I understand Ben's point... if you are using repeated approximations then Newtons method is very fast.