| ??? 10/08/08 07:44 Read: times |
#158926 - Issues with Newton-approximation Responding to: ???'s previous message |
Still, the square-root approximation by using the Newton algorithm has a number of possible issues.
First of these is convergence, which is usually good, but not guaranteed (depending on the initial guess). Another is that is needs an actual division in each step, though I think I have seen variants of the algorithm that don't (they only need a division by a power of two, if I remember correctly). The algorithm I mentioned ( http://www.embedded.com/98/9802fe2.htm ) can calculate an integer (or fixed point) square root using neither divisions nor multiplications - just with additions, subtractions and shifts. |
| Topic | Author | Date |
| code library for 24 bit sqare root in asm | 01/01/70 00:00 | |
| So, what have you tried ? | 01/01/70 00:00 | |
| Is the " subject problem" homework? | 01/01/70 00:00 | |
| code library for 24 bit sqare root in asm | 01/01/70 00:00 | |
| Algorithm is important | 01/01/70 00:00 | |
| thinking | 01/01/70 00:00 | |
| Integer? | 01/01/70 00:00 | |
| Fixed-point or integer | 01/01/70 00:00 | |
Issues with Newton-approximation | 01/01/70 00:00 | |
| Excellent link | 01/01/70 00:00 | |
| fixed point or integer method of Per Westermark | 01/01/70 00:00 |