Is this division algorithm fast?

If I have a division algorithm that resolves, on average, 3 base-10 decimal places every 5 tests (to resolve 3 digits requires anywhere from 1 to 10 tests), is that considered fast or slow, relative to other algorithms?

Is there a known method that resolves as fast as the Newton or Secant method?
Perhaps I'm misunderstanding you, but are you talking about the "keep subtracting till you hit zero" method? Doesn't that only work for ints? I'm doing operations on floats.

Dude are you trying to crack an RSA cipher???

I wrote a program about 100 years ago to try to do the same thing. If you start asking us for large prime numbers next, we're on to you! Ha..

This is great.

If you're trying to keep all the digits, try converting your numbers to binary and then using twos complement (XOR) to subtract. The bitwise operations are much faster than what the compiler will generate for you. Just remember to keep a counter. When the division yields a sign bit of 0, you're done. Depending on the accumulator logic, the other dividend is the count or count-1 depending on when you increment it.

Am I dating myself by talking Assembler? Probably, but what the heck. When I wrote something like this, it was for a Zilog Z80 processor (8 bit) running at 1 MHz so I really had no other option.

Good luck.

UPDATE: It will work with anything, you just need to employ the long division shift and keep track of the decimal point yourself. This is definitely only valid if you are willing to bypass the compiler and do the operations yourself. Do they even make assemblers any more? I honestly don't know, but if you can get to the actual CPU commands (mnemonics), your operations will go much faster.

Anyway, I guess that eliminates my assumption about RSA huh... I wish I could find my source, I'd send it to you.