4 ms·
Actually I think that roughly was my point. :-) You can’t get an exact answer in general by using fixed-size floating-point numbers, as your pseudocode does. T
by robinhouston 15y ago
Actually I think that roughly was my point. :-)
You can’t get an exact answer in general by using fixed-size floating-point numbers, as your pseudocode does. The number of bits of precision you need to get an exact answer is going to depend on the input value, presumably linearly.
- kenjackson 15y agoWhen you round you'll always get the exact integer answer, even using the precision of the golden ratio I gave in the code.
- robinhouston 15y agoThat’s really not true. You don’t have to take my word for it: try it with fib(1000), say. The answer should be: 43466557686937456435688527675040625802564660517371780402481729089536555417949051890403879840079255169295922593080322634775209689623239873322471161642996440906533187938298969649928516003704476137795166849228875
- kenjackson 15y agoOh, the old n=1000 trick. Well played.