On the maximum relative error when computing integer powers by iterated multiplications in floating-point arithmetic (Q891788): Difference between revisions

The authors improve the relative error bound for the computation of \( x^{n} \) through iterated multiplication by \( x \) in binary floating point arithmetic under mild conditions. This bound is similar and slightly better than the previous bound. They conjecture that the same bound holds in the more general case of the computation of the product of \( n \) floating-point numbers when \( n \) is not too large. Finally, examples are provided to show that the actual error can be very close to, but smaller than \( (n-1) \).