On Mon, 28 Jul 2025 17:03:48 GMT, Raffaello Giulietti <rgiulie...@openjdk.org> wrote:
>> fabioromano1 has updated the pull request incrementally with one additional >> commit since the last revision: >> >> Zimmermann suggestion > > src/java.base/share/classes/java/math/MutableBigInteger.java line 1993: > >> 1991: * non-significant bits), and the shift is performed in >> order to lose >> 1992: * the smallest number of bits in the significand if >> necessary (minimizing loss of precision). >> 1993: */ > > @fabioromano1 I've now spent several hours trying to "reverse engineer" (Java > code -> maths) what is going on in the code below that tries to come up with > a nice initial estimate of the end result, before starting Newton's loop. > I'm still not 100% sure I've got everything correctly, and whether each > detail is mathematically sound. > The same will probably happen to other readers and maintainers in the future. > > I understand the overall tactic described in the comment above. However, the > details of the different (sub)cases deserve (and need) a more detailed > documentation to help and guide future enhancements and fixes. > The best way, of course, is to use maths arguments whenever possible. @rgiulietti It would be helpful for me to know the points of the code that are still not completly clear, since there are already some comments in the code, so I can know where to add others. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/24898#discussion_r2237324405
