Final RFR 4837946: Faster multiplication and exponentiation of large integers

Brian Burkhalter brian.burkhalter at
Wed Jun 19 17:56:07 UTC 2013


On Jun 19, 2013, at 5:27 AM, Alan Eliasen wrote:

> Thanks for pushing this improvement.  As I've said before, I'm more
> than willing to put my personal e-mail address on this patch to support
> the Karatsuba and Toom-Cook and pow() and radix conversion fixes.  I'll
> put my reputation on the line.  You're free to let people e-mail me
> personally if their multiplications are incorrect.  I want to support
> this.  It makes Java, and pure math, implemented in Java, better.
>  If there's a problem, yo, I'll solve it.  (Sorry, I channeled Vanilla
> Ice there.)
>   After all, I've been publishing many of these improvements for 11+
> years, and I'm the person whose e-mail address comes up when someone
> searches about these issues.  To be frank, I'm personally involved much
> more than Sun/Oracle employees whose e-mail addresses are hidden.  I
> field a *lot* of questions about making BigInteger and BigDecimal
> faster.  I've helped a lot of people for many years integrate these
> changes into their JVMs, which isn't easy.  I try hard.

Your efforts are appreciated, for sure.

> I have many examples of fundamental problems in BigInteger being broken for 3+ years

If there are still extant problems which have no issue on file, then a bug report should be submitted for each.

> while I've fixed them in my own language within the day.  I am willing
> to support this and fix it fast.
>   I think my optimizations are simple and obvious;  after all, I posted
> a list of links to undergraduate classes who are expected to implement
> Karatsuba or Toom-Cook multiplication in a day or a week.  I don't want
> to present myself as a genius.  I'm no better than a typical
> undergraduate.

I concur that the optimizations I've reviewed so far seem to be fairly straightforward.

As for bugs for problems, enhancement requests for any new features or performance improvements in the numerics area should have corresponding issues on file. This will be of particular importance in planning for JDK 9.



More information about the core-libs-dev mailing list