8014320 ready for review

[Dmitry Nadezhin]

I'm curious to compare performance on huge numbers
between this pure Java implementation of SS and some native implementation
of SS (libgmp, for example).



On Thu, Feb 6, 2014 at 9:22 PM, Tim Buktu <tbuktu at hotmail.com> wrote:

> Hi,
>
> 8014320 is a JDK9 enhancement and everybody is probably busy with JDK8,
> but I wanted to let the BigInteger people know the patch is finished. It
> consists of 3 patched files and 1 new file; they can be found at
>
>
> https://github.com/tbuktu/bigint/raw/5142a961252b2cd217377f32e4351640086c94e2/src/main/java/java/math/BigInteger.java
>
> https://github.com/tbuktu/bigint/raw/fe807bf92b58cf8a4556b3f8473f1b6dd35412b7/src/main/java/java/math/MutableModFn.java
>
> https://github.com/tbuktu/bigint/raw/3bd7329d93540a839582afb62c6ce40e243b3e47/src/main/java/java/math/MutableBigInteger.java
>
> https://github.com/tbuktu/bigint/raw/557584cba766507fe18d0cca5e8f03f45b483223/src/test/java/BigIntegerTest.java
>
> The SS and Barrett thresholds are from timings on an Intel CPU. An AMD
> CPU gave me lower values; I didn't try any other architectures. I used
> the program at the link below to find the thresholds:
>
> https://github.com/tbuktu/bigint/blob/master/src/main/java/Tune.java
>
> Tim
>