4646474 : BigInteger.pow() algorithm slow in 1.4.0

Brian Burkhalter brian.burkhalter at oracle.com
Wed May 15 01:11:28 UTC 2013


The test in this issue report

http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4646474

is pretty much useless for comparative purposes after applying 4837946 as the computed value is just a power of two which will be handled very quickly by the new algorithm.

As a single alternate data point I modified the test instead to compute 17^13466917. The execution time on my MacBookPro5,3 before applying the 4837946 patch was 2336.975514 seconds and afterwards 79.202646 seconds: an improvement of about a factor of 30. I am sure that there are much more extreme examples than this but the result is good to see.

I linked this issue to 4837946 as a duplicate so it will be closed out at the same time.

Brian



More information about the core-libs-dev mailing list