Ah, the good 'ole switch statement earning more optimization jiras :). sent from my phone On Nov 23, 2015 7:36 PM, "John Rose" <john.r.r...@oracle.com> wrote:
> On Nov 23, 2015, at 1:42 PM, Aleksey Shipilev <aleksey.shipi...@oracle.com> > wrote: > > > > Okay, here it is (only tests changed): > > http://cr.openjdk.java.net/~shade/8136500/webrev.06/ < > http://cr.openjdk.java.net/~shade/8136500/webrev.06/> > > Thanks; that's a solid test. > > Cleanups are good too. > > You can count me as a reviewer. > > — John > > P.S. I think a promising stringSize for *random* inputs would > switch on numberOfLeadingZeros and then perform at most > one comparison per case. We'd have to improve our game > optimizing switches, though, not to hurt performance on > the naturally occurring distributions. > > Again, Integer.toString for *random* inputs is an interesting > problem that is probably amenable to divide-and-conquer, > or at least to encoding more than one digit at a time. > > P.P.S. I filed https://bugs.openjdk.java.net/browse/JDK-8143859 < > https://bugs.openjdk.java.net/browse/JDK-8143859> > to track an underlying problem with the JIT, which is that > it should be balancing those test trees for you.
