A simple optimization proposal (original) (raw)

Krystal Mok rednaxelafx at gmail.com
Wed Feb 12 12:39:32 PST 2014

Hi Martin and John,

I did a quick-and-dirty patch and it seems to work: https://gist.github.com/rednaxelafx/8964030 If it looks right then I'll refactor that code a little bit and send it in for official review.

On Wed, Feb 12, 2014 at 11:17 AM, John Rose <john.r.rose at oracle.com> wrote:

It's totally reasonable, and is already filed as an RFE (please comment on it!):

https://bugs.openjdk.java.net/browse/JDK-8003585 -- John On Feb 12, 2014, at 9:40 AM, Martin Grajcar <maaartinus at gmail.com> wrote: Most hash tables are power-of-two sized so that they can use masking for the access. It looks like the bounds check doesn't get eliminated, although it could be. Based on the equivalence a[x & (a.length - 1)] throws if and only if a.length == 0, I'm proposing this simple algorithm:

- For each array access, check if the index has been computed via a bitwise and. - If so, check if either of the operands was computed as length minus one. - If so, replace the bounds check by a zero-length check. This zero-length check can then be easily moved out of the loop by the existing optimizations. I hope I'm not talking non-sense. For more details see http://stackoverflow.com/questions/21702939/why-the-bounds-check-doesnt-get-eliminated Regards, Martin. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/attachments/20140212/ef31799d/attachment.html

More information about the hotspot-compiler-dev mailing list