On Fri, 1 Sep 2023 08:21:13 GMT, Per Minborg <pminb...@openjdk.org> wrote:

> This PR proposes adding a new method to BitSet that provides an immutable 
> snapshot of the set in the form of an `IntPredicate`.
> 
> The predicate is eligible for constant folding.
> 
> Here are some classes in the JDK that would benefit directly from 
> constant-folding of BitSets:
> 
> PoolReader (6)
> URLEncoder (1) - Updated in this PR
> HtmlTree (2)
> 
> More over, the implementation of the predicate is @ValueBased and this would 
> provide additional benefits in the future.
> 
> Initial benchmarks with the URLEncoder show encouraging results:
> 
> 
> Name                           (encodeChars) (maxLength) (unchanged) Cnt  
> Base   Error   Test   Error  Unit  Diff%
> URLEncodeDecode.testEncodeUTF8             6        1024           0  15 
> 2,371 ± 0,016  2,073 ± 0,184 ms/op  12,6% (p = 0,000*)
> URLEncodeDecode.testEncodeUTF8             6        1024          75  15 
> 1,772 ± 0,013  1,387 ± 0,032 ms/op  21,8% (p = 0,000*)
> URLEncodeDecode.testEncodeUTF8             6        1024         100  15 
> 1,230 ± 0,009  1,140 ± 0,011 ms/op   7,3% (p = 0,000*)

src/java.base/share/classes/java/util/BitSet.java line 1412:

> 1410:      * <p>
> 1411:      * Returned predicates are threadsafe and can be used without 
> external synchronisation.  Furthermore,
> 1412:      * they are eligible for constant-folding optimization by the VM.

>  Furthermore,
     * they are eligible for constant-folding optimization by the VM.

I'd probably skip this unless there's prior art to describe the 
constant-fold-eligibility of returned values from methods. :)

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/15530#discussion_r1312826139

Reply via email to