RFR: 8289908: Skip bounds check for cases when String is constructed from entirely used byte[]
Roger Riggs
rriggs at openjdk.org
Thu Jul 7 18:48:40 UTC 2022
On Thu, 7 Jul 2022 10:21:06 GMT, Сергей Цыпанов <duke at openjdk.org> wrote:
> We can skip bounds check and null check for Charset in case we use the array entirely and the Charset is either default one or proven to be non-null.
>
> Benchmark results:
>
> before
>
> Benchmark Mode Cnt Score Error Units
> StringConstructor.newStringFromArray avgt 50 4,815 ± 0,154 ns/op
> StringConstructor.newStringFromArrayWithCharset avgt 50 4,462 ± 0,068 ns/op
> StringConstructor.newStringFromArrayWithCharsetName avgt 50 8,653 ± 0,040 ns/op
> StringConstructor.newStringFromRangedArray avgt 50 5,090 ± 0,066 ns/op
> StringConstructor.newStringFromRangedArrayWithCharset avgt 50 4,550 ± 0,041 ns/op
> StringConstructor.newStringFromRangedArrayWithCharsetName avgt 50 8,080 ± 0,055 ns/op
>
> after
>
> Benchmark Mode Cnt Score Error Units
> StringConstructor.newStringFromArray avgt 50 4,595 ± 0,053 ns/op
> StringConstructor.newStringFromArrayWithCharset avgt 50 4,038 ± 0,062 ns/op
> StringConstructor.newStringFromArrayWithCharsetName avgt 50 8,035 ± 0,031 ns/op
> StringConstructor.newStringFromRangedArray avgt 50 4,084 ± 0,007 ns/op
> StringConstructor.newStringFromRangedArrayWithCharset avgt 50 4,014 ± 0,008 ns/op
> StringConstructor.newStringFromRangedArrayWithCharsetName avgt 50 7,466 ± 0,071 ns/op
src/java.base/share/classes/java/lang/String.java line 1429:
> 1427: */
> 1428: public String(byte[] bytes, int offset, int length) {
> 1429: this(bytes, offset, length, Charset.defaultCharset(), checkBoundsOffCount(offset, length, bytes.length));
Can you avoid the extra constructor by applying `checkBoundOffCount` to the offset argument; it returns the offset.
this(bytes, checkBoundsOffCount(offset, length, bytes.length), length, Charset.defaultCharset());
or call `Preconditions.checkFromIndexSize` directly.
-------------
PR: https://git.openjdk.org/jdk/pull/9407
More information about the security-dev
mailing list