RFR: 8355177: Speed up StringBuilder::append(char[]) via Unsafe::copyMemory [v5]

[Roger Riggs]

On Sat, 10 May 2025 11:20:35 GMT, Shaojin Wen <swen at openjdk.org> wrote:

>> In BufferedReader.readLine and other similar scenarios, we need to use StringBuilder.append(char[]) to build the string.
>> 
>> For these scenarios, we can Unsafe.copyMemory instead of the character copy of the char-by-char loop to improve the speed.
>> 
>> @RogerRiggs completed the optimization when the encoder is LATIN1 in PR #24967. This PR continues to complete the optimization when the encoder is UTF16.
>
> Shaojin Wen has updated the pull request with a new target base due to a merge or a rebase. The pull request now contains nine commits:
> 
>  - copyright
>  - Merge remote-tracking branch 'upstream/master' into optim_sb_append_chars_202504
>    
>    # Conflicts:
>    #	src/java.base/share/classes/java/lang/AbstractStringBuilder.java
>  - Merge remote-tracking branch 'upstream/master' into optim_sb_append_chars_202504
>    
>    # Conflicts:
>    #	src/java.base/share/classes/java/lang/AbstractStringBuilder.java
>  - Merge remote-tracking branch 'upstream/master' into optim_sb_append_chars_202504
>    
>    # Conflicts:
>    #	src/java.base/share/classes/java/lang/StringUTF16.java
>  - putCharsUnchecked
>  - copyright
>  - Using StringUTF16.compress to speed up LATIN1 StringBuilder append(char[])
>  - Using Unsafe.copyMemory to speed up UTF16 StringBuilder append(char[])
>  - add append(char[]) benchmark

Changes requested by rriggs (Reviewer).

src/java.base/share/classes/java/lang/StringUTF16.java line 1494:

> 1492:         checkBoundsBeginEnd(index, index + end - off, val);
> 1493:         checkBoundsBeginEnd(off, end, ca);
> 1494:         putCharsUnchecked(val, index, ca, off, end);

There is no reason to separate the bounds checks from the call to Unsafe.copyMemory.
Remove `putCharsUnchecked` and inline the call here. 
Do not be tempted to use it more broadly and make the bounds checks harder to audit.

src/java.base/share/classes/java/lang/StringUTF16.java line 1668:

> 1666:     private static void checkBoundsBeginEnd(int begin, int end, char[] val) {
> 1667:         String.checkBoundsBeginEnd(begin, end, val.length);
> 1668:     }

This method doesn't pull it weight, inline the check with the singular use.

test/micro/org/openjdk/bench/java/lang/StringBuilders.java line 285:

> 283:     @Benchmark
> 284:     public int appendWithCharArrayLatin1() {
> 285:         StringBuilder buf = sbLatin1;

Do not re-use the sbLatin1 buffer, its length may have changed by previous tests and perturb the test results.

test/micro/org/openjdk/bench/java/lang/StringBuilders.java line 295:

> 293:     @Benchmark
> 294:     public int appendWithCharArrayUTF16() {
> 295:         StringBuilder buf = sbUtf16;

Do not re-use the sbUtf16 buffer, its length may have changed by previous tests and perturb the test results.

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

PR Review: https://git.openjdk.org/jdk/pull/24773#pullrequestreview-2955366637
PR Review Comment: https://git.openjdk.org/jdk/pull/24773#discussion_r2164925190
PR Review Comment: https://git.openjdk.org/jdk/pull/24773#discussion_r2164930261
PR Review Comment: https://git.openjdk.org/jdk/pull/24773#discussion_r2164931643
PR Review Comment: https://git.openjdk.org/jdk/pull/24773#discussion_r2164933643