On Sun, 8 Dec 2024 08:51:34 GMT, Shaojin Wen <s...@openjdk.org> wrote:
>> This PR is a resubmission after PR #21593 was rolled back, and the unsafe >> offset overflow issue has been fixed. >> >> 1) Move getChars methods of StringLatin1 and StringUTF16 to DecimalDigits to >> reduce duplication. >> >> 2) HexDigits and OctalDigits also include getCharsLatin1 and getCharsUTF16 >> >> 3) Putting these two methods into DecimalDigits can avoid the need to expose >> them in JavaLangAccess >> Eliminate duplicate code in BigDecimal >> >> 4) This PR will improve the performance of Integer/Long.toString and >> StringBuilder.append(int/long) scenarios. This is because Unsafe.putByte is >> used to eliminate array bounds checks, and of course this elimination is >> safe. In previous versions, in Integer/Long.toString and >> StringBuilder.append(int/long) scenarios, -COMPACT_STRING performed better >> than +COMPACT_STRING. This is because StringUTF16.getChars uses >> StringUTF16.putChar, which is similar to Unsafe.putChar, and there is no >> bounds check. > > Shaojin Wen has updated the pull request with a new target base due to a > merge or a rebase. The incremental webrev excludes the unrelated changes > brought in by the merge/rebase. The pull request contains 16 additional > commits since the last revision: > > - Merge remote-tracking branch 'upstream/master' into > int_get_chars_dedup_202411 > - Merge remote-tracking branch 'upstream/master' into > int_get_chars_dedup_202411 > - fix unsafe address overflow > - add benchmark > - remove comments, from @liach > - Merge remote-tracking branch 'upstream/master' into > int_get_chars_dedup_202410 > - fix Helper > - fix Helper > - fix Helper > - unsafe putByte > - ... and 6 more: https://git.openjdk.org/jdk/compare/0f95af17...efa6e6b2 Here are the performance numbers for AMD Linux x64: ## script git remote add wenshao https://github.com/wenshao/jdk.git git fetch wenshao # checkout baseline + benchmark git checkout efa6e6b229b5f49b7080061bd4d293210c3a9e93 make test TEST="micro:java.lang.StringBuilders.appendWithInt" make test TEST="micro:java.lang.StringBuilders.appendWithLong" # checkout pr22023 git checkout efa6e6b229b5f49b7080061bd4d293210c3a9e93 make test TEST="micro:java.lang.StringBuilders.appendWithInt" make test TEST="micro:java.lang.StringBuilders.appendWithLong" ## performance numbers # baseline -Benchmark Mode Cnt Score Error Units -StringBuilders.appendWithIntLatin1 avgt 15 210.151 ± 0.139 ns/op -StringBuilders.appendWithIntUtf16 avgt 15 221.580 ± 4.940 ns/op -StringBuilders.appendWithLongLatin1 avgt 15 211.266 ± 2.063 ns/op -StringBuilders.appendWithLongUtf16 avgt 15 221.391 ± 4.943 ns/op # pr22023 +Benchmark Mode Cnt Score Error Units +StringBuilders.appendWithIntLatin1 avgt 15 191.496 ± 1.252 ns/op +9.74% +StringBuilders.appendWithIntUtf16 avgt 15 222.215 ± 0.150 ns/op -0.25% +StringBuilders.appendWithLongLatin1 avgt 15 190.951 ± 0.191 ns/op +10.63% +StringBuilders.appendWithLongUtf16 avgt 15 222.606 ± 0.434 ns/op -0.54% ## Summary PR #22023 uses Unsafe, and the performance of appendWithIntLatin1/appendWithLongLatin1 scenes has been improved by about 10%. There is no change in appendWithIntUtf16/appendWithLongUtf16. ------------- PR Comment: https://git.openjdk.org/jdk/pull/22023#issuecomment-2525950747
