On Mon, 2 Sep 2024 14:06:42 GMT, Shaojin Wen <s...@openjdk.org> wrote:
>> Use fast path for ascii characters 1 to 127 to improve the performance of >> writing Utf8Entry to BufferWriter. > > 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 21 additional > commits since the last revision: > > - Update src/java.base/share/classes/java/lang/StringCoding.java > > Co-authored-by: ExE Boss <3889017+exe-b...@users.noreply.github.com> > - vectorized countGreaterThanZero > - add comments > - optimization for none-ascii latin1 > - Revert "vectorized countGreaterThanZero" > > This reverts commit 88a77722c8f5401ac28572509d6a08b3e88e8e40. > - vectorized countGreaterThanZero > - copyright > - use JLA if length < 256 > - fix utf_len error > - code style > - ... and 11 more: https://git.openjdk.org/jdk/compare/e43c4655...2a36b443 test/micro/org/openjdk/bench/java/lang/classfile/Utf8EntryWriteTo.java line 85: > 83: for (int i = 0; i < constants.length; i++) { > 84: constants[i] = s.repeat(32); > 85: } I think this micro is testing too much irrelevant parts of classfile generation. I would recommend these changes: 1. Compute a `ConstantPoolBuilder` in `setup`, like: ConstantPoolBuilder poolBuilder; ClassEntry thisClass; poolBuilder = ConstantPoolBuilder.of(); thisClass = poolBuilder.classEntry(CLASS_DESC); for (var s : constants) { poolBuilder.utf8Entry(s); } 3. In benchmark, test like: @Benchmark public byte[] writeTo() { return cf.build(thisClass, poolBuilder, clb -> {}); } This should eradicate the cost around method or instruction generation while still ensuring the constants are written to the class file as part of the constant pool. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/20772#discussion_r1741018079
