On Thu, 27 Jun 2024 14:12:36 GMT, Shaojin Wen <d...@openjdk.org> wrote:
>> Currently, the java.util.Formatter$Conversion::isValid method is implemented >> based on switch, which cannot be inlined because codeSize > 325. This >> problem can be avoided by implementing it with ImmutableBitSetPredicate. >> >> use `-XX:+UnlockDiagnosticVMOptions -XX:+PrintInlining` to see the master >> branch: >> >> @ 109 java.util.Formatter$Conversion::isValid (358 bytes) failed to >> inline: hot method too big >> >> >> current version >> >> @ 109 java.util.Formatter$Conversion::isValid (10 bytes) inline (hot) >> @ 4 >> jdk.internal.util.ImmutableBitSetPredicate$SmallImmutableBitSetPredicate::test >> (50 bytes) inline (hot) > > Shaojin Wen has updated the pull request incrementally with one additional > commit since the last revision: > > revert & use `@ForceInline` I found using @ForceInline gave me better performance with smaller code changes, Here are the performance numbers running on Mac M1 Max: # baseline Benchmark Mode Cnt Score Error Units StringFormat.stringIntFormat avgt 15 93.299 ? 1.268 ns/op # WebRevs 00 fb5c8001 Benchmark Mode Cnt Score Error Units StringFormat.stringIntFormat avgt 15 89.164 ? 2.188 ns/op # ForceInline Benchmark Mode Cnt Score Error Units StringFormat.stringIntFormat avgt 15 84.473 ? 4.287 ns/op ------------- PR Comment: https://git.openjdk.org/jdk/pull/19926#issuecomment-2194818165
