RFR: 8343394: Make MemorySessionImpl.state a stable field [v2]
Maurizio Cimadamore
mcimadamore at openjdk.org
Thu Oct 31 17:44:28 UTC 2024
On Thu, 31 Oct 2024 17:24:23 GMT, Quan Anh Mai <qamai at openjdk.org> wrote:
> For some reasons, the compiler really dislikes the unsafe loop.
On my slower machine, the difference between the variants is bigger:
Benchmark Mode Cnt Score Error Units
LoopOverRandom.segment_loop avgt 15 3.727 ± 0.063 us/op
LoopOverRandom.segment_loop_all avgt 15 2.813 ± 0.100 us/op
LoopOverRandom.segment_loop_asUnchecked avgt 15 2.674 ± 0.051 us/op
LoopOverRandom.unsafe_loop avgt 15 2.785 ± 0.068 us/op
As you observed, `unsafe_loop` gives odd numbers. (on paper should be the fastest).
-------------
PR Comment: https://git.openjdk.org/jdk/pull/21810#issuecomment-2450460577
More information about the core-libs-dev
mailing list