On Mon, 12 Jun 2023 14:25:42 GMT, Per Minborg <pminb...@openjdk.org> wrote:
> This PR proposes adding `@Stable` to certain wrapper classes' cache arrays
> (e.g. `Integer` and `Long`).
>
> Comments are welcome as to how to improve the VM's handling now that the
> backing cache array is `@Stable`. Are there simplifications to be made or are
> there other optimizations we might add?
The benchmarks I provided above does not involve constant folding so here is
another set of benchmarks (Thanks @cl4es):
Before
Benchmark (size) Mode Cnt Score Error Units
Integers.constantFoldInteger 500 avgt 15 0.568 ± 0.013 ns/op
Integers.constantFoldInts 500 avgt 15 0.564 ± 0.005 ns/op
After
Benchmark (size) Mode Cnt Score Error Units
Integers.constantFoldInteger 500 avgt 15 0.558 ± 0.001 ns/op
Integers.constantFoldInts 500 avgt 15 0.559 ± 0.001 ns/op
@Benchmark
public int constantFoldInteger() {
return Integer.valueOf(64) * Integer.valueOf(32);
}
@Benchmark
public int constantFoldInts() {
return 64 * 32;
}
-------------
PR Comment: https://git.openjdk.org/jdk/pull/14418#issuecomment-1588823240
