RFR: 8332528: Generate code in SwitchBootstraps.generateTypeSwitch that require fewer adaptations [v3]
Claes Redestad
redestad at openjdk.org
Tue May 21 09:07:04 UTC 2024
On Tue, 21 May 2024 09:01:32 GMT, Claes Redestad <redestad at openjdk.org> wrote:
>> We can fold the call to `Objects.checkIndex` into the code generated in generateTypeSwitchSkeleton instead of doing so by filtering the MH argument. This loads 9 less classes (of which 8 generated LFs and Species classes) on a minimal test, while being neutral on a throughput sanity test:
>>
>>
>> Name Cnt Base Error Test Error Unit Change
>> SwitchSanity.switchSum 15 8,162 ± 0,117 8,152 ± 0,131 ns/op 1,00x (p = 0,800 )
>> * = significant
>> ```
>>
>> A few additional optimizations includes moving some seldom used `findStatic` calls to a holder. All in all this means a reduction by 22M cycles to bootstrap a trivial switch expression on my M1.
>
> Claes Redestad has updated the pull request incrementally with one additional commit since the last revision:
>
> Add type switch to HelloClasslist
By adding a trivial type switch to `HelloClasslist` we archive a set of classes and avoid some runtime class generation. This grows the default CDS archive by approximately 1.2Mb by pulling in a large number of Classfile API classes, which is bound to happen soon anyway.
On the SwitchSanity test, bootstrap cost drops by another 35M cycles:
Wall clock: 51.0 ms/op
.taskclock: 63.5 ms/op
.cycles: 220064872 average cycles elapsed
.instructions: 544981276 average instructions retired
-------------
PR Comment: https://git.openjdk.org/jdk/pull/19307#issuecomment-2122136668
More information about the core-libs-dev
mailing list