RFR: 8322174: RISC-V: C2 VectorizedHashCode RVV Version [v22]
Yuri Gaevsky
duke at openjdk.org
Wed Aug 6 17:07:14 UTC 2025
On Wed, 6 Aug 2025 17:02:50 GMT, Yuri Gaevsky <duke at openjdk.org> wrote:
>> The patch adds possibility to use RVV instructions for faster vectorizedHashCode calculations on RVV v1.0.0 capable hardware.
>>
>> Testing: hotspot/jtreg/compiler/ under QEMU-8.1 with RVV v1.0.0.
>
> Yuri Gaevsky has updated the pull request incrementally with one additional commit since the last revision:
>
> try m1 for grouping
`60b2d81`:
--- -XX:+UseRVV ---
Benchmark (size) Mode Cnt Score Error Units
ArraysHashCode.ints 1 avgt 30 11.298 ± 0.021 ns/op
ArraysHashCode.ints 5 avgt 30 21.475 ± 0.148 ns/op
ArraysHashCode.ints 10 avgt 30 37.011 ± 0.057 ns/op
ArraysHashCode.ints 20 avgt 30 47.146 ± 0.091 ns/op
ArraysHashCode.ints 30 avgt 30 56.377 ± 0.021 ns/op
ArraysHashCode.ints 40 avgt 30 55.116 ± 0.010 ns/op
ArraysHashCode.ints 50 avgt 30 66.537 ± 0.457 ns/op
ArraysHashCode.ints 60 avgt 30 72.110 ± 0.886 ns/op
ArraysHashCode.ints 70 avgt 30 88.029 ± 0.897 ns/op
ArraysHashCode.ints 80 avgt 30 77.739 ± 1.458 ns/op
ArraysHashCode.ints 90 avgt 30 92.164 ± 1.118 ns/op
ArraysHashCode.ints 100 avgt 30 104.852 ± 4.030 ns/op
ArraysHashCode.ints 200 avgt 30 180.402 ± 0.037 ns/op
ArraysHashCode.ints 300 avgt 30 234.598 ± 7.007 ns/op
ArraysHashCode.ints 1000 avgt 30 806.796 ± 0.126 ns/op
ArraysHashCode.ints 10000 avgt 30 8152.396 ± 79.158 ns/op
Based on above experiments it looks reasonable to use `m2` grouping.
-------------
PR Comment: https://git.openjdk.org/jdk/pull/17413#issuecomment-3160916148
PR Comment: https://git.openjdk.org/jdk/pull/17413#issuecomment-3160919289
More information about the hotspot-compiler-dev
mailing list