RFR: 8293011: riscv: Duplicated stubs to interpreter for static calls
Feilong Jiang
fjiang at openjdk.org
Mon Aug 29 02:46:07 UTC 2022
On Mon, 29 Aug 2022 01:32:26 GMT, Dingli Zhang <duke at openjdk.org> wrote:
> Follow up [JDK-8280481](https://bugs.openjdk.org/browse/JDK-8280481), it provides duplicated stubs to interpreter for static calls.
>
> ## Testing:
>
> - hotspot/jdk tier1 on unmatched board
> - hotspot/jtreg/compiler/sharedstubs/SharedStubToInterpTest.java on qemu
>
>
> ## Results
> #### Results from [Renaissance 0.14.0](https://github.com/renaissance-benchmarks/renaissance/releases/tag/v0.14.0)
> Note: 'Nmethods with shared stubs' is the total number of nmethods counted during benchmark's run. 'Final # of nmethods' is a number of nmethods in CodeCache when JVM exited.
>
> Because of the current hardware performance limitations, I selected a few cases to test and reduced the number of iterations using `-r 10`. The data is sufficient to reflect the effect of the optimization.
>
> - riscv64
>
> +------------------+-------------+----------------------------+---------------------+
> | Benchmark | Saved bytes | Nmethods with shared stubs | Final # of nmethods |
> +------------------+-------------+----------------------------+---------------------+
> | future-genetic | 47824 | 1278 | 1811 |
> | scrabble | 30616 | 738 | 1605 |
> | fj-kmeans | 23320 | 633 | 1386 |
> | par-mnemonics | 2840 | 621 | 1398 |
> | scala-kmeans | 19544 | 381 | 1365 |
> +------------------+-------------+----------------------------+---------------------+
Changes look good, except for one copyright issue.
src/hotspot/cpu/riscv/codeBuffer_riscv.cpp line 2:
> 1: /*
> 2: * Copyright Amazon.com Inc. or its affiliates. All Rights Reserved.
Missing copyright of your company for the new file.
-------------
Changes requested by fjiang (Author).
PR: https://git.openjdk.org/jdk/pull/10057
More information about the hotspot-compiler-dev
mailing list