RFR: 8365926: RISC-V: Performance regression in renaissance (chi-square) [v2]

[Hamlin Li]

On Mon, 1 Sep 2025 08:56:25 GMT, Robbin Ehn <rehn at openjdk.org> wrote:

>> Hey, please consider!
>> 
>> A bunch of info in JBS entry, please read that also.
>> 
>> I narrowed this issue down to the old jal optimization, making direct calls when in reach.
>> This patch restores them and removes this regression.
>> 
>> In essence we turn "jalr ra,0(t1)" into a "jal ra,<dest>" if reachable, and restore the jalr if a new destination is not reachable.
>> 
>> Please test on your hardware!
>> 
>> 
>> Chi Square (100 runs each, 10 fastest iterations of each run, P550)
>> JDK-23 (last version with trampoline calls)
>> Mean: 3189.5827
>> Standard Deviation: 284.6478
>> 
>> JDK-25
>> Mean: 3424.8905
>> Standard Deviation: 222.2208
>> 
>> Patch:
>> Mean: 3144.8535
>> Standard Deviation: 229.2577
>> 
>> 
>> No issues found in t1, running t2 also. Stress tested on vf2, bpi-f3, p550.
>
> Robbin Ehn has updated the pull request with a new target base due to a merge or a rebase. The incremental webrev excludes the unrelated changes brought in by the merge/rebase. The pull request contains four additional commits since the last revision:
> 
>  - Merge branch 'master' into 8365926
>  - Spelling
>  - Merge branch 'master' into 8365926
>  - draft jal<->jalr

JDK-23 (last version with trampoline calls)
Mean: 3189.5827
Standard Deviation: 284.6478

JDK-25
Mean: 3424.8905
Standard Deviation: 222.2208

Patch:
Mean: 3144.8535
Standard Deviation: 229.2577


For the performance data, do you have some data for applying this fix on top of the next commit after`JDK-23 (last version with trampoline calls)`? I think this data might be more helpful to understand the performance comparison.

-------------

PR Comment: https://git.openjdk.org/jdk/pull/26944#issuecomment-3241745418