harmony-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Aleksey Shipilev (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HARMONY-5719) [drlvm][jit][opt][performance] j.u.IdentityHashMap: inliner misses opportunities
Date Sat, 12 Apr 2008 11:03:05 GMT

    [ https://issues.apache.org/jira/browse/HARMONY-5719?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12588223#action_12588223
] 

Aleksey Shipilev commented on HARMONY-5719:
-------------------------------------------

Sure, Mikhail. Mode [2] is just proof-of-concept.
The real question I have to answer is why inliner heuristic didn't include these methods by
default.

> [drlvm][jit][opt][performance] j.u.IdentityHashMap: inliner misses opportunities
> --------------------------------------------------------------------------------
>
>                 Key: HARMONY-5719
>                 URL: https://issues.apache.org/jira/browse/HARMONY-5719
>             Project: Harmony
>          Issue Type: Sub-task
>          Components: DRLVM
>            Reporter: Aleksey Shipilev
>            Assignee: Mikhail Fursov
>
> Inliner misses performance-sensitive opportunities for inlining in j.u.IdentityHashMap.
> On ThreadLocalBench:
> mode [1]: 36600 
> mode [2]: 53700 (+46%)
> [1] 
> -Xem:server:
> [2]
> -Xem:server
> -XX:jit.arg.optimizer.inline.min_benefit_threshold=0
> -XX:jit.arg.optimizer.inline.inline_small_method_max_size=1000
> -XX:jit.arg.optimizer.inline.inline_small_method_bonus=7000

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message