[jira] Closed: (HARMONY-2769) [drlvm][jit] use per-region InlineInfo from global storage in VM to report stack frames of inlined methods

["Mikhail Fursov (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/HARMONY-2769?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Mikhail Fursov closed HARMONY-2769.
-----------------------------------

    Resolution: Fixed

Closing this issue as FIXED.
Details:
Per-call inline info mapping was removed.
Method marker insts are used after HARMONY-2114 is committed and there are several checks in code (hir2lir, emitter) that every call have valid bcmapping and inline stack.

> [drlvm][jit] use per-region InlineInfo from global storage in VM to report stack frames of inlined methods
> ----------------------------------------------------------------------------------------------------------
>
>                 Key: HARMONY-2769
>                 URL: https://issues.apache.org/jira/browse/HARMONY-2769
>             Project: Harmony
>          Issue Type: Improvement
>          Components: DRLVM
>         Environment: all
>            Reporter: Egor Pasko
>            Assignee: Mikhail Fursov
>            Priority: Minor
>
> In current implementation there are 2 mechanisms that allow to detect areas of inlined methods:
> 1. per-call-inline-info represents the whole inline nesting chain of each
>    inlined 'call' instructions. All information is stored by JIT in the
>    corresponding CodeChunkInfo->_jit_info_block and is decoded by JIT upon VM
>    requests (during filling stack traces)
> 2. per-region-inline-info is collected within VM as an action raised by the
>    CompiledMethodLoad(...) event (necessary for JVMTI) as in HARMONY-2145.
> This issue proposes to replace per-call-inline-info with per-region-inline-info
> to make a single storage and a single method to access inlining-related info in
> VM and JIT.
> there is one disadvantage to this approach:
>     Relying on per-region-inline-info introduces some constraints for JIT
>     design. Specifically, it becomes not easy to move instruction regions
>     across method boundaries.
> but per-call-inline-info has other disadvantages:
>     using per-call-inline-info is not sufficient to cover the functionality
>     necessary for CompiledMethodLoad(...) event
>     we cannot afford reporting stack frames on exceptions (because there are a
>     lot fo potentially exceptional instructions and storing inline chains for
>     all of them would be ineffective) see HARMONY-2114
> refer to the mailing list for details:
>     http://thread.gmane.org/gmane.comp.java.harmony.devel/18590/focus=18999

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