JITWatch January update

[Chris Newland]

Copy of an update I posted to the JITWatch Google group.

Additionally, JITWatch has been working fine with recent JDK 9 builds and
I'll continue to test and report to the quality outreach programme.

Cheers,

Chris

---------------------------- Original Message ----------------------------
Subject: JITWatch January update
From:    "Chris Newland" <chris at chrisnewland.com>
Date:    Tue, January 20, 2015 23:14
To:      jitwatch at googlegroups.com
--------------------------------------------------------------------------

Hi all, here's a quick update:

1) Much refactoring and bug squashing. If you've not updated in a while
then please do as the parsing is more robust around generics and the
TriView output is more reliable (Thanks to Chris Vest for a pull request
here).

2) Java 8 update 40 is out soon and this fixes a nasty bug I reported in
HotSpot's LogCompilation output when tiered compilation is used:
http://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/2014-June/014912.html

3) Added support for statically compiled Groovy (groovyc) in the Sandbox
and added an example. Groovy's bytecode for method calls is very different
to that of Java and Scala. Next on the VM language list is Nashorn which
will help me improve support for analysing lambda bytecode.

4) New report on optimized virtual_call usage (requires hsdis as this is
parsed from the assembly comments). My understanding may not be 100% here
but the idea is for the report to identify when a virtual call such as
invokeinterface has been optimised to an invokevirtual when HotSpot
observes monomorphic or bimorphic dispatch. This optimisation is only made
when a method can't be eliminated completely through inlining. It might be
able to highlight where an invokeinterface call both failed to inline and
also failed to be "sharpened" to invokevirtual so was made in the slowest
possible way.

5) Added more functionality to headless mode. You can now print the
suggestions and the optimized virtual_call table. This was requested by a
user who was having trouble parsing enormous HotSpot logs in UI mode. Have
opened an issue to do some memory profiling.

6) Interest in JITWatch is still growing and the project now has 436 stars
on GitHub with 9 code contributors. I think completing Nashorn support will
take it over 500.

Please let me know if you have any feature requests or bug reports.

Cheers,

Chris