Interpreter is slower for microbenchmarks after "backtrace" merge
-----------------------------------------------------------------
Key: JRUBY-5229
URL: http://jira.codehaus.org/browse/JRUBY-5229
Project: JRuby
Issue Type: Bug
Components: Interpreter
Affects Versions: JRuby 1.6
Reporter: Charles Oliver Nutter
Priority: Critical
Fix For: JRuby 1.6
The new backtrace branch executes compiled code faster, but has a bit more
overhead for interpreted method calls due to the extra Backtrace object setup
and stack in ThreadContext.
For microbenchmarks, like calling a method "foo" that just returns self, this
shows up as about a 20-30% degradation: https://gist.github.com/725036
However, as the workload in the method bodies grows, the degradation %
decreases, showing that it's largely tied to method call overhead:
https://gist.github.com/725037
We should make an effort to reduce this gap before 1.6, since much code will
run interpreted for at least some of its lifetime (and potentially all of its
lifetime on platforms where we can't JIT).
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://jira.codehaus.org/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe from this list, please visit:
http://xircles.codehaus.org/manage_email
