133 Members
Get! https://www.jruby.org/ | Doc http://wiki.jruby.org | Snapshots http://ci.jruby.org/ | Report at http://bugs.jruby.org18 Servers

Load older messages

Timestamp Message
21 Jan 2020
18:35:34@headius:matrix.orgheadiusI'm working on AOT...first trying to get a grip on how much ovrehead is currently involved in deserializing IRScope
18:36:02@headius:matrix.orgheadiustricky bit seems like it's going to be getting the IRScope instances in a place where the bytecode can see them
18:36:19@headius:matrix.orgheadius enebo: what's the status of your branch at this point? It may be a better place for me to work from
18:36:39@enebo:matrix.orgeneboI have to walk out the door
18:36:47@headius:matrix.orgheadiuswe'll talk later
18:37:07@enebo:matrix.orgenebobut you can look at it jit_irscope_removal
18:39:41@headius:matrix.orgheadiusrunning a bench deserializing erb.rb looks like it's sub-ms times very quickly
18:44:08@headius:matrix.orgheadiusyeah this is tricky...persisted IR and AOT bytecode both have their own tree of scopes and they need to know about each other
18:46:12@headius:matrix.orgheadiusperhaps the IRScopes could be deserialized along with method handles to point at their JVM version
18:47:25@headius:matrix.orgheadiusneed to look at branch for a bit
18:53:01@headius:matrix.orgheadiusif JVM compiler mapped scopes to method names in the resulting output it should be possible to inject those handles when deserializing
22:46:42@headius:matrix.orgheadiussoon perhaps!
22:50:49@freenode_lopex:matrix.orglopexI hope not many unstable 2.7 changes get adopted soon
22:52:39@freenode_lopex:matrix.orglopexthankfully ruby never made the python 2.7 mistake
22:52:56@freenode_lopex:matrix.orglopex headius: right ?
22:53:25@headius:matrix.orgheadiuswhat's the python 2.7 mistake?
22:53:44@freenode_lopex:matrix.orglopex headius: splitting the versions basically
22:53:56@freenode_lopex:matrix.orglopexwhich hampered adoption
22:54:30@freenode_lopex:matrix.orglopexkeeping 2.7 so long slowed down 3.x
22:55:20@freenode_lopex:matrix.orglopex they're passed that point, but that's what it was
22:56:24@headius:matrix.orgheadius1.9 was really the big move, and 1.8.7 was a bad attempt to bridge the gap
22:56:27@headius:matrix.orgheadiusin Ruby
22:56:54@freenode_lopex:matrix.orglopexyea, it was never the case in ruby
22:57:09@freenode_lopex:matrix.orglopexeven if it might seem so

There are no newer messages yet.

Back to Room List