Post closed tickets in your report. | Note: This channel is for our weekly status meetings (Tuesdays at 19:30 UTC); you probably want #parrot instead. | irclog: irclog.perlgeek.de/
Set by moderator on 23 July 2014.
06:51 basiliscos joined 07:30 basiliscos joined 12:51 lizmat_ joined 14:32 rurban1 joined 14:51 Chirag joined 15:14 basiliscos joined 15:57 Chirag joined 18:04 lizmat joined
Util Report: 19:31
# Done: Nil (Illness+Travel)
# Doing: Back on Smolder this evening.
.end
GSoC NOTICE: Monday, 11 August is the soft "pencils down" date.
Hello! Who all is in attendance?
Chirag Hey! Me 19:32
Report in 2 min
rurban me too
Util Welcome! 19:33
Chirag #Done: 19:35
*Corrected automatic headers (making methods unstatic)
*Debugged through the changes made for task# 3 (Major Blocker until today). Solved by fixing args.c
*I guess GH #1083 is now over (1.6% faster)
#TODO:
rurban DONE: 19:36
* with Chirag fixed, tested and squashed pcc-gh1083. 1.6% faster with direct CallContext calls.
we still autobox all args, there would be ~5% in my estimation.
TODO:
* test and benchmark perl6
more testing and then merge it
Chirag rurban: so there is scope for a 5% improvement if we dont autobox? 19:37
rurban I have an idea to allow fast internal pcc args handling, (no autoboxing) and only autobox it when accessed via the PMC API
This is also mentioned in the various blogposts on this subject 19:38
previously the args were stored as CELL, not as PMC
Chirag yes I remember from our first conversation I think 19:39
rurban so we could use the old CELL access to the CallContext and only autobox if needed externally
But this is probably too much for our GSOC target date, next monday 19:40
I'm fine with these changes so far now.
Chirag but 18th is hard pencils down, right? maybe I can start and stop and continue after submission 19:41
rurban sure
Chirag great 19:42
rurban we didn't get 20%, but we got 5-8% or so
Util finishes pre-typing, and sees that it is partially not needed, but posts anyway:
Google recommends that no new task be started at this point. Normal schedule:
* Finish testing and Merge to master by August 11th.
* Documentation, final blog posts, etc August 12th through 18th (the "hard" end date).
Any tasks that you planned to start this week, should be delayed until after the formal end of GSoC, and then proceed on a purely volunteer basis.
rurban yes, agreed
Util This is to avoid muddying the waters w.r.t the GSoC paid work.
rurban finish testing until monday sounds like a good idea 19:43
Chirag umm, alright... but I still have a week left
rurban esp. since there were some perl6 bugs along the same codebase (which apparently are fixed now in perl6)
you could redo some overall benchmarks to verify the numbers 19:44
with a slow machine the numbers should be better
Chirag sure! I will do that
so I finished three tasks -> 3 benchmarks 19:45
Util Chirag: when students start a new task during this week, Murphy's Law often bites, and the task that *should* take 3 days takes 8 instead, and then your work is incomplete! 19:46
rurban yes
Chirag agreed. I will continue after GSoC then
rurban esp. autoboxing looks very tricky. this works by using the pcc registers directly
Util Better to err on the side of caution, and have a squeaky-clean well-defined edge to the paid work.
rurban when you test older parrots you will see often hard crashes there 19:47
Util Is everything on track to merge by Monday? 19:54
Any blockers in the next week?
rurban not yet
but we will see with perl6. this is a very fast moving target right now
Util OK 19:55
Anything else to discuss? From anyone?
Chirag Not from me. Thanks!
Util OK. Thanks everyone! 19:58
Meeting adjourned.
20:16 rurban1 joined 21:42 rurban1 joined 23:41 lizmat joined, atrodo joined, slavorg joined, Tene joined, allison joined, kshannon joined, Util joined, cottoo joined, Coke joined 23:48 rurban joined

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!