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 18 February 2014.
06:51 denisboyun joined 09:50 denis_boyun joined 13:59 bluescreen joined 17:36 denis_boyun joined
Util Pre-report: 19:29
# Done: Added .sha256 files for old version of Parrot on ftp.parrot.org .
# Doing: $WORK
.end
Hello! Who all is in attendance? 19:30
Lots of focused progress in #parrot, and no one in #parrotsketch? 19:49
I'll take it every time over the opposite situation!
Anything before we adjourn? 20:08
rurban oops, missed it. 20:17
Util I am still here. Anything you want to discuss, or report on? 20:18
rurban work on lot of branches, 6model (works not yet mergable), jit (20%), t-nci-sig, pmc2c_orig
did some preliminary benchmarking
from 1.8.0 to now (could not build older ones for now) 20:19
perl514.cpanel.net/parrot/parrot-be...40321.data
Util rurban++
rurban will put it into a seperate repo
I had to tune the benchmark sources so that they can run from 1.0 to 6.2 without error 20:20
linux perf is the best tool for this job, also need to wait for a lower load on my machine, and get older parrots with static lib 20:21
the new mingw32 smoker are not yet stable, the rest looks good, solaris coming today
that's it 20:22
mingw64 smoker also todo 20:23
Util Great!
rurban msvc smoker maybe later
I heard that I could use more VMs from our QA dept
esp. for windows. this brings up the load on my machine so I cannot do proper benchmarks 20:24
Util Do we currently build OK on MSVC?
rurban no idea.
Util OK, thanks. 20:25
rurban And I get crazy errors on mingw32, I never saw before. So I guess it's my setup
I am also missing a BSD smoker for now
Util I am very happy to hear that MinGW and Solaris will be smoked regularly now. 20:27
rurban I esp. want timings from very old parrots with cgp and jit to compare against
Util Indeed 20:28
rurban for jit I will bring in the old perl5 Op libs and the runcores which were removed.
Esp. I need cgp (precomputed cgoto) 20:29
And then ops2c cvould be replaced by the old ops2c, and then nqp could be replaced by upstream
Util I would not think that cgp would be that hard to add to Parrot, based on the original (removed) code. 20:30
rurban but more important are the method call quirks (new runloop for each overloaded method call)
and -O3 safety to be able to do -finline
(inling of simple VTABLE methods) 20:31
probably some more volatile qualifiers are missing.
6model is not in a very good shape. 20:32
had to write docs, bring it up to our codings stds, and a lot of global static vars still (non-thread-safe)
and the whole architecture is not yet extendable. it's not a mop, it's just an perl6 instance of a mop 20:33
reprs need to be run-time addable for other HLLs
but it's already a bit faster than master 20:34
Util Are you shooting for 0 regressions before merging 6model? 20:35
rurban No, 6model is not ready yet for a long time. some months or so
0 regressions? well, We'll have to coordinate with nqp and moar, the repr model 20:36
I think it would be better to improve CallContext and sigs first
Util I understand.
rurban I'm not sure yet 20:37
Util I also agree on CallContext and sigs being earlier to be worked on.
rurban but not breaking other HLLs would be a priority
Util Yes. 20:38
Coke Do we have any other active HLLs? (aside from the bundled nqp/winxed) ?
rurban tcl was active for some time
Coke tcl has been broken for about 3 years now, IIRC. It's not going to get brokener.
rurban If parrot get's faster they will come again. they will have non-blocking threadpools 20:39
Coke I'm not worried about future stakeholders, just trying to clarify if we have any current ones you have to worry about. partcl's not on that list. 20:40
rurban nope.
A semaphore library is also terribly needed. HLLs need to be able to use simplier foreign proxy updaters (nqp e.g.)
ok, done 20:43
Util Anything before we adjourn? 20:45
OK, then. Thanks! 20:47
Meeting adjourned.
21:58 lizmat joined