github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
00:53 mgoebel joined 01:14 Kaiepi left, Kaiepi joined 01:34 mgoebel left 01:38 Kaiepi left 01:39 Kaiepi joined 01:45 Kaiepi left 01:47 Kaiepi joined 05:29 AlexDaniel joined 07:48 domidumont joined 08:37 robertle joined 09:15 domidumont left 09:17 domidumont joined 09:52 domidumont left 10:05 domidumont joined 10:09 domidumont1 joined 10:11 domidumont left 12:18 lizmat_ joined 12:20 lizmat left 12:31 domidumont1 left 12:33 MasterDuke_ joined 12:35 MasterDuke_ left 12:37 MasterDuke_ joined 12:41 MasterDuke_ left 12:58 domidumont joined
timotimo good news on the laptop. i just took it apart and removed the internal battery and now it starts when no batteries are in but it's connected to the mains 12:58
12:59 lizmat joined
nwc10 glad to hear this - this is a lot better than it could have been 13:00
13:03 lizmat_ left
masak timotimo: phew! 13:03
timotimo just now i'm told there's actually a button that allows you to disconnect the internal battery without taking the device apart 13:06
it's a tiny hole in the underside that you can push a needle through
13:08 ggoebel joined
ggoebel timotimo: for ~$300 you can pick up a thinkpad t460 w/ >8GB RAM, SSD, and 1600x900 or better display off ebay w/ close to a year left on a 3 year corporate warranty. With the thinkpad keyboard and build quality... That is better than pretty much anything you can buy new at the same price point. 13:12
masak or at least that's what the Thinkpad/eBay-industrial complex would have you believe 13:18
timotimo i'm not sure i can live with less than 16 gigs of ram :) 13:19
ggoebel there's one on ebay now w/ 16GB at $320 + $23 shipping... resolution is only 1366x768 though... It'd cost ~$75 to do the DIY 1080p upgrade. If you're patient... you can find one with both the display and ram you need for around $300. Or step back to a t450 and you'll have a better selection. 13:26
nwc10 timotimo: but you can live without a battery? 13:28
ggoebel lug around a UPS :-)
timotimo nwc10: don't have to. i already plugged it back in and it just works again 13:32
apparently it just got into some battery-backed state that was unhappy, and unplugging both batteries made it just reset to normal 13:33
nwc10 \o/ 13:35
I hope that this is a happy outcome that doesn't revert
timotimo same 13:43
if i can bridge another year with this device, surely there'll be quite cheap thinkpads that even pack a little bit of punch in the GPU department
at some point, there'll have to be GPGPU modules for perl6 13:44
though tbh i could already develop those on my desktop, surely; a GTX 1060 should support all basic things, right?
lizmat timotimo: what are you using for that ? 13:45
13:55 lizmat_ joined
timotimo using for what? 13:55
13:56 lizmat__ joined
lizmat__ timotimo: "i could already develop those on my desktop" 13:56
timotimo i'm not sure i understand the question. you mean like OpenCL vs CUDA or something like that?
lizmat__ yeah, I guess >
?
timotimo i haven't started at all :) 13:57
lizmat__ ahhh... ok :-( :-)
timotimo until then, continuing work on SSE vectorization ops could be interesting 13:58
13:59 lizmat left 14:00 lizmat_ left, lucasb joined 14:20 brrt joined
brrt ohai #moarvm 14:20
lizmat__ brrt o/
14:20 lizmat__ is now known as lizmat 14:31 Guest4810 joined, lizmat left 14:40 Ulti_ is now known as Ulti 15:13 lizmat joined 15:15 lizmat left, lizmat_ joined 15:22 lizmat_ left 15:29 robertle left 15:35 robertle joined 15:38 lizmat joined 16:03 brrt left 16:23 domidumont left 17:58 robertle left 18:01 domidumont joined 18:13 domidumont left 18:15 domidumont joined 18:19 brrt joined 18:56 domidumont left 19:15 robertle joined 19:43 ggoebel left
samcv ohai brrt 20:29
brrt ohai samcv 20:43
how is it that every time that I finally have time for moarvm, I'm already tired :-(
how are you? 20:44
hows' the rain in amsterdam? 'round here, it's wet
jnthn Where is the rain not wet? :) 20:45
brrt ohai jnthn 20:49
how was Munich?
samcv heya brrt 20:50
i feel the same since i got here. though i'm way less tired now since i'm now settled. this is the 2nd week on my real™ team
brrt what team is that? 20:51
samcv soylent. it's the team that owns the PII interface 20:52
(personally indentifiable information)
brrt oh, cool
I, well, used to know people who used to be on that team
samcv this and calango team are the two my onboarding lead recommended. also i'll get to use my unicode knowledge. since dev machines don't get real PII data 20:53
they get or should get fake/scrambled data
and i need to someday make it better
brrt I wrote the soylent bulk mapper. Is that still a thing?
samcv ah
maybe. were you on the soylent team? or before it was its own team?
brrt no, I was on EMK 20:54
we had a bunch of PII that was not mapped neatly to database records
so we needed to transform that in bulk
samcv emailk?
or what
brrt email marketing
samcv yeah. that's still like 20:55
least elegant part of soylent atm
brrt it used references and aliasing. it was slightly evil. I'm unreasonably proud of that code
hehehee
samcv s/atm/forever
brrt, they're still the biggest ab(user) of our systems 20:57
well. maybe not biggest. but like storage space i guess
brrt ouch
yeah, EMK is necessarily very PII heavy 20:58
anyway.... ehm. I've thought long and hard about the missing ancestor problem. 20:59
I've decided that if the expression IR doesn't have an implicit order, and any operator can reference any value, then from the expression IR's perspective, missing ancestors aren't a problem.
The expression IR doesn't promise any particular ordering invariant, other than that provided by `DO` 21:00
So in that sense, it is strictly a code generation bug
arguably, the code generator is well-placed to figure out if it would be emitting such a value 21:02
(or reference)
so the second thing I've also been thinking about, is to compile all the conditional structures (IF/WHEN/ALL/ANY) to a single COND structure 21:06
in the template preprocessor
the reasoning being that a single COND structure is much easier to analyze, in terms of control flow, than the combination of the above
21:07 Kaiepi left 21:09 Kaiepi joined
jnthn brrt: Munich was nice. My talks went alright. There was company, food, and beer. :) 21:18
brrt I'd expect beer at least :-) 21:19
jnthn brrt: I guess at that level one general structure is nice, in that it's less cases to consider, though maybe it makes other things harder.
brrt compilation of the DAG becomes slightly harder
because, eh, well, at that point the conditional blocks aren't a nice tree shape anymore 21:20
and the conditions need to be compiled in a series, and the code blocks after, and it's a bit of a tangle 21:22
but comparatively manageable
MasterDuke oh hey, while people are here. i mentioned this earlier, but only timotimo was around. there's a change i'll be able to go to either perlcon eu or the swiss perl workshop. any recommendations either way? one consideration is the actual conference, another is stuff to see/do at the location 21:23
brrt I don't have recommendations.
I haven't been to SPW
I've been to a (small) number of perlcons 21:24
but I'm just hoping to have time (and energy) to go this year
jnthn I'll certainly be at PerlCon. I find travel things quite tiring, and SPW is very soon after PerlCon, so I'm less likely to do that this year. Riga is a nice city; not huge, but nice. Olten is also pretty nice, though of course vastly more expensive; it's easy in Switzerland to go see other nearby places by train though, and there's lots of them. :) 21:26
MasterDuke hm, if significantly more perl6 people are going to be at perlcon eu than spw (and riga is nice) i may try to make that one work first 21:28
Geth MoarVM/jit-expr-optimizer: 4 commits pushed by (Bart Wiegmans)++ 21:46
MoarVM/jit-expr-float: 15 commits pushed by (Bart Wiegmans)++
review: github.com/MoarVM/MoarVM/compare/4...e97f4f039d		 21:48
MoarVM/expr-jit-invoke: e1870bc60c | (Bart Wiegmans)++ | 4 files
[JIT] Implement invoke sequence in expr JIT
 21:50
MoarVM/expr-jit-invoke: 763ea27151 | (Bart Wiegmans)++ | src/jit/graph.c
[JIT] Eliminate special consume_invoke opcode

Except for nativecall_invoke, we no longer need this part of the legacy JIT.
MoarVM/expr-jit-invoke: 0c91ca9ba1 | (Bart Wiegmans)++ | 7 files
[JIT] WIP on nativeinvoke in the expression JIT
brrt ^ rebases
japhb brrt: What keeps you from committing some of this partial work to master? 21:58
yoleaux 13 Mar 2019 07:53Z <ab6tract> japhb: Did you get my email about the future of Terminal::Print? I sent it to the account listed in the perl6 contributors doc.
brrt japhb: I hate buts
*bugs
and in the case of jit-expr-float, I know there are some incomplete things
but they are relatively small, so I should really get them finished 21:59
japhb .tell ab6tract No, I hadn't noticed that, though it's been a while since I checked all my mail accounts and folders. I'll see if I can find it.
yoleaux japhb: I'll pass your message to ab6tract.
brrt and in the case of the jit-expr-optimizer, I know that it introduces a very interesting bug (the one I'm supposed to blog about)
and in the case of the expr-jit-invoke, it pessimizes the jit-nativecall that nine++ built 22:00
so 'incompleteness'
and what makes me especially nervous around the floating point branch is that I have no real way of testing it 22:01
22:01 avarab is now known as avar
japhb Ah, gotcha. 22:01
brrt but the question is a fair one
I should get them merged sooner rather than later :-)
japhb Yes please! :-)
brrt (but first, sleep though :-)) 22:02
japhb Good night!
22:02 brrt left
jnthn Hoping to merge one of my postrelease-opts branches tomorrow :) 22:41
Well, my only branch with that name :)
japhb Oooh 23:29

Powered by Cro and the Raku® Programming Language.

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

Thank you!