|
github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm Set by AlexDaniel on 12 June 2018. |
|||
|
00:48
MasterDuke joined
00:49
MasterDuke left,
MasterDuke joined
01:28
zakharyas joined
03:43
squashable6 left
03:45
squashable6 joined
05:13
discord61 joined
05:14
discord6 left,
discord61 is now known as discord6
05:20
discord62 joined
05:21
discord6 left,
discord62 is now known as discord6
06:16
robertle left
06:28
domidumont joined
06:58
domidumont left
07:08
zakharyas left
07:38
domidumont joined
07:42
domidumont left
07:43
domidumont joined,
domidumont left
07:45
domidumont joined
08:08
zakharyas joined
09:16
zakharyas left
10:01
robertle joined
11:54
brrt joined
|
|||
| brrt | good * #moarvm | 11:54 | |
| nwc10 | good *, brrt | 12:12 | |
| brrt | so, the HN discussion on my post was fruitful in that it gave me this link: github.com/eclipse/omr/blob/master...ToTrees.md | 12:24 | |
| which ehm, describes an IR that is very much like mine | |||
| masak | great minds think alike | 12:28 | |
| brrt | I'd say we read the same source :-D | 12:38 | |
|
12:42
squashable6 left
12:43
squashable6 joined
13:57
robertle left
13:58
robertle joined
14:05
lizmat left,
lizmat joined
14:10
lizmat_ joined,
lizmat left,
lizmat__ joined
14:14
lizmat_ left
14:20
lizmat__ left
14:29
lizmat joined
14:39
squashable6 left
14:42
squashable6 joined
14:50
squashable6 left
14:52
lizmat left
14:56
squashable6 joined
15:35
robertle left
16:11
domidumont left
16:51
TimToady left
16:52
TimToady joined
17:35
brrt left
17:37
lucasb joined
17:46
robertle joined
17:47
lizmat joined
17:48
domidumont joined,
lizmat left,
lizmat joined
17:59
brrt joined
18:15
lizmat left
18:21
lizmat joined
18:26
brrt left
18:36
ggoebel joined
18:53
patrickb joined
19:15
patrickb left
19:21
ggoebel left
19:34
lizmat left
19:36
lizmat joined
19:40
domidumont left
19:48
brrt joined
19:57
MasterDuke left
|
|||
| brrt | ohai #moarvm | 20:11 | |
| timotimo | ohai brrt | 20:12 | |
| brrt | ohai timotimo | ||
| I have been thinking and pondering and doing more thinking | |||
| Which is dangerous | 20:13 | ||
| But here's my idea. | |||
| timotimo | i've been paralyzed by my next task on the heap snapshot profiler - which should be a trivial thing to just code up, but my brain is being a donkey | ||
| so i started a blog post | |||
| brrt | oh, I listen | ||
| :-) | |||
| and maybe I can help | |||
| timotimo | i listen, too, though | ||
| brrt | Ok! | 20:14 | |
| what if I change the expression IR | |||
| timotimo | would you say the design of the data structure has ... exp IR ed? | ||
| brrt | from a DAG, value based, unordered structure, to an instruction-based structure | ||
| underlying data structure can stay mostly the same, I think. Maybe slightly simplified | |||
| we do most of the conversion in the perl template precompiler | 20:15 | ||
| ... or maybe not, maybe at runtime | |||
| we get real basic blocks... and we can make low-level blocks over more than one basic block | 20:16 | ||
| timotimo | ah, that's probably important :) | 20:17 | |
| low-level blocks would be the kind where we have conditions inside expr templates right now, right? | |||
| and the real basic blocks are the bbs we have in spesh already? | 20:18 | ||
|
20:20
lizmat left
|
|||
| brrt | yeah | 20:23 | |
| and it'd all be the same | |||
| timotimo | so they could even overlap? | ||
| or if there's a block over more than one bb it'd just be a merged bb? or split at different points? | 20:24 | ||
|
20:25
lizmat joined
20:30
brrt left,
brrt joined,
lizmat left
|
|||
| brrt | I lost connection for a second | 20:33 | |
| No, I really just mean that an entire MoarVM routine could be in a single IR structure | |||
| rather than one tree per basic block as is the case now | |||
| timotimo | oh, ok | ||
| well, it does sound like a rosy future | 20:37 | ||
| brrt | hehe :-) | 20:38 | |
| it sounds like a bunch of work.... but nothing that can't be managed | |||
| it sounds like converging with what guile scheme also does | |||
| timotimo | but of course we can't just steal half of that, huh? | 20:44 | |
| brrt | if only | 20:53 | |
| guile has its compiler written in scheme iirc | |||
| timotimo | isn't scheme one of the easiest languages to implement? :D | 20:54 | |
| maybe forth could be easier? | |||
| brrt | forth would be easier. But scheme is fairly simple | 20:56 | |
| oh, I see your point :-D | |||
| yeah, well, if there's no GSoC student interested in the spesh/jit testing tool, I'll be writing it myself | |||
| (and it would, in a way, be a scheme) | 20:57 | ||
| or almost-a-scheme | 20:58 | ||
| timotimo | there'll have to be a pun on that for the name | 21:01 | |
| now how do i visualize my data format ... | 21:03 | ||
| the graphviz approach i tried wasn't really satisfactory | |||
| brrt | graphviz didn't work? | 21:05 | |
| :-o | |||
| discord6 | <timotimo> cdn.discordapp.com/attachments/557...nknown.png | 21:06 | |
| <timotimo> i'm not too happy with the look | 21:07 | ||
| brrt | oh, I kind of like it | 21:10 | |
| timotimo | i'm considering maybe taking a screenshot (or even html dump) of a hex editor and coloring it in or something | ||
| brrt | I'm slowly coming around to appreciating rust | 21:14 | |
| masak | it's nice to see you talking about converging on Scheme | 21:16 | |
| I think I understand the context, but even without it, I'm just a fan of the idea ;) | |||
| brrt | :-) | ||
| I guess what happens is that in both our case and in their case, we're moving slowly from a high-level language with polymorphic primitives, via a low-level language to assembly | 21:19 | ||
| timotimo | masak: got a hot tip on how to visualize the interior design of a file format? | ||
| masak | something like a collapsible hierarchy with hyperlinks in it? | 21:22 | |
| which file format are we talking about? moarvm bytecode? | |||
|
21:23
domidumont joined
21:27
domidumont left
|
|||
| timotimo | nah, the new heap snapshot format | 21:33 | |
| it should be more like a diagram of how the format is designed rather than an actual file that you can scroll through | 21:34 | ||
| brrt | I don't mind ascii graphs for that purpopse | 22:02 | |
| timotimo | i wonder what blind people would prefer in this case | ||
| an html table would let them | |||
| probably let them* navigate cell-to-cell comfortably | |||
| but a braille display might be better with an ascii diagram? | 22:03 | ||
| annoyingly, my browser recently crashed, and ghost hadn't saved a draft since i started the latest paragraph | 22:08 | ||
| brrt | :-( | 22:21 | |
| I don't think a braille display would work well with ascii. But I don't know | |||
| timotimo | since i'm randomly fascinated by blind computer usage, i should get a braille display. except those things are incredibly expensive | 22:27 | |
|
22:32
lizmat joined
|
|||
| brrt | moar blog: brrt-to-the-future.blogspot.com/20...on-is.html | 22:52 | |
| about how LuaJIT's register allocation algorithm is cool | |||
| and how we should totally steal that | |||
| timotimo: they are | |||
| I once had a program called 'zoomtext' | 22:53 | ||
| This was paid for me by.... some government agency | |||
| It was *ridiculously* expensive | |||
| fortunately zooming is built-in on most platforms now | |||
|
23:09
Geth left
23:17
brrt left
23:24
MasterDuke joined,
MasterDuke left,
MasterDuke joined
|
|||
| timotimo | oh lord | 23:25 | |
| what exactly did it do?! just hook into every kind of program and turn font sizes up? | |||
