06:23 rurban_ joined 06:30 domidumont joined 06:35 domidumont joined 07:03 domidumont joined 07:07 cognominal joined 07:17 zakharyas joined 07:26 zakharyas joined 07:33 zakharyas joined 09:03 zakharyas joined 09:27 edehont joined 10:14 ggoebel116 joined 11:22 brrt joined
brrt good * #moarvm 11:28
jnthn++ very nice writeup
jnthn o/ brrt
nwc10 good UGT heresy, #moarvm 11:29
brrt :-) 11:31
maaaybe i can start writing the tile editor today
and when i have that, i kind of know how to complete the register allocator
lizmat ++brrt :-) 11:35
jnthn nice :)
jnthn is in meetings today, alas
11:41 cognominal joined
brrt has to happen, unfortunately 11:51
nwc10 hopes meetings have coffee
jnthn Well, I'm joining them on Skype, so I can have what coffee I want :)
nwc10 encrypted.google.com/search?q=coffee+stout 11:58
(or cut to the chase, ie KahlĂşa) 11:59
12:28 brrt joined
timotimo oh, tile editor? sounds like a user interface design thing :) 13:09
13:18 zakharyas joined 13:34 domidumont joined
brrt no, it's a structure for delaying the modification of the tile list 13:49
which, among other niceties, ensures that ranges don't get shifted while the list is being processed 13:50
timotimo oh
nice
brrt it's a copy of the 'insertionset' idea from the b3 compiler backend 13:51
only, i originally had planned a (rather elegant if i say so myself?) plan wherein i'd store pairs of before-after invariants 13:52
the 'do-before-this' invariant is very powerful, of course, but it'd be most useful in the case of arglist processing 13:53
and that is easier to deal with in other ways
timotimo hm, k 13:56
13:57 rurban_ joined 14:38 edehont joined 14:52 dalek joined
timotimo there's a comment about the latest jnthn-report on reddit that seems to think the class jnthn wrote up "just uses a scalar" to store the value 14:58
www.reddit.com/r/perl/comments/4nt...rovements/
psch i like "it doesn't say who wrote this" when the about page is clear that only one persons posts there anyway 15:02
jnthn Well, Perl 6 attribute storage is done in terms of the P6opaque representation, which doesn't have a Perl 5 equivalent. 15:03
Trying to emulate it as a Scalar is sorta cute but, as noted, prevents storage of anything more than one attribute. 15:04
And mixins. :) 15:05
psch ref bless OO in perl5 can do mixins?
jnthn psch: If you blessed a Scalar though, you can't then go and tack on additional storage? 15:06
Which means you can only mix in stateless things.
psch jnthn: uh, i was just wondering if it's generally possible. all i know of perl5 core OO is skimming perlootut (i think?) a few years back :S 15:07
jnthn psch: iirc you can rebless a reference into a new package, which would be enough
psch showmetheco.de/articles/2014/12/mix...-perl.html is something i found with google, which seems to really miss the point of mixins 15:09
although i suppose if all objects are inside-out it's perfectly sensible to do it like that..? :/
jnthn But anyway, the point was to show that we're in the same ballpark for the code a typical programmer would likely write.
psch right :)
moritz
.oO( why do I want to read "show me the taco" into that domain name? )
jnthn moritz: hah, wow :) 15:10
timotimo how much more expensive would it be to go via a rw accessor?
jnthn timotimo: Why would we do that?
timotimo to make it more fair vs the perl5 code :) 15:11
jnthn Huh?
I don't go via an accessor in the Perl 5 code?
timotimo just spitballin'
jnthn I access the state directly
timotimo but it's hash-backed, so it's not direct
but ... what do i know about perl5 :)
jnthn Sure, but it's not a method call either :) 15:13
timotimo right
15:43 TimToady joined 16:41 cognominal joined 17:32 cognominal joined 17:40 domidumont joined 17:49 domidumont joined 17:54 harrow joined 18:11 rurban_ joined 18:23 FROGGS joined, vendethiel joined
FROGGS o/ 18:24
nwc10 \o 18:28
18:33 sivoais joined 19:01 cognominal joined 19:31 brrt joined
brrt ouch, much negativity on reddit 19:36
FROGGS I just wont read it 19:37
brrt good idea 19:38
nwc10 about jnthn's bloggage? 19:46
(I'm also not gong to read it. There are a lot more folks with opinions above their competance.
Sadly evidence of Dunning–Kruger keeps reappearing) 19:47
jnthn Nothing especially negative 19:58
Just a bit of missing the point and stating the obvious (e.g. "gee, you folks seem under-resourced" :P ) 19:59
brrt hmm, maybe you're right, and maybe i overreacted
yes, we are. come help :-)
FROGGS the comments are quite nice actually 20:22
dalek arVM/even-moar-jit: f0c9a86 | brrt++ | src/jit/register.c:
Add tilelist editing functionality

This will allow inserting tiles in the linear list with a fixed cost per tile and without upsetting ranges prior to the edit.
20:34
22:18 cognominal joined 23:21 cognominal joined