github.com/moarvm/moarvm | IRC logs at irclog.perlgeek.de/moarvm/today
Set by moderator on 22 May 2018.
00:11 lizmat joined 00:29 Kaiepi joined 00:30 Kaiepi joined 00:32 Kaiepi joined 00:43 Kaiepi joined 01:57 ilbot3 joined
moderator github.com/moarvm/moarvm | IRC logs at irclog.perlgeek.de/moarvm/today
02:00 Kaiepi joined 04:19 shareable6 joined 06:01 shareable6 joined 06:24 domidumont joined 06:31 domidumont joined 06:58 robertle joined 07:10 brrt joined
brrt good * 07:18
yoleaux 22 May 2018 20:54Z <samcv> brrt: want to look over docs/ChangeLog and see if any changes need to be made before release?
brrt will do 07:19
07:47 shareable6 joined 09:05 brrt joined 10:34 kiwi_14 joined
brrt i have a patch for DynASM that i'm decently sure is correct 10:39
but i don't want to put it out there just before the release 10:40
11:07 zakharyas joined
samcv i kind of want to throw out uthash.h since it's a whole mess of macros but i'm not sure yet 12:35
obv not going to throw it out before release. but i wanted a more minimal thing that used less memory 12:36
and would probably be faster
but would still be the same sort of hash table
timotimo oh? use more functions rather than putting all the code all over the place because of macros?
samcv yeah. well probably inlined
but i don't like using like one macro that has like 5 macros to go through
it's a huge pain
timotimo fair enough 12:37
samcv when i try and make code changes it's a huge headache
but time will tell if it actually happens :)
nwc10 samcv: last I knew jnthn's suggestion to investigate for a "complete replacement" (which I don't think is what you're suggesting here) was irclog.perlgeek.de/moarvm/2018-03-20#i_15943808
samcv also wanted to make it a singly linked list not a doubly linked list to save some memory and remove other data structures that we don't need for our strings
nwc10 ooh, which remdinds me 12:38
ilmari: irclog.perlgeek.de/moarvm/2018-03-20#i_15943843
samcv also i was watching this www.youtube.com/watch?v=ncHmEUmJZf4
12:38 shareable6 joined
samcv google hasn't released the source yet though. but it was interesting 12:38
nwc10 oooh
(the title) 12:39
timotimo but we don't have cpp :)
samcv yes they also have not released the source :P
but it was a very interesting watch
also i'm still not 100% sure how they implemented everything since obv it's just a talk. though they described it fairly well given the format 12:40
timotimo i'm glad he has the "restofthefuckingowl" slide in there 12:41
there is a subreddit for that concept
with that exact name
i wonder how people were using the load factor api wrong, he says at around 6:30 that every single use inside of google of that api was wrong 12:44
i'm amused by the symbol he uses for the "footnote" in some of the slides keeps changing 12:46
brrt samcv: that reminds me, maybe you've seen this link: lemire.me/blog/2018/05/09/how-quic...ode-utf-8/ 12:54
so, my plan of killing invokish and friends has stalled on two things;
one, there are some ways in which we can enter a jit frame but not have a spesh candidate 12:55
(that way being the nativecall JITting)
two, there are some ways in which we can move between frames that aren't nicely embedded in frame.c and so i missed; continuations being the chief thing i'm thinking off 12:56
there are probably more. closures, maybe 12:57
timotimo but closures are just ... ok i see from the beginning of this sentence that it's probably wrong %) 13:06
jnthn samcv: I don't think we want it as a linked list at all, but just one single blob of memory
samcv yeah of course 13:07
that would be faster
jnthn Well, also, it'd make it massively easier to deal with the current explosiveness around concurrent use of MVMHash
(We don't have to give sensible results, but we shouldn't SEGV)
timotimo i'm just beyond half-way through that talk, at a slide that's "84% true", and it looks like a flat approach already
with probing assisted by a little piece of metadata in front of "groups" of actual data where a full probe of the data can be done in ~3 sse instructions 13:08
neat. he says "it's in allcaps, and caps scare people. true story." and that's kind of how perl6 decides whether to allcaps something or not 13:09
"should the reader be slightly afraid when seeing this"
samcv i'm not sure if the microphone didn't capture anyone laughing or if the crowd was just humor dead... 13:15
but i thought some of his jokes were funny
timotimo i'd assume the mic is just very good at reducing background noise, because i'm also laughing every few slides 13:16
aha, the load factor api is "setMaxLoadFactor" it seems like 13:27
oh youtube search 13:34
i entered his name in the search bar and what do i get? first this very talk, then a whole bunch of other cppcon2017 talks by different people 13:35
i'm not quite sure how it figures those are matches; perhaps every video description contains a list of all speakers or whatever
kind of like how people would put about ten thousand words at the end of their video descriptions to make them more likely to show up for unrelated searches 13:36
13:59 domidumont joined 14:22 shareable6 joined 14:35 brrt joined
samcv brrt: also did you look at the changelog? 15:44
16:04 Kaiepi joined 16:08 shareable6 joined 16:53 robertle joined 17:18 domidumont joined 17:53 ggoebel joined 17:57 shareable6 joined 18:30 ggoebel joined 19:00 robertle joined 19:38 shareable6 joined 20:10 Ven`` joined 20:16 Ven`` joined 20:40 Ven`` joined 21:12 Kaiepi joined 21:43 Ven`` joined 22:52 ggoebel joined
AlexDaniel samcv: release? :) 23:18
samcv AlexDaniel: sounds good. don't see any people telling me there are issues with the changelog as it stands 23:45
AlexDaniel: running my final checks now then 23:50
AlexDaniel sounds good 23:51

Powered by Cro and the Raku® Programming Language.

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

Thank you!