Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2015-03-31

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
01:48 ilbot3 joined #moarvm
01:48 Topic for #moarvm is now https://github.com/moarvm/moarvm | IRC logs at  http://irclog.perlgeek.de/moarvm/today
06:37 FROGGS joined #moarvm
07:09 zakharyas joined #moarvm
07:50 rurban_ joined #moarvm
08:07 kjs_ joined #moarvm
08:08 brrt joined #moarvm
08:08 brrt \o
08:14 Ven joined #moarvm
08:19 jnthn o/
09:23 brrt joined #moarvm
09:43 brrt rudi_s - your patches Work For Me
09:44 timotimo o/
09:44 brrt \o timotimo
09:58 brrt left #moarvm
09:58 brrt joined #moarvm
09:58 brrt remind me to write a udp-proxy-for-irc one day
09:59 jnthn So you never know if your messages made it? :)
09:59 timotimo why you no irc.p6c.org? :)
10:00 brrt is that a new server?
10:00 brrt that's what message sequence numbers are for jnthn :-)
10:00 timotimo a new vm on p6c.org
10:00 brrt ah
10:00 brrt i am in general not connected to any server at p6c.org
10:00 timotimo you could be :)
10:01 timotimo you can do all your compiling and testing on hack.p6c.org
10:02 timotimo etc etc; also, host stuff on www.p6c.org
10:03 jnthn brrt: You can run irssi or so in a screen session on irc.p6c.org, which means you can be on IRC "all the time" :)
10:05 timotimo i recommend weechat actually
10:05 timotimo then you can set up glowing-bear and have a pretty frontend
10:06 tadzik glowing-bear?
10:06 tadzik that sounds fun
10:07 timotimo http://www.glowing-bear.org/
10:07 tadzik Ha, pretty cool!
10:08 timotimo as far as i know the name comes from the github repo name suggestion feature
10:09 tadzik hah, I have a project like that
10:09 timotimo panda? :P
10:09 tadzik cloaked-hipster
10:10 timotimo %)
10:10 brrt nmany wow
10:11 tadzik There's nothing in there, I just saw the suggestion and called dibs on the name :P
10:11 timotimo you called dibs on the name ... inside your own namespace?
10:11 tadzik that's right :)
10:11 tadzik you can never be too careful
10:12 masak it's cloaked. of course it appears empty.
10:12 tadzik ha!
10:12 masak the main thing is that the repo was there *before* it was mainstream.
10:12 masak if a hipster is hip under a cloak, is he really hip?
10:13 timotimo what if every country has ninjas, but the only ninjas that are bad enough to let themselves be seen every now and then are the japanese?
10:13 masak timotimo: now I'm imagining Danish ninjas
10:44 Ven joined #moarvm
11:02 rudi_s Regarding the %zu vs. %Iu discussion from yesterday, I pushed a (untested) fix for MVSC: git://git.informatik.uni-erlangen.de/he29heri/moarvm/
11:06 brrt thanks. i'm in no position to test it on msvc now
11:11 rudi_s np
11:11 jnthn On MSVC building rudi_s/master:
11:11 jnthn interp.c
11:11 jnthn src\core\interp.c(4892) : error C2094: label 'runloop' was undefined
11:11 rudi_s Hm. Then I'm confused.
11:11 rudi_s Better drop this patch then.
11:12 jnthn 642e9ca86ea7ad1 is to blame
11:13 jnthn Probably wants to be #if !MVM_CGOTO or so
11:13 jnthn err&
11:13 jnthn (bbi30)
12:35 Ven joined #moarvm
13:25 FROGGS I might get my hands on an Intel Itanium system /o/
13:25 nwc10 oh, gosh
13:26 nwc10 heavy and warm?
13:27 nwc10 what OS(es) would you run on it? VMS, HP/UX, Linux, there-must-be-a-BSD-surely, Win* still?
14:15 FROGGS I've no idea yet :o)
14:16 FROGGS bbiab
15:16 Ven joined #moarvm
15:44 FROGGS joined #moarvm
16:39 kjs_ joined #moarvm
19:58 brrt joined #moarvm
20:13 kjs_ joined #moarvm
21:48 flussence re: %z/I, take a look at inttypes.h. I dunno if MSVC has it though...
21:53 brrt anybody know a good at-t syntax assembler for linux?
21:57 flussence brrt: of the three I've got installed (llvm, nasm, gcc), only gcc's one seems to support that. I have no opinion on the relative goodness of it :)
21:58 brrt hmm
21:58 brrt gcc it will be then
21:59 flussence (I wonder what the BSDs do there, a few of them have been doing away with gcc)
22:05 rudi_s flussence: There is nothing in the standard inttypes, because %zu is the standard format for it.
22:19 vendethiel joined #moarvm
22:38 brrt ok, i estimate that in the scenario of tight numeric / integer code , the speedup factor can be as large as 6x
22:38 timotimo oh my god
22:38 timotimo for your proposed jit work?
22:38 brrt wherin estimate = measure :-)
22:38 brrt yes
22:38 timotimo that'd be *sweet*
22:38 brrt i'll put a gist online
22:38 timotimo way cool.
22:38 timotimo another factor of 6 is just what we need :)
22:39 timotimo but for code that's very "general purpose", we'll have a much smaller improvement?
22:39 brrt depends
22:39 brrt long story very short
22:39 brrt mvmarray access is basically offset + stride * position right?
22:40 brrt well, that can be a single instruction
22:40 timotimo read access, yeah
22:40 brrt and we can tell the cpu to prefetch a line
22:41 brrt but whether it matters a lot for 'real' code depends on whether we can eliminate / reduce some of the c calls
22:41 brrt or 'inline' them if they're repr ops
22:41 timotimo right
22:42 timotimo that's quite a hard thing to do until we've got the jit tree expression thingie
22:43 brrt on the other hand, i expect 'real' code to become quite a bit smaller, maybe 30% or so
22:43 brrt and that may count a lot too
22:45 timotimo mhm
22:45 timotimo the more we can keep in the instruction cache, the better
22:46 brrt right
22:47 timotimo we may not necessarily be at the point where that makes a huge difference, but we might as well reach it :)
22:57 brrt what's important to me is that we *can* reach it :-)
23:05 brrt https://gist.github.com/bdw/fef76ca07b7203e49fc2
23:06 brrt timotimo ^ if you're interested in trying it out
23:45 flussence joined #moarvm

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary