Index

Show enters and exits. Hide enters and exits.

02:58:20manveruevan: found a bug in rbx... i think
03:00:00manveruevan: http://gist.github.com/504042
03:00:51manveruactually, two bugs if you count 1.8 compatibility :)
03:52:22evanmanveru: looks almost like we fixed a bug that 1.8 has.
03:52:49evanon rbx I get "Go Go Engine Speed Up!"
03:52:59evanon 1.8 "Go Go Engine Speed Up!" and then a LocalJumpError
05:53:49boyscoutAdd aux word to header, reorganize inflation scheme for locking - ecf46b1 - Evan Phoenix (hydra)
05:53:49boyscoutAdd per object locks as Rubinius.(un)lock - e5fcd18 - Evan Phoenix (hydra)
06:39:59dbussinkevan: still there?
07:47:45manveruevan: no sure what's going on there, 1.8/1.9/rbx behave different
13:50:37goyox86morning
13:50:46jakedouglasgood morning goyox86
18:32:14dbussinkevan: ping?
18:32:27brixendbussink: I think evan is traveling today
18:34:55sbryantHow is hydra coming along?
18:35:15dbussinkbrixen: ah ok, np :)
18:38:38brixensbryant: swimmingly
18:48:16sbryantawesome.
21:57:03seydarhey all
21:57:20Defileryo!
21:57:31seydarguess what kind of CPU I'm using right now
21:58:08seydarI'll just tell you because you'll never guess in a million years
21:58:10seydaran intel!
22:00:58Defilerzomg
22:01:20evanhi guys.
22:01:47seydarhey evan
22:05:39brixenevan: did you make it to montana?
22:05:53evanI had up on wednesday
22:05:55brixenseydar: woot! now get busy son, no more excuses about running rbx
22:05:57evanlittledan is here today
22:06:01brixenahh, fun
22:06:25seydarwho's littledan?
22:06:41seydarwait
22:06:49lopexevan: how do you deoptimize on "unused" local closure var via eval ?
22:06:59seydaroh, we're facebook friends and both know mike edgar
22:07:11evanlopex: i don't follow.
22:07:34lopexevan: assume not used closure var
22:07:57evani'm not sure what you're refering to.
22:09:09lopexevan: unused local var from enclosing scope
22:11:49evanlopex: please start over
22:11:55evani'm not sure what you're asking.
22:12:27lopexevan: oh, the famous binding eval
22:13:31evanyou mean using a proc as a binding?
22:15:11lopexevan: yes ?
22:15:26evanok, what about it?
22:16:16lopexdo you keep all closure var around ?
22:16:21lopexvars
22:16:30evanyes
22:19:51lopexevan: so do you have to bail out to interpreter when eval wants to access them ?
22:20:14lopexI assume they're not native compiled..
22:24:56evanno, why would I?
22:25:09evanthe locals that eval can access are in a heap data structure
22:25:14evanit's got nothing to do with the JIT
22:25:58evanJIT'd code can't assume anything about a local after an external call
22:26:02evanso that eval could change it
22:26:06evanand the JIT won't break.
22:27:37lopexso the jit still refers to the locals via local table ?
22:28:08evanyep.
22:28:25lopexok then
22:29:46lopexevan: but with further optz they could end up registers right ?
22:29:54evanno.
22:29:58evanthey have to be GC'd
22:29:59lopexwith deopt mechanism...
22:30:05evanthey can be in registers temporarily
22:30:17lopexwrt AA ?
22:30:18evanbut they have to be flushed to memory between calls
22:30:24lopexoh
22:30:25evanso that the GC can see them.
22:30:59lopexso it seems I have misunderstood a principal here :/
22:31:15evanok
22:31:40evanthe JIT has to be conservative wrt globally visible memory
22:31:45evanbecause the GC is accurate.
22:31:53lopexyeah I see that now
22:32:15lopexand it's not just threads
22:34:07lopexevan: you'd need a hardware event that is called when cache is flushed os something
22:34:50lopexs/os/or/
22:35:16evani guess?
22:36:05lopexwouldnt it be easier ?
22:37:42lopexI wonder how long it takes to be as smart as mental
22:37:58lopexand how much IQ you need :)
22:38:56seydarwell perspective is worth 80 IQ points, so that's a good place to start
22:39:16brixenwow, look at the smarts on seydar :D
22:39:22brixen^5 seydar
22:39:33seydar^5 brixen
22:39:55brixenseydar: you are wize beyond your years, or you've been playing us :)
22:39:59brixener wise
22:40:00lopexperspective ?
22:40:04brixenI can haz spelling
22:40:16seydarrunning kicked my ass into shape. now i'm no longer 15
22:40:20brixenlopex: knowing how to look at a problem
22:40:46lopexbrixen: but that requires knowledge
22:40:51slavahi brixen
22:40:55brixenslava: hi!
22:41:07lopexbrixen: that couldnt be posessed without certain IQ level
22:41:08brixenslava: I was wondering if the goog had absorbed you
22:41:16brixenlopex: perhaps
22:41:55brixenlopex: I agree that wisdom often comes from experience, but also, thinking differently about a problem can expose valuable avenues of discovery
22:42:13lopexoh yes
22:44:02seydarflees to run
22:44:16seydarpeace children. I'll install rbx tonight and start getting back into the swing of it shortly
22:44:29brixenseydar: sweet!
22:44:31seydarit's tough with a 9-5 job though
22:44:42seydarjob + running kills time
22:44:49brixenseydar: yeah, remember, being a kid is more important at this point
22:44:53brixenenjoy yourself!
22:45:32BrianRice-workbrixen, that reminds me about your harley tweet. that happens often?
22:45:54brixenBrianRice-work: omfg, yesterday at brunch in the NW was *straight* out of the south park episode
22:46:02brixenI was getting literally pissed off
22:46:04BrianRice-workhuh
22:46:21BrianRice-workwell yeah that's pretty life-threatening
22:46:27brixenBrianRice-work: nw 23rd is a magnet for that shit
22:46:29lopexslava: oh, with your intelligence you should be working on HGP or something
22:46:36slavaHGP?
22:46:45lopexhuman genome project
22:46:55BrianRice-workI dunno what to tell you except that harleys seem to correlate with asshole behavior. see also south park, of course.
22:47:10lopexslava: very hard problem related to proteomics and geen mapping
22:47:14BrianRice-workFactor as bioinformatics language?
22:47:16lopex"gene"
22:47:22brixenBrianRice-work: the encroaching on the bike lane on hwy 30 was so blatant though
22:47:49brixenBrianRice-work: and not one of the groups of sport riders in packs of 3 or more were anywhere near me when they passed
22:48:27brixenBrianRice-work: one harley douche actually reved his engine right next to me while going a const 50 mi clip, no one was shifting, slowing down, speeding up
22:48:31brixeneffen douche bags
22:48:38lopexslava: are you interested in evolutionary studies ?
22:48:42BrianRice-workugh
22:58:16lopexBrianRice-work: the thing is that we still know almost nothing after revealing the genome
23:00:19lopexand bionics could help designing computer systems too
23:00:57lopexincluding the work that's being done here
23:02:04brixenlopex: where would one start investigating that?
23:02:33brixenlopex: do you have a good primer on bioinformatics?
23:04:04lopexbrixen: bioinformatics seems to be just a tool to analyze the genes, but there's much more
23:06:32lopexbrixen: take that one for example: http://en.wikipedia.org/wiki/Evolvability
23:06:52lopexanyone looks at it as a sequence of genes
23:07:01lopexwhich is too simplistic
23:07:22lopexif a thing evolves any aspect of it evolves
23:07:41lopexincluding the whole aparatus that seems to drive that
23:08:29lopexso it's not as simple as genetic algorithms
23:09:23brixenhm, that evolvability is a characteristic of systems that evolve isn't exactly surprising
23:10:56lopexwell, yeah, but imagine (whith it) there's no easily defined information that undergoes evolution
23:11:42lopexbrixen: for example the genes that encode the protein include the knowlegde of the chemistry
23:11:55lopexand real world we live in
23:12:10lopexand their constraints
23:12:23lopexer, it's
23:12:42brixenyes, very interesting
23:14:10lopexbrixen: then we cannot draw any knowlengde from here since we know knothing why the real world behaves as it does
23:14:30lopexlike "why gravity works"
23:14:34brixenlopex: ParseError
23:14:38brixenoh, I see
23:14:41brixenwhy it works...
23:14:46lopexyeah "why"
23:14:50lopexwe're on to it
23:14:56lopexbut we're not there yet
23:14:58brixenI had a physics prof that loved to ask the why question to trick us
23:15:23brixen"why is the inverse square law having a 2 here?"
23:15:53lopexbut evolutionary systems might converge on "evolvibility"
23:16:11brixeninteresting
23:16:29lopexlike http://en.wikipedia.org/wiki/Canalisation_%28genetics%29
23:16:58lopexlike producing distinguished organgs
23:17:03lopexorgans
23:17:57lopexthere's a lot of such phenomena
23:19:26brixenyeah, interesting
23:20:05brixenI was very interested in this in the mid '90s http://en.wikipedia.org/wiki/Tierra_%28computer_simulation%29
23:20:25brixenbut I've done about zero with anything like this sinces then
23:20:28brixen-s
23:21:26lopexI got interested in all that when working on a degree comparing ant systems and genetic algorithms
23:21:34BrianRice-workearly 90's a-life was a fun field
23:24:59lopexyeah, it was a fun thing to do back then
23:26:10lopexthough very few realize that gnetic information carrier has undergone the same evolutionary process that the information that it carries
23:26:45lopexthe genetic code has evolved itself too
23:26:52lopexand so on
23:27:14lopexwith more or less harsh constraints, but anyways..
23:28:50brixenlopex: so is there a good text to get up to speed on all this?
23:29:26lopexbrixen: it's a bit scattered, but starting on wiki would be a good I thing
23:30:16lopexbrixen: also http://www.illigal.uiuc.edu/web/ has good papers on it wrt CS
23:31:08lopexbrixen: depends what youre interested in
23:31:12brixenok, cool
23:31:35brixenilligal looks like a good start
23:31:39lopexbrixen: many papers are from the viewpoint of biologists
23:31:47brixenstarts doing some illigal research :)
23:31:49lopexbut they seem to converge
23:32:59lopexbrixen: when genetics alrorithms started it was a Holland disertation
23:33:13lopexnow it's all changed just like with genomics
23:33:24lopexbut from information theory standpoint
23:33:30evanback.
23:36:14evanlopex: did I answer your questions about eval?
23:36:22evanhardware caches don't really have anything to do with it.
23:36:24lopexevan: yeah, thanks
23:43:02lopexbrixen: though, evolved systems will never be as good as designed ones - what counts it that they can discover things/rules in real world we can miss
23:43:11lopexthey had a lot of time to do it
23:43:43brixenlopex: seems that would be very hard to prove: "evolved systems will never be as good as designed ones"
23:44:28brixenlopex: seems that evolved systems are beter at evolving than designed ones
23:44:37brixen+t
23:44:57lopexbrixen: yeah, well, I'm using colloquial "disigner" term which you'd have to define first
23:45:09brixenindeed :)
23:45:37brixenlopex: after all, we designers are just evolved systems :D
23:45:37lopexand the term meaning "designed" as well
23:46:05brixenlopex: so design is merely a process of certain evolvable systems
23:47:00lopexbrixen: I think it's a term created by purpose-driven thinking
23:47:14lopexbrixen: which again itself is subjective
23:47:32lopexand ultimately meaningless
23:47:53brixenindeed, trying to exclude subjectivity leats to tail chasing
23:48:00brixener leads
23:48:13brixenhate coffee house wifi
23:48:23brixeninvariably sucks
23:51:03lopexbrixen: things are just as they are, we're no different than a falling stone ultimately, just to maximize entropy :)
23:51:26lopexwhat we'd have to define first is complexity
23:56:22brixenlopex: can we use a constructive definition? universe :)
23:56:53evanNEWFLASH: Threads are tricky. More at 11.
23:58:04brixenevan: wait wait, what?! :)
23:58:12evantotally.
23:58:15evanwho would have thought?
23:58:50slavahi evan
23:59:03evanhey there.
23:59:06lopexbrixen: to compress earlier statements ?
23:59:13slavais rubyspec set up so that you can run specs in multiple threads at the same time?
23:59:18slavathat would be a good way to shake out VM thread safety issues
23:59:32evanyeah, it would be