2014-09-22T01:06:11Z zacts quit (Quit: leaving)
2014-09-22T01:16:37Z heddwch quit (Read error: Connection reset by peer)
2014-09-22T01:17:28Z heddwch joined #sbcl
2014-09-22T01:29:45Z wmarvel quit (Quit: Lost terminal)
2014-09-22T02:23:12Z kanru joined #sbcl
2014-09-22T02:38:30Z christoph_debian quit (Ping timeout: 258 seconds)
2014-09-22T02:52:46Z christoph_debian joined #sbcl
2014-09-22T02:53:23Z echo-area joined #sbcl
2014-09-22T02:55:18Z oleo__ joined #sbcl
2014-09-22T02:56:44Z oleo is now known as Guest6629
2014-09-22T02:58:05Z Guest6629 quit (Ping timeout: 260 seconds)
2014-09-22T03:30:15Z LiamH quit (Quit: Leaving.)
2014-09-22T05:23:44Z drmeister_ quit (Remote host closed the connection)
2014-09-22T05:54:46Z oleo__ quit (Quit: Verlassend)
2014-09-22T06:17:10Z pranavrc joined #sbcl
2014-09-22T06:46:32Z gingerale joined #sbcl
2014-09-22T06:47:52Z DGASAU quit (Read error: Connection reset by peer)
2014-09-22T06:48:26Z DGASAU joined #sbcl
2014-09-22T06:49:09Z drmeiste_ joined #sbcl
2014-09-22T06:50:22Z DGASAU quit (Read error: Connection reset by peer)
2014-09-22T06:50:56Z DGASAU joined #sbcl
2014-09-22T06:53:50Z drmeiste_ quit (Ping timeout: 244 seconds)
2014-09-22T07:08:30Z DGASAU quit (Read error: Connection reset by peer)
2014-09-22T07:09:07Z DGASAU joined #sbcl
2014-09-22T07:21:44Z prxq joined #sbcl
2014-09-22T07:25:22Z edgar-rft quit (Quit: you are not expected to understand this message)
2014-09-22T08:04:08Z leo2007 quit (Ping timeout: 260 seconds)
2014-09-22T08:07:15Z angavrilov joined #sbcl
2014-09-22T08:11:41Z leo2007 joined #sbcl
2014-09-22T08:19:17Z alchemis7 left #sbcl
2014-09-22T08:19:17Z alchemis7 joined #sbcl
2014-09-22T08:34:55Z attila_lendvai joined #sbcl
2014-09-22T08:34:55Z attila_lendvai quit (Changing host)
2014-09-22T08:34:55Z attila_lendvai joined #sbcl
2014-09-22T08:37:20Z drmeiste_ joined #sbcl
2014-09-22T08:38:17Z stassats joined #sbcl
2014-09-22T08:38:55Z stassats: cute, "no transport function for object 0x0641800f (widetag 0x53)"
2014-09-22T08:39:07Z stassats: during type.impure.lisp
2014-09-22T08:42:25Z drmeiste_ quit (Ping timeout: 272 seconds)
2014-09-22T08:52:54Z stassats: (that was because of my changes to destructuring-bind)
2014-09-22T08:55:55Z attila_lendvai quit (Quit: Leaving.)
2014-09-22T09:05:07Z segv- joined #sbcl
2014-09-22T09:32:50Z pranavrc quit (Read error: Connection reset by peer)
2014-09-22T09:33:05Z pranavrc joined #sbcl
2014-09-22T09:33:06Z pranavrc quit (Changing host)
2014-09-22T09:33:06Z pranavrc joined #sbcl
2014-09-22T09:41:24Z davazp joined #sbcl
2014-09-22T10:25:28Z drmeiste_ joined #sbcl
2014-09-22T10:26:37Z echo-area quit (Remote host closed the connection)
2014-09-22T10:30:00Z drmeiste_ quit (Ping timeout: 246 seconds)
2014-09-22T10:39:39Z hlavaty joined #sbcl
2014-09-22T10:45:08Z pranavrc_ joined #sbcl
2014-09-22T10:46:09Z pranavrc quit (Ping timeout: 272 seconds)
2014-09-22T10:51:17Z pranavrc_ quit (Remote host closed the connection)
2014-09-22T11:36:48Z minion quit (Disconnected by services)
2014-09-22T11:36:49Z specbot quit (Disconnected by services)
2014-09-22T11:37:33Z minion joined #sbcl
2014-09-22T11:37:35Z specbot joined #sbcl
2014-09-22T11:37:36Z minion quit (Disconnected by services)
2014-09-22T11:38:06Z specbot quit (Disconnected by services)
2014-09-22T11:38:10Z specbot joined #sbcl
2014-09-22T11:41:30Z specbot quit (Remote host closed the connection)
2014-09-22T11:41:57Z yacks quit (Ping timeout: 245 seconds)
2014-09-22T11:42:00Z minion joined #sbcl
2014-09-22T11:42:00Z specbot joined #sbcl
2014-09-22T12:33:24Z davazp quit (Ping timeout: 260 seconds)
2014-09-22T12:36:21Z nyef joined #sbcl
2014-09-22T12:39:56Z yacks joined #sbcl
2014-09-22T13:25:48Z eudoxia joined #sbcl
2014-09-22T13:28:12Z oleo joined #sbcl
2014-09-22T13:33:45Z eudoxia_ joined #sbcl
2014-09-22T13:36:57Z eudoxia quit (Ping timeout: 260 seconds)
2014-09-22T13:37:46Z eudoxia_ quit (Client Quit)
2014-09-22T13:50:29Z Xach joined #sbcl
2014-09-22T13:51:37Z Xach: hi friends. Any chance of my fix to https://bugs.launchpad.net/sbcl/+bug/1364413 slipping in?
2014-09-22T13:55:36Z stassats: as it's not a regression and doesn't affect anything important, i don't think it's worth including it now
2014-09-22T13:57:06Z Xach: ok.
2014-09-22T13:57:22Z Xach: I will be back when things thaw
2014-09-22T13:57:32Z Xach: in the sbcl spring
2014-09-22T13:57:41Z nyef: So... April?
2014-09-22T13:57:54Z Xach: sbcl spring happens 12 times per year
2014-09-22T13:59:30Z nyef: Most years, at least.
2014-09-22T13:59:42Z Xach left #sbcl
2014-09-22T14:22:07Z stassats quit (Ping timeout: 272 seconds)
2014-09-22T14:24:41Z drmeiste_ joined #sbcl
2014-09-22T14:25:58Z drmeiste_ is now known as drmeister
2014-09-22T14:27:03Z DGASAU quit (Remote host closed the connection)
2014-09-22T14:28:34Z DGASAU joined #sbcl
2014-09-22T14:30:28Z attila_lendvai joined #sbcl
2014-09-22T14:35:54Z eudoxia joined #sbcl
2014-09-22T14:48:20Z eudoxia quit (Quit: Lost terminal)
2014-09-22T14:50:55Z Krystof: Xach: ITYM "Look to my coming, at first light, on the fifth day"
2014-09-22T15:04:52Z slyrus quit (Ping timeout: 240 seconds)
2014-09-22T15:20:40Z karswell` quit (Read error: Connection reset by peer)
2014-09-22T15:30:19Z flip214: ahh, good ol' LoTR references.... why not LoΛ instead? ;)
2014-09-22T15:32:17Z flip214: I could offer "As we now know, nobody truly understands macros."
2014-09-22T15:41:02Z oleo is now known as Guest99867
2014-09-22T15:41:07Z DGASAU quit (Ping timeout: 245 seconds)
2014-09-22T15:41:51Z oleo__ joined #sbcl
2014-09-22T15:43:57Z Guest99867 quit (Ping timeout: 246 seconds)
2014-09-22T15:45:00Z DGASAU joined #sbcl
2014-09-22T15:55:16Z nyef: ... fd-stream can occasionally break on cheneygc? That's neat.
2014-09-22T15:57:55Z oleo__ quit (Quit: Verlassend)
2014-09-22T16:03:43Z oleo__ joined #sbcl
2014-09-22T16:03:58Z oleo__ is now known as oleo
2014-09-22T16:04:23Z prxq: I liked "come in, you can share my mess". Said somewhere in Gondor.
2014-09-22T16:09:26Z edgar-rft joined #sbcl
2014-09-22T16:11:50Z DGASAU quit (Ping timeout: 244 seconds)
2014-09-22T16:14:37Z DGASAU joined #sbcl
2014-09-22T16:17:00Z Bike: `#(#x3 #x5 ,@(loop repeat 4 collect 1)) = (apply #'vector `(#x3 #x5 ,@(loop repeat 4 collect 1))) => #(3 5 1 1 1 1), right?
2014-09-22T16:55:57Z DGASAU quit (Remote host closed the connection)
2014-09-22T16:57:52Z DGASAU joined #sbcl
2014-09-22T17:03:36Z nyef: Bike: Per CLHS 2.4.6, bullet 8, it should be, yes.
2014-09-22T17:08:46Z nyef: (Specifically, bullet 8 describes the input in terms of bullet 5, which is in terms of bullet 4, which... yeah.
2014-09-22T17:08:49Z nyef: )
2014-09-22T17:10:45Z Bike: well, you saw the #lisp discussion, i just thought it might be easily fixable enough that i could mention it here instead of filing a bug
2014-09-22T17:13:15Z DGASAU quit (Remote host closed the connection)
2014-09-22T17:15:19Z DGASAU joined #sbcl
2014-09-22T17:25:44Z nicdev quit (Read error: Connection reset by peer)
2014-09-22T18:21:56Z nicdev joined #sbcl
2014-09-22T18:46:23Z fikusz joined #sbcl
2014-09-22T18:47:11Z Krystof: nyef: only since we moved to memmove() in byte-blt
2014-09-22T18:47:18Z Krystof: when we did our own memmove it was fine
2014-09-22T18:52:11Z nyef: Ahh.
2014-09-22T18:53:16Z sdemarre joined #sbcl
2014-09-22T19:01:27Z pkhuong: nyef: re cheney and the write barrier
2014-09-22T19:01:47Z pkhuong: we should probably make sure to hit each page when allocating arrays.
2014-09-22T19:03:23Z pkhuong: We also had some bug reports on syscalls repeatedly failing on write protected arrays. Doesn't really happen now that we only mprotect boxed pages, but still a huge interop problem if we do pinned pages.
2014-09-22T19:03:30Z pkhuong: *pinned vectors
2014-09-22T19:03:47Z pkhuong: which is perhaps why pinned pages aren't mprotected after writes.
2014-09-22T19:07:59Z nyef: Yeah, there's that, but cheneygc is even more limited.
2014-09-22T19:10:30Z foom: what's the status of soft-dirty experiment?
2014-09-22T19:12:07Z nyef: Would we consider removing cheneygc if all of our backends supported gencgc?
2014-09-22T19:16:25Z fitzsim joined #sbcl
2014-09-22T19:22:11Z pkhuong: foom: it works
2014-09-22T19:22:55Z pkhuong: foom: but I don't have any reason to merge it in.
2014-09-22T19:24:32Z pkhuong: nyef: not sure. is cheney easier for a new port?
2014-09-22T19:26:56Z nyef: Yes, cheney is easier for a new port.
2014-09-22T19:27:19Z nyef: How often are we going to be creating a new compiler backend, though?
2014-09-22T19:28:05Z nyef: (I can name the last two times, being ARM and x86-64, but the third escapes me.)
2014-09-22T19:28:18Z jsnell: pkhuong: there's no performance improvement?
2014-09-22T19:28:57Z pkhuong: jsnell: nope.
2014-09-22T19:29:06Z pkhuong: or nothing on simple stuff like building cl-ppcre
2014-09-22T19:30:38Z Krystof: times before that were ports from cmucl
2014-09-22T19:30:46Z Krystof: we will someday create aarch64, though
2014-09-22T19:31:00Z jsnell: huh, that's unexpected. would've expected the trap cost to be significant, and the kernel's handling of fragmented memory maps hasn't exactly been optimal
2014-09-22T19:31:33Z jsnell: 10 years of wanting a kernel feature, and it turns out not to be useful after all :-(
2014-09-22T19:31:36Z pkhuong: last time I microbenchmarked this, the write barrier was ~1k-2k cycles per GC "page"
2014-09-22T19:32:02Z Krystof: do your benchmarks have "enough" heap fragmentation, whatever "enough" means?
2014-09-22T19:32:24Z pkhuong: soft dirty pages are probably slightly faster (one less user space <-> kernel transition), but can *only* do 4K pages
2014-09-22T19:32:28Z pkhuong: Krystof: probably not
2014-09-22T19:32:48Z pkhuong: I also had to switch to MADV_FREE to get a contiguous memory map.
2014-09-22T19:32:49Z foom: jsnell/pkhuong: i'd expect it'd allow us to use smaller pages.
2014-09-22T19:33:05Z pkhuong: foom: yup, switched to 4K GC pages, because there's no reason not to
2014-09-22T19:33:15Z foom: I believe that'll be useful.
2014-09-22T19:33:29Z pkhuong: that might have made allocation a bit slower; I'll have to play with the allocation region size
2014-09-22T19:33:49Z foom: what linux version does this first work in?
2014-09-22T19:33:55Z pkhuong: 3.11
2014-09-22T19:36:12Z foom: Great, I think I could actually use that, even. :)
2014-09-22T19:36:33Z Krystof: pff, 3.11 is so last year
2014-09-22T19:36:36Z pkhuong: unfair. 3.8.13 is the bleeding edge here ;)
2014-09-22T19:37:18Z jsnell: rocking on 2.6.32 ;-)
2014-09-22T19:37:27Z foom: So, I guess the downside of merging that support is that it requires there be two codepaths for the next 5 years or so, one with soft dirty and one with signals.
2014-09-22T19:37:47Z pkhuong: foom: we'd need signals for !linux/x86oids anyway
2014-09-22T19:37:50Z foom: Er right.
2014-09-22T19:38:05Z foom: Unless all the other OSes get soft dirty too, it'll be needed indefinitely.
2014-09-22T19:38:09Z Bicyclidine joined #sbcl
2014-09-22T19:43:03Z Krystof: jsnell: did you get any feedback on sb-eval2?
2014-09-22T19:43:26Z pkhuong: I'll dedicate some time to div by mul tonight.
2014-09-22T19:43:39Z foom: Did doug mention his new eval again?
2014-09-22T19:43:42Z jsnell: Krystof: no
2014-09-22T19:45:23Z Krystof: foom: in mail to me, yes -- in the context of new codewalker
2014-09-22T19:46:01Z Krystof: we talked about most excellent hacks to implement vtables for environments (and streams)
2014-09-22T19:46:34Z foom: Well, I meant in context of adding evaluators to sbcl.
2014-09-22T19:47:33Z Krystof: I gave him a heads-up that I wanted to do have lotso evaluators
2014-09-22T19:48:06Z Krystof: I don't want to cause him unnecessary maintenance pain, but...
2014-09-22T19:51:03Z segv- quit (Ping timeout: 246 seconds)
2014-09-22T19:51:08Z foom: Sorry, I didn't parse that
2014-09-22T19:53:57Z Krystof: he mentioned his new evaluator in the context of my making bugfixes to our current codewalker
2014-09-22T19:54:49Z Krystof: he also had a proposal for how to allow multiple evaluators (and associated support such as environments) to coexist without substantial efficiency hits
2014-09-22T19:55:28Z Krystof: which was particularly helpful in the light of lack of agreement on how to proceed with *either* current out-of-tree new evaluator
2014-09-22T19:55:39Z Krystof: and then I prodded jsnell
2014-09-22T19:55:56Z Krystof: to see if we could make progress on that one
2014-09-22T19:56:27Z foom: ah.
2014-09-22T19:56:52Z Krystof: I would like to see both of them merged, eventually; it seemed to me, unfair though it is, that since Google has a solid use case for dougk's evaluator, it was more likely to survive if the other was merged first
2014-09-22T19:57:02Z Krystof: and I acknowledge the pain that this will cause
2014-09-22T19:57:09Z foom: I guess I don't see what the purpose of merging both is.
2014-09-22T19:57:18Z angavrilov quit (Remote host closed the connection)
2014-09-22T19:57:38Z foom: I mean, it's quite unfortunate that doug didn't do his work like 2 months earlier, or at least tell anyone he was going to do that
2014-09-22T19:57:38Z scymtym_ joined #sbcl
2014-09-22T19:57:41Z Krystof: I think the purpose is to allow them to be tried out
2014-09-22T19:58:23Z Bicyclidine quit (Ping timeout: 240 seconds)
2014-09-22T19:58:29Z Krystof: my understanding is that they have different characteristics: that neither strictly dominates the other, nor indeed strictly dominates the existing one in all circumstances
2014-09-22T20:00:43Z foom: Well, I'd (perhaps naively) suggest that SBCL ought to have one evaluator, not two or three.
2014-09-22T20:01:02Z nyef: Like it only has one GC?
2014-09-22T20:01:09Z Krystof: it's already got 2½
2014-09-22T20:01:12Z foom: that'd be good, wouldn't it. :)
2014-09-22T20:01:17Z Krystof: we could remove 1½ of them, sure
2014-09-22T20:03:03Z Krystof: I'm put in mind of the N implementations in Lisp in Small Pieces
2014-09-22T20:03:37Z pkhuong: the way I've been approaching the evaluator situation is to try and merge them in trunk asap, let them see some real world use
2014-09-22T20:03:58Z Krystof: but even if you say SBCL should have one evaluator, and by that you mean one plus the compiler plus a codewalker, I still don't know which one
2014-09-22T20:04:04Z pkhuong: and hopefully one clearly dominates the other, after a couple tweaks
2014-09-22T20:05:07Z Krystof: Google has a use case.  Lots of other people have other use cases, and finding out about them and that they're not compatible with Google's after support for Google's has been made the only option would be really unpleasant
2014-09-22T20:05:08Z jsnell: back in spring of last year when we were considering the gsoc proposals I think we counted the partial and full compilers and interpreters sbcl already contained, and it was a substantial number
2014-09-22T20:05:19Z Krystof: oh yeah, I forgot the fop interpreter
2014-09-22T20:05:35Z Krystof: and I don't even know what cfasls are but they're probably something
2014-09-22T20:05:54Z nyef: cfasls are scary, and an indication of a poorly-designed build process.
2014-09-22T20:08:17Z jsnell: ignoring performance, the thing sb-eval2 does is debugger integration, while sb-new-eval does on-the-fly rather than up front macroexpansion
2014-09-22T20:08:42Z jsnell: (ignoring performance since dougk said it was about even between the two)
2014-09-22T20:09:42Z jsnell: it'd be a shame to lose the debugger support, but I understand the latter is an absolute requirement for google
2014-09-22T20:10:28Z jsnell: that said, I don't actually expect that a merge of one would be destructive and make merging or maintaining the other any harder
2014-09-22T20:11:15Z foom: It's just a maintenance burden to have N+2 evaluators, instead of having them replace one of the N. I'm surprised that you're all willing to take that on.
2014-09-22T20:12:18Z nyef: What if the interface were by way of the contrib mechanism or similar?
2014-09-22T20:13:24Z foom: Alternately, one could ask the contributors to do the remaining work to subsume the features of the others, like, "doug, you can merge your eval if you add <whatever is missing vs the existing eval, if anything> and debugger support."
2014-09-22T20:15:07Z foom: I mean -- I'm not doing either that work or the work of maintaining the 3 evaluators if they all get merged. So, if you really want 3, sure, whatever. I just wouldn't want that. :)
2014-09-22T20:15:17Z pkhuong: foom: I'm only willing to consider adding more evaluators if there's some deadline to prune the experiments back.
2014-09-22T20:24:33Z sdemarre quit (Ping timeout: 258 seconds)
2014-09-22T20:31:22Z Krystof: one of the nice things about multiple evaluators in particular is that they can be part of the great differential testing lab
2014-09-22T20:31:44Z Krystof: they're all meant to have the same semantics
2014-09-22T20:32:32Z Krystof: foom: debugger support was mentioned a year ago when this first came up
2014-09-22T20:33:18Z Krystof: then everything sort of stalled, at least from the point of view of my visibility
2014-09-22T20:33:37Z Krystof: wow that was a horrible broken metaphor
2014-09-22T20:34:52Z Krystof: (don't think I'm not grateful that people are contributing, and that Google in particular is allowing people to make their internal modifications public)
2014-09-22T20:38:37Z stassats joined #sbcl
2014-09-22T20:54:00Z Bicyclidine joined #sbcl
2014-09-22T21:22:52Z stassats quit (Ping timeout: 244 seconds)
2014-09-22T21:27:43Z yacks quit (Ping timeout: 272 seconds)
2014-09-22T21:35:47Z foom: Krystof: I think if it was clear that there was a technical issue/set of issues that could be resolved, doug would work towards that.
2014-09-22T21:36:19Z foom: I believe he stopped working on it earlier, because it was "good enough" for the limited internal need, and it did not seem clear that it'd ever be mergable.
2014-09-22T21:57:57Z drmeister quit (Remote host closed the connection)
2014-09-22T22:01:25Z Bicyclidine quit (Ping timeout: 260 seconds)
2014-09-22T22:02:21Z Bicyclidine joined #sbcl
2014-09-22T22:05:11Z drmeister joined #sbcl
2014-09-22T22:27:05Z prxq quit (Ping timeout: 260 seconds)
2014-09-22T22:38:28Z drmeister quit (Read error: Connection reset by peer)
2014-09-22T22:46:55Z drmeister joined #sbcl
2014-09-22T23:04:27Z Bicyclidine quit (Ping timeout: 245 seconds)
2014-09-22T23:09:29Z Bicyclidine joined #sbcl
2014-09-22T23:19:39Z davazp joined #sbcl