2016-09-08T00:01:53Z jrm left #sbcl
2016-09-08T00:18:18Z abbe joined #sbcl
2016-09-08T00:21:13Z sjl joined #sbcl
2016-09-08T00:34:06Z sjl quit (Read error: Connection reset by peer)
2016-09-08T00:36:56Z cromachina joined #sbcl
2016-09-08T00:42:21Z abbe quit (Ping timeout: 276 seconds)
2016-09-08T00:43:23Z abbe joined #sbcl
2016-09-08T00:53:39Z nyef`: stassats: If you're still poking around at that substitute-let-funargs thing, it looks like lp#1203260 might be relevant.
2016-09-08T00:53:55Z stassats: i've moved onto float/integer stuff
2016-09-08T00:53:58Z stassats: more exciting
2016-09-08T00:55:41Z nyef`: Fair enough.
2016-09-08T00:56:47Z stassats: i'll try the patch anyway
2016-09-08T00:56:53Z stassats: i may have tried it before
2016-09-08T00:58:08Z stassats: no, doesn't do anything
2016-09-08T01:02:46Z stassats: even doing (sb-bignum:bignum-compare bignum (ash int exp)) is faster than the current way
2016-09-08T01:02:55Z stassats: but i'd like to avoid the (ash int exp) step
2016-09-08T01:03:05Z stassats: though it makes things more difficult
2016-09-08T01:04:42Z abbe quit (Ping timeout: 258 seconds)
2016-09-08T01:06:34Z abbe joined #sbcl
2016-09-08T01:29:02Z abbe quit (Ping timeout: 250 seconds)
2016-09-08T01:30:08Z stassats: if i shift it all the way too the left while it's still within a word, it'll make things easier
2016-09-08T01:30:46Z abbe joined #sbcl
2016-09-08T01:31:09Z stassats: and if it's split, do the same thing, then check the remaining lower digits for zero or non zero
2016-09-08T01:31:16Z stassats: sounds like a plan
2016-09-08T01:45:48Z em1l_ joined #sbcl
2016-09-08T01:48:52Z em1l quit (Ping timeout: 240 seconds)
2016-09-08T01:55:17Z abbe quit (Read error: Connection reset by peer)
2016-09-08T01:55:40Z abbe joined #sbcl
2016-09-08T02:10:29Z DGASAU quit (Read error: Connection reset by peer)
2016-09-08T02:11:08Z DGASAU joined #sbcl
2016-09-08T02:14:15Z nyef`: Oh, wow. lp#1531253 looks shockingly familiar in its defined-reproducible cause and effects. (-:
2016-09-08T02:14:54Z nyef`: I think I'm going to mark this one as fix committed.
2016-09-08T02:18:20Z stassats: nyef`: re 1036962, i think some internal variable name just got change
2016-09-08T02:18:21Z stassats: d
2016-09-08T02:18:51Z stassats: that's how it tracks visible code and whether to report the note
2016-09-08T02:19:04Z abbe quit (Ping timeout: 258 seconds)
2016-09-08T02:19:39Z nyef`: Ah, okay.
2016-09-08T02:19:47Z abbe joined #sbcl
2016-09-08T02:19:59Z stassats: the question is, what variable was that
2016-09-08T02:20:12Z nyef`: ... and can we cause it to trigger again, somehow?
2016-09-08T02:21:36Z stassats: trying
2016-09-08T02:21:42Z stassats: i'm sure it's possible
2016-09-08T02:27:09Z nyef`: ... finished reviewing the bug tracker for core compiler bugs. One major theme found, and one possible outlier that looks interesting but I don't yet understand sufficiently to be able to tell if it really is a bug or not.
2016-09-08T02:27:22Z nyef`: The major theme is escape analysis for non-local exits.
2016-09-08T02:28:44Z nyef`: Something like four or five bugs for silliness involved in returning a closure over an exit that has reached the end of its extent.
2016-09-08T02:29:13Z nyef`: The outlier is some sort of possible DX issue.
2016-09-08T02:29:44Z nyef`: The other theme is two or three on the subject of inlining.
2016-09-08T02:30:05Z nyef`: And that's pretty much it, everything else seems to be target-specific or feature requests.
2016-09-08T02:35:55Z stassats: well, i can get a deletion note, but not that strange anymore
2016-09-08T02:36:07Z stassats: if i rename the SAP variable to STREAM
2016-09-08T02:36:26Z stassats: then it's able to find it at the top-level
2016-09-08T02:36:35Z nyef`: I'm an odd mix of impressed and disappointed by this. Impressed that it seems like we've done, as SBCL maintainers, a heck of a job maintaining the compiler, and disappointed that there aren't that many interesting core bugs to tackle at the moment.
2016-09-08T02:37:02Z stassats: but that's no longer influence by THE (UNSIGNED-BYTE 8)
2016-09-08T02:37:19Z stassats: probably there was something sticking out of that cast
2016-09-08T02:40:42Z stassats: nyef`: but there are optimization opportunities to the horizon
2016-09-08T02:40:54Z stassats: and new bugs with them
2016-09-08T02:41:06Z nyef`: Mmm. Not the same, though, really.
2016-09-08T02:42:16Z abbe quit (Ping timeout: 250 seconds)
2016-09-08T02:43:29Z stassats: nyef`: you can tackle https://bugs.launchpad.net/sbcl/+bug/1523149 any time
2016-09-08T02:43:47Z nyef`: Was thinking about it.
2016-09-08T02:44:16Z nyef`: Basically, there's a small cluster of bugs in the tracker around that.
2016-09-08T02:44:34Z nyef`: The other one, as I said, was NLX escape analysis.
2016-09-08T02:44:39Z stassats: where i left off, FUN gets inlined after it's already let converted
2016-09-08T02:45:36Z stassats: actually, i haven't really checked that it gets let converted
2016-09-08T02:45:43Z abbe joined #sbcl
2016-09-08T02:45:44Z nyef`: I don't think that I'm going to do much more tonight, really.
2016-09-08T02:46:23Z nyef`: More getting a feel for what the next major things are for tomorrow, or later on.
2016-09-08T02:46:50Z stassats: it's actually let converted in a XEP, from #'fun
2016-09-08T02:47:14Z stassats: at least it says #<COMPONENT :NAME (XEP (FLET FUN :IN FOO)) :REANALYZE T {100D3E62C3}>
2016-09-08T02:47:24Z stassats: but i don't trust component names lately
2016-09-08T02:47:43Z nyef`: Yeah, component names are junk in the presence of inlining.
2016-09-08T02:47:47Z stassats: but it's believable
2016-09-08T02:48:37Z nyef`: It seems like it always picks the "wrong" component name when merging components, but I'm not sure why, if it's really deterministic, or if the other options were/are worse.
2016-09-08T02:48:42Z stassats: oh, new failure mode!
2016-09-08T02:48:48Z nyef`: Oh?
2016-09-08T02:48:57Z stassats: The value NIL is not of type SB-C::IR2-LVAR
2016-09-08T02:49:14Z nyef`: Neat.
2016-09-08T02:49:27Z stassats: https://bugs.launchpad.net/sbcl/+bug/1523149/comments/3
2016-09-08T02:49:34Z stassats: tried to make the XEP escape
2016-09-08T02:50:36Z stassats: that one is actually newish
2016-09-08T02:53:21Z DGASAU quit (Read error: Connection reset by peer)
2016-09-08T02:53:26Z stassats: allowing it to proceed
2016-09-08T02:53:39Z stassats: (foo ) #<CLOSURE (FLET FUN :IN FOO) {100991DA3B}> => 9
2016-09-08T02:53:41Z stassats: 9?
2016-09-08T02:53:47Z stassats: ok, so, really broken
2016-09-08T02:54:31Z stassats: it was broken in the past too, even though it compiled
2016-09-08T02:54:40Z stassats: (foo) #<CLOSURE (FLET FUN :IN FOO) {1002BBB4DB}> => FOO
2016-09-08T02:54:52Z stassats: the correct return value would be NIL, not FOO or 9
2016-09-08T02:55:56Z stassats: and now it's not influenced by %coerce-callable-to-fun
2016-09-08T02:57:23Z stassats: i bet it's something simple, but it breaks everything
2016-09-08T02:57:31Z nyef`: That sounds like a good thing for me to try to dig into the next time I have a couple of days free.
2016-09-08T02:58:02Z stassats: i'll rename the bug
2016-09-08T03:00:27Z DGASAU joined #sbcl
2016-09-08T03:00:29Z DGASAU quit (Read error: Connection reset by peer)
2016-09-08T03:03:42Z stassats: for the ir2-conversion thing, i don't actually need BAR
2016-09-08T03:04:21Z DGASAU joined #sbcl
2016-09-08T03:04:42Z stassats: https://bugs.launchpad.net/sbcl/+bug/1523149/comments/4
2016-09-08T03:06:26Z stassats: so i think it's the same issue, the block gets deleted but the XEP still refers to it
2016-09-08T03:06:43Z stassats: the BLOCK NIL block
2016-09-08T03:07:31Z DGASAU quit (Read error: Connection reset by peer)
2016-09-08T03:08:32Z DGASAU joined #sbcl
2016-09-08T03:09:15Z abbe quit (Ping timeout: 276 seconds)
2016-09-08T03:10:00Z abbe joined #sbcl
2016-09-08T03:11:51Z stassats quit (Ping timeout: 244 seconds)
2016-09-08T03:55:01Z DavidGu joined #sbcl
2016-09-08T04:39:16Z DavidGu quit (Ping timeout: 250 seconds)
2016-09-08T05:01:57Z shka_ joined #sbcl
2016-09-08T05:11:36Z nzambe quit (Quit: http://www.kiwiirc.com/ - A hand crafted IRC client)
2016-09-08T05:13:39Z les quit (Ping timeout: 265 seconds)
2016-09-08T05:13:59Z les joined #sbcl
2016-09-08T05:46:43Z gingerale joined #sbcl
2016-09-08T06:04:22Z scymtym quit (Ping timeout: 255 seconds)
2016-09-08T06:16:27Z shka_ quit (Ping timeout: 244 seconds)
2016-09-08T06:41:58Z DavidGu joined #sbcl
2016-09-08T07:04:31Z DavidGu quit (Ping timeout: 244 seconds)
2016-09-08T07:08:21Z gingerale quit (Remote host closed the connection)
2016-09-08T07:55:02Z angavrilov joined #sbcl
2016-09-08T08:13:27Z scymtym joined #sbcl
2016-09-08T08:31:58Z Bike quit (Quit: leaving)
2016-09-08T08:34:30Z rudolfochrist joined #sbcl
2016-09-08T08:35:40Z ktt9 joined #sbcl
2016-09-08T09:03:52Z sh1r0 joined #sbcl
2016-09-08T09:04:30Z sh1r0 quit (Client Quit)
2016-09-08T09:04:50Z shirokage joined #sbcl
2016-09-08T09:05:38Z shirokage left #sbcl
2016-09-08T09:53:17Z attila_lendvai joined #sbcl
2016-09-08T09:56:33Z Xof quit (Ping timeout: 240 seconds)
2016-09-08T09:58:04Z rudolfochrist quit (Ping timeout: 264 seconds)
2016-09-08T10:04:07Z chris_l joined #sbcl
2016-09-08T10:38:50Z Xof joined #sbcl
2016-09-08T10:53:20Z rudolfochrist joined #sbcl
2016-09-08T10:58:39Z rudolfochrist quit (Ping timeout: 260 seconds)
2016-09-08T11:54:06Z rudolfochrist joined #sbcl
2016-09-08T11:57:46Z PosterdatiMobile joined #sbcl
2016-09-08T11:57:49Z _PosterdatiMobil joined #sbcl
2016-09-08T11:58:55Z rudolfochrist quit (Ping timeout: 252 seconds)
2016-09-08T12:10:28Z sjl joined #sbcl
2016-09-08T12:23:21Z _PosterdatiMobil quit (Ping timeout: 250 seconds)
2016-09-08T12:23:36Z PosterdatiMobile quit (Ping timeout: 276 seconds)
2016-09-08T12:25:29Z sjl quit (Quit: WeeChat 1.3)
2016-09-08T12:26:09Z sjl joined #sbcl
2016-09-08T12:29:33Z sjl quit (Client Quit)
2016-09-08T12:30:38Z sjl joined #sbcl
2016-09-08T12:31:18Z madbub joined #sbcl
2016-09-08T12:39:20Z nzambe joined #sbcl
2016-09-08T12:59:40Z stassats joined #sbcl
2016-09-08T13:03:29Z DavidGu joined #sbcl
2016-09-08T13:19:14Z DavidGu quit (Ping timeout: 260 seconds)
2016-09-08T14:00:07Z cromachina quit (Read error: Connection reset by peer)
2016-09-08T14:09:48Z DavidGu joined #sbcl
2016-09-08T14:29:38Z nyef`: Can we close lp#1187879? The test has long since been changed to copy the stack-allocated structure, and the memory fault comes from trying to print the stack-allocated result at the REPL.
2016-09-08T14:29:46Z nyef`: (Oh, and G'morning all.)
2016-09-08T14:33:34Z rudolfochrist joined #sbcl
2016-09-08T14:39:31Z rudolfochrist quit (Ping timeout: 252 seconds)
2016-09-08T14:57:14Z stassats: i have no idea what's going on in 1187879, too much code
2016-09-08T15:01:28Z nyef`: The test case was "fixed" in commit debae3c18d31b5222be4d5de8dcb2601336e24a4 which is one of yours. You had some idea at some point. d-:
2016-09-08T15:02:50Z nyef`: I could go to the trouble of verifying that all of the stack analysis bits worked out right, but given that this bug pre-dates the reworking of the exit deletion and various stack analysis fixes I'm really not inclined to.
2016-09-08T15:03:33Z nyef`: Maybe the point of the bug now is to rewrite the test case in terms of ALLOCATION-INFORMATION?
2016-09-08T15:05:15Z chris_l quit (Quit: Ex-Chat)
2016-09-08T15:24:58Z mgodshall quit (Quit: leaving)
2016-09-08T15:33:16Z Bike joined #sbcl
2016-09-08T15:36:59Z stassats: hell, (defmethod meth (x y) (meth)) doesn't report argument mismatch
2016-09-08T15:38:54Z stassats: that wasted me some time
2016-09-08T15:39:42Z scymtym_ joined #sbcl
2016-09-08T15:40:31Z scymtym quit (Remote host closed the connection)
2016-09-08T15:42:55Z scymtym__ joined #sbcl
2016-09-08T15:45:06Z scymtym_ quit (Ping timeout: 276 seconds)
2016-09-08T15:46:27Z DavidGu quit (Quit: DavidGu)
2016-09-08T16:37:25Z stassats: now i have a :default-initargs initarg ignored
2016-09-08T16:39:54Z stassats: probably some rogue initialize-instance method
2016-09-08T16:56:06Z rumbler31 joined #sbcl
2016-09-08T17:04:21Z nzambe quit (Ping timeout: 244 seconds)
2016-09-08T17:11:33Z scymtym__: what does slime do while executing SLIME-EDIT-USES? (slime-edit-uses "1-") leads to all kinds of compiler output as if the compiler was invoked while building the xref list
2016-09-08T17:47:28Z gingerale joined #sbcl
2016-09-08T17:50:20Z Cooler_ quit (Read error: Connection reset by peer)
2016-09-08T17:53:54Z chris_l joined #sbcl
2016-09-08T18:07:22Z PosterdatiMobile joined #sbcl
2016-09-08T18:08:07Z Cooler_ joined #sbcl
2016-09-08T18:08:22Z _PosterdatiMobil joined #sbcl
2016-09-08T18:11:45Z scymtym__: this is probably as good as it gets: http://paste.lisp.org/display/325420#2
2016-09-08T18:20:03Z attila_lendvai: scymtym__: nice! I've seen many people shredding off much smaller pieces of the core than that.
2016-09-08T18:22:46Z rumbler31 quit (Remote host closed the connection)
2016-09-08T18:27:11Z shka_ joined #sbcl
2016-09-08T18:30:14Z scymtym__: attila_lendvai: only applies to cores with xref info, though. and x86_64. savings are smaller for 32-bit words
2016-09-08T18:30:51Z attila_lendvai: oh, I see. but still, it must have been a fun hack!
2016-09-08T18:32:05Z scymtym__: it was
2016-09-08T18:35:35Z stassats: xrefs are enabled for user code by default
2016-09-08T18:35:53Z stassats: scymtym__: x86-64 and not 64-bit?
2016-09-08T18:36:00Z nyef`: Four and a half megs? Nice!
2016-09-08T18:40:56Z scymtym__: stassats: i don't understand the question
2016-09-08T18:41:36Z stassats: why is it limited to x86-64?
2016-09-08T18:43:13Z nyef`: Because it's the only 64-bit platform that we have that anyone has ever heard of? (-:
2016-09-08T18:44:52Z scymtym__: stassats: it is not. but the space savings mainly come from avoiding one element in an unspecialized vector for each referenced symbol. thus the savings will be less significant for 32-bit words
2016-09-08T18:45:16Z nyef`: Clearly, we need to hurry up and implement HPPA 2.0 (64-bit) support, and SPARC64, and a true 64-bit alpha port...
2016-09-08T18:45:34Z stassats: nyef`: ppc64 is probably more useful
2016-09-08T18:45:47Z nyef`: Hence why I didn't put it on that list, just like
2016-09-08T18:45:51Z nyef`: MIPS64. d-:
2016-09-08T18:46:30Z stassats: what about x32?
2016-09-08T18:47:24Z nyef`: That's the memory model where the address space is 32 bits wide, but the register set is 64 bits?
2016-09-08T18:47:35Z stassats: yes
2016-09-08T18:48:16Z nyef`: Hrm.
2016-09-08T18:48:20Z _PosterdatiMobil quit (Quit: KVIrc 4.2.0 Equilibrium http://www.kvirc.net/)
2016-09-08T18:48:37Z nyef`: I don't know if that should be a separate backend, or if it should be a variant on the 64-bit backend.
2016-09-08T18:49:14Z nyef`: Probably the latter, actually.
2016-09-08T18:49:27Z stassats: i wanted to have one backend for ppc and ppc64
2016-09-08T18:49:32Z chris_l quit (Quit: Ex-Chat)
2016-09-08T18:50:02Z nyef`: ... because the base instruction set is the same?
2016-09-08T18:50:07Z nyef`: Well, largely the same?
2016-09-08T18:50:07Z stassats: yeah
2016-09-08T18:50:21Z sjl quit (Ping timeout: 276 seconds)
2016-09-08T18:50:23Z nyef`: Can we re-unify the x86oids, or are they too different?
2016-09-08T18:50:33Z stassats: register difference?
2016-09-08T18:50:43Z stassats: and all the REX stuff
2016-09-08T18:50:57Z stassats: probably if you write with that stuff in mind you can share a lot
2016-09-08T18:51:05Z nyef`: Not much of a register difference at the assembler level, you "just" lock out the additional SC-offsets.
2016-09-08T18:51:11Z stassats: but now i think it's too late
2016-09-08T18:51:39Z scymtym__: any ideas for doing https://github.com/scymtym/sbcl/blob/compact-xref-storage/src/compiler/xref.lisp#L215 more elegantly? (that alone saves 1.5 MB, so i would like to have something like that, if possible)
2016-09-08T18:53:15Z nyef`: ... There are literally "List" and "List*" symbols?
2016-09-08T18:53:34Z stassats: yeah, backqoute
2016-09-08T18:53:39Z nyef`: Ah.
2016-09-08T18:53:46Z nyef`: That's a little confusing.
2016-09-08T19:05:01Z rumbler31 joined #sbcl
2016-09-08T19:55:28Z sjl joined #sbcl
2016-09-08T20:21:10Z scymtym__ quit (Ping timeout: 255 seconds)
2016-09-08T20:26:01Z PosterdatiMobile quit (Quit: KVIrc 4.2.0 Equilibrium http://www.kvirc.net/)
2016-09-08T20:30:18Z PosterdatiMobile joined #sbcl
2016-09-08T20:30:28Z _PosterdatiMobil joined #sbcl
2016-09-08T20:31:08Z _PosterdatiMobil quit (Client Quit)
2016-09-08T20:50:07Z gingerale quit (Remote host closed the connection)
2016-09-08T21:02:25Z madbub quit (Remote host closed the connection)
2016-09-08T21:09:46Z shka_ quit (Ping timeout: 255 seconds)
2016-09-08T21:33:36Z ktt9 quit (Quit: leaving)
2016-09-08T21:44:52Z angavrilov quit (Remote host closed the connection)
2016-09-08T21:46:15Z scymtym joined #sbcl
2016-09-08T22:04:00Z ASau joined #sbcl
2016-09-08T22:05:56Z PosterdatiMobile quit (Quit: KVIrc 4.2.0 Equilibrium http://www.kvirc.net/)
2016-09-08T23:09:55Z DeadTrickster quit (Ping timeout: 255 seconds)
2016-09-08T23:10:20Z DeadTrickster joined #sbcl
2016-09-08T23:31:48Z slyrus joined #sbcl
2016-09-08T23:34:18Z cromachina joined #sbcl
2016-09-08T23:34:56Z rumbler31 quit (Remote host closed the connection)