00:05:25 -!- segv- [~mb@95-91-241-45-dynip.superkabel.de] has quit [Remote host closed the connection]
01:21:47 <pkhuong> Krystof: and you're right, the aver is redundundant
01:34:24 slyrus [~chatzilla@107-1-159-131-ip-static.hfc.comcastbusiness.net] has joined #sbcl
01:41:03 -!- attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Ping timeout: 256 seconds]
01:49:21 -!- Bike [~Glossina@67-5-239-215.ptld.qwest.net] has quit [Ping timeout: 248 seconds]
01:51:11 Bike [~Glossina@174-25-51-29.ptld.qwest.net] has joined #sbcl
01:52:09 -!- ASau [~user@p5797F654.dip0.t-ipconnect.de] has quit [Read error: Connection reset by peer]
01:52:52 ASau [~user@87.151.246.84] has joined #sbcl
01:55:27 attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has joined #sbcl
01:55:59 drmeister [~drmeister@pool-71-185-168-200.phlapa.fios.verizon.net] has joined #sbcl
02:04:07 -!- attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Ping timeout: 252 seconds]
02:05:32 Hydan`` [~hydan@ip-89-103-110-5.net.upcbroadband.cz] has joined #sbcl
02:07:23 attila_lendvai [~attila_le@92.47.189.202] has joined #sbcl
02:07:23 -!- attila_lendvai [~attila_le@92.47.189.202] has quit [Changing host]
02:07:23 attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has joined #sbcl
02:08:47 -!- Hydan` [~hydan@ip-89-103-110-5.net.upcbroadband.cz] has quit [Ping timeout: 252 seconds]
02:18:48 -!- drmeister [~drmeister@pool-71-185-168-200.phlapa.fios.verizon.net] has quit [Remote host closed the connection]
02:21:23 Fare [~fare@173-9-65-97-NewEngland.hfc.comcastbusiness.net] has joined #sbcl
02:39:06 -!- christoph4 [~christoph@ppp-188-174-86-199.dynamic.mnet-online.de] has quit [Ping timeout: 264 seconds]
02:52:21 christoph4 [~christoph@ppp-188-174-93-255.dynamic.mnet-online.de] has joined #sbcl
03:03:04 -!- Fare [~fare@173-9-65-97-NewEngland.hfc.comcastbusiness.net] has quit [Remote host closed the connection]
03:03:46 -!- attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Quit: Leaving.]
03:04:04 attila_lendvai [~attila_le@92.47.189.202] has joined #sbcl
03:04:04 -!- attila_lendvai [~attila_le@92.47.189.202] has quit [Changing host]
03:04:04 attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has joined #sbcl
03:04:06 -!- attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Client Quit]
03:04:29 Guest22221 [~attila_le@92.47.189.202] has joined #sbcl
03:05:41 -!- Guest22221 is now known as attila_lendvai1
03:05:49 -!- attila_lendvai1 [~attila_le@92.47.189.202] has quit [Changing host]
03:05:49 attila_lendvai1 [~attila_le@unaffiliated/attila-lendvai/x-3126965] has joined #sbcl
05:05:49 *|3b|* gets "Failure:            load.impure.lisp / PARALLEL-FASL-LOAD" after a few runs of tests
05:13:43 benkard [~benkard@mnch-5d868f73.pool.mediaWays.net] has joined #sbcl
05:24:50 sdemarre [~serge@133.83-64-87.adsl-dyn.isp.belgacom.be] has joined #sbcl
05:25:47 -!- benkard [~benkard@mnch-5d868f73.pool.mediaWays.net] has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz]
05:28:33 -!- psilord [~psilord@c-69-180-173-249.hsd1.mn.comcast.net] has quit [Ping timeout: 248 seconds]
05:30:52 foreignFunction [~niksaak@ip-4761.sunline.net.ua] has joined #sbcl
05:31:17 -!- sdemarre [~serge@133.83-64-87.adsl-dyn.isp.belgacom.be] has quit [Ping timeout: 248 seconds]
05:39:48 <|3b|> running just load.impure.lisp in a loop failed once after 350 times
05:41:00 <|3b|> 2 of the three threads got "R2: The function COMMON-LISP-USER::TWO-FUN is undefined."
05:41:34 *|3b|* notes that the error calls both r2 and r3 R2
05:43:15 psilord [~psilord@c-69-180-173-249.hsd1.mn.comcast.net] has joined #sbcl
06:05:26 benkard [~benkard@mnch-5d868f73.pool.mediaWays.net] has joined #sbcl
06:19:36 prxq [~mommer@mnhm-5f75eec9.pool.mediaWays.net] has joined #sbcl
06:38:25 -!- ehaliewicz [~user@50-0-51-11.dsl.static.sonic.net] has quit [Ping timeout: 248 seconds]
06:39:58 -!- Bike [~Glossina@174-25-51-29.ptld.qwest.net] has quit [Remote host closed the connection]
06:40:24 Bike [~Glossina@174-25-51-29.ptld.qwest.net] has joined #sbcl
06:44:45 -!- benkard [~benkard@mnch-5d868f73.pool.mediaWays.net] has quit [Read error: Connection reset by peer]
06:50:22 benkard [~benkard@mnch-5d868f73.pool.mediaWays.net] has joined #sbcl
06:52:52 johnc [~johnc@61.135.169.73] has joined #sbcl
06:58:06 teggi [~teggi@113.173.4.86] has joined #sbcl
07:04:01 -!- johnc [~johnc@61.135.169.73] has quit [Quit:  HydraIRC -> http://www.hydrairc.com <- In tests, 0x09 out of 0x0A l33t h4x0rz prefer it :)]
07:04:14 -!- benkard [~benkard@mnch-5d868f73.pool.mediaWays.net] has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz]
07:10:51 -!- attila_lendvai1 [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Quit: Leaving.]
07:29:43 stassats [~stassats@wikipedia/stassats] has joined #sbcl
07:34:57 -!- brucem [~bmitchene@waywardmonkeys.com] has quit [*.net *.split]
07:44:24 attila_lendvai [~attila_le@87.247.13.177] has joined #sbcl
07:44:24 -!- attila_lendvai [~attila_le@87.247.13.177] has quit [Changing host]
07:44:24 attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has joined #sbcl
07:49:20 brucem [~bmitchene@waywardmonkeys.com] has joined #sbcl
07:55:04 <stassats> what's the point of having both define-source-transform funcall and deftransform funcall ((function &rest args) * *)? the second never seem to be triggered
07:59:06 -!- jdz [~jdz@85.254.212.34] has quit [Quit: Byebye.]
08:02:07 jdz [~jdz@85.254.212.34] has joined #sbcl
08:08:04 -!- jdz [~jdz@85.254.212.34] has quit [Remote host closed the connection]
08:11:33 jdz [~jdz@85.254.212.34] has joined #sbcl
08:16:38 *stassats* also sees no point in legal-fun-name-p which just does (values (valid-function-name-p name))
08:52:09 attila_lendvai1 [~attila_le@87.247.13.177] has joined #sbcl
08:52:09 -!- attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Disconnected by services]
08:52:09 -!- attila_lendvai1 [~attila_le@87.247.13.177] has quit [Changing host]
08:52:09 attila_lendvai1 [~attila_le@unaffiliated/attila-lendvai/x-3126965] has joined #sbcl
09:19:29 -!- scymtym__ [~user@ip-5-147-122-209.unitymediagroup.de] has quit [Ping timeout: 256 seconds]
09:36:18 <Krystof> stassats: the second one comes because (I think) of uncertainty around designing the extension to valid function names
09:36:26 <Krystof> is it public?  What is the interface?  etc
09:48:51 *stassats* is not sure that having (defun (defmacro x) ()) is a good idea
10:07:21 -!- foreignFunction [~niksaak@ip-4761.sunline.net.ua] has quit [Quit: Leaving.]
10:22:48 -!- yacks [~py@180.151.36.168] has quit [Read error: Connection reset by peer]
10:24:16 -!- danlentz [~danlentz@c-68-37-70-235.hsd1.nj.comcast.net] has quit [Remote host closed the connection]
10:35:41 davazp [~user@213.37.200.68.dyn.user.ono.com] has joined #sbcl
10:43:51 danlentz [~danlentz@c-68-37-70-235.hsd1.nj.comcast.net] has joined #sbcl
10:45:19 <pkhuong> you want both a source transform and an ir1 transform to let the transform happen as early as possible
10:46:21 <pkhuong> because the ir1 transform machinery involves some slow and heavy duty IR1 frobbing to convert full calls back into useful known calls.
10:47:03 <stassats> but i can't make up a form which would invoke the transform
10:47:19 <stassats> if you want to trigger it early, why not just leave the source transform?
10:48:37 <pkhuong> because sometimes the ir1 transform is needed
10:49:06 <pkhuong> e.g. (let ((foo ...)) (funcall foo ...)), or inlined higher-order functions
10:50:59 <stassats> the let goes through the source transform
10:52:53 <stassats> and i actually don't understand what the transform transforms the code into, `(lambda (function ,@arg-names) `(%funcall ,(ensure-lvar-fun-form function 'function) ,@arg-names)), who calls this lambda?
10:53:20 <pkhuong> that's what transform do: they return a lambda form that's ir1-converted and called instead of the transformed function.
10:54:06 <pkhuong> there's a convenience thing where, if you return a non-lambda form, you get wrapped in a lambda + ignorable (you can return declarations as a second return value)
10:54:24 <stassats> ok, that part is now clear
10:55:29 <pkhuong> make the let-bound form more complex, e.g. so that it depends on type analysis and DCE to have a known value, and you'll need the ir1 transform
10:59:50 <stassats> i can see that %coerce-callable-to-fun can have both
11:01:06 <stassats> the only way for the ir1-transform to get triggered, is not to expand the source-transform
11:01:26 <stassats> %coerce-callable-to-fun does that, but the funcall one just always goes to %funcall
11:03:42 *stassats* inserts a break and tries to compile a lot of code
11:09:49 <pkhuong> (funcall funcall ...) and that sort of stuff
11:13:42 -!- davazp [~user@213.37.200.68.dyn.user.ono.com] has quit [Read error: Connection reset by peer]
11:13:49 -!- automaciej [~automacie@cl-1008.ham-02.de.sixxs.net] has quit [Ping timeout: 245 seconds]
11:15:07 automaciej [~automacie@quinoa.blizinski.pl] has joined #sbcl
11:28:36 <stassats> no success so far
11:28:40 <stassats> or rather, no failure
11:30:57 <pkhuong> notinline, maybe.
11:31:27 <stassats> notinling funcall just makes a full call
11:32:06 <stassats> maybe my selection of libraries is thin, i'll try to compile everything in quicklisp
11:32:37 <pkhuong> you're not likely to see the sort of strange control flow necessary for the deftransform in the wild.
11:33:01 <pkhuong> ISTR something about forms returned by ir1 transform not always going through source transforms.
11:33:35 <stassats> well, if another transform returns a call to funcall, maybe, but when would that happen?
11:33:52 <stassats> (i compile sbcl with the break there)
11:33:56 <stassats> compiled
11:40:02 <scymtym> on #+win32, why does USER-HOMEDIR-PATHNAME use (sb!win32::get-folder-namestring sb!win32::csidl_profile) but USER-HOMEDIR-NAMESTRING does not?
11:40:31 <scymtym> i think, this can lead to a TYPE-ERROR on Windows when the HOME envvar is not set
11:43:23 yacks [~py@180.151.36.168] has joined #sbcl
11:43:30 davazp [~user@213.37.200.68.dyn.user.ono.com] has joined #sbcl
12:13:07 <stassats> oops, the random tester reached up to my changes too
12:35:24 drmeister [~drmeister@pool-71-185-168-200.phlapa.fios.verizon.net] has joined #sbcl
12:35:38 <stassats> gah, made a typo in "subtype" and it should have been supertype
12:35:41 -!- davazp [~user@213.37.200.68.dyn.user.ono.com] has quit [Remote host closed the connection]
12:39:07 -!- drmeister [~drmeister@pool-71-185-168-200.phlapa.fios.verizon.net] has quit [Remote host closed the connection]
12:42:23 <stassats> huh, (defconstant error 25) => success
12:49:16 <stassats> 11.1.2.1.2 doesn't say you can't do that, but it's a good way to mess things up
12:58:36 <stassats> all other implementations seem also to allow it, but it doesn't make any sense to block from defining special variables, but allow constants to go
12:59:14 <stassats> especially since it breaks any otherwise portable code which uses ERROR as a lexical variable
13:01:03 <stassats> all other don't seem to have package locks for variables either, but lispworks signals errors on both constants and variables
13:03:09 <stassats> (cl-gearman, which does that very thing, breaks a half of other systems)
13:14:25 segv- [~mb@95-91-241-83-dynip.superkabel.de] has joined #sbcl
13:22:31 wbooze [~wbooze@xdsl-78-35-137-140.netcologne.de] has joined #sbcl
13:23:23 nyef [~nyef@c-50-157-244-41.hsd1.ma.comcast.net] has joined #sbcl
13:23:31 <nyef> G'morning all.
13:40:29 drmeister [~drmeister@farnsworth.chem.temple.edu] has joined #sbcl
13:56:34 rpg [~rpg@216.243.156.16.real-time.com] has joined #sbcl
13:58:55 Odyessus [~odyessus@089144192114.atnat0001.highway.a1.net] has joined #sbcl
14:09:13 -!- Odyessus [~odyessus@089144192114.atnat0001.highway.a1.net] has quit [Quit: Colloquy for iPad - http://colloquy.mobi]
14:44:22 -!- attila_lendvai1 [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Quit: Leaving.]
15:08:41 -!- Vivitron [~Vivitron@pool-98-110-213-33.bstnma.fios.verizon.net] has quit [Ping timeout: 252 seconds]
15:15:13 -!- slyrus [~chatzilla@107-1-159-131-ip-static.hfc.comcastbusiness.net] has quit [Ping timeout: 248 seconds]
15:22:57 *stassats* sees a lot of special variables in the compiler, which appear to be unused
15:23:13 <stassats> *code-vector* *next-location* *result-fixups* *continuation-number* *continuation-numbers*
15:24:15 <stassats> scratch the last two
15:26:24 slyrus [~chatzilla@23-25-20-209-static.hfc.comcastbusiness.net] has joined #sbcl
15:31:45 -!- teggi [~teggi@113.173.4.86] has quit [Remote host closed the connection]
15:31:51 <nyef> stassats: Nice catch. I wouldn't entirely be surprised if they weren't used in CMUCL either.
15:32:07 <stassats> that's correct, just checked
15:32:28 <stassats> the build succeeds without them
15:33:17 <stassats> those three were on the same line, perhaps they are related and were removed some time ago
15:33:26 <stassats> i need to find csv logs for cmucl
15:34:13 <stassats> google for "cmucl git mirror" returns repo.or.cz/w/sbcl.git
15:35:03 <stassats> and "SBCL" is highlighted
15:36:02 <nyef> Looking at the cmucl git that I have locally, the last commit to touch that line is from commit fa542ac2. by "pw", "Change most PROCLAIMs to DECLAIMs."
15:36:22 <nyef> Don't look for a mirror, look for their master source repository.
15:36:28 <stassats> yeah, already found
15:36:53 <stassats> despite google being unhelpful
15:37:15 <stassats> i guess it's right, who would want to use cmucl anyway
15:38:00 <nyef> rtoy?
15:38:17 <stassats> yeah, except rtoym
15:39:26 <nyef> Now that you have access to the CMUCL source, you too can be horrified by the garbage collectors. (-:
15:40:04 <stassats> i had cmucl sources already, just a snapshot
15:40:18 <stassats> useful for bug archaeology
15:43:02 <stassats> found it
15:45:37 -!- slyrus [~chatzilla@23-25-20-209-static.hfc.comcastbusiness.net] has quit [Remote host closed the connection]
15:46:13 <stassats> http://common-lisp.net/gitweb?p=projects/cmucl/cmucl.git;a=blobdiff;f=compiler/main.lisp;h=fb04e72da9d479f79e301da4128bb061513f42e7;hp=bfc8fa9cb86c12752a39d140545de011491c5300;hb=76b5e40b38bd9783170844c5900e7bb1869ae1db;hpb=9b2a483feb196a95a39c0070868353fa3e268c01;js=1
15:46:43 <stassats> looks like it was renamed
15:47:00 <stassats> on 24 Aug 1990
15:47:53 <stassats> or rather, changed how it works, *code-segment* replaced the three of them
15:48:06 <nyef> Yeah, was just catching up to that.
15:48:49 <nyef> Very good find for once we're out of code freeze (or are we out of freeze already? I've been too busy to pay attention...)
15:49:00 <stassats> ok, almost 13 years
15:49:04 <stassats> we're not yet
15:49:31 <stassats> don't know what prompted me to grep that variable
15:49:56 <stassats> ah, i was grepping for "fixup" first
15:50:20 <stassats> found *result-fixups*, tried to see where it leads
15:50:47 <stassats> and a good thing that we're not out of the freeze, Eric Marsden keeps finding bugs
15:51:06 <stassats> regressions, specifically
15:54:17 <brucem> 23 years.
15:54:35 <stassats> of course
15:54:55 <nyef> brucem: Yeah, our BCD carry flag is broken.
15:55:26 <stassats> it was so easy to calculate dates from the 90s, it's stuck
15:56:07 <stassats> and 23 years is more impressive than 13
15:56:35 <stassats> the previous unused thing i removed was just 9 years old
15:57:17 <stassats> now, compiling a half of quicklisp, i still couldn't trigger the FUNCALL deftransform
15:59:09 *stassats* already has 5-6 commits to be cleaned up and go after the freeze
15:59:23 <stassats> mostly trivial things, though
16:00:14 <stassats> would a transform for constant folding encode-universal-time when possible be useful?
16:00:39 <nyef> Just declare it inline?
16:00:58 <stassats> it will increase code size
16:01:02 <nyef> FUNCALL deftransform?
16:01:29 <stassats> nyef: yes, funcall has both define-source-transform and deftransform, and i can't figure how to trigger the latter/why is it there
16:02:15 <stassats> i didn't actually get the idea for folding encode-universal-time from real code, just saw that it can be done
16:02:31 <stassats> and now i'm not sure whether it's worth it
16:03:33 <stassats> grepping some third party sources, it can be used, maybe not in a hot path
16:03:49 <stassats> the most often is (encode-universal-time 0 0 0 1 1 1970 0)
16:03:57 <stassats> most frequent
16:05:27 <stassats> ok, that settles it, I'll add it
16:09:55 <stassats> many variables in the compiler can be replaced by defglobal
16:10:51 <nyef> Does defglobal work for cross-compilation from a non-SBCL build host?
16:11:25 <stassats> i presume it'll just make it an ordinary special
16:11:32 <Krystof> yes, in cross-misc.lisp
16:11:47 <stassats> because it's already used
16:30:12 *stassats* should institute a prize of some sort for anybody who can come up with a piece of code which triggers the funcall deftransform
16:30:32 <stassats> or can prove that it can't happen
16:32:29 easy-iPad [~easyipad@61.Red-2-137-44.dynamicIP.rima-tde.net] has joined #sbcl
16:40:24 <nyef> I suspect that it might not be able to happen, but I can in no way prove it at this point.
16:43:58 davazp [~user@213.37.200.68.dyn.user.ono.com] has joined #sbcl
16:44:19 <stassats> even if i expand another transform into funcall, it still goes through the source-transform
16:45:10 <stassats> what if %coerce-callable-to-fun is made a static function
16:45:26 <stassats> it's used quite often
16:50:53 <nyef> Please be aware that some platforms have a bit of a limit for how many static objects they can address.
16:51:27 <stassats> how many or how large?
16:51:34 <stassats> both?
16:52:03 <nyef> It's actually a constraint on how much of a displacement they can generate from NIL.
16:53:13 <nyef> Keeping in mind that, if I recall correctly, NIL has LIST-POINTER-LOWTAG, and most of the related objects have OTHER-POINTER-LOWTAG.
16:53:55 <nyef> Oh, and in some cases we try to hit a slot within one of the static objects directly, such as the function-raw-addr of an fdefinition.
16:55:12 <stassats> it's offset by 3 here
16:56:43 *stassats* is wondering a way where %coerce-callable-to-fun (something else in reality) can actually call the function
16:56:56 <stassats> probably won't work well with argument passing
17:03:48 attila_lendvai [~attila_le@92.47.189.202] has joined #sbcl
17:03:48 -!- attila_lendvai [~attila_le@92.47.189.202] has quit [Changing host]
17:03:48 attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has joined #sbcl
17:07:43 -!- attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Client Quit]
17:08:33 leuler [~user@p548FC5BC.dip0.t-ipconnect.de] has joined #sbcl
17:10:56 <stassats> the only thing i can think of is returning the address, not the fdefn
17:36:07 <stassats> so, making it static saves 1 byte in the code, removes a constant, becomes very slightly faster
17:38:07 Vivitron [~Vivitron@pool-98-110-213-33.bstnma.fios.verizon.net] has joined #sbcl
17:38:41 -!- easy-iPad [~easyipad@61.Red-2-137-44.dynamicIP.rima-tde.net] has quit [Quit: Outta here?]
17:39:32 <stassats> (apply x y) could probably use a %coerce-callable-to-fun which calls it, since it's the arguments are already in a list
17:51:33 -!- yacks [~py@180.151.36.168] has quit [Ping timeout: 248 seconds]
17:58:43 -!- davazp [~user@213.37.200.68.dyn.user.ono.com] has quit [Remote host closed the connection]
18:19:34 <pkhuong> don't think replacing specials with globals is such a good idea. It'll just complicate things when someone tries to make the compiler re-entrant.
18:26:17 <stassats> yeah, thought about that, but some things which really won't be rebound can go
18:28:42 -!- milosn_ is now known as milosn
18:28:43 <pkhuong> it's even more true for symbols that will have to be bound to NIL to allow things to be GCed, rather than simply undoing a special binding.
18:29:19 <stassats> i'm more about tables which are static
18:29:51 <stassats> like *backend-template-names*, etc.
18:29:52 <foom> you don't need a deftransform for constant folding
18:29:54 Bike_ [~Glossina@75-164-170-20.ptld.qwest.net] has joined #sbcl
18:29:56 <foom> just defknown
18:30:15 <stassats> foom: encode-universal-time can't be constant folded just like that
18:30:40 <stassats> it depends on the timezone argument, and on the year, if it's 0-99
18:30:59 <foom> oh, crud, I hate APIs that do shit to years when they're 0-99.
18:31:22 -!- Bike [~Glossina@174-25-51-29.ptld.qwest.net] has quit [Ping timeout: 252 seconds]
18:31:36 <foom> I seem to have one of those in my codebase too. :(
18:32:12 -!- Bike_ is now known as Bike
18:32:17 <foom> It's a terrible terrible thing to do, because it's nearly impossible to tell what code is depending on the current time.
18:32:46 <stassats> it's more for interactive things, i would suppose
18:33:54 <foom> It's for handling 2-digit years in various data formats.
18:34:58 <stassats> i mean that you can't defer the handling, i.e. handle it now, then it's deterministic
18:35:12 <stassats> (unless done around the new year's eve)
18:37:13 <foom> I once tried to untangle the usage of current time from the code, and gave up before finishing trying to track down everything that might decode a time with a 2 digit year.
18:37:38 <stassats> handling time-zones is as tricky
18:37:46 <stassats> especially with dst
19:05:41 *|3b|* left load.impure.lisp running overnight, out of 2000 runs, it got a total of 2 failures, one mentioned above in parallel-fasl-load, and another apparently in same test that just stopped there and got "Invalid exit status: load.impure.lisp"
19:05:59 <|3b|> running just that test in a loop didn't seem to fail after 10k tries
19:13:05 <stassats> some static functions aren't annotated in disassembly, because they aren't present in sb-fasl:*assembler-routines*
19:13:43 <stassats> "no problem, says i, i'll just go through *static-funs* and call static-fun-offset"
19:14:17 <stassats> only that now i have to know the value of NIL, and it has a different name for all backends, and the offset may be not from NIL at all
19:14:46 <nyef> But you know what NIL is for each backend, surely?
19:15:04 LiamH [~none@pool-173-48-164-128.bstnma.fios.verizon.net] has joined #sbcl
19:16:14 <stassats> i mixed up null-tn with nil-value in my head
19:16:21 <stassats> yeah, nil-value seems to be common
19:20:35 <stassats> ok, that seems to work
19:20:52 <pkhuong> http://arxiv.org/abs/1305.5800 has really interesting numbers on contention handling for compare-and-swap.
19:21:09 <stassats> i can also annotate safepoints checking that way as well
19:27:52 -!- leuler [~user@p548FC5BC.dip0.t-ipconnect.de] has quit [Quit: ERC Version 5.1.2 $Revision: 1.796.2.6 $ (IRC client for Emacs)]
19:29:06 <stassats> rebroadcast: TEST AL, [#x200FF000]      ; safepoint looks nicer
19:32:41 *stassats* doesn't know what else to tweak
19:33:28 <stassats> nyef: btw, non-consing sleep would go in after the freeze
19:33:51 <nyef> Why not detect that specific combination of arguments to TEST and replace the whole thing with SAFEPOINT ?
19:34:10 <stassats> i like looking at real code
19:35:04 <stassats> s/real/actual/
19:35:20 <pkhuong> avoid arm and ppc ;)
19:35:40 <stassats> maybe one can do that from the binary instructions, i can't
19:36:45 <stassats> for some reason, i can't do CALL [static-address] directly
19:36:52 <stassats> goes to a memory fault
19:39:13 <stassats> and it should be sign-extended in this case
19:39:16 <stassats> shouldn't be
19:39:59 <pkhuong> you're calling the address stored in static-address.
19:40:42 <stassats> by [] i meant ea
19:41:37 sdemarre [~serge@109.134.148.136] has joined #sbcl
19:41:44 <pkhuong> that's what it is
19:41:50 <stassats> tried it again, now it seems to work
19:41:57 <pkhuong> you only have absolute indirect calls and direct relative calls
19:42:20 <stassats> maybe i tried it before with something else
19:46:34 <stassats> the comment says ;; Need to load the target address into a register, since ;; immediate call arguments are just a 32-bit displacement, ;; which obviously can't work with >4G spaces., but if the address is in the upper half, wouldn't that matter anyway (i may be utterly confused)
19:49:17 <pkhuong> the real issue isn't sign extension as much as the fact that it's a *displacement*, not an address
19:50:06 <stassats> ah, ok, what from is it displaced?
19:50:19 <pkhuong> address of the next instruction
19:51:07 <stassats> even for far calls?
19:51:20 <stassats> (no)
19:56:28 <stassats> so,  CALL QWORD PTR [#x201011B8] should only worry about sign extension?
19:56:59 <pkhuong> some OSes have flags to mmap pages in the lower 32 bit. Other ABI/loader combos have tricks to make calls to nearby (within +/- 2GB) addresses quick at the expense of some trampolines for calls to far-away pages
19:57:54 <nyef> And who's going to have a code segment larger than 2GB anyway?
19:57:58 <stassats> replacing MOV RAX, [#x201011B8] CALL RAX  with CALL QWORD PTR [#x201011B8] appears to work
19:58:35 <stassats> are static function guaranteed to be less than 32 bits?
19:59:13 -!- LiamH [~none@pool-173-48-164-128.bstnma.fios.verizon.net] has quit [Quit: Leaving.]
19:59:16 <nyef> They're guaranteed to be in static space. Anything beyond that is, presumably, not guaranteed.
19:59:37 <pkhuong> We assume the address of NIL and T are < 2 GB, but that's it.
19:59:42 <stassats> well, the vop can check and decide what to use
19:59:50 <pkhuong> right.
20:00:15 <stassats> maybe using a function
20:00:45 <pkhuong> nyef: C++, add some huge static data members, etc. it seems *plausible*.
20:11:43 <stassats> so, static funs have just a pointer to the function in the static space, and assembly routines actually live in the static space?
20:16:38 -!- redline6561 [~redline65@li69-162.members.linode.com] has left #sbcl
20:16:46 redline6561 [~redline65@li69-162.members.linode.com] has joined #sbcl
20:17:50 <pkhuong> sounds right.
20:18:38 <pkhuong> if this is where you're going: yes, it would be possible to prefix each assembly routine with its address, and thus enable call [assembly-routine-prefix]
20:19:02 <stassats> no, i was just trying to understand things
20:19:23 <stassats> i guess i tried to do call [routine] previously, and it failed
20:20:28 <stassats> and this explains it. i don't think adding prefixing is worth it, just for space savings
20:24:06 <stassats> using R11 for temporary isn't optimal, since for it takes more to encode
20:25:04 <stassats> i'll leave this to the person who will be doing register allocation
20:28:54 foreignFunction [~niksaak@ip-4761.sunline.net.ua] has joined #sbcl
20:38:43 ASau` [~user@p5797E91E.dip0.t-ipconnect.de] has joined #sbcl
20:42:16 -!- ASau [~user@87.151.246.84] has quit [Ping timeout: 252 seconds]
20:51:12 <pkhuong> yeah... I don't know what to think about temp-reg-tn.
21:00:45 -!- ASau` is now known as ASau
21:01:57 -!- stassats [~stassats@wikipedia/stassats] has quit [Ping timeout: 248 seconds]
21:07:18 Quadrescence [~quad@unaffiliated/quadrescence] has joined #sbcl
21:09:44 fisxoj [~fisxoj@c-24-12-190-29.hsd1.il.comcast.net] has joined #sbcl
21:10:59 -!- nyef [~nyef@c-50-157-244-41.hsd1.ma.comcast.net] has quit [Quit: G'night all.]
21:15:08 LiamH [~none@pool-173-48-164-128.bstnma.fios.verizon.net] has joined #sbcl
21:20:51 -!- sdemarre [~serge@109.134.148.136] has quit [Ping timeout: 256 seconds]
21:21:20 -!- psilord [~psilord@c-69-180-173-249.hsd1.mn.comcast.net] has quit [Quit: Leaving.]
21:47:17 benkard [~benkard@mnch-5d868f73.pool.mediaWays.net] has joined #sbcl
21:49:19 <pkhuong> oh wow... our >/< VOPs seem to assume the constant argument is second... but we can't guarantee that at all.
21:55:12 -!- drmeister [~drmeister@farnsworth.chem.temple.edu] has quit [Remote host closed the connection]
22:03:30 -!- maxm [~user@openchat.com] has quit [*.net *.split]
22:08:10 -!- benkard [~benkard@mnch-5d868f73.pool.mediaWays.net] has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz]
22:09:39 drmeister [~drmeister@166.137.104.76] has joined #sbcl
22:10:15 -!- Quadrescence [~quad@unaffiliated/quadrescence] has quit [Quit: This computer has gone to sleep]
22:19:58 maxm- [~user@75.99.81.170] has joined #sbcl
22:22:20 Quadrescence [~quad@unaffiliated/quadrescence] has joined #sbcl
22:25:37 -!- LiamH [~none@pool-173-48-164-128.bstnma.fios.verizon.net] has quit [Ping timeout: 248 seconds]
22:33:39 -!- drmeister [~drmeister@166.137.104.76] has quit [Remote host closed the connection]
22:35:24 LiamH [~none@pool-173-48-164-128.bstnma.fios.verizon.net] has joined #sbcl
22:44:03 -!- LiamH [~none@pool-173-48-164-128.bstnma.fios.verizon.net] has quit [Quit: Leaving.]
22:46:00 -!- Quadrescence [~quad@unaffiliated/quadrescence] has quit [Quit: This computer has gone to sleep]
22:50:03 -!- segv- [~mb@95-91-241-83-dynip.superkabel.de] has quit [Remote host closed the connection]
23:14:04 -!- Hydan`` is now known as hydan
23:23:00 -!- prxq [~mommer@mnhm-5f75eec9.pool.mediaWays.net] has quit [Remote host closed the connection]
23:32:39 drmeister [~drmeister@pool-71-185-168-200.phlapa.fios.verizon.net] has joined #sbcl
23:37:46 milosn_ [~milosn@user-5af502ec.broadband.tesco.net] has joined #sbcl
23:41:01 -!- milosn [~milosn@user-5af5072d.broadband.tesco.net] has quit [Ping timeout: 276 seconds]
23:42:45 milosn [~milosn@user-5af502ee.broadband.tesco.net] has joined #sbcl
23:42:58 -!- milosn_ [~milosn@user-5af502ec.broadband.tesco.net] has quit [Ping timeout: 276 seconds]
23:44:47 -!- maxm- is now known as maxm
23:47:17 Quadrescence [~quad@unaffiliated/quadrescence] has joined #sbcl
23:58:43 -!- wbooze [~wbooze@xdsl-78-35-137-140.netcologne.de] has quit [Read error: Operation timed out]