06:48:40 Modius_ [n=Modius@99.179.103.172] has joined #ccl 07:08:06 -!- Modius [n=Modius@ppp-70-243-82-104.dsl.austtx.swbell.net] has quit [Read error: 110 (Connection timed out)] 12:08:26 H4ns1 [n=Hans@p57BBA9E7.dip0.t-ipconnect.de] has joined #ccl 12:08:40 -!- H4ns [n=hans@p57A0EF06.dip.t-dialin.net] has quit [Nick collision from services.] 12:08:43 -!- H4ns1 is now known as H4ns 12:09:25 H4ns1 [n=hans@p57A0EF06.dip.t-dialin.net] has joined #ccl 12:12:18 -!- alms [n=alms@209-150-48-250.c3-0.bkl-ubr1.sbo-bkl.ma.cable.rcn.com] has quit [] 12:35:32 -!- H4ns [n=Hans@p57BBA9E7.dip0.t-ipconnect.de] has quit ["Leaving."] 12:49:50 H4ns [n=Hans@p57BBA9E7.dip0.t-ipconnect.de] has joined #ccl 13:13:04 trebor_dki [n=user@mail.dki.tu-darmstadt.de] has joined #ccl 13:15:56 hello. i am new to clozure cl. i checked out ccl via svn and did (rebuild-ccl :full t), but i got errors: Clozure Common Lisp Version 1.2-r10552 (LinuxX8664) ... Error(s) during kernel compilation ... in gc-common.c:22 ... While executing: REBUILD-CCL, in process listener(1). ... 13:17:02 is there a different version i should check out? 13:19:59 -!- H4ns [n=Hans@p57BBA9E7.dip0.t-ipconnect.de] has left #ccl 13:20:12 H4ns [n=Hans@p57BBA9E7.dip0.t-ipconnect.de] has joined #ccl 13:55:36 -!- segv_ [n=mb@p4FC1EFD8.dip.t-dialin.net] has quit [Read error: 110 (Connection timed out)] 13:56:07 segv_ [n=mb@p4FC1F553.dip.t-dialin.net] has joined #ccl 14:37:42 Modius__ [n=Modius@adsl-66-143-165-178.dsl.austtx.swbell.net] has joined #ccl 14:43:09 -!- Modius_ [n=Modius@99.179.103.172] has quit [Read error: 104 (Connection reset by peer)] 15:35:26 trebor_dki: http://trac.clozure.com/openmcl/wiki/ReleaseNotes/1.3 15:41:47 gz: thanks, checked it out - installing. 15:49:42 milanj [n=milan@91.148.92.61] has joined #ccl 15:53:12 H4ns2 [n=hans@p57A0C7B0.dip.t-dialin.net] has joined #ccl 16:10:31 -!- H4ns1 [n=hans@p57A0EF06.dip.t-dialin.net] has quit [Read error: 110 (Connection timed out)] 17:02:35 -!- H4ns2 is now known as H4ns1 17:03:02 -!- H4ns [n=Hans@p57BBA9E7.dip0.t-ipconnect.de] has quit [Nick collision from services.] 17:03:05 -!- H4ns1 is now known as H4ns 17:04:10 H4ns1 [n=Hans@p57BBA9E7.dip0.t-ipconnect.de] has joined #ccl 17:16:00 hm. i am doing (apropos "save-lisp-") but i do not see sth like save-lisp-and-die - how is that function named under ccl? 17:17:21 sorry, i found http://ccl.clozure.com/manual/chapter4.7.html 17:26:17 anRch [n=markmill@98.98.44.134] has joined #ccl 17:54:55 omg - even after using #+/- to replace all sb-... calls with ccl... calls - nothing works. i never thought, that lisp-implementations are this kind of incompatible (or i am such a bad lisp-beginner).... 17:55:22 (it is not due to ccl:... calls, just simple clos does it). 17:57:41 -!- anRch [n=markmill@98.98.44.134] has quit [] 18:06:06 is there a way of stepping through code? trying resulted in SWANK-BACKEND:ACTIVATE-STEPPING not implementated - i hope there is an alternative 18:10:44 trebor_dki, usually you know what's implementation specific and what's portable ansi common lisp 18:11:06 ccl does have a stepper, but it's not integrated with slime 18:11:29 (i'm not sure if the stepper is in trunk or if it's still in another branch. there were, at the time, performance/memory use issues with it) 18:13:56 anRch [n=markmill@nmd.sbx07283.medfoma.wayport.net] has joined #ccl 18:14:38 milanj: i found it, as expected it was my fault. sbcl didn't complain about a slot having a wrong :type. ccl. does ;) 18:20:57 btw: thanks milanj & segv_ for your answers. i think i am a worst case of a user (a beginner who has finnished two apps under a different lisp-implemenation and now cries about not being able to run them under ccl) 18:27:48 now the app is working (slowly but no errors) ;) 18:33:25 -!- anRch [n=markmill@nmd.sbx07283.medfoma.wayport.net] has quit [] 18:52:22 yes, that is strange regarding sbcl, if you have in mind their "declarations as assertions" attitude 19:25:01 hm. i think this one is my fault, too. but maybe someone can give me some hints: running the sam application (particle-simulation) on the same machine - i see the following: 19:25:37 time ellapsed: sbcl 91sec ccl 420sec 19:26:00 (same number of particles) 19:27:24 maximum number of particles computed before memory error (stack exhausted): sbcl: 70000 ccl: 16000. 19:29:49 what do i have to take care of, when using ccl? do i have to be more declarative (types, optimization)? 19:32:55 i have to immagine it's numerically intensive? 19:33:03 sbcl's compiler is a lot better at that than ccl 19:33:19 though i didn't think it was 4 times better... 19:36:27 yes, it computes a lot of geometric-properties and relations. a lesser speed is something i could live with, but i wonder what makes the heap exploding so fast. (sbcl gc-time fraction is <10%, ccl gc-time fraction is ~30%) 19:40:46 can you preallocate the data? 19:40:53 (you'd lose the gc time that way) 19:41:09 ccl has a precise gc but i don't know how it compares to sbcl 19:41:34 sellout [n=greg@cpe-67-241-206-65.maine.res.rr.com] has joined #ccl 19:42:04 if gbyers is around he'd be able to tell everything you'd need to know (and then some) 19:43:27 mikesch [n=axel@xdsl-87-79-178-149.netcologne.de] has joined #ccl 19:44:52 hm. maybe somehow, for the sake of flexibility ... that would help to avoid heap-explosion, too. i suppose. 19:46:00 i don't think it's normal that you're runnig out of heap 19:46:05 (though i have no idea what you're doing) 19:49:17 -!- sellout [n=greg@cpe-67-241-206-65.maine.res.rr.com] has quit [Read error: 60 (Operation timed out)] 19:50:57 i have a list of list of particles which do touch each other (which is computed in the first step). i combine those lists to get the clusters. like: ((1) (2) (3 1) (4) (5 2) (6) (7 2) (8 1)) -> ((1 3 8) (2 5 7) (4) (6)) - i think this cluster-computation makes the heap-crashing. 19:51:07 i always experienced less memory consuption with ccl, thought i didnt do much of heavy numeric things .. 19:51:57 well, i don't know how many particles you're simulating, nor if the number increases with time, but if you're running out of heap maybe you're not freeing some memory you could be freeing? 19:52:09 though i'd really suggest using arrays for this instead of lists. 19:52:13 -!- Modius__ [n=Modius@adsl-66-143-165-178.dsl.austtx.swbell.net] has quit [SendQ exceeded] 19:52:25 arrays will be faster on sbcl as well. 19:52:37 and you can just preallocate a huge array at the beginning and reuse it 19:53:35 Modius__ [n=Modius@adsl-66-143-165-178.dsl.austtx.swbell.net] has joined #ccl 19:54:13 in the first step lots of geometric-objects are generated and tested if they fit to the current particle configuration. if they do fit - their contact-partners are stored in a list. after some "puts" the clusters are calculated. with time the number of particles increases, the cluster-number is increasing (and then decreasing) and the cluster-sizes are increasing until an upper limit (simulation-box). 19:54:37 ok, then i guess you just need a lot of ram 19:54:40 oh well. 19:56:25 on this machine i do have 4gb, which allows sbcl to calculate 70000 particles (i set a 1gb-ram-limit) - ccl crashes at 17000 (i set no ram-limit). 19:56:39 how big is the code? 19:56:52 ~3kLOC 19:57:01 can i try running it here? 19:57:51 you can - but it is still under development. 19:57:51 well, that actually wouldn't mean much. 19:57:59 right, and i don't have a whole lot of time right now. 19:58:04 i can send it to you permail 19:58:29 it is primitive-code - i am a lisp beginner. 19:58:31 have you tried profiling it? 19:59:06 i used some (time ...) to check several parts. 19:59:21 no, i mean something like oprofile or sb-sprof 19:59:28 per function time/memory usage 20:00:03 not really - i read in norvigs book about it - but as speed was ok for me - i didnt .... 20:00:21 well, it'll profile memory as well as speed 20:00:27 so you know where all your ram is going 20:00:50 try using the profiler in ccl/tools/profilere.lisp 20:00:55 (deterministic profiler) 20:00:59 i was running top in parallel - there was plenty of ram left. 20:01:13 didn't you start off saying you were running out of heap? 20:01:13 (but i fear this was the wron tool ;) 20:01:34 segv_: yes, ccl said so. 20:01:36 for some reason your ccl code is using a lot more ram than your sbcl code right? 20:01:51 the only way i can think of to track that down is to profile the code. 20:01:59 and see which functions allocate the most memory 20:02:10 ok. i will do so. 20:02:40 i thought, there may be a difference in running out of memory and running out of heap-space. 20:02:47 not really. 20:02:59 ccl grabs a chunk of memory when it starts up, that's all the ram you get. 20:04:30 will the profiler produce the same output as (time ...)? i ask because i do not really understand its output. 20:04:48 no, it will produce much more detailed output. 20:04:55 i'd really suggest giving it a try. 20:05:21 sellout [n=greg@cpe-67-241-206-65.maine.res.rr.com] has joined #ccl 20:07:13 trebor_dki: 20:07:16 http://ccl.clozure.com/ccl-documentation.html#Profiling-Using%20the%20Linux%20oprofile%20system-level%20profiler 20:07:52 ah, thanks, i was just about to ask ((apropos "profile") results in (swank-backend:...)) - thanks reading 20:09:07 you will need to apt-get install oprofile 20:09:21 or you can load the .asd in ccl/tools/advice-profiler/ 20:11:21 (that asd is for another, pure lisp, profiler) 20:11:30 oh, this is far more complicated as in "paradigms of artificial intelligence" (norvig). i have to ask the admin to install these tools. would the tools from norvig give a first glue, too? 20:11:44 i have no idea how norvig's profiler works. 20:11:59 if you want a pure lisp tool there's the advice-profiler i was talking about before. 20:12:03 i wasn't reading while typing - 20:13:00 i think the .asd-advice will be the thing i can try first (i am not root on either machine, here). 20:13:17 thanks. 20:13:42 good luck. 20:13:53 oprofile is a nice tool (and has far less overhead) 20:13:58 but it's something of a pain to setup 20:14:56 i have to buy some things, before shops are closing (otherwise my wife will be ... you know ;) thanks for your help. i will be back tomorrow (reporting) - ciao. 20:16:04 tschuss 20:28:26 -!- mikesch [n=axel@xdsl-87-79-178-149.netcologne.de] has quit [] 21:31:26 rme [n=rme@pool-70-105-127-222.chi.dsl-w.verizon.net] has joined #ccl 22:07:06 mikesch [n=axel@xdsl-87-79-178-149.netcologne.de] has joined #ccl 22:42:46 anRch [n=markmill@nmd.sbx07258.melroma.wayport.net] has joined #ccl 23:12:19 -!- anRch [n=markmill@nmd.sbx07258.melroma.wayport.net] has quit [] 23:34:26 sellout- [n=greg@cpe-67-241-206-65.maine.res.rr.com] has joined #ccl 23:35:54 -!- sellout [n=greg@cpe-67-241-206-65.maine.res.rr.com] has quit [Read error: 110 (Connection timed out)] 23:49:49 -!- mikesch [n=axel@xdsl-87-79-178-149.netcologne.de] has quit []