irclog2html for #brlcad on 20050829

01:02.07 Twingy hrm
01:10.59 Twingy almost ready to test my kd caching
01:11.26 Twingy shiping this data across the network and removing peak memory issue causes lots of grunt work
01:15.44 Twingy heh
01:15.53 Twingy 15 downloads of my audio data transfer library to date
01:16.45 Twingy 1 per month, woot
03:35.17 *** join/#brlcad PKMOBILE (n=Apathy@resnet-235-118.resnet.UMBC.EDU)
04:59.41 *** join/#brlcad CIA-5 (n=CIA@flapjack.navi.cx)
06:04.11 Twingy good deal, kd-tree caching works now
06:05.11 silvap do i get to quit m3 now?
06:06.00 Twingy we should talk about some of that memory alignment stuff
06:06.08 Twingy now that I can do clever stuff with the cache
06:06.26 silvap any way to abstract a few parameters out?
06:06.28 Twingy I still have a full day of work to do to the cache though
06:06.30 silvap to say a properties file
06:06.37 Twingy I want to pack in the data that was used to build the cache
06:06.38 silvap and then do some batch benches
06:06.44 Twingy and some other stuff to prevent wrong cache with wrong geometry
06:06.47 Twingy as well as endian
06:07.19 Twingy once I finish tacking on that other stuff I need to integrate the memory resizing stuff to keep peak memory from going any higher
06:07.31 Twingy abstract?
06:07.52 silvap pull the params out
06:07.56 silvap and read it at runtime
06:07.58 Twingy for the kd-tree?
06:08.04 silvap cache
06:08.13 Twingy well here's the deal
06:08.13 silvap at least temporarily
06:08.43 Twingy in my opinion a kd-tree builder is not complete until you can completely throw away the "depth" term
06:08.52 Twingy that is to say, it won't run on forever
06:09.04 Twingy so, I think once I get that one last piece of logic in
06:09.16 Twingy it should be building optimal kd-trees without any sort of 'tuning' for a scene
06:09.36 Twingy i.e the optimality of the kd-tree becomes a level of agressiveness
06:09.47 Twingy which I might expose
06:09.55 silvap k
06:09.56 Twingy right now adrt_bench is in charge of building the cache
06:10.16 Twingy by passing it the '-c' argument
06:10.21 Twingy it dumps out kdtree.cache
06:10.36 Twingy and if you put in: kdtree_cache_file,kdtree.cache
06:10.41 silvap make it so that u can variate align boundaries, packing size, etc
06:10.41 Twingy in the whatever.env file
06:10.45 silvap at least in bench
06:10.47 Twingy then it'll pull that and distribute it to the nodes
06:11.00 Twingy well, I wanted to talk to you about that before I do that
06:11.39 Twingy I haven't even optimized the cache loader yet
06:11.44 Twingy it's bloated and barely working
06:11.49 Twingy an the t62 cache loads in under a second
06:12.01 Twingy so I should have this thing screaming in no time
06:12.22 silvap ok
06:12.27 Twingy that still translates to 1 minute for stryker
06:12.39 Twingy so I should be able to very easily get that down to 10 - 15 seconds
06:12.59 Twingy I suspect I'll do a 3 hour build of the kd-tree for stryker
06:13.08 Twingy tommorrow before lunch
06:13.25 silvap i forget, is there a min tris bound for leaves now?
06:13.36 Twingy no, I cannot guarantee that
06:13.41 Twingy I wish I could though
06:13.48 Twingy cause then
06:13.52 Twingy I could pack the # of triangles
06:13.54 silvap but the tree is balanced ya?
06:13.57 Twingy into the lower bits of a ptr I have
06:14.02 Twingy that could store 0 - 7 triangles
06:14.10 Twingy that would be so sweet
06:14.23 silvap what does that buy u
06:14.26 Twingy but like I said, I have some nodes with thousands still left in them
06:14.39 Twingy that saves me 4 bytes of an int in every geometry node
06:14.51 Twingy 5 - 10% savings in memory
06:14.58 silvap on the avg case, does that save u anything
06:14.59 silvap k memory
06:15.22 Twingy mmm
06:15.30 Twingy so much fun stuff to work on now that this ugly part is over with
06:16.06 silvap cant have u two versions of tie_work running that take the num_tris packed value and the seperate num_tris value
06:16.19 Twingy atleast now if my pre-alpha version of this shot selection tool crashes while using it, I can reload in seconds instead of minutes
06:16.24 silvap hmm how to determine which to call
06:16.45 Twingy I don't understand
06:16.57 silvap nm confusing runtime with memory here
06:16.58 silvap carry on
06:17.00 Twingy k
06:17.07 Twingy I think for the time being
06:17.14 Twingy until I 'perfect' my kd-tree algorithm
06:17.19 Twingy I'm going to 'tune' a kd-tree for stryker
06:17.23 Twingy and just use that for a while
06:17.41 Twingy I'm seeing 1.2 - 2.5 million rays/sec on my desktop with t62 now
06:17.57 silvap what are the criterion being minimized/maximized?
06:17.57 Twingy using an optimal kd-tree with the new algorithm
06:18.06 silvap for this tuning process
06:18.10 Twingy I explaind the algorithm right?
06:18.11 silvap just load time?
06:18.41 silvap yea but not this tuning process
06:18.46 Twingy right now there are 5 values you can use
06:19.00 Twingy the number of initial slices from each orthogonal axis
06:19.06 Twingy known as max slices
06:19.14 Twingy which gets smaller as the # of triangles in a cell get smaller
06:19.18 Twingy min slices
06:19.27 Twingy as nodes get down to a fistful of triangles
06:19.40 Twingy min density, used to determine geometry span gaps
06:19.43 Twingy as a %
06:19.54 Twingy the min span a gap must be wrt the node dimension
06:20.17 Twingy and a scale coefficient which is the bias a node receives wrt to the center when no gaps are found
06:20.55 Twingy I can explain them more tommorrow
06:21.13 silvap k
06:21.16 Twingy there's going to be 2 more heuristics to replace 1 soon
06:21.21 Twingy as I make it even more aggressive
06:21.27 Twingy I think for the API
06:21.33 Twingy I'll have like a LOW, MED, and HIGH
06:21.41 Twingy in terms of how agressive it will be
06:21.50 Twingy and amount of memory that can be used
06:22.02 Twingy as a percentage of the geometry size
06:22.10 Twingy i.e 10% mem wrt geometry size
06:22.14 Twingy or 50%
06:22.16 Twingy etc.
06:22.26 Twingy so basically 2 arguments to engine
06:22.53 silvap too late
06:22.55 silvap talk tomorrow
06:22.58 Twingy k
06:23.02 silvap nite
06:23.07 Twingy if I'm in tommorrow :)
06:23.12 Twingy I should get to bed
06:37.47 CIA-5 BRL-CAD: 03twingy * 10brlcad/src/adrt/ (11 files in 3 dirs):
06:37.47 CIA-5 BRL-CAD: Integrated kd-tree caching.
06:37.47 CIA-5 BRL-CAD: It's generated with adrt_bench by supplying the -c argument.
06:37.47 CIA-5 BRL-CAD: kdtree_cache_file,kdtree.cache goes into project.env file to make it work.
06:38.58 Twingy mmm
06:47.10 Twingy ah
06:47.12 Twingy you stink
06:47.29 Twingy allz I got is my wimpy altec lansings
06:47.59 Twingy btw, connie called me yesterday and wants to know if you and I would go visit her in reno, heh
06:48.29 Twingy I think she was serious :)
06:51.36 Twingy you camping out at the office tonight?
06:55.40 brlcad nah, just working through the night
06:55.52 brlcad no smores
06:55.55 Twingy so should I tell her you'll be visiting reno in the near future? :)
06:56.01 brlcad hmm.. smores in the toaster
06:56.07 Twingy go fer it
06:56.18 Twingy just don't plug it into the ups
06:56.24 Twingy or my ups anyways
06:56.26 brlcad hehe
06:57.08 Twingy hehe
06:57.15 brlcad actually, I'll probably burn up half my vacation this week
06:57.24 brlcad going down south later this week
06:57.24 Twingy we could go gambling and lose lots of money, it could be fun!
06:57.33 Twingy :}
06:57.41 brlcad if I wanted to do that, atlantic city would be far quicker ;)
06:57.46 Twingy indeed
06:57.55 Twingy nicer too
06:57.59 Twingy with the ocean and all
06:58.11 brlcad trump plaza
06:58.15 brlcad trump taco bell
06:58.19 Twingy I won $150 from uncle trump
06:58.24 brlcad trump exxon
06:58.31 Twingy trump donuts
06:58.36 brlcad mm.. donuts
06:58.43 brlcad i knew i should have stopped on the way in
06:58.52 Twingy I went running today
06:58.56 brlcad i didn't
06:58.59 Twingy even though my knee is feeling all fucked up
06:59.05 brlcad eek
06:59.22 Twingy so if I'm hobbling everytime I stand up tommorrow you'll know why
06:59.33 Twingy dunno if the knee thing has anything to do with running
06:59.36 Twingy it's just off and on
06:59.44 brlcad i should go workout before everyone shows up in the morning
06:59.46 Twingy a few weeks it bothers me, a few weeks it's fine
07:00.02 brlcad Twingy's getting old
07:00.09 Twingy nah
07:00.15 brlcad heh
07:00.21 Twingy ooh
07:00.26 Twingy I get lower car insurance next month
07:03.29 brlcad lex is nice to use
07:03.33 brlcad bison is a pita
07:04.24 Twingy *sniffle* I just cannot get this crappy 3ghz p4 to go above 750k rays /sec
07:04.58 brlcad maybe if you kick it hard enough
07:05.43 Twingy maybe...
07:05.54 Twingy maybe I should overclock it to 80 gigglehurts
07:06.20 brlcad supercool it to 200F
07:06.40 brlcad watch the bits just ooze all over the place
07:08.18 Twingy blah
07:08.18 Twingy groove salad just died
07:08.33 Twingy ah, there we go
07:10.22 *** join/#brlcad Maloeran (n=alexis@modemcable233.164-81-70.mc.videotron.ca)
07:10.45 Twingy welp, I think I'm gonna hit the sack now
07:10.58 brlcad cya
07:31.36 CIA-5 BRL-CAD: 03twingy * 10brlcad/src/adrt/libtie/define.h: defaults at biggest bang for the buck right now.
11:36.55 *** join/#brlcad ibot (i=ibot@pdpc/supporter/active/TimRiker/bot/apt)
11:58.46 *** join/#brlcad d_rossberg (n=c28bf505@bz.bzflag.bz)
12:04.51 brlcad hello d_rossberg!
12:08.36 d_rossberg hello brlcad!
12:18.15 brlcad the windows release seems to be coming together at last
12:18.43 brlcad between your changes, Bob's, and mine it looks like it won't be long before binaries are posted
12:19.10 brlcad possibly even next iteration/month
12:20.02 d_rossberg NOW libbn, libbu, librt, libsysv and libwdb should work with MS Visual Studio
12:20.19 brlcad yeah, I noticed the commits
12:20.22 d_rossberg at least my viewer application works with it
12:20.37 brlcad CIA bot is offline right now or we'd see here too :/
12:20.42 brlcad excellent
12:21.55 d_rossberg what's the difference between regex.h and brlregex.h? (the last one isnt't used any more)
12:22.39 brlcad depending on your version of studio the regex.h would conflict with a system header
12:22.53 brlcad there were also a few unix boxes that had the problem apparently
12:23.00 brlcad but I never encountered the problem myself
12:23.06 brlcad it "should" go away if it can
12:23.32 brlcad the "real" regex sources are in src/other/libregex
12:23.51 d_rossberg may i remove brlregex.h from the repository?
12:23.58 brlcad sysv files will disappear some day soon
12:24.51 brlcad i believe it's safe to delete brlregex.h
12:25.00 brlcad checking where it's referenced now
12:26.11 brlcad hmm.. interesting, only the old sysv files
12:26.17 brlcad sure, go ahead :)
12:27.15 d_rossberg until today brlregex.h was only used within ifdef (_WIN32) and only in files not contained in the MS project files, i'll remove the header
12:27.43 brlcad yeah, that goes back several years
12:27.56 brlcad vc6 iirc
12:32.00 d_rossberg all the adaptation for version 7 are in regex.h, so it looks like nobody has used brlregex.h since then
12:37.50 brlcad yeah, it should work fine
12:38.28 brlcad i use the same files in another popular project that is cross-platform to windows and works just fine there
12:59.21 *** join/#brlcad ChanServ (ChanServ@services.)
12:59.21 *** mode/#brlcad [+o ChanServ] by irc.freenode.net
13:00.20 *** mode/#brlcad [+o brlcad] by ChanServ
13:15.44 *** join/#brlcad archivist_ (n=archivis@host217-35-76-52.in-addr.btopenworld.com) [NETSPLIT VICTIM]
13:15.44 *** join/#brlcad Maloeran (n=alexis@modemcable233.164-81-70.mc.videotron.ca) [NETSPLIT VICTIM]
13:15.44 *** join/#brlcad d_rossberg (n=c28bf505@bz.bzflag.bz) [NETSPLIT VICTIM]
13:16.23 *** join/#brlcad learner (n=brlcad@pdpc/supporter/silver/brlcad) [NETSPLIT VICTIM]
13:16.23 *** join/#brlcad Twingy (n=justin@pcp0011647505pcs.aberdn01.md.comcast.net) [NETSPLIT VICTIM]
13:16.23 *** join/#brlcad brlcad (n=brlcad@pdpc/supporter/silver/brlcad) [NETSPLIT VICTIM]
13:16.23 *** mode/#brlcad [+oo learner brlcad] by irc.freenode.net
13:16.24 *** join/#brlcad ChanServ (ChanServ@services.)
13:16.24 *** mode/#brlcad [+o ChanServ] by irc.freenode.net
13:16.28 *** join/#brlcad Erroneous (n=Daniel_R@DHCP-170-143.caltech.edu) [NETSPLIT VICTIM]
13:16.28 *** join/#brlcad archivist (n=archivis@host217-35-103-47.in-addr.btopenworld.com) [NETSPLIT VICTIM]
13:17.33 *** join/#brlcad silvap_ (n=prasad@141.157.107.76) [NETSPLIT VICTIM]
13:17.33 *** join/#brlcad Semhirage (i=semhirag@unaffiliated/semhirage) [NETSPLIT VICTIM]
13:20.30 *** join/#brlcad ChanServ (ChanServ@services.)
13:20.30 *** join/#brlcad archivist_ (n=archivis@host217-35-76-52.in-addr.btopenworld.com) [NETSPLIT VICTIM]
13:20.30 *** join/#brlcad Maloeran (n=alexis@modemcable233.164-81-70.mc.videotron.ca) [NETSPLIT VICTIM]
13:20.30 *** join/#brlcad d_rossberg (n=c28bf505@bz.bzflag.bz) [NETSPLIT VICTIM]
13:20.30 *** mode/#brlcad [+o ChanServ] by irc.freenode.net
13:21.02 *** join/#brlcad learner (n=brlcad@pdpc/supporter/silver/brlcad) [NETSPLIT VICTIM]
13:21.02 *** join/#brlcad Twingy (n=justin@pcp0011647505pcs.aberdn01.md.comcast.net) [NETSPLIT VICTIM]
13:21.02 *** join/#brlcad brlcad (n=brlcad@pdpc/supporter/silver/brlcad) [NETSPLIT VICTIM]
13:21.02 *** mode/#brlcad [+oo learner brlcad] by irc.freenode.net
13:21.13 *** join/#brlcad silvap_ (n=prasad@141.157.107.76) [NETSPLIT VICTIM]
13:21.13 *** join/#brlcad Semhirage (i=semhirag@unaffiliated/semhirage) [NETSPLIT VICTIM]
13:22.54 *** join/#brlcad Erroneous (n=Daniel_R@DHCP-170-143.caltech.edu) [NETSPLIT VICTIM]
13:22.54 *** join/#brlcad archivist (n=archivis@host217-35-103-47.in-addr.btopenworld.com) [NETSPLIT VICTIM]
13:23.02 *** join/#brlcad archivist_ (n=archivis@host217-35-76-52.in-addr.btopenworld.com) [NETSPLIT VICTIM]
13:23.26 *** join/#brlcad Maloeran (n=alexis@modemcable233.164-81-70.mc.videotron.ca) [NETSPLIT VICTIM]
13:23.56 *** join/#brlcad archivist_ (n=archivis@host217-35-76-52.in-addr.btopenworld.com) [NETSPLIT VICTIM]
13:24.00 *** join/#brlcad Maloeran (n=alexis@modemcable233.164-81-70.mc.videotron.ca) [NETSPLIT VICTIM]
13:55.31 *** join/#brlcad d_rossberg (n=c28bf505@bz.bzflag.bz)
14:27.00 *** join/#brlcad Twingy_ (n=justin@pcp0011647505pcs.aberdn01.md.comcast.net)
14:36.13 *** join/#brlcad TheLastSpartan (i=guu@myth.gibbscam.com)
14:39.14 *** join/#brlcad DTRemenak (n=DTRemena@DHCP-170-143.caltech.edu)
14:44.58 *** join/#brlcad ChanServ (ChanServ@services.)
14:44.58 *** join/#brlcad Maloeran (n=alexis@modemcable233.164-81-70.mc.videotron.ca) [NETSPLIT VICTIM]
14:44.58 *** join/#brlcad d_rossberg (n=c28bf505@bz.bzflag.bz) [NETSPLIT VICTIM]
14:44.58 *** join/#brlcad Twingy_ (n=justin@pcp0011647505pcs.aberdn01.md.comcast.net) [NETSPLIT VICTIM]
14:44.58 *** join/#brlcad DTRemenak (n=DTRemena@DHCP-170-143.caltech.edu)
14:44.58 *** join/#brlcad archivist_ (n=archivis@host217-35-76-52.in-addr.btopenworld.com) [NETSPLIT VICTIM]
14:44.58 *** mode/#brlcad [+o ChanServ] by irc.freenode.net
14:45.02 *** join/#brlcad learner (n=brlcad@pdpc/supporter/silver/brlcad) [NETSPLIT VICTIM]
14:45.02 *** join/#brlcad Twingy (n=justin@pcp0011647505pcs.aberdn01.md.comcast.net) [NETSPLIT VICTIM]
14:45.02 *** join/#brlcad brlcad (n=brlcad@pdpc/supporter/silver/brlcad) [NETSPLIT VICTIM]
14:45.02 *** mode/#brlcad [+oo learner brlcad] by irc.freenode.net
14:45.11 *** join/#brlcad archivist (n=archivis@host217-35-103-47.in-addr.btopenworld.com) [NETSPLIT VICTIM]
14:45.38 *** join/#brlcad silvap_ (n=prasad@141.157.107.76) [NETSPLIT VICTIM]
14:45.39 *** join/#brlcad Semhirage (i=semhirag@unaffiliated/semhirage) [NETSPLIT VICTIM]
15:09.00 *** join/#brlcad archivist_ (n=archivis@host217-35-76-52.in-addr.btopenworld.com) [NETSPLIT VICTIM]
15:09.00 *** join/#brlcad DTRemenak (n=DTRemena@DHCP-170-143.caltech.edu)
15:09.07 *** join/#brlcad Twingy_ (n=justin@pcp0011647505pcs.aberdn01.md.comcast.net) [NETSPLIT VICTIM]
15:09.07 *** join/#brlcad Maloeran (n=alexis@modemcable233.164-81-70.mc.videotron.ca) [NETSPLIT VICTIM]
15:09.32 *** join/#brlcad polyspin (n=butler@bz.bzflag.bz)
15:30.35 *** join/#brlcad Erroneous (n=Daniel_R@DHCP-170-143.caltech.edu)
16:18.42 *** join/#brlcad CIA-4 (n=CIA@flapjack.navi.cx)
16:25.47 CIA-4 BRL-CAD: 03twingy * 10brlcad/src/adrt/doc/binary_spec.txt: updated.
18:05.09 *** join/#brlcad Metthyn (n=metintin@85.101.255.212)
18:05.14 *** part/#brlcad Metthyn (n=metintin@85.101.255.212)
18:56.49 CIA-4 BRL-CAD: 03lbutler * 10brlcad/src/librt/vlist.c: Fixed bug in overlay code when EOF encountered
18:58.15 CIA-4 BRL-CAD: 03lbutler * 10brlcad/NEWS: overlay command in mged fixed
19:26.26 CIA-4 BRL-CAD: 03twingy * 10brlcad/src/adrt/ (5 files in 2 dirs):
19:26.26 CIA-4 BRL-CAD: kd-tree caching is now tuned to the point where full start-up time takes less than
19:26.26 CIA-4 BRL-CAD: 1/4 the time as it did with the fast mid-split algorithm. This number should approach
19:26.26 CIA-4 BRL-CAD: 1/8 as the various parts of the packing/unpacking and caching algorithms are tuned
19:26.26 CIA-4 BRL-CAD: even further.
19:53.07 *** join/#brlcad TheLastSpartan (i=guu@myth.gibbscam.com)
20:33.27 *** join/#brlcad docelic (n=docelic@195.246.23.200)
20:44.14 CIA-4 BRL-CAD: 03lbutler * 10brlcad/regress/nightly.sh: Added checks for host not known, and build failure
21:33.18 CIA-4 BRL-CAD: 03lbutler * 10brlcad/src/libbu/linebuf.c: Cannot return a void explicitly on Irix
22:54.52 *** join/#brlcad ChanServ (ChanServ@services.)
22:54.52 *** join/#brlcad brlcad (n=brlcad@pdpc/supporter/silver/brlcad)
22:54.53 *** join/#brlcad Twingy (n=justin@pcp0011647505pcs.aberdn01.md.comcast.net)
22:54.53 *** join/#brlcad archivist (n=archivis@host217-35-103-47.in-addr.btopenworld.com) [NETSPLIT VICTIM]
22:54.53 *** join/#brlcad learner (n=brlcad@pdpc/supporter/silver/brlcad)
22:54.53 *** mode/#brlcad [+ooo ChanServ brlcad learner] by irc.freenode.net
22:55.18 *** join/#brlcad docelic (n=docelic@195.246.23.200) [NETSPLIT VICTIM]
22:55.18 *** join/#brlcad silvap_ (n=prasad@141.157.107.76) [NETSPLIT VICTIM]
22:55.19 *** join/#brlcad Semhirage (i=semhirag@unaffiliated/semhirage) [NETSPLIT VICTIM]
22:56.02 *** join/#brlcad polyspin (n=butler@bz.bzflag.bz) [NETSPLIT VICTIM]
22:56.04 *** join/#brlcad TheLastSpartan (i=guu@myth.gibbscam.com)
22:56.04 *** join/#brlcad DTRemenak (n=Daniel_R@DHCP-170-143.caltech.edu)
22:56.04 *** join/#brlcad archivist_ (n=archivis@host217-35-76-52.in-addr.btopenworld.com) [NETSPLIT VICTIM]
22:56.32 *** join/#brlcad CIA-4 (n=CIA@flapjack.navi.cx)
22:56.32 *** join/#brlcad Twingy_ (n=justin@pcp0011647505pcs.aberdn01.md.comcast.net) [NETSPLIT VICTIM]
22:56.32 *** join/#brlcad Maloeran (n=alexis@modemcable233.164-81-70.mc.videotron.ca) [NETSPLIT VICTIM]
22:57.54 *** join/#brlcad brlcad (n=brlcad@bz.bzflag.bz) [NETSPLIT VICTIM]
22:58.02 *** join/#brlcad archivist (n=archivis@host217-35-103-47.in-addr.btopenworld.com) [NETSPLIT VICTIM]
22:58.48 *** join/#brlcad Twingy (n=justin@pcp0011647505pcs.aberdn01.md.comcast.net)
22:58.59 *** join/#brlcad learner (n=brlcad@pcp0011649376pcs.aberdn01.md.comcast.net) [NETSPLIT VICTIM]
22:58.59 *** mode/#brlcad [+o learner] by ChanServ
23:19.24 *** join/#brlcad archivist (n=archivis@host217-35-103-47.in-addr.btopenworld.com) [NETSPLIT VICTIM]

Generated by irclog2html.pl by Jeff Waugh - find it at freshmeat.net! Modified by Tim Riker to work with blootbot logs, split per channel, etc.