irclog2html for #brlcad on 20050829

01:02.07Twingyhrm
01:10.59Twingyalmost ready to test my kd caching
01:11.26Twingyshiping this data across the network and removing peak memory issue causes lots of grunt work
01:15.44Twingyheh
01:15.53Twingy15 downloads of my audio data transfer library to date
01:16.45Twingy1 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)
06:04.11Twingygood deal, kd-tree caching works now
06:05.11silvapdo i get to quit m3 now?
06:06.00Twingywe should talk about some of that memory alignment stuff
06:06.08Twingynow that I can do clever stuff with the cache
06:06.26silvapany way to abstract a few parameters out?
06:06.28TwingyI still have a full day of work to do to the cache though
06:06.30silvapto say a properties file
06:06.37TwingyI want to pack in the data that was used to build the cache
06:06.38silvapand then do some batch benches
06:06.44Twingyand some other stuff to prevent wrong cache with wrong geometry
06:06.47Twingyas well as endian
06:07.19Twingyonce I finish tacking on that other stuff I need to integrate the memory resizing stuff to keep peak memory from going any higher
06:07.31Twingyabstract?
06:07.52silvappull the params out
06:07.56silvapand read it at runtime
06:07.58Twingyfor the kd-tree?
06:08.04silvapcache
06:08.13Twingywell here's the deal
06:08.13silvapat least temporarily
06:08.43Twingyin my opinion a kd-tree builder is not complete until you can completely throw away the "depth" term
06:08.52Twingythat is to say, it won't run on forever
06:09.04Twingyso, I think once I get that one last piece of logic in
06:09.16Twingyit should be building optimal kd-trees without any sort of 'tuning' for a scene
06:09.36Twingyi.e the optimality of the kd-tree becomes a level of agressiveness
06:09.47Twingywhich I might expose
06:09.55silvapk
06:09.56Twingyright now adrt_bench is in charge of building the cache
06:10.16Twingyby passing it the '-c' argument
06:10.21Twingyit dumps out kdtree.cache
06:10.36Twingyand if you put in: kdtree_cache_file,kdtree.cache
06:10.41silvapmake it so that u can variate align boundaries, packing size, etc
06:10.41Twingyin the whatever.env file
06:10.45silvapat least in bench
06:10.47Twingythen it'll pull that and distribute it to the nodes
06:11.00Twingywell, I wanted to talk to you about that before I do that
06:11.39TwingyI haven't even optimized the cache loader yet
06:11.44Twingyit's bloated and barely working
06:11.49Twingyan the t62 cache loads in under a second
06:12.01Twingyso I should have this thing screaming in no time
06:12.22silvapok
06:12.27Twingythat still translates to 1 minute for stryker
06:12.39Twingyso I should be able to very easily get that down to 10 - 15 seconds
06:12.59TwingyI suspect I'll do a 3 hour build of the kd-tree for stryker
06:13.08Twingytommorrow before lunch
06:13.25silvapi forget, is there a min tris bound for leaves now?
06:13.36Twingyno, I cannot guarantee that
06:13.41TwingyI wish I could though
06:13.48Twingycause then
06:13.52TwingyI could pack the # of triangles
06:13.54silvapbut the tree is balanced ya?
06:13.57Twingyinto the lower bits of a ptr I have
06:14.02Twingythat could store 0 - 7 triangles
06:14.10Twingythat would be so sweet
06:14.23silvapwhat does that buy u
06:14.26Twingybut like I said, I have some nodes with thousands still left in them
06:14.39Twingythat saves me 4 bytes of an int in every geometry node
06:14.51Twingy5 - 10% savings in memory
06:14.58silvapon the avg case, does that save u anything
06:14.59silvapk memory
06:15.22Twingymmm
06:15.30Twingyso much fun stuff to work on now that this ugly part is over with
06:16.06silvapcant have u two versions of tie_work running that take the num_tris packed value and the seperate num_tris value
06:16.19Twingyatleast now if my pre-alpha version of this shot selection tool crashes while using it, I can reload in seconds instead of minutes
06:16.24silvaphmm how to determine which to call
06:16.45TwingyI don't understand
06:16.57silvapnm confusing runtime with memory here
06:16.58silvapcarry on
06:17.00Twingyk
06:17.07TwingyI think for the time being
06:17.14Twingyuntil I 'perfect' my kd-tree algorithm
06:17.19TwingyI'm going to 'tune' a kd-tree for stryker
06:17.23Twingyand just use that for a while
06:17.41TwingyI'm seeing 1.2 - 2.5 million rays/sec on my desktop with t62 now
06:17.57silvapwhat are the criterion being minimized/maximized?
06:17.57Twingyusing an optimal kd-tree with the new algorithm
06:18.06silvapfor this tuning process
06:18.10TwingyI explaind the algorithm right?
06:18.11silvapjust load time?
06:18.41silvapyea but not this tuning process
06:18.46Twingyright now there are 5 values you can use
06:19.00Twingythe number of initial slices from each orthogonal axis
06:19.06Twingyknown as max slices
06:19.14Twingywhich gets smaller as the # of triangles in a cell get smaller
06:19.18Twingymin slices
06:19.27Twingyas nodes get down to a fistful of triangles
06:19.40Twingymin density, used to determine geometry span gaps
06:19.43Twingyas a %
06:19.54Twingythe min span a gap must be wrt the node dimension
06:20.17Twingyand a scale coefficient which is the bias a node receives wrt to the center when no gaps are found
06:20.55TwingyI can explain them more tommorrow
06:21.13silvapk
06:21.16Twingythere's going to be 2 more heuristics to replace 1 soon
06:21.21Twingyas I make it even more aggressive
06:21.27TwingyI think for the API
06:21.33TwingyI'll have like a LOW, MED, and HIGH
06:21.41Twingyin terms of how agressive it will be
06:21.50Twingyand amount of memory that can be used
06:22.02Twingyas a percentage of the geometry size
06:22.10Twingyi.e 10% mem wrt geometry size
06:22.14Twingyor 50%
06:22.16Twingyetc.
06:22.26Twingyso basically 2 arguments to engine
06:22.53silvaptoo late
06:22.55silvaptalk tomorrow
06:22.58Twingyk
06:23.02silvapnite
06:23.07Twingyif I'm in tommorrow :)
06:23.12TwingyI should get to bed
06:37.47CIA-5BRL-CAD: 03twingy * 10brlcad/src/adrt/ (11 files in 3 dirs):
06:37.47CIA-5BRL-CAD: Integrated kd-tree caching.
06:37.47CIA-5BRL-CAD: It's generated with adrt_bench by supplying the -c argument.
06:37.47CIA-5BRL-CAD: kdtree_cache_file,kdtree.cache goes into project.env file to make it work.
06:38.58Twingymmm
06:47.10Twingyah
06:47.12Twingyyou stink
06:47.29Twingyallz I got is my wimpy altec lansings
06:47.59Twingybtw, connie called me yesterday and wants to know if you and I would go visit her in reno, heh
06:48.29TwingyI think she was serious :)
06:51.36Twingyyou camping out at the office tonight?
06:55.40brlcadnah, just working through the night
06:55.52brlcadno smores
06:55.55Twingyso should I tell her you'll be visiting reno in the near future? :)
06:56.01brlcadhmm.. smores in the toaster
06:56.07Twingygo fer it
06:56.18Twingyjust don't plug it into the ups
06:56.24Twingyor my ups anyways
06:56.26brlcadhehe
06:57.08Twingyhehe
06:57.15brlcadactually, I'll probably burn up half my vacation this week
06:57.24brlcadgoing down south later this week
06:57.24Twingywe could go gambling and lose lots of money, it could be fun!
06:57.33Twingy:}
06:57.41brlcadif I wanted to do that, atlantic city would be far quicker ;)
06:57.46Twingyindeed
06:57.55Twingynicer too
06:57.59Twingywith the ocean and all
06:58.11brlcadtrump plaza
06:58.15brlcadtrump taco bell
06:58.19TwingyI won $150 from uncle trump
06:58.24brlcadtrump exxon
06:58.31Twingytrump donuts
06:58.36brlcadmm.. donuts
06:58.43brlcadi knew i should have stopped on the way in
06:58.52TwingyI went running today
06:58.56brlcadi didn't
06:58.59Twingyeven though my knee is feeling all fucked up
06:59.05brlcadeek
06:59.22Twingyso if I'm hobbling everytime I stand up tommorrow you'll know why
06:59.33Twingydunno if the knee thing has anything to do with running
06:59.36Twingyit's just off and on
06:59.44brlcadi should go workout before everyone shows up in the morning
06:59.46Twingya few weeks it bothers me, a few weeks it's fine
07:00.02brlcadTwingy's getting old
07:00.09Twingynah
07:00.15brlcadheh
07:00.21Twingyooh
07:00.26TwingyI get lower car insurance next month
07:03.29brlcadlex is nice to use
07:03.33brlcadbison is a pita
07:04.24Twingy*sniffle* I just cannot get this crappy 3ghz p4 to go above 750k rays /sec
07:04.58brlcadmaybe if you kick it hard enough
07:05.43Twingymaybe...
07:05.54Twingymaybe I should overclock it to 80 gigglehurts
07:06.20brlcadsupercool it to 200F
07:06.40brlcadwatch the bits just ooze all over the place
07:08.18Twingyblah
07:08.18Twingygroove salad just died
07:08.33Twingyah, there we go
07:10.22*** join/#brlcad Maloeran (n=alexis@modemcable233.164-81-70.mc.videotron.ca)
07:10.45Twingywelp, I think I'm gonna hit the sack now
07:10.58brlcadcya
07:31.36CIA-5BRL-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)
12:04.51brlcadhello d_rossberg!
12:08.36d_rossberghello brlcad!
12:18.15brlcadthe windows release seems to be coming together at last
12:18.43brlcadbetween your changes, Bob's, and mine it looks like it won't be long before binaries are posted
12:19.10brlcadpossibly even next iteration/month
12:20.02d_rossbergNOW libbn, libbu, librt, libsysv and libwdb should work with MS Visual Studio
12:20.19brlcadyeah, I noticed the commits
12:20.22d_rossbergat least my viewer application works with it
12:20.37brlcadCIA bot is offline right now or we'd see here too :/
12:20.42brlcadexcellent
12:21.55d_rossbergwhat's the difference between regex.h and brlregex.h? (the last one isnt't used any more)
12:22.39brlcaddepending on your version of studio the regex.h would conflict with a system header
12:22.53brlcadthere were also a few unix boxes that had the problem apparently
12:23.00brlcadbut I never encountered the problem myself
12:23.06brlcadit "should" go away if it can
12:23.32brlcadthe "real" regex sources are in src/other/libregex
12:23.51d_rossbergmay i remove brlregex.h from the repository?
12:23.58brlcadsysv files will disappear some day soon
12:24.51brlcadi believe it's safe to delete brlregex.h
12:25.00brlcadchecking where it's referenced now
12:26.11brlcadhmm.. interesting, only the old sysv files
12:26.17brlcadsure, go ahead :)
12:27.15d_rossberguntil 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
12:27.43brlcadyeah, that goes back several years
12:27.56brlcadvc6 iirc
12:32.00d_rossbergall the adaptation for version 7 are in regex.h, so it looks like nobody has used brlregex.h since then
12:37.50brlcadyeah, it should work fine
12:38.28brlcadi 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)
16:25.47CIA-4BRL-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)
18:56.49CIA-4BRL-CAD: 03lbutler * 10brlcad/src/librt/vlist.c: Fixed bug in overlay code when EOF encountered
18:58.15CIA-4BRL-CAD: 03lbutler * 10brlcad/NEWS: overlay command in mged fixed
19:26.26CIA-4BRL-CAD: 03twingy * 10brlcad/src/adrt/ (5 files in 2 dirs):
19:26.26CIA-4BRL-CAD: kd-tree caching is now tuned to the point where full start-up time takes less than
19:26.26CIA-4BRL-CAD: 1/4 the time as it did with the fast mid-split algorithm. This number should approach
19:26.26CIA-4BRL-CAD: 1/8 as the various parts of the packing/unpacking and caching algorithms are tuned
19:26.26CIA-4BRL-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)
20:44.14CIA-4BRL-CAD: 03lbutler * 10brlcad/regress/nightly.sh: Added checks for host not known, and build failure
21:33.18CIA-4BRL-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.