IRC log for #brlcad on 20080304

02:37.55brlcadnope
03:17.28*** join/#brlcad Twingy (n=justin@74.92.144.217)
06:32.13*** join/#brlcad Z80-Boy (n=clock@217-162-111-154.dclient.hispeed.ch)
06:32.38Z80-BoyMy infamous model: Frame  0:    4720128 rays   in  26011.79 sec =       181.46 rays/sec (wallclock)
06:32.46Z80-BoyAnd that's a COre 2 Duo 2.2 GHz
07:33.36*** join/#brlcad Axman6 (n=Axman6@pdpc/supporter/student/Axman6)
07:59.53*** join/#brlcad clock_ (n=clock@zux221-122-143.adsl.green.ch)
08:00.45clock_hi all
10:47.28*** join/#brlcad elite01 (n=elite01@dslb-088-070-001-048.pools.arcor-ip.net)
11:25.58*** join/#brlcad d_rossberg (n=rossberg@bz.bzflag.bz)
11:34.52brlcadwow, that is incredibly slow
11:38.47clock_brlcad: I hope you can look at the model which causes this and see the cause
11:38.53clock_now I am leaving for lunch sorry
12:28.30*** join/#brlcad illethal (n=oden@c-69-137-199-63.hsd1.fl.comcast.net)
12:49.06*** join/#brlcad Elperion (n=Bary@p54877627.dip.t-dialin.net)
13:39.18illethalHello fellow humans of Earth!
13:40.48``Erik /det
13:41.11``Erikheh
13:41.19``Erikno hello for me, illethal? *cry* :D
14:02.42*** join/#brlcad elite01_ (n=elite01@dslb-088-070-001-048.pools.arcor-ip.net)
14:05.25*** join/#brlcad clock_ (n=clock@zux221-122-143.adsl.green.ch)
14:46.08*** join/#brlcad MinuteEl1ctron (n=MinuteEl@bz.bzflag.bz)
14:58.25*** join/#brlcad MinuteElectron (n=MinuteEl@silentflame/member/pdpc.base.minuteelectron)
15:04.30*** join/#brlcad poolio_ (n=poolio@bz.bzflag.bz)
15:04.48*** join/#brlcad MinuteElectron (n=MinuteEl@silentflame/member/pdpc.base.minuteelectron)
15:08.21*** join/#brlcad cad84 (n=91486201@bz.bzflag.bz)
15:13.49*** join/#brlcad d_rossberg (n=rossberg@bz.bzflag.bz)
15:16.58CIA-32BRL-CAD: 03d_rossberg * r30449 10/brlcad/trunk/src/irprep/Makefile.am: put irdisp neatly under WITH_X11 condition
15:23.36*** join/#brlcad elite01 (n=elite01@dslb-088-070-001-048.pools.arcor-ip.net)
15:37.46*** join/#brlcad brlcad (n=sean@bz.bzflag.bz)
15:49.14*** join/#brlcad tofu (n=sean@bz.bzflag.bz)
15:51.55``Erikquit pissing peer off, dude, he'll jack you up
15:53.09*** join/#brlcad jgay (n=jgay@fsf/staff/jgay)
15:56.45clock_brlcad: yes I have 118 pixels per second
15:57.00clock_brlcad: it's the strange model that causes 210x slowdown
15:57.28clock_http://ronja.twibright.com/3d/headcut.g object name headcut
15:58.59brlcadclock_: I'm aware of the model, and had looked at it
15:59.02brlcadand commented on it
15:59.08brlcador have you forgotten?? :)
15:59.17``Erikneat, bus error
15:59.20clock_I faintly remember you have commented on it
15:59.22clock_but not what you said
15:59.30clock_A fact is it's still hell slow
15:59.36clock_is it gonna stay this slow?
16:00.44brlcadit's slow for several reasons -- the biggest of which relates to how it's modeled -- you can get the same shape with a different hierarchy and probably get back much of the performance
16:00.49brlcadyou're more than 210x slower
16:00.55brlcadyou're mixing two different issues
16:01.30clock_can't the program transform the hierarchy into the more efficient one automatically?
16:01.42``Erikcuz that's insanely complex to do?
16:01.49brlcadclock_: go for it
16:02.21brlcad:)
16:02.22clock_So now I have to do a insanely complex thing with my model to get back some of the performance?
16:02.41``Erikinsanely complex to automate correctly
16:02.58clock_how should I change the hierarchy?
16:03.09``Erikif we could do that reasonably, we could also convert polygon soup to csg reasonably
16:04.36brlcadusually just balancing the csg tree will do wonders particularly if it's deep
16:05.05brlcadbut the better opt involves looking at the shapes and bounding boxes of the shapes being used
16:05.05clock_No it's shallow
16:05.07clock_like 3 or 4
16:05.32brlcader, not that I recall..
16:06.22brlcadgah, it's hard as heck to chat right now
16:06.30brlcadisp is being attacked
16:06.34``Erikwell, it may be disguised
16:06.51``Erikif you, say, union 16 things all together at once, it looks like one level, but it's actually at least 3
16:07.10``Eriker, 5
16:07.32clock_But it has a lot of things in one node
16:07.36``Erikthe actual csg structure is all leaf, unary or binary
16:07.38brlcadand depending on the ops and the way they're input, you might even end up with 16 levels
16:08.00clock_Balancing a tree isn't an insanely complex operation
16:08.11clock_It has the insane complexity of second year of the university
16:08.14``Erikbut it's an ordered tree
16:08.25clock_what's an ordered tree?
16:08.26``Erikif you change the order or even position, you change the semantic
16:08.47clock_not for unions
16:09.04brlcadclock_: so then seriously .. code it up, you're being given the simple explanation just for sake of explaining it
16:09.04clock_not for logical ANDs
16:09.26``Erikok, imagine if you have, say, a union, then you use that union to cut from another object, that makes one shape... but if you cut, then union, say through a tree balancing, it's the same operations, but you get a different shape
16:09.48clock_I have an impression now that BRL-CAD contains some kind of simple straightforward unoptimized implementation that goes off on heavily unbalanced trees
16:09.52``Erikand I think for simple union, we do balance?
16:10.44brlcadclock_: actually the csg processing is *heavily* optimized, one of the tightest you'll find anywhere
16:10.56brlcadbut in this particular instance, there is other stuff going on
16:11.20``Erikthe next step in optimizing is worth a fair number of doctorates, I'd imagine :/
16:11.22brlcadseveral issues, one being the construction hierarchy, other issues being the bounding volumes iirc
16:11.40clock_Do I imagine CSG correctly as firing a ray, intersecting with all primitives their bounding boxes the ray hits, and then finding the first visible intersect according to the logical operations?
16:11.44``Erikperhaps np complete, even
16:12.21clock_My scene has no reason to be difficult
16:12.28clock_It's a stack of cylindrical slices
16:12.36clock_and the stack is cut in half with a large primitive
16:12.42brlcadclock_: no, it's considerably more complex than that -- there is spatial partitioning to determine which bounding boxes to test against
16:12.46clock_Every cylindrical slice has a tiny bounding box
16:13.00brlcadif you did what you suggest of "intersecting with all primitives their bounding boxes the ray hits" it would be horrendously slow
16:13.14clock_so every ray hits only through small amount of slices unless you are looking close to the axis
16:13.22brlcaddepending on how you actually determine which bounding boxes it intersects
16:13.26clock_The deadly picture isn't AFAIK looking along the axis
16:14.00``Erikwhoa
16:14.11brlcadclock_: I believe itss that large primitive that is killing it
16:14.26clock_is it so complex to calculate an intersection with a cube?
16:14.28brlcadby doing that, it basically has to evaluate every primitive every time
16:14.33``Erikdoes  'head' raytrace fast?
16:14.40clock_yes
16:14.42clock_fast as hell
16:14.44``Erik(and did anyone do a tree on head? jfc)
16:15.43clock_Or maybe the bounding boxes are evaluated too liberal?
16:15.58clock_Like saying a tiny slice of metal with a huge negative box give a bounding box of the negative box?
16:16.22``ErikFrame  0:     263441 rays   in     12.24 sec =     21527.96 rays/sec (RTFM)
16:16.25clock_For an AND it's good to at least take the bounding box of the smaller part
16:17.27``Erikdefinitely the massive boolean overload from having the slice at that position :/
16:17.46clock_I can't put the slice elsewhere sorry
16:17.57clock_The idea is to show where the bolts are going through
16:18.05``Erikwithout the cutter
16:18.06``ErikFrame  0:     262165 rays   in      0.04 sec =   6802545.46 rays/sec (RTFM)
16:18.18clock_why do you get 21527 rays and me 118?
16:18.23clock_I have 2.2GHz Core Duo
16:18.28``Erikcuz my machine is moar awesomer
16:18.31clock_What do you have?
16:18.35``Erik8 3ghz cores
16:18.41illethalThat is awesome.
16:18.44illethalOctcore?
16:18.53clock_ZX Spectrum with a secret tape of Manic Miner picked from the wreckage of US 193?
16:19.02clock_octopus
16:19.02``Eriktwo quadcore xeons, beefy mac pro
16:19.20``Erikmakes a nice footrest, too
16:19.21clock_OK
16:19.44clock_Do you understand the mechanism of the slowdown?
16:19.47``Erikif you made, say, a hundred small cutting boxes, the scnee would go much faster
16:19.48``Erikyes
16:20.07clock_why would making hundred small ones help?
16:20.25``Erikbecause right now, your big cutting box is in the tree making the bounding volume include everything
16:20.34``Erikso every single pixel must evaluate every single primitive
16:20.41``Erikand weave all of them together
16:20.51``Erikyou've effectively eliminated space partitioning
16:21.00clock_so if you have a 3mm part and cut it with a 1m cube, the bounding box inflates to 1m?
16:21.07``Erikyes
16:21.24clock_Whoa! What algorithmic motorization! Secret trick, dude: if you subtract from something, it never gets bigger
16:21.49``Erikyeah, but what if that 1m cube not only clips out part of that 3mm part, but ALSO part of the other 3mm part over there
16:22.12``Erikuntil evaluated, there's no way to know what primitives that subtraction impacts
16:22.22``ErikYOU know, because you're modelling it,but the software CANNOT know
16:22.47clock_OK jokes about pinnacle of development back let's assume it's really serious
16:23.12clock_Let's say I have a scene with zillion tiny slices and one big cube that subtracts (eats off) half of each slice
16:23.29clock_I fire a ray
16:23.49clock_Now I take one primitive after another, check it's bounding box and possibly calculate intersection
16:28.03clock_``Erik: what the software knows is that the cube is always the same cube so it doesn't have to calculate it million times
16:35.33clock_What the software also know is that zillion parts cut with one cutter equals zillion results of (part cut with the cutter)
16:36.02clock_or more precisely can know, if programmed thusly
16:38.10illethalDang how do I compile this.
16:52.18clock_illethal: do you have an error?
16:52.22illethalI type make install
16:52.26illethalbut it does nothing, I'm su.
16:52.33illethalIt says no targets.
16:52.36clock_did you do make before?
16:52.38illethalWhat directory do I have to be in?
16:52.46clock_in the toplevel brl-cad
16:52.50illethalDoesn't do anything either
16:53.19clock_What I do is unpack tgz, configure, make, make install
16:53.20clock_it works
16:53.23clock_on Linux and OpenBSD
16:53.27clock_what system do you have?
16:53.29illethalLinux
16:53.33illethalUbuntu 64
16:53.37clock_oops 64
16:53.54clock_heard that a lots of things don't work on 64
16:54.14illethalAhhhh a lot do actually.
16:54.25illethalJust Flash is meh.
16:54.40illethalnspluginwrapper crap. It works for 30 mins them breaks, then I have to recompile.
16:55.29illethalroot@valhall:/home/oden/Desktop/progs/usr# make
16:55.31illethalmake: *** No targets specified and no makefile found.  Stop.
16:55.40illethalarg
16:55.42illethalwrong dir
16:55.54illethalStill
16:55.58illethalDoesn't do anything wtf.
16:56.05clock_maybe it's a demo version
16:56.10clock_after 30 seconds you have to buy a licence?
16:56.21clock_My Firefox crashes
16:56.29clock_That's normal that Firefox is a piece of crap
16:56.34illethalEverything else works fine except flash for me.
16:56.38clock_oh 30 minutes you wrote not seconds sorry
16:56.45clock_My flash works but often crashes
16:56.58clock_And brings down the whole Firefox. It hangs and stops refreshing the window
16:56.58illethalAdobe needs to burn.
16:57.03clock_I have to kill it lose all my open tabs
16:57.15clock_Adobe is piece of crap
16:57.16illethalWhat distro you on?
16:57.22clock_Linux From Scratch
16:57.34illethalFrom Scrath?
16:57.39clock_yes
16:57.40illethalLike you made your own os?
16:57.46illethalOr modified/whatever?
16:57.49clock_compiled
16:58.08clock_LFS is a web instructions where they tell you what to type to compile everything from scratch
16:58.18illethalCool.
16:58.19clock_kernel, gcc, glibc,...bash,...,firefox,...
16:58.26illethalUbuntu is pretty unstable.
16:58.28illethalImo.
16:58.37illethalI think I'm going to switch to gentoo.
16:58.45clock_Friend had an Ubuntu and he typed locate and locate segfaulted
16:58.53illethalYeah.
16:59.03clock_I had gentoo but every time I upgraded it blew up about 5 times and I had to try secret tricks from google
16:59.14illethalhaha
16:59.16clock_sometimes the tricks didn't work so I had to google up a different remedy etc.
16:59.19clock_It's a load of bullshit
16:59.25clock_Then I ditched gentoo and tried OpenBSD
16:59.30clock_Also exploded when I upgrded
16:59.31illethalWell my bro uses gentoo more than ubuntu. So if I need help I've got it.
16:59.41clock_So now I have LFS. I had LFS for years before and I was basically happy
16:59.45clock_So now I am happy again :)
16:59.49illethalI havn't upgraded in a long ass time.
16:59.52illethalI'm still on feisty.
17:00.17illethalUbuntu has such gay names.
17:00.21illethalGutsy Gibbon lol
17:00.28clock_Software today is piece of crap. Load of bullshit. Heap of trash.
17:00.48clock_I am gay
17:00.59clock_you shouldn't use the word "gay" as a synonym for something bad or laughable
17:01.14illethalYou're a homosexual.
17:01.22illethalNot gay =P
17:01.22clock_Yes that too
17:01.28clock_No I am gay
17:01.34illethalYou could be a manic depressed homosexual!
17:01.44clock_you could be Manic Miner
17:02.39illethalDon't mean to offend you buddy.
17:02.39clock_Yeah sure I know
17:02.39clock_you just didn't care to offend all gays in the world
17:03.23illethalYou mean homos.
17:03.33clock_I know what I mean
17:03.35clock_you can't know it
17:03.50illethalHomos are gays. Heteros are Sads.
17:04.38clock_yeah you have all those wifes shopping babies etc.
17:04.45clock_changing diapers
17:04.49clock_kindergartens
17:05.41illethalGays like to shop too I thought.
17:05.50clock_Especially the camp ones
17:05.53illethalCould just be a stereotype.
17:06.17clock_Buy a lot offancy crap to hook up on their HIV-soaked promiscuous bodies
17:06.36illethalExactly.
17:06.43clock_I don't like promiscuous gays
17:06.45illethalTo each his own.
17:06.53clock_I read 1 of 6 gays in the "scene" in Zurich are HIV+
17:07.36clock_sounds like I should walk only with rubber gloves protective goggles and a preservative already on in the street just for the case I passed by some of them
17:08.20clock_sounds like a promiscuity is a ticket to cemetery here
17:09.27illethalLol
17:09.36illethalyeah that's groooooooz
17:09.52clock_1 in 6 that's like the Russian Roulette
17:10.45clock_If they invent a vaccine for HIV they need to excavate a lot of earth for the storage tanks in the factory
17:10.58illethalHaha
17:11.05clock_maybe build a pipeline
17:11.06*** join/#brlcad MinuteElectron (n=MinuteEl@silentflame/member/pdpc.base.minuteelectron)
17:11.42illethalIsn't Zurich in Switzerland?
17:11.51clock_``Erik: did I offend you with my comment on pinnacle of 30 years of development?
17:11.56clock_yes Switzerland
17:12.15clock_in the news today in teh paper new in the train
17:12.21clock_they wrote the price of prostitution dropped
17:12.33clock_like 30 CHF (35 USD?) for a sex in the street
17:12.47illethalhttp://micah.noobgrinder.com/builda.png
17:12.58clock_What's happening with our Zurich?
17:13.23illethalSomething I was working on a while ago.
17:13.27clock_Autodesk that's not BRL-CAD
17:13.35clock_do you want to use BRL-CAD for that?
17:13.59illethalIt's maya.
17:14.07illethal8.5
17:14.18illethalI could probably never do that in BRL-CAD.
17:14.35illethalAll I can do in BRL-CAD is make a cube and move it's faces lol
17:14.44clock_yeah
17:14.59clock_nothing against BRL-CAD I like BRL-CAD even with the limitations it has
17:15.05clock_But I wanted to do a polygonal plate
17:15.16clock_So I asked brlcad and he advised me some secret command that was undocumented
17:15.22clock_Described the syntax in detail
17:15.32clock_I typed it I believe exactly as he said and it didn't work
17:15.59clock_Every now and then I run into some error in the doc, ambiguity or some topic not being covered at all
17:16.04clock_I also found a lot of segfaults
17:16.26clock_Like - it's still great - but these can be serious obstackles during work, especially for a novice
17:16.38clock_Now I have a problem if I do a cutaway view the rendering alsmost grinds to a halt
17:17.28clock_BRL-CAD people are great they fix all or most of the problems shortly after I find them
17:17.31clock_Especially segfaults
18:07.09``Eriknah, didn't offend me, I was out at lunch
18:08.05``ErikI'm hoping my contributions have removed a lot of the suck and modernized things a lot int he last few years, but *shrug* I'm only one codegod :D and smothered in bs politics and crap so I don't have much time to commit
18:15.26clock_``Erik: the problem is my model is logically built
18:15.34clock_that means I first make a device like the optical head
18:15.46clock_and then I say now we want to look what's inside .... cut!
18:16.01clock_Thinking about the cut at the moment I make the screw is not appropriate.
18:16.07clock_Also because the script is made by an automatic script.
18:16.25clock_However, would it help if the cutting thing was composed of a larger amount of smaller subthings?
18:17.21clock_the script is made -> the bolt (screw) is made
18:46.26CIA-32BRL-CAD: 03bob1961 * r30450 10/brlcad/trunk/include/raytrace.h: struct dg_obj and supporting cast has been moved to dg.h
18:46.50*** join/#brlcad docelic (n=docelic@212.15.184.79)
18:47.02CIA-32BRL-CAD: 03bob1961 * r30451 10/brlcad/trunk/include/dg.h: Initial check-in.
18:48.34*** join/#brlcad poolio (n=poolio@bz.bzflag.bz)
18:48.54CIA-32BRL-CAD: 03bob1961 * r30452 10/brlcad/trunk/src/ (69 files in 5 dirs): Mods related to bio.h and the new dg.h wrt getting things compiled on Windows.
18:54.14CIA-32BRL-CAD: 03bob1961 * r30453 10/brlcad/trunk/include/raytrace.h: Move "drawable geometry" related function declarations to dg.h
19:02.44CIA-32BRL-CAD: 03bob1961 * r30454 10/brlcad/trunk/include/dg.h: Added function declarations.
19:50.43*** join/#brlcad Elperion (n=Bary@p54877627.dip.t-dialin.net)
20:55.25*** join/#brlcad elite01_ (n=elite01@dslb-088-070-003-070.pools.arcor-ip.net)
21:00.30CIA-32BRL-CAD: 03bob1961 * r30455 10/brlcad/trunk/src/librt/opennurbs_ext.cpp: Move include for bio.h after common.h
21:05.58CIA-32BRL-CAD: 03bob1961 * r30456 10/brlcad/trunk/src/libfb/if_wgl.c: Move include for bio.h after common.h
21:13.49*** join/#brlcad jgay (n=jgay@fsf/staff/jgay)
22:18.54``Erik1http://bash.org/?10626
22:24.48``Erikminus that first 1, of course
22:25.09archivistcooking, whats wrong with beans on cheese on toast?
23:48.16*** join/#brlcad Elperion (n=Bary@p548777EC.dip.t-dialin.net)
23:48.57*** mode/#brlcad [+o brlcad] by ChanServ

Generated by irclog2html.pl Modified by Tim Riker to work with infobot.