| 00:26.08 | *** join/#brlcad dtidrow (n=dtidrow@c-69-255-182-248.hsd1.va.comcast.net) |
| 01:17.34 | *** join/#brlcad Twingy (n=justin@74.92.144.217) |
| 01:21.44 | *** join/#brlcad ewilhelm_ (n=ewilhelm@pool-71-111-49-155.ptldor.dsl-w.verizon.net) |
| 06:27.12 | poolio | siggraph: AHA! AHA!!! |
| 06:27.47 | poolio | siggraph: I finally figured out why the GA wasn't really working and now I want to shoot myself. It was actually working pretty well, the issue is that it has a hard time dealing with source objects with a lot of empty space |
| 06:28.21 | poolio | Because there was so much empty space in my two seperated spheres trials, the network had a hard time. It was being rewarded largely for not matching at all (lots of similar empty space) |
| 06:28.53 | poolio | Testing it on a cluster of 3 spheres works out nicely, the GA does produce better results as the generations increase :D |
| 06:29.55 | poolio | I want to shoot myself for just sticking to using that one example as a benchmark for the past few weeks. |
| 06:30.08 | poolio | I always assumed it was the code that was at fault and not the test data. |
| 06:45.56 | poolio | siggraph: pretty graph if you ever read this: http://poolio.org/files/working.png |
| 07:14.33 | *** join/#brlcad Laniakea (n=clock@zux221-122-143.adsl.green.ch) |
| 11:03.01 | *** join/#brlcad elite01 (n=elite01@dslb-088-070-031-031.pools.arcor-ip.net) |
| 12:03.56 | *** join/#brlcad b0ef (n=b0ef@062016141081.customer.alfanett.no) |
| 12:25.40 | *** join/#brlcad thing0 (n=ric@203-59-167-160.dyn.iinet.net.au) |
| 12:39.47 | *** part/#brlcad thing0 (n=ric@203-59-167-160.dyn.iinet.net.au) |
| 13:36.38 | *** join/#brlcad b0ef (n=b0ef@062016141081.customer.alfanett.no) |
| 14:11.02 | *** join/#brlcad poolio (n=poolio@c-69-251-3-107.hsd1.md.comcast.net) |
| 14:11.55 | poolio | mornin |
| 14:17.19 | ``Erik | oi |
| 14:18.50 | poolio | oy vey? |
| 14:24.40 | ``Erik | heh, I spell mine japanese style, sorry :) lots of cultures have the same sound with roughly the same meaning |
| 14:30.14 | poolio | oy. |
| 14:57.35 | *** join/#brlcad CIA-4 (n=CIA@208.69.182.149) |
| 15:09.41 | *** join/#brlcad Elperion (n=Bary@p548763F1.dip.t-dialin.net) |
| 15:35.17 | *** join/#brlcad CIA-4 (n=CIA@208.69.182.149) |
| 15:40.30 | *** join/#brlcad CIA-4 (n=CIA@208.69.182.149) |
| 15:54.27 | poolio | hey ``Erik, do you know if there's a way to find a kind of "bogus" union...between an object and another object that fits inside that other object? |
| 15:59.59 | *** join/#brlcad CIA-4 (n=CIA@208.69.182.149) |
| 16:00.18 | poolio | The CIA is going crazy. |
| 16:21.22 | *** join/#brlcad CIA-4 (n=CIA@208.69.182.149) |
| 16:31.50 | *** join/#brlcad CIA-4 (n=CIA@208.69.182.149) |
| 16:48.47 | ``Erik | poolio:not mathematically, but it could be sampled with buttloads of rays |
| 16:49.29 | ``Erik | brlcad had a slew of csg optimizations scribbled on his whiteboard at one point, that may've been part of the tree reduction stuff :/ *shrug* |
| 16:50.41 | poolio | heh alright> i've though about it and decided that I actually want to allow that behavior |
| 16:50.44 | poolio | so yeah :) |
| 17:04.44 | *** join/#brlcad b0ef (n=b0ef@062016141081.customer.alfanett.no) |
| 17:15.44 | poolio | ``Erik: stupid question, how can I save the framebuffer from mged? |
| 17:23.08 | poolio | err nevermind, I just took a screenshot |
| 17:23.16 | poolio | http://poolio.org/files/cone.png |
| 17:27.59 | *** join/#brlcad CIA-4 (n=CIA@208.69.182.149) |
| 17:48.08 | *** join/#brlcad CIA-27 (n=CIA@208.69.182.149) |
| 18:01.59 | *** join/#brlcad ertugerata (n=Ertugrul@88.230.240.109) |
| 18:02.05 | ertugerata | salut |
| 18:07.36 | ertugerata | enfin j'ai fini compile et install brlcad |
| 18:08.22 | ``Erik | heh, 'shitty ftpd'... you had will to write code in java at some point??? :) |
| 18:08.36 | poolio | ``Erik: I had a class in it ... |
| 18:08.41 | poolio | It made me stupider. |
| 18:09.06 | ``Erik | striving for mediocrity, yes |
| 18:09.19 | poolio | always. |
| 18:09.30 | poolio | jmario: 2d mario-clone in Java. died with my will to write code in Java. |
| 18:09.54 | ``Erik | for real, right? not the bass ackwards subset they teach in classes? |
| 18:10.29 | ertugerata | en CVS, je ne peux pas build opennurbs ? |
| 18:11.31 | ``Erik | hum, babelfish kinda ate that one |
| 18:11.32 | poolio | ``Erik: errr, I'm learning it on my own. I have to take a data structures class in Java sometime though |
| 18:12.02 | poolio | ertugerata: tu peut mais ce n'est pas "active" en la "default" versione |
| 18:12.20 | ertugerata | hmm |
| 18:12.30 | poolio | faire comme "--enable-opennurbs-build |
| 18:12.32 | poolio | " |
| 18:12.39 | ``Erik | sicp is a good book, um, and free video lectures by the dudes at http://swiss.csail.mit.edu/classes/6.001/abelson-sussman-lectures/ |
| 18:12.43 | ertugerata | en config.log il ya |
| 18:12.44 | poolio | ./configure --enable-opennurbs-build |
| 18:12.55 | poolio | ``Erik: that's what I'm working trhough, I'll check out the lectures later |
| 18:13.01 | poolio | I really hvae no time now though :\ |
| 18:13.29 | ``Erik | the videos are the lectures to go with the sicp, might as well do 'em at the same time :) |
| 18:13.39 | ``Erik | which scheme are you using? drscheme out of rice? |
| 18:13.48 | ertugerata | configure:45666: result: Build openNURBS.......................: no (doing without, BREP unsupported) poolio |
| 18:15.45 | ``Erik | ertugerata: do you have emmintrin.h ? |
| 18:16.27 | ertugerata | dans quel package a emmintrin.h ? |
| 18:17.06 | ``Erik | <-- bsd user, does not know linux |
| 18:17.40 | ertugerata | in bsd quel source ? |
| 18:18.59 | ``Erik | /usr/include/emmintrin.h (it comes from gcc) |
| 18:19.45 | ertugerata | /usr/lib/gcc/i686-pc-linux-gnu/3.4.6/include/emmintrin.h |
| 18:19.46 | ertugerata | /usr/lib/gcc/i686-pc-linux-gnu/3.4.6/install-tools/include/emmintrin.h |
| 18:20.05 | ertugerata | in gcc ici |
| 18:23.06 | ``Erik | redhat... ./configure --enable-optimized --prefix=/usr/brlcad/HEAD --enable-opennurbs-build |
| 18:23.07 | ``Erik | Build openNURBS.......................: yes |
| 18:23.16 | ``Erik | $ locate emmintrin.h |
| 18:23.19 | ``Erik | /usr/lib/gcc/x86_64-redhat-linux/3.4.3/include/emmintrin.h |
| 18:23.24 | ``Erik | /usr/lib/gcc/x86_64-redhat-linux/4.1.1/include/emmintrin.h |
| 18:24.13 | ``Erik | :/ |
| 18:24.33 | poolio | ``Erik: mit-scheme |
| 18:24.35 | ``Erik | $ grep header_emmintrin config.log |
| 18:24.35 | ``Erik | ac_cv_header_emmintrin_h=yes |
| 18:25.46 | ``Erik | poolio: I wasn't thrilled by mit-scheme... drscheme is kinda a nice ide thing going and a funky 'trace' thing that actually draws arrows on the code... 'gauche' and 'chicken' are neat |
| 18:27.59 | ``Erik | oh, and 'slime48' looks interesting if you're an emacs user who could tolerate scheme48 |
| 18:29.00 | poolio | nah, not yet an emacs user |
| 18:29.34 | ``Erik | <-- vimmer and proud... tried to learn emacs for slime... *shrug* I guess I'm editor retarded or something |
| 18:31.07 | poolio | oh yeah, I did try emacs + slime way back when |
| 18:33.42 | ertugerata | poolio: je dois faire patch configure.ac for openurbs ou je coprends mal ? |
| 18:35.11 | poolio | ``Erik: what were you trying to tell ertugerata about open nurbs? |
| 18:35.43 | ``Erik | it's trying to enable it, the only test that could fail that would be if emmintrins.h isn't happy |
| 18:35.58 | ``Erik | emmintrin.h rather |
| 18:36.11 | ``Erik | I think |
| 18:36.54 | ``Erik | (don't know why anyone would want the open nurbs stuff yet, it's not very wired in...) |
| 18:37.58 | ertugerata | opennurbs a quel fonction? si il n'existe pas mged ne marche pas bien ? |
| 18:39.38 | ``Erik | ok, babel fish made a mess of that one, too :) |
| 18:45.42 | ertugerata | poolio: dans ton system ilya une file en /etc/env.d nomme comme 50brlcad ? |
| 18:45.45 | poolio | ertugerata: mged marche bien sans opennurbs. opennurbs ne marche pas maintenant, c'est normale que opennurbs est disable |
| 18:46.00 | poolio | err je n'ai pas /etc/env.d |
| 18:46.15 | ertugerata | poolio: je fait enable opennurbs |
| 18:47.35 | ``Erik | (just recently, brlcad changed the opennurbs default from 'auto' to 'no' in cvs, jlowenz has moved to cooler projects in a cooler group) |
| 18:47.41 | ertugerata | dans /etc il ya une dossier pour environment ? |
| 18:47.44 | poolio | ertugerata: non non, opennurbs ne fonctionne pas bien, il est meilleur si tu n'enable pas |
| 18:48.04 | poolio | errr non? |
| 18:48.11 | poolio | il y a un ~/.bashrc |
| 18:48.29 | ertugerata | paste moi ton bashrc |
| 18:48.59 | ertugerata | poolio: il y aune lisence problem pour opennurbs ? |
| 18:49.08 | poolio | non ? |
| 18:49.16 | poolio | opennurbs ne marche pas pour tout le monde |
| 18:49.34 | ertugerata | hmm |
| 18:49.36 | poolio | la seule chose de BRL-CAD en .bashrc est |
| 18:49.43 | poolio | PATH="/usr/brlcad/bin:$PATH" |
| 18:49.52 | poolio | ertugerata: mged marche bien sans opennurbs |
| 18:50.03 | ertugerata | merci bien je vais faire disable |
| 18:50.10 | ``Erik | MANPATH, too... and sometimes LD_LIBRARY_PATH, though that can easily be done to ld.so.conf as well |
| 18:50.10 | poolio | c'est rien |
| 18:50.18 | ertugerata | pour lib ? |
| 18:50.29 | ertugerata | LDPATH ? |
| 18:51.09 | ``Erik | on some systems, the rpath in the binaries doesn't get there or isn't respected, so it complains about missing libraries. I'm sure linux doesn't have that issue... :) |
| 18:51.53 | ``Erik | but the MANPATH is good to have, BRL-CAD has pretty decent man pages |
| 18:53.17 | poolio | ertugerata: il n y a pas plus de changed, seulement PATH, et si tu veut les "man pages" ajouter MANPATH |
| 18:54.16 | ``Erik | poor ben, stuck translating :D |
| 18:55.02 | poolio | eh it's nothing, glad I can actually do something helfpul :) |
| 18:55.31 | poolio | also makes me feel like those years spent studying french weren't such a waste after all |
| 18:56.04 | ``Erik | my years were studying japanese, but I'd be effed if someone wanted me to translate, I only remember a few choice ... phrases... :) |
| 19:33.18 | ``Erik | HAH |
| 19:33.23 | ``Erik | http://drj11.wordpress.com/2007/04/02/subclassing-javaawtpaint/ |
| 19:34.45 | ``Erik | about halfway down, he snarfed the camo algorithm out of BRL-CAD |
| 19:39.01 | ertugerata | ./usr/brlcad/bin/rt: MGED database not specified ? |
| 19:40.20 | poolio | ``Erik: atleast he gives BRL-CAD credit |
| 19:41.14 | ``Erik | /usr/brlcad/bin/rt -F/dev/Xl /usr/brlcad/share/db/havoc.g havoc |
| 19:42.02 | ertugerata | ``Erik: hmm je vais faire essayer |
| 19:42.41 | ertugerata | hmm |
| 19:42.56 | *** join/#brlcad dtidrow (n=dtidrow@host131.objectsciences.com) |
| 19:43.03 | ertugerata | my database ici: /usr/share/db |
| 19:43.29 | ``Erik | rt dossier.g objets... |
| 19:44.21 | ertugerata | tout et dans /usr/share/db |
| 19:45.02 | ertugerata | 30 dossier.g existe |
| 19:46.21 | ``Erik | "havoc.g" has an object named "havoc". "sphflake.g" has an object named "scene.r" |
| 19:46.53 | ``Erik | "rt /path/to/havoc.g havoc" "rt /path/to/sphflake.g scene.r" |
| 19:46.54 | ertugerata | comment je peux savoir |
| 19:47.37 | ``Erik | mged -c /path/to/dossier.g tops |
| 19:48.37 | ``Erik | or: "mged", "file->open", choose, type "tops" in the command window |
| 19:49.23 | ertugerata | hmm |
| 19:50.22 | poolio | ertugerata: quel est la probleme? |
| 19:51.03 | ``Erik | "-c" means classic mode and allows MGED commands to be issued from the shell command line |
| 19:51.07 | ertugerata | poolio: benchmark ne trouve a-pas databese and rt aussi |
| 19:53.41 | poolio | quel est l'erreur avec benchmark? |
| 19:54.41 | ertugerata | ng [/usr/brlcad/bin/rt] for RT |
| 19:54.42 | ertugerata | ERROR: Could not find the BRL-CAD database directory |
| 19:55.36 | ``Erik | if that's about "make benchmark", I always have to set variables like DBDIR and PIXDIR |
| 19:55.56 | poolio | ertugerata: c'est neccesaire de modifier les variables DBDIR et PIXDIR |
| 19:55.59 | ertugerata | in bashrc = |
| 19:56.07 | ertugerata | ou ? |
| 19:56.09 | poolio | si tu veut |
| 19:56.12 | poolio | ou comme |
| 19:56.20 | poolio | DBDIR="/usr/share/db" benchmark |
| 19:56.28 | ertugerata | hmm |
| 19:57.00 | ``Erik | erm |
| 19:57.02 | ``Erik | sorry |
| 19:57.03 | ``Erik | DB |
| 19:57.05 | poolio | thats wrong |
| 19:57.06 | poolio | yeah |
| 19:57.10 | ertugerata | wrong |
| 19:57.23 | *** join/#brlcad Elperion (n=Bary@p548763F1.dip.t-dialin.net) |
| 19:57.30 | ``Erik | it's inconsistent :( PIXDIR and DB both talking about directories? |
| 19:58.36 | ``Erik | I build out of dir (in /usr/tmp/username/brlcadbuild) and this is why my benchmark attempt looks like... $ PIXDIR=~/src/brlcad/pix CMP=/usr/brlcad/bin/pixcmp RT=/usr/brlcad/bin/rt DB=$PWD/db/ ~/src/brlcad/bench/run.sh |
| 19:58.37 | ``Erik | O.o |
| 19:59.21 | poolio | what about 'benchmark' ? |
| 19:59.32 | ertugerata | <PROTECTED> |
| 19:59.53 | poolio | ``Erik: is DB pointing to where the databases are stored |
| 20:00.09 | ``Erik | 'make benchmark' is just (cd bench && ./run.sh) or something |
| 20:00.47 | ``Erik | the .g files have to be in $DB/ ... so 'make clean' would have broken that |
| 20:02.27 | poolio | ertugerata: est-ce que tu as fait 'make clean' |
| 20:02.58 | ertugerata | poolio: j'ai compris maintenant |
| 20:03.19 | ertugerata | pix est danl le source ? |
| 20:03.21 | poolio | d'accord? |
| 20:03.28 | poolio | quoi? |
| 20:03.28 | ertugerata | oui |
| 20:03.47 | poolio | est-ce que benchmark marche maintenant? |
| 20:03.55 | ertugerata | je vais essayer |
| 20:04.00 | poolio | alors |
| 20:04.11 | ertugerata | je crosi que ça marche |
| 20:07.30 | poolio | bien |
| 21:52.17 | *** join/#brlcad docelic (n=docelic@77.237.96.193) |
| 21:59.54 | dtidrow | anybody watching the shuttle launch? |
| 22:08.36 | poolio | eeek. the shuttle is launching? |
| 22:08.42 | *** join/#brlcad iraytrace (n=iraytrac@sdcc-66-78-210-221.smartcity.com) |
| 22:27.22 | dtidrow | poolio: yep, launches in about 9 minutes |
| 22:34.22 | poolio | dtidrow: what channel/station? |
| 22:37.17 | dtidrow_work | www.nasa.gov |
| 22:37.41 | dtidrow_work | sorry, was distracted |
| 22:38.58 | yukonbob | http://www.nasa.gov/multimedia/nasatv/index.html |
| 22:40.33 | poolio | up up and away. Missed take off, just read it :\ |
| 22:42.49 | louipc | how long does it take to exit the atmosphere? |
| 22:47.38 | dtidrow_work | poolio: sorry about that :-\ |
| 22:48.11 | dtidrow_work | they usually do a replay of the launch in a half-hour or so |
| 22:48.49 | poolio | eh I'll catch it on the 11 o' clock news or something |
| 22:49.51 | ``Erik | so it's in stable orbit and opened up now? |
| 22:50.01 | dtidrow_work | pfft - they'll hardly mention it |
| 22:50.15 | dtidrow_work | ``Erik: they're in orbit |
| 22:50.33 | dtidrow_work | still taking pics of the tank |
| 22:50.47 | ``Erik | they launch a lot and have an astoundingly good track record given how experimental and archaic it all is, but it's always good to hear that they made it up safe |
| 22:50.53 | dtidrow_work | they open the doors in an hour or so |
| 22:51.53 | louipc | archaic? |
| 22:52.16 | poolio | hmm...anyone have a fast box they want to lone to me? |
| 22:52.37 | dtidrow_work | louipc: yeah, most of the Shuttle was designed in the mid-70's |
| 22:53.27 | louipc | so was BRLCAD eh? /haha |
| 22:53.38 | dtidrow_work | heh |
| 22:54.01 | louipc | they've upgraded things I hope |
| 22:56.28 | ``Erik | they're buying used 70's microchips off of ebay to keep the shuttles running |
| 22:56.53 | ``Erik | (now true, they WORK and it's awfully hard to get a rad/temp proofed microchip) |
| 22:56.59 | louipc | :O |
| 22:57.31 | dtidrow_work | poolio: they're doing some launch replays now |
| 23:25.28 | siggraph | PIXDIR? |
| 23:25.34 | siggraph | you're making up variables ``Erik |
| 23:26.06 | siggraph | it's just PIX and DB and ... man benchmark |
| 23:27.24 | siggraph | you should only have to set them if benchmark isn't installed, and maybe if you're building out of dir |
| 23:28.11 | siggraph | also on cvs head opennurbs was turned off by default (and unless they're writing code, there's absolutely no reason for them to enable it) |
| 23:35.03 | poolio | siggraph: ahoy! |
| 23:35.24 | poolio | siggraph: did you read my "aha!" moment? |
| 23:37.03 | siggraph | yep |
| 23:37.11 | siggraph | like the cone :) |
| 23:37.28 | siggraph | that's great |
| 23:38.24 | poolio | just finished adding support for scaling to both bounding boxes |
| 23:38.26 | poolio | well it kind of cheats |
| 23:38.32 | poolio | actually it's an awful hackish solution, but it works |
| 23:38.50 | CIA-27 | BRL-CAD: 03poolio * 10brlcad/src/gtools/beset/ (beset.c fitness.c population.c beset.h fitness.h): fitness now scales to union of bounding boxes. other slight modifications |
| 23:39.03 | poolio | Instead of shooting rays to find the union of the bounding boxes, I just create a box that encloses the two bounding boxes |
| 23:39.39 | poolio | and divides the number of similar rays by the rays shot in the (overly large encompassing) bounding box |
| 23:39.44 | poolio | does that make any sense? |
| 23:40.45 | siggraph | "shooting rays to find the union of the bounding boxes" -- what's that mean? |
| 23:41.01 | poolio | siggraph: I doubt you have any time now, but I'd like to try some larger-scale tests on more complex data and am short on CPU power. Maybe early next week you could perform some trial runs or give me access to a more powerful box. It's taking a quite large population and a lot of generations to even closely match the source |
| 23:41.15 | poolio | siggraph: well, I actually wouldn't need to shoot rays, sorry |
| 23:41.24 | poolio | but finding the volume of the union of the two bounding boxes |
| 23:41.34 | siggraph | what two bounding boxes? |
| 23:41.40 | siggraph | i'm missing some context |
| 23:41.41 | poolio | the input and the individuals |
| 23:42.03 | poolio | Alright so when I raytrace, I used a fixed bounding box that is the same for all shapes (that bounding box is the bounding box of the original source object) |
| 23:42.21 | siggraph | sure |
| 23:42.52 | poolio | As you talked about earlier, it's kind of cropping the individual. So for example lets say your source is an rpp, any giant shape would pretty much be a perfect fit |
| 23:43.16 | siggraph | yep |
| 23:43.36 | poolio | so to counteract that I'm now scaling it by the union of the source bounding box and the individuals bounding box |
| 23:43.39 | poolio | ideally they'd be the same |
| 23:43.48 | poolio | and if it's smaller there is no effect |
| 23:43.58 | siggraph | scaling what? |
| 23:44.05 | poolio | the fitness value |
| 23:44.29 | poolio | heh I am having issues finding the wording |
| 23:45.13 | poolio | but if the rays from the source and the individual are the same, I increment the fitness by the length of the ray which they are similar in |
| 23:45.41 | siggraph | so you're still generating a fitness based on rays fired on source to rays fired on sample, but then scaling that fitness based on the ratio of their respective bounding boxes |
| 23:45.49 | poolio | well no |
| 23:46.04 | poolio | orignally it was similar_value/#rays fired |
| 23:46.09 | poolio | sorry no |
| 23:46.17 | poolio | similar_value/(#rays fired * length of rays) |
| 23:46.29 | poolio | so basically similar rays / (volume or maximum value it could be) |
| 23:46.49 | poolio | now it's similar rays / (volume of large bounding box that encompasses the union of both bounding boxes) |
| 23:47.17 | poolio | Does that make sense? |
| 23:48.09 | siggraph | hm, I think so |
| 23:48.22 | poolio | Ideally it'd be similar rays / (volume of union of both bounding boxes) |
| 23:48.35 | poolio | where volume is going to heavily depend on # rays fired...not the actual volume of the object |
| 23:48.42 | poolio | moreso the maximum fitness that could be attained |
| 23:48.46 | siggraph | I think that might actually be the same end-result to what I'd said |
| 23:48.49 | poolio | (if there was an exact atch) |
| 23:48.54 | siggraph | just operations applied in a different order |
| 23:48.56 | poolio | Yeah probably :) |
| 23:49.09 | poolio | I can't think of a good way to decompose similar rays / (volume of union of both bounding boxes) |
| 23:49.54 | poolio | The code for it is pretty hackish. I was just trying to get it working. It works pretty well. See: the cone |
| 23:50.08 | siggraph | so .. does it actually converge on one/two/three sphere(s)? |
| 23:50.25 | siggraph | the cone is looking MUCH better |
| 23:50.39 | siggraph | sort of like it's starting to work, though still not quite there :) |
| 23:50.41 | poolio | Also, if you don't have the right number of shapes or the GA has proposed a good solution with less shapes than the individual is made out of it essentially "hides" those spheres. By either shrinking them or putthing them within another sphere |
| 23:50.52 | poolio | It doesn't really converge |
| 23:51.04 | poolio | It finds solutions that visually make sense, but it's not waht you want. |
| 23:51.21 | poolio | And the main reason it wasn't working well is that it doesn't deal very well when a large % of the input is empty space |
| 23:52.15 | poolio | I'm not sure if it would converge, I've run stuff for ~1 hour and it's slowly getting better. The issue is my laptop goes through a meltdown... I have to run to dinner now but I'll start up a longer run |
| 23:52.34 | poolio | and one sphere does converge. |
| 23:53.04 | poolio | Two spheres seems to like having one large ellipse cover both spheres and then there are a lot of tiny spheres around it that have yet to be cleaned up (but probably would be if the network ran longer) |
| 23:53.14 | poolio | I'll try a run on two spheres slightly seperated while I'm at dinner. |
