| 00:43.16 | *** join/#brlcad clock (~clock@77-58-143-135.dclient.hispeed.ch) | |
| 01:30.03 | starseeker | huh - Embree had a SIGGRAPH paper this year |
| 01:36.49 | starseeker | cool: http://www.cs.princeton.edu/~vk/projects/Shape2Pose/ |
| 01:42.36 | starseeker | wishes the Mitsuba guys hadn't gone with full-up GPL... |
| 01:43.01 | brlcad | starseeker: bu_same_file() |
| 01:43.46 | brlcad | starseeker: I guess you missed when I posted all of the siggraph tech paper first pages? |
| 01:44.27 | starseeker | brlcad: er - guess so :/ |
| 01:44.33 | starseeker | sorry |
| 01:45.43 | brlcad | http://s2014.siggraph.org/sites/default/files/firstpages.default.pdf |
| 01:46.09 | brlcad | and of course, you found that intel posted the full embree paper |
| 01:46.17 | brlcad | has a copy on his desk he's been digesting |
| 01:46.25 | starseeker | how's it look? |
| 01:46.48 | brlcad | not finished ;) |
| 01:47.15 | starseeker | heh - fair enough |
| 01:47.28 | brlcad | most of the fundamental techniques they use for performance are very well known |
| 01:48.16 | starseeker | so then the key becomes are they very well *implemented*, I guess |
| 01:48.35 | brlcad | a lot of their zazz is that they implement global illumination, which requires a coherent approach for handling a ton of incoherent rays |
| 01:49.52 | brlcad | meh, if pretty pictures is what we were after, there are several other renderers we could feed our geometry to that give stunning pictures |
| 01:50.24 | Notify | 03BRL-CAD:starseeker * 62666 brlcad/trunk/src/gtools/gdiff2/gdiff2.c: Thanks Sean - use bu_same_file when doing this operation in a file context... |
| 01:51.08 | brlcad | our focus remains on solid full-path tracing, V&V'd correctness (at least rigorously well-defined), and getting that going as fast as we can (implement the well-known techniques) |
| 01:52.54 | Notify | 03BRL-CAD:starseeker * 62667 brlcad/trunk/src/gtools/gdiff2/gdiff2.c: Might as well have an informative message for all the cases... |
| 01:53.37 | Stragus | Anything fun or interesting about their approach to make coherent bundles from incoherent rays? |
| 01:54.11 | brlcad | starseeker: prior gdiff had a decent startup where it'd check bu_file_exists on both, then bu_same_file to give a helpful message |
| 01:54.35 | starseeker | ok, can do |
| 01:55.16 | Stragus | I played with 3D grids for origins and a bunch of buckets covering 8 general ray angles or such, to make coherent bundles of rays. The gain was very small if any |
| 01:55.38 | Stragus | The overhead from totally random memory access (to sort the rays) on CUDA didn't play well |
| 01:56.40 | Stragus | Also, CUDA hardware is very good at pulling from memory, but absolutely not for writing to memory in an incoherent manner |
| 02:32.33 | starseeker | brlcad: I do have bu_file_exists checks in now - are they not adequate? |
| 02:35.30 | brlcad | what happens when you feed it a file that does not exist? |
| 02:35.53 | starseeker | "Cannot stat file <filename>" |
| 02:36.15 | starseeker | think I lifted that message from gdiff.c, actually |
| 02:36.52 | brlcad | if you already got that bit, sounds like you're good to go then |
| 02:36.58 | brlcad | yeah, that's exactly what the old did |
| 02:37.09 | brlcad | terrible message, I probably wrote that |
| 02:37.21 | brlcad | but technically accurate |
| 02:37.48 | brlcad | saying the file does not exist might not "actually" be true .. no way of really knowing, all we know is we cannot stat the file ;) |
| 02:38.53 | brlcad | yay, almost cleared out my queue far enough that I can get to reviewing patches! |
| 03:13.13 | starseeker | Stragus: do you think this approach (http://www.cs.jhu.edu/~misha/Code/PoissonRecon) might do better than ball-pivot for point clouds? |
| 03:26.46 | Stragus | starseeker, I think the main issue is that for your needs (the way I understand it), you absolutely need closed, watertight shapes |
| 03:27.08 | Stragus | It might be better but you need an algorithm that absolutely guarantees closed shapes |
| 03:28.26 | Stragus | I had proposed some kind of "Collapsing box" technique, it seems to be a close cousin to some "Shrinkwarp" technique |
| 03:45.46 | *** join/#brlcad ries (~ries@D979EA84.cm-3-2d.dynamic.ziggo.nl) | |
| 07:06.04 | *** join/#brlcad ries (~ries@D979EA84.cm-3-2d.dynamic.ziggo.nl) | |
| 07:59.22 | *** join/#brlcad teepee- (bc5c2133@gateway/web/freenode/ip.188.92.33.51) | |
| 08:23.06 | *** join/#brlcad Ch3ck (~localhost@195.24.220.134) | |
| 08:48.26 | *** join/#brlcad localhost__ (~localhost@195.24.220.134) | |
| 11:35.35 | starseeker | Stragus: so something like the powercrust algorithm or http://www.ahornung.net/files/pub/Hornung_SGP06.pdf |
| 11:42.03 | *** join/#brlcad ries (~ries@D979EA84.cm-3-2d.dynamic.ziggo.nl) | |
| 11:44.33 | *** join/#brlcad Zhao_Anqing (~clouddrif@218.79.166.196) | |
| 11:46.38 | *** join/#brlcad konrado (~konro@41.205.22.54) | |
| 12:14.53 | *** join/#brlcad teepee- (bc5c2133@gateway/web/freenode/ip.188.92.33.51) | |
| 12:15.26 | *** join/#brlcad Ch3ck (~localhost@195.24.220.134) | |
| 12:40.57 | *** join/#brlcad teepee- (bc5c2133@gateway/web/freenode/ip.188.92.33.51) | |
| 13:12.05 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) | |
| 13:23.13 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) | |
| 14:10.45 | Notify | 03BRL-CAD:starseeker * 62668 (brlcad/branches/dm-work/include/dm/bview.h brlcad/branches/dm-work/include/ged.h and 15 others): Put back solid list, but do it as part of struct gedp rather than as a global. |
| 14:24.53 | Notify | 03BRL-CAD:starseeker * 62669 (brlcad/branches/dm-work/CHANGES brlcad/branches/dm-work/NEWS and 13 others): Sync with trunk through r62668 |
| 14:25.26 | *** join/#brlcad FreezingCold (~FreezingC@CPE602ad06bea2a-CM602ad06bea27.cpe.net.cable.rogers.com) | |
| 14:32.20 | Notify | 03BRL-CAD:starseeker * 62670 (brlcad/trunk/include/bn.h brlcad/trunk/include/dm/bview.h and 80 others): Having gone that far with the merge back into trunk from dm-work, we might as well go the rest of the way. This is mostly reworking the display_list/solid list handling as a preliminary to shifting the low-level details down the stack and ideally out of the applications. The freesolid list has been restored, |
| 14:32.22 | Notify | although it is no longer a global in either MGED or libged. |
| 14:33.14 | starseeker | brlcad: well, now we'll find out how much damage I did |
| 14:38.26 | *** join/#brlcad archivist (~archivist@host81-149-189-98.in-addr.btopenworld.com) | |
| 15:02.03 | *** join/#brlcad clock (~clock@77-58-143-135.dclient.hispeed.ch) | |
| 15:12.32 | Notify | 03BRL-CAD:carlmoore * 62671 (brlcad/trunk/include/dm.h brlcad/trunk/src/libdm/dm-generic.c): remove trailing blanks/tabs; fix spelling |
| 18:20.30 | *** join/#brlcad Ch3ck (~localhost@195.24.220.134) | |
| 18:40.56 | Notify | 03BRL-CAD:carlmoore * 62672 brlcad/trunk/src/nirt/nirt.c: move -M description in Usage; this matches what is in the man page |
| 19:38.45 | *** join/#brlcad ries_nicked (~ries@D979EA84.cm-3-2d.dynamic.ziggo.nl) | |
| 20:43.52 | Notify | 03BRL-CAD:carlmoore * 62673 (brlcad/trunk/doc/docbook/system/man1/en/nirt.xml brlcad/trunk/src/nirt/nirt.h): shift the -E description for uniformity; fix punctuation (including putting ending punctuation outside quote marks) |
| 21:01.43 | Notify | 03BRL-CAD:n_reed * 62674 brlcad/branches/brep-debug/src/libbrep/boolean.cpp: discard degenerate loops |
| 21:19.24 | Notify | 03BRL-CAD:starseeker * 62675 (brlcad/branches/dm-work/doc/docbook/system/man1/en/nirt.xml brlcad/branches/dm-work/doc/html/ReleaseNotes/email4.0.html and 4 others): Make sure we're synced up with trunk as of r62674 - time to figure out MGED's illumination mechanisms |
| 21:26.41 | *** join/#brlcad Stragus (~alexis@modemcable008.161-160-184.mc.videotron.ca) | |
| 21:39.56 | Notify | 03BRL-CAD:starseeker * 62676 brlcad/trunk/src/gtools/gdiff2/gdiff2.c: We want gdiff2 to return 0 in either of these two cases, not just in the empty case. |
| 21:41.50 | Notify | 03BRL-CAD:starseeker * 62677 brlcad/trunk/src/gtools/CMakeLists.txt: rename gdiff2 directory |
| 21:42.53 | Notify | 03BRL-CAD:starseeker * 62678 brlcad/trunk/src/gtools/gdiff/CMakeLists.txt: Rename individual gdiff2 files |
| 21:46.16 | Notify | 03BRL-CAD:starseeker * 62679 (brlcad/trunk/src/gtools/gdiff/diff3_summary.c brlcad/trunk/src/gtools/gdiff/diff_summary.c and 3 others): fix includes |
| 21:58.09 | Notify | 03BRL-CAD:starseeker * 62680 brlcad/branches/dm-work/src/gtools/CMakeLists.txt: Merge with trunk thru r62679 so make regress works |
| 22:01.32 | Notify | 03BRL-CAD:n_reed * 62681 brlcad/branches/brep-debug/src/libbrep/intersect.cpp: close parallel segments don't necessarily overlap; handle the case where the intersect at a point |
| 22:50.00 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) | |
| 23:01.45 | *** join/#brlcad clock (~clock@77-58-143-135.dclient.hispeed.ch) | |
| 23:09.26 | Notify | 03BRL-CAD:n_reed * 62682 brlcad/branches/brep-debug/src/libbrep/intersect.cpp: need to return non-normalized parameters in event |
| 23:24.51 | Notify | 03BRL-CAD:n_reed * 62683 brlcad/branches/brep-debug/src/libbrep/intersect.cpp: fix typo |
| 23:57.40 | Notify | 03BRL-CAD:starseeker * 62684 brlcad/trunk/src/libdm/axes.c: Move the header inclusion to the top of the file. |