| 00:01.41 | *** join/#brlcad Twingy (n=justin@74.92.144.217) [NETSPLIT VICTIM] | |
| 00:15.13 | Maloeran | Big rayforce commit, Erik, feel free to inform me if I broke anything... Performance should be more acceptable ( SSE and other goodies, still no quad volume tracing, etc. ) |
| 00:17.29 | Maloeran | Would it be possible for autoconf to put -msse -msse -m3dnow -m3dnowext -fomit-frame-pointer -ffast-math in CFLAGS by itself? |
| 00:18.01 | Maloeran | And it detects a missing log2() function systematically since it checks without -std=c99 |
| 01:23.18 | brlcad | Maloeran: it is possible to add those flags, easiest is to probably just set CFLAGS="$CFLAGS -msse -msse -m3dnow -m3dnowext -fomit-frame-pointer -ffast-math" somewhere in the configure.ac file somewhere after AM_INIT_AUTOMAKE |
| 01:24.41 | brlcad | ideally, you'd save your cflags (PRECFLAGS="$CFLAGS"), set a new cflags (CFLAGS="$CFLAGS ..."), and then try a compilation test to make sure they actually work and if not restoring the saved cflags (CFLAGS="$PRECFLAGS") as those options are certainly not universal |
| 01:25.41 | Maloeran | I see, thanks. Though, the flags should only be present if presently supported |
| 01:28.01 | Maloeran | The compiled test will have to be executed, running SSE instructions... and wait for a SIGILL. I have much to read about autoconf, unless Erik wants to commit that |
| 01:28.35 | brlcad | that's pretty easy to add, plenty such examples in brl-cad's .ac |
| 01:28.53 | brlcad | look for AC_TRY_RUN |
| 01:29.09 | brlcad | or AC_TRY_LINK if you just want to test compilation and not runnability |
| 01:29.23 | Maloeran | Oh? I see, nice |
| 01:30.26 | brlcad | AC_TRY_RUN([ source here ], [what to do if succeeded], [what to do if failed]) |
| 01:31.18 | brlcad | usually set a variable for the if succeeded/failed part and then handle the result outside the AC_TRY_RUN |
| 01:34.13 | Maloeran | *nod* I'll experiment with that, thanks |
| 14:07.36 | *** join/#brlcad danfalck (n=dan@pool-71-111-76-8.ptldor.dsl-w.verizon.net) | |
| 14:26.36 | Maloeran | Erik, another quick note if you happen to try it out : delete any old cache file lying around, or you won't get proper performance |
| 17:05.46 | *** join/#brlcad Amir (i=AMiR@h170n1-t-gr100.ias.bredband.telia.com) | |
| 17:06.08 | *** part/#brlcad Amir (i=AMiR@h170n1-t-gr100.ias.bredband.telia.com) | |
| 18:49.12 | Maloeran | Hey Erik, did you see my note about reorganizing milestones?.. |
| 19:00.49 | Maloeran | On the truck_bots file Lee uploaded, I also get a flood of errors such as db_walk_subtree() FAIL on '/all/fuel_pump.rc.bot from his framework for read .g files, investigating |
| 19:05.56 | Maloeran | Nevermind, it still works despites the flood on stderr |
| 20:46.43 | CIA-5 | BRL-CAD: 03brlcad * 10brlcad/NEWS: prepare for release, numbering it 7.8.4 as planned even though it's borderline 7.10 feature-wise. last news item, doug howard added (via bob) support for handling larger hierarchies in archer. |
| 20:51.49 | CIA-5 | BRL-CAD: 03brlcad * 10brlcad/ (configure.ac ChangeLog include/config_win.h): prepare for release 7.8.4, standard updates |
| 21:00.39 | Maloeran | Raytracer detects sectors of 86 and 93 triangles which aren't possible to split up. I located the sectors and there's nothing out of the ordinary visible, any possibility of overlapping triangle mess? |
| 21:01.00 | Maloeran | That came out of the CSG->triangle converter |
| 21:13.33 | Maloeran | I'm asking since this kind of situation would have been very apparent in ADRT as well, I don't suppose it was noticed |
| 21:19.55 | ``Erik | the truck was never done in adrt |
| 21:20.05 | ``Erik | <-- doesn't know about the tesselation stuff... |
| 21:20.43 | brlcad | it's not supposed to output overlapping triangles, but it's certainly possible if there was a degenerate case of some sort I bet |
| 21:20.56 | Maloeran | Okay I got it in play sight, very long thin triangles going from an edge of the sector to the opposite edge of the other sector |
| 21:21.18 | Maloeran | The not yet written permutation pass of the prep will take care of that |
| 21:24.21 | Maloeran | It's a fairly bad case, performance collapse by 70% there, down to 1m/s. on slow laptop |
| 21:24.52 | ``Erik | heh, the geometry was a better test case than expected :> |
| 21:25.32 | Maloeran | Did you quickly test the new code? Performance is getting decent |
| 21:30.57 | ``Erik | nope, was installing openbsd on a machine at work, and my time ended at 3, so now I'm home cleaning house |
| 21:31.44 | Maloeran | Gah. I can't spend weeks to complete the prep as planned, with the milestones coming up. I suppose that will wait until the end of the project |
| 23:35.20 | Maloeran | Erik, are all these questions and statements welcome or are you occupied with other matters these days? I could use your thoughts on certain matters related to features, milestones and the work to be done. It can wait though |
| 23:36.03 | Twingy | Like Suicide |
| 23:39.23 | ``Erik | mal: your timing is off :) I like to leave work at work... talk to me tomorrow before, oh, say, 5pm, when I have access to all the files and people |
| 23:41.20 | Maloeran | Right. So noted, thanks |
| 23:54.43 | Twingy | das goot! |
| 23:54.50 | Twingy | pan done, not for ruler |
| 23:54.53 | Twingy | *now |
| 23:56.36 | ``Erik | http://img.photobucket.com/albums/v350/Pinecone/Cavalier2.jpg |
| 23:59.02 | Twingy | your new car? |
