| 00:15.31 | ``Erik | hm |
| 00:34.38 | brlcad | how was dinner with la chica loca? |
| 00:36.03 | starseeker | raises eyebrow |
| 00:38.03 | ``Erik | la what now? |
| 00:38.57 | ``Erik | oh |
| 00:39.13 | ``Erik | heh, I didn't get out of there until too late, so I called them and let them know I wouldn't make it |
| 00:46.24 | starseeker | is bemused - the stix fonts are still not released, but at least they finally settled on the SIL Open Font License |
| 00:46.52 | starseeker | also, they have "discovered many missing glyphs in the non-Unicode fonts" |
| 00:47.01 | starseeker | wonders how that happened... |
| 00:50.17 | AirBender | what is bwish in brlcad sources? |
| 00:51.36 | AirBender | uhmm something like a customized tcl/tk ? |
| 00:51.50 | AirBender | is it needed? can I use just Tcl/Tk? |
| 00:53.25 | starseeker | you can build with a system Tcl/Tk if it's the right version, but IIRC bwish is still built as a wrapper? I'd have to check |
| 00:53.41 | AirBender | ok |
| 00:53.42 | starseeker | is bwish causing a problem? |
| 00:53.51 | starseeker | can usually treat it like wish |
| 00:53.55 | AirBender | I'm having undefined references with bwish |
| 00:54.05 | starseeker | blinks |
| 00:54.09 | starseeker | starting MGED? |
| 00:54.17 | AirBender | tkImgFmtPIX |
| 00:54.29 | starseeker | which program are you running? |
| 00:54.30 | AirBender | when compiling brlcad |
| 00:54.33 | starseeker | oh |
| 00:54.36 | AirBender | from svn |
| 00:54.56 | starseeker | um - you might try updating - there've been a lot of commits today... |
| 00:55.05 | starseeker | dunno if any of theim hit that part |
| 00:55.06 | AirBender | ok |
| 00:55.12 | starseeker | which directory is it failing in? |
| 00:55.24 | AirBender | the symbol is required in cmd.c from bwish |
| 00:55.39 | AirBender | src/bwish/cmd.c |
| 00:55.44 | AirBender | lne 76 |
| 00:55.48 | AirBender | line* |
| 00:56.05 | starseeker | erm |
| 00:56.26 | starseeker | yeah, try an svn up first and see if that pulls any fixes |
| 00:56.35 | AirBender | yeap, doing that |
| 00:56.49 | AirBender | 36338 |
| 00:56.58 | AirBender | let's see |
| 01:03.24 | AirBender | lots of warning huh... |
| 01:06.32 | AirBender | by far the most warning-full compilation I've ever done... =D |
| 01:06.46 | starseeker | hmm, font article: http://mihmo.livejournal.com/45152.html |
| 01:06.57 | starseeker | AirBender: yeah, it's a bit noisy in some areas |
| 01:07.06 | starseeker | particularly on the newest gcc compilers |
| 01:08.00 | AirBender | yeap |
| 01:08.34 | CIA-22 | BRL-CAD: 03starseeker * r36339 10/brlcad/trunk/configure.ac: If we aren't ready to build step yet, can't have the Makefile in configure.ac without breaking distcheck. |
| 01:13.05 | *** join/#brlcad matthewmpp_ (n=chatzill@wsip-98-172-82-189.ph.ph.cox.net) |
| 01:13.54 | brlcad | AirBender: what are your configure flags? |
| 01:13.55 | *** part/#brlcad matthewmpp (n=chatzill@wsip-98-172-82-189.ph.ph.cox.net) |
| 01:15.12 | brlcad | a couple dirs are chatty, but you really shouldn't be getting a lot of warnings unless you enabled additional warnings |
| 01:15.13 | AirBender | mmm |
| 01:15.32 | brlcad | more than a dozen of the core are even completely warning-free |
| 01:15.34 | AirBender | --enable-optimized --widh-ogl --prefix=/usr/local/brlcad |
| 01:15.47 | brlcad | huh, interesting |
| 01:16.13 | AirBender | may be the reason is I'm using Gcc 4.x.x |
| 01:16.32 | AirBender | may be the reason is I'm using Gcc 4.4.1 |
| 01:16.45 | brlcad | nah, I've run many a build on 4.4 already |
| 01:16.50 | AirBender | ok |
| 01:17.29 | brlcad | also, fwiw, the bwish issue is something pretty recent from one of the 100+ commits today |
| 01:17.46 | AirBender | I think it's ok now |
| 01:17.46 | brlcad | if you want a stable build, can grab the stable branch instead |
| 01:18.13 | AirBender | I want to be synced with svn |
| 01:18.22 | AirBender | there's no problem |
| 01:18.25 | brlcad | so you were just out of sync? |
| 01:18.31 | AirBender | it's building ok now |
| 01:18.40 | brlcad | or did you edit something? |
| 01:18.53 | AirBender | and the warning were on the first 10 minutes of compiling... |
| 01:19.04 | brlcad | oh, heh |
| 01:19.05 | brlcad | yeah |
| 01:19.16 | brlcad | everything in src/other is not our code, so it's chocked full of warnings |
| 01:19.24 | brlcad | that compiles fire |
| 01:19.26 | brlcad | *first |
| 01:19.33 | AirBender | I see... |
| 01:19.41 | AirBender | well it's my first time with brlcad |
| 01:19.58 | brlcad | src/other are our external dependencies, they compile when it doesn't detect a suitable system-installed version of that dependency |
| 01:20.09 | brlcad | e.g., libpng, libz, tcl/tk, etc |
| 01:20.20 | AirBender | yes, I've read it some minutes ago... |
| 01:21.03 | AirBender | interesting approach... instead of installing every dependency on avery error |
| 01:22.46 | brlcad | *nod*, mostly download/distribution convenience but also for controlled compilation testing |
| 01:22.58 | AirBender | ok |
| 01:23.00 | brlcad | lets us turn everything on or everything off or individually, etc |
| 01:24.22 | AirBender | by the way, do you see brlcad as a replacing for CATIA? |
| 01:25.07 | AirBender | replacement* |
| 01:25.11 | AirBender | sorry for my english |
| 01:25.33 | brlcad | that's certainly a desirable achievement and a package that covers a similar domain |
| 01:25.34 | CIA-22 | BRL-CAD: 03starseeker * r36340 10/brlcad/trunk/src/libdm/dm-rtgl.c: Well, can now refresh with B - but doing a draw of a second item when the first one is working without either letting the first one finish or clearing causes a crash. |
| 01:26.07 | brlcad | but I more see ourselves as just doing the best at whatever our users need, more niche requirements, more flexible customization |
| 01:27.03 | brlcad | CATIA employs more than 1000 developers -- we have quite a ways to reach that scale to be considered an outright replacement for all their features |
| 01:27.16 | AirBender | I know, and I understand that... |
| 01:27.39 | AirBender | I just wanted to read an opinion of the current state |
| 01:28.04 | ``Erik | seen the 'industry diagram' on the web site? |
| 01:28.24 | AirBender | uhmm in the screenshots? |
| 01:29.11 | ``Erik | in the 'diagrams' chunk of the gallery, yes |
| 01:29.22 | ``Erik | http://brlcad.org/gallery/s/diagrams/Industry_Diagram.png.html |
| 01:30.23 | ``Erik | that kinda helps explain our... niche :) |
| 01:30.23 | AirBender | interesting... |
| 01:31.15 | starseeker | hmm, caps only, but looks OK small (not clearly licensed though :-() http://www.kottke.org/plus/type/silkscreen/ |
| 01:31.28 | brlcad | we're slowly expanging towards the left |
| 01:32.54 | AirBender | excellent |
| 01:33.05 | brlcad | starseeker: not bad.. not as good as profont though :) |
| 01:33.17 | brlcad | plus the all-caps is rather annoying |
| 01:33.26 | starseeker | nods |
| 01:33.55 | starseeker | downloads profont to see if he can dig the author's name out of the font file... |
| 01:35.35 | AirBender | cool just finished the build process... installing now |
| 01:35.56 | AirBender | Many thanks for the help |
| 01:36.37 | starseeker | thanks for trying it out! |
| 01:36.41 | AirBender | hope brlcad meets my partner's requirements... |
| 01:36.47 | AirBender | in fact they will be the users... |
| 01:36.57 | starseeker | run /usr/local/brlcad/bin/mged to get going |
| 01:37.10 | AirBender | I'm just building it |
| 01:37.17 | AirBender | ok |
| 01:37.29 | starseeker | you saw the docs page? |
| 01:37.53 | AirBender | not too much |
| 01:37.56 | starseeker | http://brlcad.org/wiki/Documentation |
| 01:38.14 | starseeker | specifically, Introduction to MGED and the MGED Quick Reference Card |
| 01:39.10 | AirBender | ok |
| 01:39.23 | brlcad | AirBender: yeah, that tutorial book is required reading -- just like any CAD system, it's very complex with a lot to learn for new users |
| 01:40.18 | brlcad | if you want to try something fun, can run the "/usr/local/brlcad/bin/benchmark" command to test your system performance |
| 01:40.20 | AirBender | I know... but that's for my partners... I have enough to do with my part(communications/electronics) |
| 01:40.47 | AirBender | will try that |
| 01:42.13 | AirBender | Wow, nice to see good documentation |
| 01:43.57 | AirBender | seems like your specialty is in the cmputer graphics/rendering area... |
| 01:44.30 | AirBender | well it's the most atractive info to publish anyway =D |
| 01:47.37 | starseeker | brlcad: I take it most of the profont bzflag discussion took place on irc? |
| 01:47.46 | brlcad | starseeker: mostly |
| 01:47.59 | brlcad | starseeker: try emailing the profont addrs |
| 01:48.09 | brlcad | there should be a couple in the license file |
| 01:48.36 | brlcad | the problem is the license file is mostly just poorly worded |
| 01:48.54 | starseeker | nods - ooooold ones - figured those had been tried already, but I suppose it can't hurt |
| 01:48.56 | brlcad | I remember it having a couple serious problems |
| 01:50.13 | starseeker | hopes his gmail addy won't be spam flagged... |
| 01:50.14 | brlcad | the terms "maybe" added an additional restriction and didn't explicitly allow derivative works |
| 01:50.30 | brlcad | both those being lgpl/gpl incompatibilities |
| 01:50.40 | starseeker | nods |
| 01:51.01 | brlcad | otherwise the terms are almost cc-by or cc-by-nc (the latter of which would be a problem) |
| 01:51.25 | brlcad | looks up the name |
| 01:51.32 | starseeker | what do you make of the SIL open font license? |
| 01:51.52 | brlcad | so yeah, Carl Osterwald or Steve Gilardi would probably suffice |
| 01:53.13 | AirBender | is the benchmark usable for the developers? |
| 01:53.19 | AirBender | or just for my interest? |
| 01:53.27 | AirBender | it finished |
| 01:53.52 | AirBender | 6651 times faster than reference |
| 01:54.42 | brlcad | AirBender: it explains what the number means, but yeah it's useful to know |
| 01:55.11 | starseeker | raises his eyebrow - never seen a one letter prefix to a mac.com email address |
| 01:55.19 | brlcad | plus it gives you a metric you can run system to system, compilation to compilation, etc .. and gives you a directly comparable metric |
| 01:56.30 | brlcad | starseeker: er, don't think they're valid.. three is the min |
| 01:59.24 | starseeker | brlcad: hmm. must be a red herring |
| 01:59.48 | AirBender | well, hope to be able to cooperate with this great project in the near future |
| 02:00.02 | brlcad | AirBender: likewise! |
| 02:00.02 | starseeker | "Carl R. Osterwald" <i.DeleteThis@mac.com> wrote... |
| 02:00.20 | AirBender | now we will install it on the other computers |
| 02:00.47 | brlcad | we're always looking for new devs |
| 02:00.51 | brlcad | if you're a developer |
| 02:01.01 | AirBender | I've read that there's a C++ abstraction layer in development? |
| 02:01.15 | brlcad | yeah, geometry engine akin to the ACIS engine |
| 02:01.27 | AirBender | ok |
| 02:01.53 | brlcad | we have a C API over geometry services right now |
| 02:02.19 | brlcad | http://brlcad.org/BRL-CAD_Priorities.png |
| 02:02.39 | brlcad | high-level marketing priority talk |
| 02:03.16 | brlcad | starseeker: can't hurt to e-mail/cc every address you find :) |
| 02:04.33 | brlcad | or contacting folks they used to know |
| 02:04.39 | brlcad | we didn't go that far last time |
| 02:04.48 | brlcad | (e.g., http://bishop.mc.duke.edu/bolo/guides/mapedit.html ) |
| 02:05.56 | starseeker | suspects anything short of sending out paid detectives would be easier than finding brlcad a satisfactory replacement font :-) |
| 02:09.02 | *** join/#brlcad maes (n=maes@190.163.31.17) |
| 02:10.58 | *** part/#brlcad maes (n=maes@190.163.31.17) |
| 02:12.28 | *** join/#brlcad maes (n=maes@190.163.31.17) |
| 02:12.50 | brlcad | one e-mail to a dozen addresses is pretty easy :P |
| 02:12.59 | starseeker | hehe |
| 02:13.16 | brlcad | they'll either come back failures or get ignored |
| 02:13.22 | starseeker | it doesn't get serious until we pick up the phone and start calling people |
| 02:15.05 | brlcad | would totally be worth it for that font |
| 02:15.19 | brlcad | tens of thousands of geeks would rejoice |
| 02:16.40 | starseeker | well, distcheck now passes again on the Mac |
| 02:16.57 | starseeker | sans step-g, sadly :-( |
| 02:17.22 | starseeker | really doesn't get that, what it is saying is undefined IS defined in an included header, and Linux can find it... |
| 02:20.38 | starseeker | should probably head home now |
| 02:21.30 | starseeker | saddles up and moves out |
| 02:38.46 | CIA-22 | BRL-CAD: 03brlcad * r36341 10/brlcad/trunk/src/bwish/cmd.c: |
| 02:38.46 | CIA-22 | BRL-CAD: document the Tk_PhotoImageFormat tkImgFmtPIX structure, that it comes from |
| 02:38.46 | CIA-22 | BRL-CAD: libtclcad and providing PIX image processing support to Tk's image subsystem |
| 02:38.46 | CIA-22 | BRL-CAD: (even though it only seems to be used by bwish and not mged oddly enough). |
| 02:38.46 | CIA-22 | BRL-CAD: restructure to avoid decls while we're at it. |
| 03:07.04 | *** part/#brlcad maes (n=maes@190.163.31.17) |
| 04:18.39 | CIA-22 | BRL-CAD: 03brlcad * r36342 10/brlcad/trunk/src/conv/step/OpenNurbsInterfaces.cpp: our nurbs headers have to come after the step/sdai headers because we also define DEBUG_OFF |
| 04:19.37 | CIA-22 | BRL-CAD: 03brlcad * r36343 10/brlcad/trunk/src/conv/step/RepresentationItem.h: missing sdai.h for the SCLP23wahtever wrappage. |
| 04:20.51 | CIA-22 | BRL-CAD: 03brlcad * r36344 10/brlcad/trunk/configure.ac: readd step to build, it should generate the makefile regardless of compilability |
| 04:24.01 | CIA-22 | BRL-CAD: 03brlcad * r36345 10/brlcad/trunk/src/conv/Makefile.am: re-enable traversal into the step dir. critical piece seems to be missing here that you have to define DIST_SUBDIRS so we traverse all dist dirs regardless of SUBDIRS (which isn't the same as EXTRA_DIST'ing a subtree) |
| 04:24.48 | CIA-22 | BRL-CAD: 03brlcad * r36346 10/brlcad/trunk/src/conv/step/Makefile.am: minor cleanup, one-per-line, shouldn't tab empty lines, declare built sources. |
| 04:25.42 | PrezKennedy | brlcad, guess what my current title at the pentagon is |
| 04:26.06 | starseeker | grins evilly... must resist temptation... |
| 04:26.24 | starseeker | brlcad: thanks alot for looking at that step stuff |
| 04:26.32 | starseeker | is grateful |
| 04:26.44 | brlcad | there are compilation failures/assumptions galore that still have to get fixed |
| 04:27.03 | brlcad | PrezKennedy: Monkey Butler? |
| 04:27.21 | PrezKennedy | close... computer programmer |
| 04:27.30 | brlcad | cool |
| 04:39.53 | CIA-22 | BRL-CAD: 03brlcad * r36347 10/brlcad/trunk/include/ (dvec.h vector.h): rename vector.h to dvec.h so we can avoid a name clash with the old stl compatibility header of same name |
| 04:39.53 | CIA-22 | BRL-CAD: 03brlcad * r36348 10/brlcad/trunk/include/dvec.h: rename to dvec.h; need raytrace.h for fastf_t |
| 04:39.53 | CIA-22 | BRL-CAD: 03brlcad * r36349 10/brlcad/trunk/src/conv/step/PullbackCurve.cpp: include dvec.h instead of vector.h so we get the right header |
| 04:39.53 | brlcad | so that gets it to compile for me (untested runtime) |
| 04:39.53 | brlcad | check distcheck and see if it still fails |
| 04:39.55 | PrezKennedy | its all good, contract will be over long before we ever have the tools we need to code :) |
| 04:40.12 | brlcad | excellent |
| 04:46.28 | brlcad | starseeker: mini todo list for the mess in there... 1) copyright headers to all files (sh/header.sh, you or indianlarry since it was proxy should be fine), 2) standard footers, 3) no using namespace std;, 4) headers are a mess (see HACKING) but at least need common.h first, then system, then whatever else everywhere for portability, 5) svn:ignore dir products (docs need that too), 6) indentation (sh/indent.sh) |
| 04:50.05 | starseeker | brlcad: sounds good - I'll hit it first thing tomorrow morning (will sleep momentarily) |
| 04:50.38 | starseeker | sorry about docs - been building out of dir too much and forgot the reorg wiped out the old settings |
| 04:51.22 | starseeker | muchas gracias :-) |
| 04:53.58 | *** join/#brlcad Ralith_ (n=ralith@69.90.49.189) |
| 04:54.24 | *** join/#brlcad kanzure (i=bryan@dhcp-84-36.me.utexas.edu) |
| 06:39.57 | *** join/#brlcad Axman6 (n=Axman6@pdpc/supporter/student/Axman6) [NETSPLIT VICTIM] |
| 06:40.48 | *** join/#brlcad kanzure (i=bryan@dhcp-84-36.me.utexas.edu) [NETSPLIT VICTIM] |
| 07:04.28 | *** join/#brlcad cpc26 (n=cpc26@72.170.156.242) |
| 07:52.45 | *** join/#brlcad cpc26 (n=cpc26@72.170.156.242) |
| 08:53.26 | CIA-22 | BRL-CAD: 03brlcad * r36350 10/brlcad/trunk/include/Makefile.am: vector.h was renamed to dvec.h |
| 09:48.40 | CIA-22 | BRL-CAD: 03d_rossberg * r36351 10/brlcad/trunk/src/librt/ (4 files in 2 dirs): vector.h was renamed to dvec.h because of possible a conflict with a STL header file name |
| 10:13.16 | CIA-22 | BRL-CAD: 03d_rossberg * r36352 10/brlcad/trunk/src/conv/obj-g.c: because of MSVC compiler error: replaced c99 idiom with c89 compatible one (all declarations have to be on top of a block) |
| 12:05.14 | *** join/#brlcad _clock_ (n=_sushi_@80-218-244-105.dclient.hispeed.ch) |
| 12:11.32 | CIA-22 | BRL-CAD: 03brlcad * r36353 10/brlcad/trunk/src/librt/brep_test.cpp: another vector.h to dvec.h conversion |
| 13:32.11 | *** join/#brlcad parigaudi (n=quassel@pd95b7f5e.dip0.t-ipconnect.de) |
| 13:44.40 | *** join/#brlcad Elrohir (n=kvirc@p5B14EE36.dip.t-dialin.net) |
| 13:50.23 | CIA-22 | BRL-CAD: 03brlcad * r36354 10/brlcad/trunk/src/util/Makefile.am: need to put the source in extra dist manually if it's not compiled, just like the others |
| 13:51.39 | CIA-22 | BRL-CAD: 03brlcad * r36355 10/brlcad/trunk/doc/docbook/Makefile.am: handful of other files and one dir missing from the dist. nirt fig20, authors xml, mged05 image, spec image dir, and manpage readme missing. |
| 13:53.36 | starseeker | brlcad: ah, thanks |
| 13:55.21 | brlcad | do you get a slew of warnings in the step dir about "warning: ignoring old commands for target"? |
| 13:55.33 | starseeker | yes |
| 13:55.44 | starseeker | on Mac |
| 13:55.44 | brlcad | is that from a clean checkout? |
| 13:55.46 | starseeker | not sure about Linux |
| 13:55.50 | starseeker | I believe so |
| 13:56.00 | brlcad | not one that "upgraded" to having it enabled |
| 13:56.48 | starseeker | oh, wait |
| 13:57.12 | brlcad | i think it's the two variable rules causing it, but needs to be tested clean first in case it's red herring |
| 13:57.14 | starseeker | I think I saw those warnings only when conv/step was enabled |
| 13:58.04 | starseeker | does a quick svn status |
| 13:58.46 | brlcad | it's not how it is now -- it's whether you had a plain checkout/build with it already on, or if you reran autogen.sh at some point |
| 13:59.20 | brlcad | only way to test clean is: sh autogen.sh && ./configure && make distclean && sh autogen.sh && ./configure --whatevers.... |
| 13:59.28 | brlcad | or just check out clean again |
| 13:59.54 | brlcad | hits the road |
| 14:00.05 | starseeker | brlcad: I've been doing part of that cleaning process, but not full - I was just getting set up for a distcheck build so I'll give it a go |
| 14:00.49 | brlcad | you only need the double autogen/configure when you suspect its automake being pissy .. which those double-rules could be |
| 14:01.03 | brlcad | course, that ${}: rule is screwy too |
| 14:01.49 | brlcad | that probably needs to be changed, plus the other change, and hopefully we're good to go with a few doc updates |
| 14:02.10 | brlcad | need to get larry to document his change in TODO |
| 14:02.13 | brlcad | er NEWS |
| 14:02.28 | starseeker | freudian slip ;-) |
| 14:02.47 | brlcad | not really, TODO needs to be updated too :) |
| 14:03.26 | starseeker | wishes indianlarry had committed his recent work - merging in any changes after I get done with indents and header/footers will be hell |
| 14:03.44 | starseeker | oh, well - burnt hand teaches best :-P |
| 14:05.32 | *** join/#brlcad KingofCSU (n=king@222.247.247.227) |
| 14:11.00 | brlcad | word of caution that indent.sh results need to be visually inspected |
| 14:11.10 | brlcad | particularly if there are #ifdefs or namespaces |
| 14:11.44 | brlcad | easier if there are headers/footers first as the footer will end up being indented indicating there are extra braces |
| 14:37.11 | starseeker | hrm make[2]: *** No rule to make target `html/specifications/en/images', needed by `distdir' |
| 14:37.14 | starseeker | checks |
| 14:38.37 | starseeker | oh - brlcad, the reason spec images weren't there yet is that there aren't any (yet) |
| 14:43.04 | starseeker | the empty directory was causing problems, iirc |
| 14:45.05 | starseeker | doesn't remember which ones |
| 14:56.54 | starseeker | deletes the spec images directory until needed |
| 14:59.51 | CIA-22 | BRL-CAD: 03starseeker * r36356 10/brlcad/trunk/doc/docbook/ (Makefile.am specifications/en/images/): Don't put an empty directory in the svn repository - will re-enable this logic once there are actual specification images. |
| 15:08.08 | starseeker | yeah, it's overriding commands in step dir on a clean build |
| 15:13.08 | CIA-22 | BRL-CAD: 03starseeker * r36357 10/brlcad/trunk/src/conv/step/ (344 files): Add footers to all step conv files that don't appear to be static copies of NIST files - need to discuss what has to happen with the NIST code with indianlarry in order to use the step libs in src/otheer/step |
| 15:17.45 | CIA-22 | BRL-CAD: 03starseeker * r36358 10/brlcad/trunk/src/conv/step/ (STEPEntity.cpp STEPEntity.h STEPWrapper.cpp STEPWrapper.h): Get a few more files that aren't NIST copies. |
| 15:57.10 | starseeker | hmm, interesting: http://pastebin.bzflag.bz/m3e5e2b63 |
| 16:10.00 | starseeker | does Mac have emmintrin.h I wonder... |
| 16:10.38 | starseeker | ok it does... |
| 16:14.21 | starseeker | const double *, that looks ok... |
| 16:39.48 | *** join/#brlcad parigaudi (n=quassel@pd95b7f5e.dip0.t-ipconnect.de) |
| 17:23.48 | ``Erik | ppc mac? didja try building a universal binary? |
| 18:26.46 | ``Erik | "The greeks invented sex. Later, the italians invented it with women." |
| 18:32.52 | starseeker | brlcad: yep, you called it - the fpu implementation works |
| 18:39.54 | brlcad | you should compmile on linux with fpu, then compare sse vs fpu |
| 18:40.22 | starseeker | wouldn't fpu pretty much always be slower? |
| 18:40.31 | starseeker | shudders to think about raytracing |
| 18:42.00 | CIA-22 | BRL-CAD: 03starseeker * r36359 10/brlcad/trunk/src/conv/step/ (PullbackCurve.cpp PullbackCurve.h): Put headers on PullBackCurve files |
| 18:51.27 | CIA-22 | BRL-CAD: 03starseeker * r36360 10/brlcad/trunk/src/conv/step/ (6 files): Switch from using std namespace to explicitly prefixing with std:: (per brlcad's advice) |
| 18:59.28 | *** join/#brlcad Ralith (n=ralith@d142-058-090-093.wireless.sfu.ca) |
| 18:59.38 | starseeker | hey Ralith :-) |
| 19:00.01 | Ralith | hullo |
| 19:00.07 | starseeker | how goes it? |
| 19:00.28 | Ralith | much to my dismay, univ involves work ;_; |
| 19:00.35 | starseeker | heh |
| 19:00.53 | starseeker | yeah, they do like that |
| 19:01.17 | Ralith | at least it's friday |
| 19:02.47 | starseeker | Ralith: do you recall if we're using quaterniuns for rotation in ogre or is it the yaw, pitch, etc. stuff? |
| 19:07.13 | *** join/#brlcad Ralith (n=ralith@d142-058-090-093.wireless.sfu.ca) |
| 19:08.31 | Ralith | and the wifi sucks, to boot. |
| 19:14.07 | starseeker | distcheck passes on Linux |
| 19:17.51 | starseeker | looks at the TODO file for items to do before next release and winces |
| 19:25.48 | Maloeran | starseeker, fpu can be faster if you are doing a lot of cos(), sin(), pow() and other stuff that the SSE instructions won't do |
| 19:26.13 | Maloeran | Besides that... stick to SSE of course :) |
| 19:40.14 | ``Erik | I thought amd had trig 'n shtuff built into theirs? just intel was r-tarded? :) |
| 19:40.25 | ``Erik | (mebbe needs explicit 3dnow stuff) |
| 19:53.35 | Maloeran | There was no trigonometry stuff in 3dnow either |
| 19:54.08 | Maloeran | But it had horizontal instructions, the fun stuff that Intel took 7 years to think about and they managed not to get it quite right either |
| 19:56.07 | Maloeran | Go USD, a recovery of 5% in just a week! |
| 19:56.19 | Maloeran | It's amazing for such a major currency to be so unstable... |
| 20:07.33 | *** join/#brlcad Ralith_ (n=ralith@d142-058-090-093.wireless.sfu.ca) |
| 20:13.58 | brlcad | starseeker: no, the fpu is very often faster -- it depends on a lot of factors |
| 20:14.07 | brlcad | how much work you can keep feeding the pipeline |
| 20:14.48 | brlcad | there's a cost to transfer data into and out of the simd registers |
| 20:15.11 | brlcad | so that cummulative overhead has to be recovered with enough computation and minimal/no interruptions as much possible |
| 20:15.36 | ``Erik | mal: what're you talking about? CAD fluxes like mad, sure, but I'd hardly call it a major ... :> *duck* |
| 20:35.32 | starseeker | makes notes on personal todo... email profont guys, benchmark fpu vs sse, check out ttkdraw... |
| 20:37.11 | Maloeran | ``Erik, that's a good point! You can see it both ways... Either the USD is fluctuating wildly, or it is stable but every other currency in the world is! ;) |
| 20:38.10 | Maloeran | brlcad, cost to puttign stuff in and out of the simd registers is about the same as for the fpu |
| 20:38.50 | Maloeran | Technically, older archs like Athlon64 were faster with fpu stuff ( a movss from memory requires zeroing out the remaining 96 bits, so it was slower ), but recent chips like Core 2 crunch the SSE stuff better |
| 20:39.40 | Maloeran | FPU requieres the code to fxchg all the time to bring any value you want to work on at the top of the stack, SSE lets you access any register directly in a sane way |
| 20:40.28 | brlcad | not just the load |
| 20:40.36 | brlcad | you still have to wait for the pipeline to flush to get your result |
| 20:40.46 | brlcad | that's part of the "load" overhead if you can't fill it |
| 20:40.59 | Maloeran | The latency and throughoutput of the SSE stuff is better for all recent chips, as far as I know |
| 20:41.14 | Maloeran | If you benchmark an Athlon64 or a Pentium 4, then it's a very different story |
| 20:41.55 | brlcad | we tested just two years ago and took a pretty big hit for incorrent work |
| 20:42.32 | brlcad | thinks it was two, a xeon either way |
| 20:42.35 | Maloeran | Besides all the "basic" instructions, you can do stuff like fmin(), fmax(), fabs() without branching on SSE, it's a single instruction |
| 20:42.52 | Maloeran | I guess it was a Xeon based on the Pentium 4 arch |
| 20:43.57 | brlcad | well we have a perfect way to test that theory on most chips with our vmath interface |
| 20:44.22 | Maloeran | You can eliminate a lot of branches if you begin doing bitwise stuff with SSE, but that'll take some use of SSE intrinsics |
| 20:44.59 | brlcad | those are pretty specific cases from what I'm talking about |
| 20:45.16 | brlcad | there are lots of things sse/simd does way better |
| 20:45.19 | brlcad | otherwise what'd be the point |
| 20:45.31 | Maloeran | *nods* Right, I'm not too sure what your benchmark was doing of course |
| 20:45.43 | brlcad | the general commpute case though, isn't beneficial generally speaking though |
| 20:46.25 | Maloeran | Well I would be very surprised, I thought that was only right for chips of the P4 or Athlon64 generation |
| 20:46.41 | brlcad | otherwise the compiler could just do it for all math ops and we'd be good to go -- some compilers can do it for sections of code where there is a loop of ops that translate well, but still not for general case |
| 20:47.04 | Maloeran | nods |
| 20:48.06 | brlcad | if you're writing the algorithm fresh and can keep the pipeline filled, great :) |
| 20:49.02 | Maloeran | I guess I'm a bit biased, I write the code with a good idea in mind of what GCC will output... when I'm not just using SSE intrinsics directly |
| 20:49.03 | brlcad | the test case we were using was a surface solver iirc |
| 20:51.10 | brlcad | the only point I was making was that you can't just take something like a simple single vector multiply, feed it to the simd unit, and expect it to speed up |
| 20:52.25 | Maloeran | Right, although the throughoutput should be higher with SSE on some very recent chips |
| 20:52.42 | Maloeran | x87 is getting rather deprecated |
| 20:52.44 | ``Erik | be amusing to twist vmath up with sse stuff to see what the difference is there, though |
| 20:52.56 | brlcad | throughput should be higher even on older ones |
| 20:53.20 | brlcad | if you have more than random food for the pipeline, you can keep it busy and get some gain |
| 20:53.38 | brlcad | that's just not the general case unless you design for it |
| 20:53.54 | brlcad | ``Erik: yeah, jason actually did that with those vector headers in include/ |
| 20:53.56 | Maloeran | You can keep the x87 busy as well, the Athlon FX and Athlon 64 were doing very intense register renaming under the hood from sequences of fxchg |
| 20:54.06 | brlcad | not vmath directly, but several of the same routines are implemented |
| 20:54.12 | brlcad | which is what he's using in the solver |
| 20:54.50 | brlcad | yeah, l1/l2 cache on the chip is just as important, but likewise you have to plan for it |
| 20:55.03 | brlcad | keeping data coherent and fit in cachelines doesn't just happen |
| 20:55.36 | Maloeran | I never wrote any x87 assembly, tracking where your stuff is on the fpu stack looks painful |
| 20:55.38 | brlcad | if you're already coherent and fit, then yeah sure .. you'll translate pretty simply to simd and can g et even more gains |
| 20:55.45 | Maloeran | nods |
| 20:56.43 | brlcad | totally would love to see all our CSG prims have a coheret shot() routine with a coherent boolean eval to leverage the same techniques that are used for triangle tracing |
| 20:57.00 | brlcad | at least an order of magnitude to be harvested there |
| 20:57.29 | brlcad | even with lots of data validation branches remaining sprinkled throughout |
| 20:57.42 | Maloeran | Sounds like there are other things to worry about before getting into fpu versus SSE :) |
| 20:58.11 | brlcad | everyone likes faster tracing ;) |
| 20:58.18 | brlcad | that'd make a great paper too |
| 20:58.39 | Maloeran | shivers at the mention of writing papers |
| 20:58.45 | brlcad | heh |
| 20:58.59 | brlcad | you shoulda written one on rayforce at the time :) |
| 20:59.13 | Maloeran | It looks most uninteresting to me, really |
| 20:59.19 | brlcad | think most have caught up by now though too |
| 20:59.36 | brlcad | "most" being ill-defined of course |
| 20:59.46 | brlcad | the top five tracer impls |
| 20:59.51 | Maloeran | Perhaps, not too sure if they are still stuck in their BSP ages, or other tree-based techniques |
| 21:00.50 | Maloeran | I code for fun really, I just hate the thought of reading or writing papers |
| 21:06.13 | Maloeran | Such as this rather fast "tricubic weighted b-splines" image filtering I just wrote today for the CFD visualization... I tend to think it would have been rather boring to just implement what some paper might have said |
| 21:09.25 | Maloeran | ( although if the algorithm has an official name, I don't have a clue what it is ) |
| 21:10.42 | ``Erik | damn you, emacs |
| 21:17.33 | brlcad | you might get joy/satisfaction just out of solving the problem, but I hate to spend time solving a problem that has already been solved |
| 21:17.47 | brlcad | there are plenty of unsolved problems to resolve every problem that is simply 'new' to me |
| 21:17.59 | brlcad | to each their own :) |
| 21:20.51 | ``Erik | nice http://jalemanyf.files.wordpress.com/2008/05/microsoft-fail.jpg |
| 21:21.29 | ``Erik | (look at the laptop they're driving with) |
| 21:28.47 | yukonbob | heh |
| 21:38.33 | *** join/#brlcad Ralith (n=ralith@69.90.49.189) |
| 22:28.10 | starseeker | ``Erik: is that legit or a photoshop job? |
| 22:28.43 | ``Erik | dunno, had the same thought myself |
| 22:34.22 | Maloeran | True brlcad, I guess sometimes it just takes as long to understand what someone else did and re-create it, than just create something yourself :) |
| 22:34.59 | Maloeran | Eheh Erik, nice one |
| 22:45.51 | *** join/#brlcad IriX64 (n=Warlock@bas2-sudbury98-1177726213.dsl.bell.ca) |
| 22:46.05 | brlcad | Maloeran: that's often true, but even if it's faster to figure it out myself, to me that is still wasted effort as it builds on nothing but my own experiences (which as impressive as they may be will never be as much as everyone else currently and previously alive) |
| 22:50.00 | brlcad | plus in my fatalistic view, I'll be dead soon enough and everything I've learned is moot, so there's value in knowing the things I've done are actually 'improvements' in some regard as they build on those before; for whatever qualification of 'better' (faster, easier, more maintainable, foss-style free, innovative, new, whatever) |
| 22:51.57 | *** join/#brlcad louipc (n=louipc@archlinux/trusteduser/louipc) |
| 22:52.23 | ``Erik | just like exercise, you'll be dead soon enough, and any muscle you have will just rot away, and there're other people with more muscle :D so exercise would be pointless |
| 22:52.26 | ``Erik | *duck* |
| 22:52.48 | brlcad | if muscle were the goal, that would be completely true :) |
| 22:53.08 | ``Erik | is getting ready to code up something overdone and useless just to get the brain exercise from it :) |
| 22:53.48 | ``Erik | provided, of course, she decides to let me O.o cats seem to know that their proper place is on the laptops keyboard. |
| 22:53.49 | brlcad | I do it for the "drugs" .. the endorphin rushes, the runner's highs, the 'feel good' and 'feel fantastic' times |
| 22:54.24 | ``Erik | endorphines trigger the opiate receptors, there're other ways to trigger those :D not nearly as healthy, but easier and quicker |
| 22:54.26 | brlcad | makes me happy, might as well enjoy my limited time here |
| 22:54.39 | brlcad | right |
| 22:54.56 | ``Erik | or; bind to, rather |
| 22:55.11 | ``Erik | brain chemistry is nutty stuff |
| 22:56.19 | brlcad | my way doesn't actually endanger others (generally speaking), shorten my lifespan (generally speaking), require interacting with nefarious individuals, or make an impact on my ability to buy toys (which also just make me happy) |
| 22:56.46 | ``Erik | whoa, I don't need to know about your "toys" there, dude |
| 22:57.05 | brlcad | my toys rock |
| 22:57.12 | brlcad | :) |
| 22:57.48 | ``Erik | speaking of, I've been thinking about picking up the sheevaplugs bigger brother |
| 22:58.11 | yukonbob | laughs about ``Erik going on about "toys" and "plugs" |
| 22:58.39 | brlcad | sheep plugs? ew |
| 22:58.56 | ``Erik | sshoot, had the url the other day |
| 22:59.17 | ``Erik | yukonbob: I thought you liked debian? O.o |
| 22:59.31 | brlcad | this thing? http://www.treehugger.com/files/2009/02/marvell-sheevaplug-plug-pc-computer-wall-wart.php |
| 23:00.14 | ``Erik | that's the $100 wall wart version, there's a $250 one with loads more to it |
| 23:00.25 | yukonbob | ``Erik: /me -used- to be a debian person, before I evolved to *BSD. |
| 23:00.58 | ``Erik | ahhh, you saw the light, hehehe |
| 23:00.58 | ``Erik | the sheeva stuff all comes with debian, I'm hoping to get fbsd going on one |
| 23:01.22 | louipc | wow that's awesome |
| 23:03.48 | yukonbob | as far a Linux goes, debian seems sane to me. I'm inclinded to say same about slack, but these days I only _read_ about it (infrequently). Been ~14 years since I've _run_ it. |
| 23:06.34 | ``Erik | http://www.globalscaletechnologies.com/p-24-openrd-client-openrd-client-board-with-enclosure.aspx |
| 23:06.37 | ``Erik | there we go |
| 23:06.52 | ``Erik | hah, yowza, that's quite a while :) |
| 23:06.57 | ``Erik | I went to bsd about 10 years ago |
| 23:09.06 | yukonbob | <- FreeBSD dabbling ~10 years ago, NetBSD permanently since NetBSD 1.6 |
| 23:09.51 | ``Erik | never tried netbsd. fbsd is my bread&butter, done some obsd, and I've even dug up old 43bsd and 44bsdlite for simh |
| 23:10.05 | ``Erik | (I think the 43 used the 'tahoe' set) |
| 23:10.30 | ``Erik | on a vax 780... was working on installing BRL-CAD 4.4 on it... :D |
| 23:23.45 | ``Erik | *snrkt* http://www.collegehumor.com/picture:1923263 nice |
| 23:56.11 | *** join/#brlcad Axman6 (n=Axman6@pdpc/supporter/student/Axman6) |
