IRC log for #brlcad on 20091106

00:00.13Ralithbrlcad: you might remember doing a glass render of it?
00:02.15``Erikrecalls the impeller, vagually recalls some discussion about having a draft of it available for people to test
00:49.11brlcadRalith: course I remember.. asn't that long ago :)
00:50.15brlcadhe had a script, no writeup iirc, though
00:50.49Ralithaw.
00:50.54Ralithscript'd be interesting nonetheless
00:50.56Ralithknow where I can find it?
00:51.15Ralithis, to his surprise, starting to hear from reprappers interested in mged
01:46.07brlcadi'd have to dig through the logs
01:46.17brlcadmaybe someone else is motivated ;)
01:54.10``Erikhttp://www.motivatedphotos.com/?id=22920
03:03.40CIA-37BRL-CAD: 03brlcad * r36424 10/brlcad/trunk/src/libbu/ (parse.c ptbl.c rb_insert.c semaphore.c): quell extra compilation warnings
03:05.40brlcadheh
03:12.35brlcadmmm.. tasty: http://www.motivatedphotos.com/?id=1374
03:12.45brlcadwonder if you can get that toasted
03:43.51CIA-37BRL-CAD: 03brlcad * r36425 10/brlcad/trunk/NEWS: include a note that there have been further tweaks and enhancements that affect/improve NURBS ray-tracing via keith.
03:49.35CIA-37BRL-CAD: 03brlcad * r36426 10/brlcad/trunk/NEWS: incomplete for plate-style non-solid bspline/nurbs geometries, but I made old bspline/nurbs use the new/improved opennurbs ray-tracing.
07:22.32*** join/#brlcad d_rossberg (n=rossberg@BZ.BZFLAG.BZ)
11:42.04``Erikyou'd just end up with tuna breath
12:02.23*** join/#brlcad mafm (n=mafm@217.Red-83-38-34.dynamicIP.rima-tde.net)
12:09.53*** join/#brlcad b0ef (n=b0ef@157.26.202.84.customer.cdi.no)
13:17.17CIA-37BRL-CAD: 03bob1961 * r36427 10/brlcad/trunk/misc/win32-msvc8/ (3 files in 3 dirs): Update windows build to accomodate adding/deleting source files from build.
13:41.39*** join/#brlcad parigaudi (n=quassel@217.91.127.94)
15:21.10CIA-37BRL-CAD: 03erikgreenwald * r36428 10/brlcad/trunk/src/librt/primitives/metaball/metaball.c: add/use a function two search for the surface intersection between two points.
15:27.04``Erikheh, "two search", I'm teh smrtt
15:42.20*** join/#brlcad Elrohir (n=kvirc@p5B14FEEB.dip.t-dialin.net)
15:57.27*** join/#brlcad samrose (n=samrose@68.73.192.212)
16:07.59brlcadsenses a slight disturbance in the force
16:08.39*** join/#brlcad poolio (n=poolio@63.246.136.16)
16:08.40*** join/#brlcad Maloeran (n=maloeran@glvortex.net) [NETSPLIT VICTIM]
16:08.53*** join/#brlcad indianlarry (n=indianla@BZ.BZFLAG.BZ) [NETSPLIT VICTIM]
16:12.23*** join/#brlcad ibot (i=ibot@rikers.org)
16:12.23*** topic/#brlcad is BRL-CAD Open Source Solid Modeling || http://brlcad.org || http://sf.net/projects/brlcad || Release 7.16.0 posted (20091006) || Logs at http://ibot.rikers.org/%23brlcad/
16:22.02brlcadheh, a recursive bitcount function
16:24.56brlcad``Erik: there's a high-performance 32-bit bit-counter in src/libbu/bitv.c
16:25.20brlcadas well as other general bit vector facilities if that's what you were doing with that..
16:26.25brlcadstarseeker: do you think you will be able to sync and tag today?
16:46.17PrezKennedybrlcad, what type of monitors do you have at work these days?
17:00.36brlcadlcd?
17:00.38brlcadflat ones?
17:01.10brlcad30"'s and 24"'s
17:24.46*** join/#brlcad louipc (n=louipc@archlinux/trusteduser/louipc)
17:45.34CIA-37BRL-CAD: 03brlcad * r36429 10/brlcad/trunk/src/util/pix-ppm.c: cleanup formatting but more importantly improve the usage statement to indicate that output is on stdout
18:14.59``Erik*burp*
18:16.33``Erikrecursive ftw
18:17.43``Erikwas also looking at doing some inline assembly to use SSE shtuff in that file O.O
18:46.10brlcadthere's a handful of places I've thought about using some inline too
18:46.35brlcadbut didn't want to do the configure foo to make sure it's portable at the time, and the payoff was minimal given other stuff
18:47.14brlcada libbn/vmath sse toggle would be pretty interesting to revisit
18:49.57starseekerbrlcad: I'll try - I'll see if I can get all the distcheck builds done
18:50.09brlcadk
18:50.43brlcadreally needs to at least get tagged today if we can so stable nightly will pick it up tonight
18:50.51starseekernods
19:00.51*** join/#brlcad samrose (n=samrose@66.93.0.189)
19:05.53``Erik2hm
19:06.09``Erikon x86, runtime check if SSE is available and have both codepaths available in the binary?
19:06.17``Eriklibrtsse.so ?
19:23.11``Erik"I'm not lost, I'm spacially displaced"
19:43.02*** join/#brlcad _sushi_ (n=_sushi_@84-73-204-88.dclient.hispeed.ch)
19:44.15CIA-37BRL-CAD: 03brlcad * r36430 10/brlcad/trunk/include/ (config_win.h dm_xvars.h): tk for windows provides the X11 types that are part of the tk api. unbreak the windows build and improve the typedeffage we don't typedef if we already have them.
19:44.49brlcadthat'd be cool, but even more work :)
19:44.58*** join/#brlcad Ralith (n=ralith@d142-058-091-091.wireless.sfu.ca)
19:48.27CIA-37BRL-CAD: 03bob1961 * r36431 10/brlcad/trunk/src/archer/archer: Modify where to look for archer.ico.
19:51.52CIA-37BRL-CAD: 03bob1961 * r36432 10/brlcad/trunk/misc/win32-msvc8/tclsh/library/installTree.tcl: Added lines for attrGroupsDisplayUtilityP. No longer copying entire doc directory to install directory (i.e. it breaks nsis)
19:55.41CIA-37BRL-CAD: 03bob1961 * r36433 10/brlcad/trunk/src/tclscripts/ (archer/Archer.tcl lib/Ged.tcl): Added code to reduce the number of unnecessary redraws in Archer. More to follow.
19:58.30CIA-37BRL-CAD: 03bob1961 * r36434 10/brlcad/trunk/src/libged/ (dg_obj.c draw.c): Fixed a bug that potentially tries to use a bu_list as a solid (i.e. when the list is empty).
19:59.46CIA-37BRL-CAD: 03bob1961 * r36435 10/brlcad/trunk/src/libtclcad/ged_obj.c: Tweak the png command a bit.
20:04.48CIA-37BRL-CAD: 03brlcad * r36436 10/brlcad/trunk/src/util/ (pix-png.1 pix-png.c): improve pix-png output and add a '-o filename' option so that a file redirect is not required
20:05.26CIA-37BRL-CAD: 03brlcad * r36437 10/brlcad/trunk/src/util/pix-ppm.c: make sure pix-ppm doesn't blather raw ppm data directly to a terminal..
20:29.37CIA-37BRL-CAD: 03brlcad * r36438 10/brlcad/trunk/src/util/ (pix-png.1 pix-png.c): refactor so that you can actually send image data out to both a file and a specified redirect, in case some crazy user actually tries it.
20:32.17CIA-37BRL-CAD: 03brlcad * r36439 10/brlcad/trunk/src/util/pix-png.c: match case ordering
20:35.03CIA-37BRL-CAD: 03brlcad * r36440 10/brlcad/trunk/src/util/pix-ppm.c: make pix-ppm work a lot like pix-png, stubbing in support for a -o option and adding additional tty protections for input/output checking.
21:27.30CIA-37BRL-CAD: 03bob1961 * r36441 10/brlcad/trunk/include/opennurbs_ext.h: Sean changed size_t to int in a few places to iterate over BA node min/max indices. Also includes a WS update
21:43.42CIA-37BRL-CAD: 03starseeker * r36442 10/brlcad/trunk/ (ChangeLog NEWS include/conf/PATCH): OK, here we go - update version numbers, ChangeLog
21:43.47CIA-37BRL-CAD: 03brlcad * r36443 10/brlcad/trunk/src/util/pix-ppm.c: eliminate two globals for width/height
21:44.34*** join/#brlcad R0b0t1 (n=Enigma@unaffiliated/r0b0t1)
21:44.53brlcad~nickometer R0b0t1
21:44.59brlcadheh
21:45.21R0b0t1wat
21:45.24brlcadR0b0t1: howdy!
21:45.28brlcadignore that :)
21:45.32R0b0t1:D
21:45.32starseekercool :-)
21:45.38starseeker~nickometer starseeker
21:45.45R0b0t1What.
21:45.47starseekerhehe
21:45.55louipc~nickometer Irix64
21:46.18starseekerbrlcad: I hadn't see that feature before
21:47.13louipcadding 1 to the end of my nick makes me 22% lame
21:47.28brlcadso don't do that ;)
21:47.57louipcheheh
21:48.00brlcadthe algorithm has it's flaws .. some lameness it can't detect
21:48.14CIA-37BRL-CAD: 03brlcad * r36444 10/brlcad/trunk/src/util/ (pix-png.c pix-ppm.c): no need for usage to be static
21:48.31brlcad~nickometer thisissomereallyannoyinglylonglamenick
21:48.39brlcadooh, it's gotten better
21:48.50louipcshould be in the 90s
21:48.58brlcadheh
21:50.36R0b0t1~nickometer a
21:50.57louipc~nickometer 1
21:51.02louipchahah
21:51.45R0b0t1~nickometer p3n1l31n5t7rum3nt0ft0r7ur3
21:58.54yukonbob~nickometer yukonbob
21:59.02yukonbobah, thanks ibot
22:00.10brlcad~nickometer 123
22:00.21brlcadtricksy that one
22:00.41brlcad~nickometer 1234567890
22:00.56CIA-37BRL-CAD: 03brlcad * r36445 10/brlcad/trunk/src/util/pix-ppm.c: refactor the ppm output into a single write function and fix the header always going to stdout instead of the specified output.
22:06.16CIA-37BRL-CAD: 03brlcad * r36446 10/brlcad/trunk/src/util/pix-ppm.c: cleanup defines no longer used
22:13.42yukonbob~nickometer brlcads
22:13.47yukonbob~nickometer brlcad
22:14.02yukonbob~nickometer ``Erik
22:14.09yukonbobohhhhhhhhh
22:15.28louipcHAHH
22:35.28``Erikoh, hey, I'm WAY lamer than that
22:35.30``Erikdamn bot
22:36.11``Erik<-- pets his stealth marks
22:36.32``Erik~nickometer `````````````
22:36.50``Eriklime heuristic :)
22:37.04``Eriklame, even
22:37.59Maloeranscratches his head as he just can't find GCC's floating point software emulation in the whole source tree
22:38.33MaloeranBut really, I'm only very curious what the algorithm for pow() looks like. Anyone got some pointers?
22:40.10CIA-37BRL-CAD: 03brlcad * r36447 10/brlcad/trunk/src/util/pix-ppm.c: enable the -o output option and also allow redirected output per pix-png
22:40.32``Erikwell, on yoru fbsd machine, cd /usr/src, fire up cscope and find it :D
22:42.13MaloeranYes well, I'm grepping abundantely, I just... don't find it!
22:43.32MaloeranAlternatively, anyone of you who actually studied that computer science stuff could lead me to the algorithm for computing floating point powers? :)
22:43.47``Erikfold_builtin_pow() ?
22:43.54``Erikin builtins.c ?
22:44.04CIA-37BRL-CAD: 03brlcad * r36448 10/brlcad/trunk/NEWS:
22:44.06CIA-37BRL-CAD: per confusion by a particular user, improved the pix-ppm tool (and pix-png as
22:44.08CIA-37BRL-CAD: collatoral damage) by adding a -o output option so you don't have to redirect
22:44.10CIA-37BRL-CAD: stdout if you don't want to. even better, you can redirect AND use -o output.
22:45.26MaloeranWoah, that actually leads somewhere
22:45.39MaloeranThanks Erik, not sure how you did it but I'm impressed
22:45.56``Erikmy grep-fu is better than yours. :D
22:46.01MaloeranYes :(
22:46.25``Erikblames his college education :> *duck*
22:48.05MaloeranThat function doesn't actually compute pow(), it tries to simplify it, but I hope I can backtrack from that point
22:48.39``Erikit passes the simplified tree to an expression folder that copes with it
22:48.51``Erikif I understood correctly from my quick skim
22:49.25MaloeranRight, it tries to simplify the pow() call to something cheaper to compute
22:49.33MaloeranI'm looking for the big expensive version
22:51.14MaloeranAh well, that doesn't actually lead to the floating point software emulation, it's for optimization of pow() with constants
22:52.27``Erikhm
22:52.28``Erik*shrug*
22:54.49``Erikmebbe gmp would be a better target?
22:54.52MaloeranSo from all these years in college, you wouldn't know something about the pow() algorithm? :)
22:55.19MaloeranYes maybe. I assumed I could find that easily in GCC's source code, it's supposed to have a complete floating point emulation layer
22:55.24Maloeran... somewhere...
22:55.50``ErikI'm sure I saw something about it, but the point of university isn't applied knowledge as much as the skill to research topics and have a breadth of understanding, from what I've seen
22:56.01``Erikthat's what differentiates university from a tech school
22:59.06MaloeranFound it in GMP. Something tells me the algorithm to compute a fixed 32 or 64 bits would have been simpler... but that will do, thanks!
22:59.30``Erikgcc is hardly a good place to look for clean algorithms... :D
23:00.29MaloeranI don't know, I like their source code design
23:05.53MaloeranGah. Looks like that in GMP, you can't do float^float?
23:06.12MaloeranThe exponent must be an integer, that's terrible
23:09.56MaloeranThat's really disappointing form a big number math library, I'm amazed...
23:11.51*** join/#brlcad Ralith (n=ralith@69.90.49.189)
23:19.35``Erik(could always compile -march=i386 and disassemble?)
23:19.46``Erikexpmed.c seems ... screwy O.o heh
23:23.49MaloeranSure, I could compile -msoft-float ... All right let's try that
23:24.17MaloeranI was googling online wondering how could GMP not support powers and logarithms with floats
23:25.26``Erikheh, decNumberExponent in gcclib raises to an int, too
23:26.32Maloeran-msoft-float makes a call to "pow" in glibc instead of using the fpu, not much better!
23:26.50``Erikheh
23:26.58MaloeranAnd glibc doesn't provide floating point software emulation
23:27.31``Erikis starting to wonder if powf() still works on reasonably modern gcc's O.O
23:27.36``Erikon fpu-less cpu's
23:28.33MaloeranThere has got to be a software implementation of pow() *somewhere*
23:29.36MaloeranDisassembled libm.so, pow() just runs the x87 instruction
23:34.08CIA-37BRL-CAD: 03brlcad * r36449 10/brlcad/trunk/src/nirt/CMakeLists.txt: create the showshot binary just so the cmake file is in sync with the Makefile.am
23:34.31``Erikhttp://wurstcaptures.untergrund.net/assembler_tricks.html ?
23:35.53CIA-37BRL-CAD: 03brlcad * r36450 10/brlcad/trunk/sh/cmakecheck.sh:
23:35.55CIA-37BRL-CAD: update the script to not really care where/how a file is referenced for cmake
23:35.57CIA-37BRL-CAD: since the SOURCES thing it keys on for our libraries is just a convention.
23:35.59CIA-37BRL-CAD: assume if a source file is referenced, that it's good enough to consider it in
23:36.01CIA-37BRL-CAD: sync. this lets us add in the intaval, iges, and nirt directories to the list.
23:38.21MaloeranMeh, it's an approximation for values strictly in the range [0,1] and [1,128]
23:38.41MaloeranBut I think I found something in GCC's source, gcc/real.c is their floating point software emulation stuff
23:40.10MaloeranIt only supports integer exponents too! This... can't be...
23:44.47MaloeranLooks like no programmer on earth knows how to compute pow() with floats, since no one implements it ( GMP, gcc, etc. ). Only hardware designers know its secrets
23:45.23``Erikwell
23:45.27``Erikthere it is in msun
23:45.37``Erike_powf.c
23:45.58``Erikdoes leenewx have msun stuff?
23:47.11``Erikhttp://www.freebsd.org/cgi/cvsweb.cgi/src/lib/msun/src/e_expf.c?rev=1.14.2.1.2.1;content-type=text%2Fx-cvsweb-markup
23:47.24MaloeranOh, nice
23:48.59MaloeranYou gave me e_expf.c but I found e_powf.c, nice
23:49.32``Erikoh, srry, found it on my machine, was trying to find the cgi one real fast
23:49.47Maloerannods
23:49.51Maloeran:) Thanks

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