IRC log for #brlcad on 20091106

00:00.13 Ralith brlcad: you might remember doing a glass render of it?
00:02.15 ``Erik recalls the impeller, vagually recalls some discussion about having a draft of it available for people to test
00:49.11 brlcad Ralith: course I remember.. asn't that long ago :)
00:50.15 brlcad he had a script, no writeup iirc, though
00:50.49 Ralith aw.
00:50.54 Ralith script'd be interesting nonetheless
00:50.56 Ralith know where I can find it?
00:51.15 Ralith is, to his surprise, starting to hear from reprappers interested in mged
01:46.07 brlcad i'd have to dig through the logs
01:46.17 brlcad maybe someone else is motivated ;)
01:54.10 ``Erik http://www.motivatedphotos.com/?id=22920
03:03.40 CIA-37 BRL-CAD: 03brlcad * r36424 10/brlcad/trunk/src/libbu/ (parse.c ptbl.c rb_insert.c semaphore.c): quell extra compilation warnings
03:05.40 brlcad heh
03:12.35 brlcad mmm.. tasty: http://www.motivatedphotos.com/?id=1374
03:12.45 brlcad wonder if you can get that toasted
03:43.51 CIA-37 BRL-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.
03:49.35 CIA-37 BRL-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 ``Erik you'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)
13:17.17 CIA-37 BRL-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)
15:21.10 CIA-37 BRL-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 ``Erik heh, "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)
16:07.59 brlcad senses 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/
16:22.02 brlcad heh, a recursive bitcount function
16:24.56 brlcad ``Erik: there's a high-performance 32-bit bit-counter in src/libbu/bitv.c
16:25.20 brlcad as well as other general bit vector facilities if that's what you were doing with that..
16:26.25 brlcad starseeker: do you think you will be able to sync and tag today?
16:46.17 PrezKennedy brlcad, what type of monitors do you have at work these days?
17:00.36 brlcad lcd?
17:00.38 brlcad flat ones?
17:01.10 brlcad 30"'s and 24"'s
17:24.46 *** join/#brlcad louipc (n=louipc@archlinux/trusteduser/louipc)
17:45.34 CIA-37 BRL-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 ``Erik recursive ftw
18:17.43 ``Erik was also looking at doing some inline assembly to use SSE shtuff in that file O.O
18:46.10 brlcad there's a handful of places I've thought about using some inline too
18:46.35 brlcad but 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
18:47.14 brlcad a libbn/vmath sse toggle would be pretty interesting to revisit
18:49.57 starseeker brlcad: I'll try - I'll see if I can get all the distcheck builds done
18:50.09 brlcad k
18:50.43 brlcad really needs to at least get tagged today if we can so stable nightly will pick it up tonight
18:50.51 starseeker nods
19:00.51 *** join/#brlcad samrose (n=samrose@66.93.0.189)
19:05.53 ``Erik 2hm
19:06.09 ``Erik on x86, runtime check if SSE is available and have both codepaths available in the binary?
19:06.17 ``Erik librtsse.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)
19:44.15 CIA-37 BRL-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.
19:44.49 brlcad that'd be cool, but even more work :)
19:44.58 *** join/#brlcad Ralith (n=ralith@d142-058-091-091.wireless.sfu.ca)
19:48.27 CIA-37 BRL-CAD: 03bob1961 * r36431 10/brlcad/trunk/src/archer/archer: Modify where to look for archer.ico.
19:51.52 CIA-37 BRL-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)
19:55.41 CIA-37 BRL-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.
19:58.30 CIA-37 BRL-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).
19:59.46 CIA-37 BRL-CAD: 03bob1961 * r36435 10/brlcad/trunk/src/libtclcad/ged_obj.c: Tweak the png command a bit.
20:04.48 CIA-37 BRL-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
20:05.26 CIA-37 BRL-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..
20:29.37 CIA-37 BRL-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.
20:32.17 CIA-37 BRL-CAD: 03brlcad * r36439 10/brlcad/trunk/src/util/pix-png.c: match case ordering
20:35.03 CIA-37 BRL-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.
21:27.30 CIA-37 BRL-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
21:43.42 CIA-37 BRL-CAD: 03starseeker * r36442 10/brlcad/trunk/ (ChangeLog NEWS include/conf/PATCH): OK, here we go - update version numbers, ChangeLog
21:43.47 CIA-37 BRL-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)
21:44.53 brlcad ~nickometer R0b0t1
21:44.59 brlcad heh
21:45.21 R0b0t1 wat
21:45.24 brlcad R0b0t1: howdy!
21:45.28 brlcad ignore that :)
21:45.32 R0b0t1 :D
21:45.32 starseeker cool :-)
21:45.38 starseeker ~nickometer starseeker
21:45.45 R0b0t1 What.
21:45.47 starseeker hehe
21:45.55 louipc ~nickometer Irix64
21:46.18 starseeker brlcad: I hadn't see that feature before
21:47.13 louipc adding 1 to the end of my nick makes me 22% lame
21:47.28 brlcad so don't do that ;)
21:47.57 louipc heheh
21:48.00 brlcad the algorithm has it's flaws .. some lameness it can't detect
21:48.14 CIA-37 BRL-CAD: 03brlcad * r36444 10/brlcad/trunk/src/util/ (pix-png.c pix-ppm.c): no need for usage to be static
21:48.31 brlcad ~nickometer thisissomereallyannoyinglylonglamenick
21:48.39 brlcad ooh, it's gotten better
21:48.50 louipc should be in the 90s
21:48.58 brlcad heh
21:50.36 R0b0t1 ~nickometer a
21:50.57 louipc ~nickometer 1
21:51.02 louipc hahah
21:51.45 R0b0t1 ~nickometer p3n1l31n5t7rum3nt0ft0r7ur3
21:58.54 yukonbob ~nickometer yukonbob
21:59.02 yukonbob ah, thanks ibot
22:00.10 brlcad ~nickometer 123
22:00.21 brlcad tricksy that one
22:00.41 brlcad ~nickometer 1234567890
22:00.56 CIA-37 BRL-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.
22:06.16 CIA-37 BRL-CAD: 03brlcad * r36446 10/brlcad/trunk/src/util/pix-ppm.c: cleanup defines no longer used
22:13.42 yukonbob ~nickometer brlcads
22:13.47 yukonbob ~nickometer brlcad
22:14.02 yukonbob ~nickometer ``Erik
22:14.09 yukonbob ohhhhhhhhh
22:15.28 louipc HAHH
22:35.28 ``Erik oh, hey, I'm WAY lamer than that
22:35.30 ``Erik damn bot
22:36.11 ``Erik <-- pets his stealth marks
22:36.32 ``Erik ~nickometer `````````````
22:36.50 ``Erik lime heuristic :)
22:37.04 ``Erik lame, even
22:37.59 Maloeran scratches his head as he just can't find GCC's floating point software emulation in the whole source tree
22:38.33 Maloeran But really, I'm only very curious what the algorithm for pow() looks like. Anyone got some pointers?
22:40.10 CIA-37 BRL-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 ``Erik well, on yoru fbsd machine, cd /usr/src, fire up cscope and find it :D
22:42.13 Maloeran Yes well, I'm grepping abundantely, I just... don't find it!
22:43.32 Maloeran Alternatively, anyone of you who actually studied that computer science stuff could lead me to the algorithm for computing floating point powers? :)
22:43.47 ``Erik fold_builtin_pow() ?
22:43.54 ``Erik in builtins.c ?
22:44.04 CIA-37 BRL-CAD: 03brlcad * r36448 10/brlcad/trunk/NEWS:
22:44.06 CIA-37 BRL-CAD: per confusion by a particular user, improved the pix-ppm tool (and pix-png as
22:44.08 CIA-37 BRL-CAD: collatoral damage) by adding a -o output option so you don't have to redirect
22:44.10 CIA-37 BRL-CAD: stdout if you don't want to. even better, you can redirect AND use -o output.
22:45.26 Maloeran Woah, that actually leads somewhere
22:45.39 Maloeran Thanks Erik, not sure how you did it but I'm impressed
22:45.56 ``Erik my grep-fu is better than yours. :D
22:46.01 Maloeran Yes :(
22:46.25 ``Erik blames his college education :> *duck*
22:48.05 Maloeran That function doesn't actually compute pow(), it tries to simplify it, but I hope I can backtrack from that point
22:48.39 ``Erik it passes the simplified tree to an expression folder that copes with it
22:48.51 ``Erik if I understood correctly from my quick skim
22:49.25 Maloeran Right, it tries to simplify the pow() call to something cheaper to compute
22:49.33 Maloeran I'm looking for the big expensive version
22:51.14 Maloeran Ah well, that doesn't actually lead to the floating point software emulation, it's for optimization of pow() with constants
22:52.27 ``Erik hm
22:52.28 ``Erik *shrug*
22:54.49 ``Erik mebbe gmp would be a better target?
22:54.52 Maloeran So from all these years in college, you wouldn't know something about the pow() algorithm? :)
22:55.19 Maloeran Yes maybe. I assumed I could find that easily in GCC's source code, it's supposed to have a complete floating point emulation layer
22:55.24 Maloeran ... somewhere...
22:55.50 ``Erik I'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 ``Erik that's what differentiates university from a tech school
22:59.06 Maloeran Found 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 ``Erik gcc is hardly a good place to look for clean algorithms... :D
23:00.29 Maloeran I don't know, I like their source code design
23:05.53 Maloeran Gah. Looks like that in GMP, you can't do float^float?
23:06.12 Maloeran The exponent must be an integer, that's terrible
23:09.56 Maloeran That'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 ``Erik expmed.c seems ... screwy O.o heh
23:23.49 Maloeran Sure, I could compile -msoft-float ... All right let's try that
23:24.17 Maloeran I was googling online wondering how could GMP not support powers and logarithms with floats
23:25.26 ``Erik heh, decNumberExponent in gcclib raises to an int, too
23:26.32 Maloeran -msoft-float makes a call to "pow" in glibc instead of using the fpu, not much better!
23:26.50 ``Erik heh
23:26.58 Maloeran And glibc doesn't provide floating point software emulation
23:27.31 ``Erik is starting to wonder if powf() still works on reasonably modern gcc's O.O
23:27.36 ``Erik on fpu-less cpu's
23:28.33 Maloeran There has got to be a software implementation of pow() *somewhere*
23:29.36 Maloeran Disassembled libm.so, pow() just runs the x87 instruction
23:34.08 CIA-37 BRL-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 ``Erik http://wurstcaptures.untergrund.net/assembler_tricks.html ?
23:35.53 CIA-37 BRL-CAD: 03brlcad * r36450 10/brlcad/trunk/sh/cmakecheck.sh:
23:35.55 CIA-37 BRL-CAD: update the script to not really care where/how a file is referenced for cmake
23:35.57 CIA-37 BRL-CAD: since the SOURCES thing it keys on for our libraries is just a convention.
23:35.59 CIA-37 BRL-CAD: assume if a source file is referenced, that it's good enough to consider it in
23:36.01 CIA-37 BRL-CAD: sync. this lets us add in the intaval, iges, and nirt directories to the list.
23:38.21 Maloeran Meh, it's an approximation for values strictly in the range [0,1] and [1,128]
23:38.41 Maloeran But I think I found something in GCC's source, gcc/real.c is their floating point software emulation stuff
23:40.10 Maloeran It only supports integer exponents too! This... can't be...
23:44.47 Maloeran Looks 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 ``Erik well
23:45.27 ``Erik there it is in msun
23:45.37 ``Erik e_powf.c
23:45.58 ``Erik does leenewx have msun stuff?
23:47.11 ``Erik http://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
23:47.24 Maloeran Oh, nice
23:48.59 Maloeran You gave me e_expf.c but I found e_powf.c, nice
23:49.32 ``Erik oh, srry, found it on my machine, was trying to find the cgi one real fast
23:49.47 Maloeran nods
23:49.51 Maloeran :) Thanks

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