| 00:34.30 | ``Erik | ../../rtirenderblend4sse.c: In function `elemcallb': |
| 00:34.30 | ``Erik | ../../rtirenderblend4sse.c:251: warning: implicit declaration of function `_mm_castps_si128' |
| 00:34.30 | ``Erik | ../../rtirenderblend4sse.c:251: error: incompatible type for argument 2 of `_mm_and_si128' |
| 00:39.08 | Maloeran | Hum. Okay |
| 00:39.31 | Maloeran | Do you have SSE2 on that box? |
| 00:40.22 | ``Erik | <PROTECTED> |
| 00:42.20 | Maloeran | On FreeBSD, right? Does it find emmintrin.h? Don't you have these pseudo function calls in it? |
| 00:43.18 | Maloeran | Perhaps I used non-standard SSE intrinsics by mistake, I'm just a bit surprised |
| 00:43.48 | ``Erik | $ ls -l /usr/include/emmintrin.h |
| 00:43.48 | ``Erik | -r--r--r-- 1 root wheel 34528 Sep 16 2005 /usr/include/emmintrin.h |
| 00:44.23 | Maloeran | Can you grep it for castps_si128? |
| 00:44.44 | Maloeran | Must be a non-standard SSE intrinsic, it's just a cast anyway |
| 00:45.22 | ``Erik | http://www.freebsd.org/cgi/cvsweb.cgi/src/contrib/gcc/config/i386/emmintrin.h?rev=1.1.1.3&content-type=text/x-cvsweb-markup |
| 00:47.01 | Maloeran | Okay. I'll fix that soon |
| 00:47.53 | *** join/#brlcad brlcad (n=sean@bz.bzflag.bz) | |
| 00:48.04 | *** mode/#brlcad [+o brlcad] by ChanServ | |
| 00:48.50 | brlcad | "There was a problem with the switch that has been resolved by the IP Engineering department, your server was not rebooted but is currently online and responsive to my ping requests." |
| 00:49.08 | brlcad | for anyone on or connected to .bz, that is |
| 00:52.04 | ``Erik | O.o |
| 00:56.17 | ``Erik | ld: .libs/job.o has external relocation entries in non-writable section (__TEXT,__text) for symbols: |
| 00:56.17 | ``Erik | jobThreadWork |
| 00:56.19 | ``Erik | neat |
| 00:57.57 | Maloeran | What is that? |
| 00:58.47 | Maloeran | Oh, some platform doesn't like my assembly hack to align the stack on 16 bytes on ia32 |
| 00:59.18 | Maloeran | Have a look at RF/job.c, I'm not sure what that error message means |
| 01:03.46 | Maloeran | Any thoughts or further insight? |
| 01:05.05 | *** join/#brlcad IriX64 (n=mario_du@bas3-sudbury98-1168049077.dsl.bell.ca) | |
| 01:08.36 | *** join/#brlcad brlcad (n=sean@bz.bzflag.bz) | |
| 01:08.45 | *** mode/#brlcad [+o brlcad] by ChanServ | |
| 01:09.13 | ``Erik | couldn't you do something like an alignment attribute? O.o |
| 01:09.51 | Maloeran | No, it's the stack itself that is not aligned |
| 01:10.06 | Maloeran | The compiler can't preserve stack alignment if the thread stack wasn't aligned to begin with |
| 01:10.37 | ``Erik | hrm *shrug* |
| 01:10.37 | IriX64 | what the hey does the compiler care about a users thread? |
| 01:10.37 | Maloeran | So when launching a thread, first thing I do is fixing the alignment, and the compiler preserves the alignment from that point |
| 01:10.46 | IriX64 | it's all code. |
| 01:10.57 | IriX64 | unless you have thread optimizations. |
| 01:10.59 | ``Erik | the compiler doesn't, but if you are misaligned when you try to execute an sse op, shit blows up |
| 01:11.11 | IriX64 | runtime? |
| 01:11.14 | Maloeran | Yes |
| 01:11.15 | ``Erik | yeah |
| 01:11.17 | IriX64 | obviously duh. |
| 01:11.43 | IriX64 | auto assigned stack or user determined? |
| 01:11.56 | Maloeran | What's the fix, Erik? Is the symbol incorrect on fbsd, does it need an underscore or something? |
| 01:12.04 | Maloeran | Automatically assigned stack |
| 01:12.23 | IriX64 | system *should keep track of it then. |
| 01:12.28 | IriX64 | alas |
| 01:12.37 | IriX64 | all systems are not created equeal. |
| 01:12.44 | Maloeran | An automatic stack will be allocated in the proper memory bank for the processor, for higher performance. If I were to provide it, I would need to do some non-portable libNUMA fu and Erik would hate me |
| 01:12.53 | IriX64 | ergo code generated for one system etc... |
| 01:13.13 | IriX64 | heh beat eric with a noodle. |
| 01:13.43 | IriX64 | btw am i breaking anything by having Mr. Muss's face on my space? |
| 01:13.59 | IriX64 | I mean is it allowed i should have asked first. |
| 01:14.39 | ``Erik | what symbol? the job breakage was osX on an x86 core duo |
| 01:14.58 | Maloeran | The jobThreadWork symbol that I'm calling from assembly |
| 01:15.34 | ``Erik | that's a mac, not fbsd |
| 01:16.04 | ``Erik | <-- doesn't know the fix |
| 01:16.05 | Maloeran | <-- neither |
| 01:16.06 | ``Erik | <-- thinks only linux has the alignment broken stuff, though |
| 01:16.32 | Maloeran | Erik, test it. Disable the RF_I386_STACK_ALIGNMENT_HACK switch |
| 01:16.35 | ``Erik | I'm fairly sure you can just call jobThreadWork() on fbsd and it self-aligns |
| 01:16.47 | ``Erik | but your sse stuff is broken on fbsd right now *shrug* |
| 01:17.58 | ``Erik | heh, I have my reasons. |
| 01:18.25 | Maloeran | If you want to test Erik, just replace _mm_castps_si128(x) with (__m128i)x |
| 01:20.02 | ``Erik | I deleted RF_I386_STACK_ALIGNMENT_HACK from config.h and it works fine on the mac |
| 01:20.21 | ``Erik | I think RF_I386_STACK_ALIGNMENT_HACK should only be if __linux__ && __x86__ |
| 01:20.27 | Maloeran | Agreed |
| 01:22.57 | ``Erik | hrm, ok, changing the castps let it compile... but it's not decently headed, so rfdemo won't do well |
| 01:23.11 | ``Erik | what can I run to see if it executes without crashing with no head? :D |
| 01:23.48 | Maloeran | rfregtest? :) |
| 01:23.55 | ``Erik | failed to create context |
| 01:24.03 | ``Erik | does it assume files in pwd? |
| 01:24.10 | Maloeran | Yes |
| 01:24.20 | ``Erik | bah |
| 01:24.45 | Maloeran | Context creation failed is about rt.conf and the .so |
| 01:30.18 | ``Erik | log2f issues again, heh |
| 01:31.38 | Maloeran | Your last "fix" would never detect log2() in any circumstance, because it's C99 and requires -lm |
| 01:31.49 | Maloeran | And the configure thing tested C89 without -lm |
| 01:33.27 | ``Erik | heh, quit using c99? :D *duck* |
| 01:35.40 | ``Erik | ok, hacked the log2 defs back in and it's cooking away now |
| 01:35.42 | ``Erik | no crashes so far |
| 01:36.00 | Maloeran | Do you think you could test a big endian platform? |
| 01:37.05 | ``Erik | pretty easily |
| 01:37.11 | ``Erik | but SDL gives you that for free O.o |
| 01:37.40 | Maloeran | Ah well, there's the geometry group file format to test too |
| 01:38.03 | Maloeran | No big deal, it should work unless I made a stupid typo somewhere |
| 01:38.42 | brlcad | heh, sdl don't give jack for free |
| 01:38.44 | ``Erik | ok, the hack is off and regtest is running on opteron/fbsd now |
| 01:38.47 | brlcad | there's a price *somewhere* |
| 01:38.54 | ``Erik | heh, yeah, you have to include it to get the macros |
| 01:39.04 | ``Erik | which penalize the unpreferred format |
| 01:39.30 | Maloeran | The hack is only for ia32 platforms by the way |
| 01:39.41 | brlcad | providing seamless maintainability integration can be a pita with them, but doable |
| 01:39.45 | Maloeran | The switch is ignored on amd64, as it's unecessary and it's ia32 assembly anyway |
| 01:39.57 | ``Erik | it was on in my compile :/ |
| 01:40.08 | Maloeran | Sure it's on, but ignored in the code |
| 01:40.16 | ``Erik | *shrug* |
| 01:40.31 | ``Erik | how long does it take for results agan? |
| 01:40.54 | Maloeran | It takes several minutes, just kill it ;) |
| 01:41.03 | ``Erik | woops, there went a core dump |
| 01:41.12 | Maloeran | Woah. |
| 01:41.57 | ``Erik | #0 0x000000080144b89c in pthread_testcancel () from /lib/libpthread.so.2 |
| 01:42.35 | Maloeran | Can you backtrace a bit more? |
| 01:43.05 | ``Erik | I bailed out of the machine and closed the screen, sorry... there were only three symbols, the first and last were pthread |
| 01:43.38 | Maloeran | Actually... This was refregtest, right? I think I forgot to fix something there |
| 01:43.53 | ``Erik | yeah, rfregtest |
| 01:44.03 | brlcad | woot, autogen.sh now works from subconf dirs |
| 01:44.13 | ``Erik | given the number of machines I was jumping through and the slow link, rfdemo wouldn't have worked well :) |
| 01:47.08 | CIA-7 | BRL-CAD: 03brlcad * 10brlcad/autogen.sh: |
| 01:47.08 | CIA-7 | BRL-CAD: add a function to locate the configure template so that we find the right files |
| 01:47.08 | CIA-7 | BRL-CAD: when running autogen.sh from an ac_config_subdirs directory, perhaps being |
| 01:47.08 | CIA-7 | BRL-CAD: configured/compiled in isolation (without making a copy of this script). also |
| 01:47.09 | CIA-7 | BRL-CAD: add a sanity check that we did indeed end up with a configure script when all |
| 01:47.09 | CIA-7 | BRL-CAD: was said and done. |
| 01:51.44 | brlcad | IriX64: ehm.. that's pretty impossible at this point |
| 01:51.59 | brlcad | the channel is publicly logged and indexed by google daily |
| 01:52.13 | IriX64 | ahh well. |
| 01:52.29 | IriX64 | <*> |
| 01:54.27 | brlcad | search for "IriX64 BRL-CAD" and you can see a few on the first page alone |
| 01:56.34 | Maloeran | Cool, logs from the channel 14 months ago |
| 01:57.03 | IriX64 | really? explains something |
| 01:58.11 | ``Erik | heh http://paste.lisp.org/display/30558 |
| 01:59.06 | IriX64 | jack?.... Jack Shit, I know him well. ;) |
| 01:59.30 | IriX64 | what are you testing? |
| 01:59.54 | IriX64 | ermf for get I said that or you'll be asking me same question. |
| 02:01.11 | Maloeran | IriX64, grab a C guide online, get dev-c++ and get started |
| 02:01.18 | ``Erik | heh, squeak |
| 02:01.48 | IriX64 | the mouse that roared? |
| 02:02.11 | ``Erik | no, the smalltalk environment |
| 02:02.22 | ``Erik | irix: check out http://mitpress.mit.edu/sicp/ |
| 02:02.23 | ``Erik | it's nifty |
| 02:02.29 | IriX64 | why smalltalk why not crosstalk? |
| 02:03.06 | IriX64 | ermf xbox? |
| 02:04.46 | Maloeran | Seriously, IriX64, learn to code ; I think it's the kind of hobby you would enjoy |
| 02:05.31 | ``Erik | the exercised in structured thinking is a good growth opportunity |
| 02:05.46 | ``Erik | think of it like sudoku on steroids :) |
| 02:06.06 | IriX64 | problem is the environment insists on stunting my growth :) |
| 02:06.51 | ``Erik | heh |
| 02:07.01 | Maloeran | The environment? |
| 02:07.32 | IriX64 | structured programming environment, I seriously don't think like that. |
| 02:07.48 | Maloeran | No one does, it's an acquired skill |
| 02:07.59 | Maloeran | Just give it a try |
| 02:08.04 | IriX64 | old dog, new tricks thing. |
| 02:08.20 | ``Erik | what? old john, new tricks? |
| 02:08.24 | ``Erik | sorry, I'll behave O:-) |
| 02:08.29 | IriX64 | heheh |
| 02:08.54 | IriX64 | when you get that old the trick is in keeping her :) |
| 02:09.18 | Maloeran | Are you retired or so, IriX64? |
| 02:09.46 | IriX64 | disability. (does it show, mentally challenged :)) |
| 02:11.01 | Maloeran | Really? You seem joyful and not too knowledgeable about computers but I would not have thought that. Are you serious? |
| 02:11.26 | IriX64 | just the disability part. but i'm getting well. |
| 02:11.57 | Maloeran | Well, clearly you can type, therefore you can code |
| 02:12.09 | ``Erik | heh |
| 02:12.13 | IriX64 | used to be a DEC fse before i took work hhere at the plant (Algoma Steel Inc) |
| 02:12.27 | ``Erik | plenty of hunt&peck coders out there |
| 02:12.28 | IriX64 | trvth |
| 02:13.06 | Maloeran | It's a disability acquired at a steel mill?.. |
| 02:13.14 | IriX64 | yes |
| 02:13.29 | ``Erik | if you spend more time typing than thinking, I'd assert that either you're using an inappropriate language for the task, or approaching the task in a silly manner |
| 02:13.42 | Maloeran | Ouch. Would you... share details, if you want to? |
| 02:14.07 | IriX64 | re: coding? I tend to disagree, you could just be a hunt and pecker :) |
| 02:14.29 | IriX64 | Maloeran not publicly sorry. |
| 02:14.50 | ``Erik | heh, mal is a pecker :D |
| 02:14.50 | Maloeran | Okay, right. |
| 02:15.09 | IriX64 | ``Erik is a hunter :) |
| 02:15.19 | ``Erik | spyhunter, yo |
| 02:15.30 | IriX64 | wait ive got dialtone. |
| 02:15.50 | IriX64 | bond james bone. |
| 02:15.56 | ``Erik | bondo? |
| 02:15.58 | IriX64 | :) |
| 02:16.15 | IriX64 | stuff used to patch cars get with it. |
| 02:16.16 | ``Erik | car repair paste of international superspies? O.o |
| 02:16.52 | IriX64 | paste? as in pasting teeth? :) |
| 02:17.36 | IriX64 | tell me ``Erik if it's common sense, why isn't there more of it around? |
| 02:18.19 | ``Erik | heh, don't make me bust out my big books of steven wright quotes |
| 02:18.42 | ``Erik | or mitch hedberg |
| 02:18.59 | IriX64 | steven wright? I turn to calvin and hobbes. |
| 02:20.29 | Maloeran | Hey Justin, received mail from Mark forwarding yours |
| 02:20.51 | Maloeran | The voxel stuff was dropped on the way, apparently the people who thought they needed it were actually using triangles |
| 02:21.49 | Twingy | we'll see about that |
| 02:29.54 | IriX64 | Maloeran, I wasn't really kidding about that don't start the thread bit, if it gives you hassle incorporate the code into whatever thinks it needs to start a thread. |
| 02:30.44 | IriX64 | forr what its worth :) |
| 02:34.24 | Maloeran | Threads shouldn't be avoided, they are used to distribute the work over multiple processors |
| 02:35.24 | IriX64 | not saying avoid them bit if one particular piece of code is giving you grief why fight it. |
| 02:38.11 | Maloeran | It's fine, just OSX doing things a bit differently, no big deal |
| 02:41.43 | ``Erik | heh |
| 02:48.04 | louipc | what's the ideology behind that feature? |
| 02:49.42 | Maloeran | That ia32's largest data type is "double", which is 8 bytes, therefore stacks only require 8 bytes alignment |
| 02:50.51 | Maloeran | Then SSE arrived, which required 16 bytes. The glibc and kernel crowd kept putting the blame on the other camp, I think it got fixed at some point |
| 02:52.25 | Maloeran | Compilers like Intel would align the stack of every function dynamically whenever SSE was used to work around that issue |
| 02:59.15 | louipc | wowzers this processor mumbo jumbo is beyond me |
| 02:59.57 | Maloeran | Oops I'm sorry, I assumed you were a programmer |
| 03:07.59 | louipc | no prob, I'm just reading generally about it now. It makes programming at the processor level look like a real pain |
| 03:09.15 | louipc | IriX64: you have a *nix? which one? |
| 03:10.29 | IriX64 | DEC :P |
| 03:12.04 | louipc | hmm |
| 03:13.17 | IriX64 | pdp11/70 louipc :) |
| 03:14.21 | louipc | haha no way |
| 03:15.37 | IriX64 | sysV runs well on it tho. |
| 03:17.32 | IriX64 | btw if anybody cares, ./m4 *seems tied to 1.9.6. |
| 03:35.52 | IriX64 | louipc, recepricol question which unixen do you have? |
| 03:38.50 | IriX64 | can i do it that way? |
| 13:20.05 | *** join/#brlcad docelic (n=docelic@212.91.114.160) | |
| 14:14.46 | ``Erik | *cookcookcook* |
| 15:10.46 | *** join/#brlcad lg_ (n=lg_@88.234.13.222) | |
| 15:10.51 | lg_ | hey... |
| 15:11.22 | lg_ | i am here with a mged modeling question... something i would do with an extrude along path in another app... |
| 15:12.18 | lg_ | i am building a segment of a domes with ribs, but everything is rounded |
| 15:15.54 | lg_ | ok, thank you... |
| 15:16.03 | Maloeran | If you were to idle for a while, you'll get an answer eventually, but I wouldn't count on anything in a short time |
| 15:16.27 | lg_ | i think i have first to cut away the corner and than past in a part of a torus, which is awful |
| 15:17.07 | lg_ | maloeran, do you know if it is possible to specify geometry by angle in mged? it is a rather basic question, but i could not find an answer yet |
| 16:31.11 | *** join/#brlcad dtidrow (n=dtidrow@c-69-255-182-248.hsd1.va.comcast.net) | |
| 16:45.38 | *** join/#brlcad docelic (n=docelic@212.91.114.160) | |
| 17:56.21 | *** join/#brlcad brlcad (n=sean@bz.bzflag.bz) | |
| 17:56.31 | *** mode/#brlcad [+o brlcad] by ChanServ | |
| 18:28.53 | CIA-7 | BRL-CAD: 03brlcad * 10brlcad/configure.ac: be a little more liberal on the -all_load libtool script fix to avoid escaping the dollarsign |
| 19:24.02 | *** join/#brlcad clock_ (i=clock@84-72-63-213.dclient.hispeed.ch) | |
| 20:08.29 | CIA-7 | BRL-CAD: 03brlcad * 10brlcad/configure.ac: loop over potential locations instead of just testing for one. this helps subdir configures clean up too if added to their configure. |
| 20:30.14 | Maloeran | Cool, 3 minutes to compile BRL-CAD, optimized and almost-everything |
| 20:36.01 | brlcad | nifty |
| 20:36.29 | louipc | I'm extremely jealous |
| 20:36.50 | brlcad | Maloeran: how long for a ./configure --enable-all compile (i.e. non-optimized, almost-everything) |
| 20:37.17 | brlcad | both with "make -jwhatever" and "make -jwhatever fast" |
| 20:37.43 | brlcad | might be in contention for fastest compile to date |
| 20:52.00 | *** join/#brlcad IriX64 (n=mario_du@bas3-sudbury98-1168050480.dsl.bell.ca) | |
| 20:53.32 | louipc | IriX64: I'm running linux, archlinux specifically |
| 20:53.36 | louipc | ;) |
| 20:54.48 | IriX64 | louipc: great what have you photon mapped (it's such fun) |
| 20:54.48 | Maloeran | brlcad, will test |
| 20:54.48 | Maloeran | I'm not too sure what the optimal -jX is though, yet |
| 20:55.41 | Maloeran | --enable-all is 1 minute 54 seconds |
| 20:56.18 | Maloeran | Trying this "make fast" now |
| 20:56.19 | brlcad | Maloeran: with or without 'fast' |
| 20:56.23 | Maloeran | Without |
| 20:56.23 | brlcad | ahh, k |
| 20:56.35 | brlcad | cool, that would be a record regardless ;) |
| 20:56.40 | louipc | IriX64: nothing worth noting |
| 20:56.53 | Maloeran | Neat, I like my new toy :) |
| 20:56.54 | brlcad | is this cvs head too, or somewhere thereabouts? |
| 20:57.08 | Maloeran | The latest release, not CVS |
| 20:57.21 | brlcad | ahh, okay |
| 20:57.36 | brlcad | still a record, though latest will be a bit slower |
| 20:57.41 | IriX64 | but have you played with the examples? They're truly works of art, I especially like castle. |
| 20:58.05 | *** join/#brlcad cad53 (n=51f8d09e@bz.bzflag.bz) | |
| 20:58.37 | Maloeran | Hum. brlcad, it didn't tell me the time it took for "make -j12 fast", I guess I should have clocked |
| 21:00.08 | brlcad | Maloeran: ah, no matter then |
| 21:01.39 | IriX64 | http://www.pastebin.ca/370750 (*shrug*) |
| 21:02.26 | Maloeran | time make -j12 fast : real 2m3s user 6m46s sys 2m57 |
| 21:02.38 | Maloeran | Oh well, it's adding the time of all cores |
| 21:04.18 | ``Erik | huh? |
| 21:05.12 | IriX64 | bwahahah I can now shut down windows from my cygwin environment. thats hysterical louipc you wondered what I'm using, I just told you ;) |
| 21:08.38 | IriX64 | http://www.pastebin.ca/370763 |
| 21:08.51 | IriX64 | louipc: check it. |
| 21:14.33 | IriX64 | http://www.pastebin.ca/370770 (I'll stop now :)) |
| 22:44.32 | *** join/#brlcad docelic (n=docelic@212.91.113.94) | |
| 23:05.15 | louipc | IriX64: I wouldn't call that *nix, it's more like GNU/windows hah |
| 23:19.00 | ``Erik | indeed, even less unixy than linux O.O |
| 23:21.35 | IriX64 | puhleaze, I can do anything unix can do try me. |
| 23:21.40 | IriX64 | :) |
| 23:21.47 | IriX64 | i log in too. |
| 23:22.29 | IriX64 | btw my little x works very well. |
| 23:22.52 | IriX64 | has to or i would not be able to run BRL-CAD. |
| 23:23.47 | IriX64 | wonders too. |
| 23:24.33 | ``Erik | ok |
| 23:24.38 | ``Erik | run on ppc. |
| 23:24.43 | ``Erik | or rs6k. |
| 23:24.45 | ``Erik | or h9k |
| 23:24.55 | ``Erik | or vax or pdp11 or m68k or ... |
| 23:25.12 | IriX64 | now you're speaking my langugae |
| 23:25.27 | IriX64 | pdp8s best machine ever. |
| 23:26.03 | ``Erik | or support hundreds of concurrent users |
| 23:26.11 | IriX64 | doing a no debugging build want to see how fast havoc photon maps. |
| 23:26.30 | ``Erik | or provide services to the internet with no antivirus software chewing cycles and not get infected or cracked |
| 23:26.43 | IriX64 | huh? |
| 23:27.19 | IriX64 | ``Erik i respect you're opinion no matter how wrong you may be ;) |
| 23:27.34 | ``Erik | <-- doens't think he's all that wrong O:-) |
| 23:27.50 | IriX64 | heh neither do i :) |
| 23:28.43 | ``Erik | and photon mapping is a two phase operation... it raytraces as fast as it does with no lighting... and the photo map pass totally depends on how many photons you specify and a few other parameters |
| 23:29.10 | ``Erik | path tracing seems a bit more interesting, though... have you managed to get adrt to build? it has something in it called "rise" which is pretty neat |
| 23:29.16 | IriX64 | sure but you can get a guesstimate if you leave all the settings the same. |
| 23:29.34 | IriX64 | missing sdl gotta tend to that. |
| 23:29.40 | ``Erik | libsdl.org |
| 23:29.44 | IriX64 | ty |
| 23:33.19 | IriX64 | yum thanks ``Erik i'll let you know if I ever get it going btw is it fraught? |
| 23:34.26 | IriX64 | you're supposed to say "there's just no arguing with a word like fraught" :) |
| 23:35.25 | IriX64 | python is also needed. |
| 23:36.38 | ``Erik | ... yes, sdl and python are (sorta) needed, though I think rise itself doesn't use either |
| 23:36.58 | IriX64 | getting python 2.5 now. |
| 23:37.36 | ``Erik | n/m, rise observer uses sdl |
| 23:37.44 | IriX64 | ah |