| 00:49.31 | starseeker | makes a note that there is now a bold version of the inconsolata font... https://code.google.com/p/googlefontdirectory/source/browse/ofl/inconsolata/ |
| 01:38.03 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) | |
| 02:00.18 | vasc | ah great |
| 02:18.28 | vasc | brlcad, i rewrote rt_booleval. tell me what you think |
| 02:18.34 | vasc | later |
| 02:18.47 | vasc | put a patch on the tracker |
| 02:19.56 | vasc | man the new one is so much easier to read |
| 02:20.54 | vasc | no gotos and uses a lot more less mem |
| 02:20.59 | vasc | and it llinearizes and compact the tree |
| 03:14.36 | starseeker | vasc: did you do any performance/correctness testing? |
| 03:15.24 | starseeker | reflects we should probably put together a test-all-branches suite of geometry for that code... |
| 03:16.40 | starseeker | wonder if some sort of segment-input generating fuzzer could be written just for the boolean code (http://lcamtuf.coredump.cx/afl/ style...) |
| 03:18.19 | vasc | well i ran a couple of runs with operators.g and i got slightly better speeds actually |
| 03:18.28 | vasc | rendering speeds |
| 03:18.33 | vasc | not counting pre-preprocessing |
| 03:19.03 | vasc | it doesn't have the xor because the xor code we use is whacked |
| 03:19.11 | vasc | i don't know how to implement it without breaking something |
| 03:19.35 | vasc | couldn't find any test scenes with it either. allegedly to be removed as operator in 8.90 |
| 03:19.38 | vasc | 8.0 |
| 03:20.10 | vasc | i examined temporary outputs, they looked fine |
| 03:20.19 | vasc | the final output looks fine too |
| 03:20.33 | vasc | correctness testing.... ask a computational maths guy |
| 03:20.43 | vasc | had enough of form proofs of correction in college |
| 03:20.59 | vasc | let alone against legacy code with unused legacy options and gotos |
| 03:21.20 | vasc | more test scenes would be appreciated. |
| 03:21.48 | vasc | or you try it and tell me |
| 03:21.54 | vasc | the patch is in the sf.net |
| 03:22.01 | vasc | https://sourceforge.net/p/brlcad/patches/417/ |
| 03:22.58 | vasc | XOR in the old code was implemented with access with these crap operations |
| 03:23.04 | vasc | additional ones |
| 03:23.40 | vasc | #define OP_GUARD MKOP(9) /**< @brief Unary: not L, or else! */ |
| 03:23.40 | vasc | #define OP_XNOP MKOP(10) /**< @brief Unary: L, mark region */ |
| 03:23.59 | vasc | it's crap. you don't need that with the new code |
| 03:24.06 | vasc | its just A^B and that's it |
| 03:24.31 | vasc | it's all insane anyway |
| 03:24.53 | vasc | coz XOR can be defined as (A/B) u (B/A) |
| 03:24.59 | vasc | so why need new operators for XOR? |
| 03:25.02 | vasc | bonkers |
| 03:26.24 | vasc | was it to save on tree size or runtime stack space? |
| 03:26.37 | vasc | well its useless with my code |
| 03:32.27 | Stragus | Perhaps processing could be simplified with a single XOR operation? |
| 03:32.52 | Stragus | Or just decompose in 3 operations as you suggested, eh |
| 03:34.33 | vasc | it's trivial to add XOR |
| 03:34.42 | vasc | the advantage is less stack space and opcode space |
| 03:35.13 | vasc | my problem is how not to break compact with old XOR. all edge cases. whatever those are supposed to be. |
| 03:35.18 | vasc | compat |
| 03:35.36 | vasc | like i said its like 7 lines of code |
| 03:36.25 | vasc | s/opcode/machine code/ |
| 03:36.30 | vasc | as in virtual RPN machine code |
| 03:36.33 | vasc | snatch |
| 03:36.49 | vasc | i made an RPN calculator for computing the CSG |
| 03:38.15 | vasc | well |
| 03:38.20 | vasc | a small piece of the CSG code |
| 03:38.25 | vasc | rt_booleval |
| 03:38.35 | vasc | its probably applicable to other places as well |
| 03:38.46 | vasc | i'll try to extend its use later |
| 03:39.24 | vasc | but that was the perf critical bit |
| 03:46.05 | vasc | i guess i can make my own test scene |
| 03:51.20 | vasc | it just blows up |
| 03:51.27 | vasc | even the stock version |
| 03:51.51 | vasc | maybe its a different UTF-8 character.... |
| 03:52.28 | vasc | just a sec |
| 03:55.26 | vasc | segmentation fault again |
| 03:56.47 | vasc | you can't even do 'c xor obj1 ^ obj2' in mged without it crashing stupidly |
| 03:56.52 | vasc | and that's with head |
| 03:56.59 | vasc | so uh. |
| 03:57.09 | vasc | i mean trunk |
| 03:57.17 | vasc | trunk doesn't work with xor input by user |
| 03:57.23 | vasc | so why should my code work with that? |
| 03:57.39 | vasc | your thoughts brlcad |
| 03:57.44 | vasc | i'm gonna sleep |
| 06:15.26 | *** join/#brlcad dracarys983 (dracarys98@nat/iiit/x-uiigltxouomsqhzs) | |
| 08:14.58 | *** join/#brlcad ries (~ries@D979C47E.cm-3-2d.dynamic.ziggo.nl) | |
| 08:23.16 | *** join/#brlcad dracarys983 (dracarys98@nat/iiit/x-fsoiiqocuuwwaquv) | |
| 08:36.21 | *** join/#brlcad Boquete (~piotr@83.5.96.244) | |
| 09:06.27 | *** join/#brlcad Boquete (~piotr@aano244.neoplus.adsl.tpnet.pl) | |
| 10:51.30 | *** join/#brlcad Boquete (~piotr@aano244.neoplus.adsl.tpnet.pl) | |
| 10:54.04 | *** join/#brlcad LordOfBikes (~armin@dslb-088-066-150-176.088.066.pools.vodafone-ip.de) | |
| 11:21.31 | *** join/#brlcad vasc (~vasc@bl8-194-96.dsl.telepac.pt) | |
| 11:46.35 | ``Erik | huh, cmake --build . |
| 13:31.06 | starseeker | ``Erik: something broken? |
| 13:38.44 | ``Erik | no, cmake has a --build argument that calls whatever build system is appropriate |
| 13:39.09 | ``Erik | (make, ninja, msvc, xcodebuild...) |
| 13:47.12 | starseeker | ah |
| 13:47.24 | starseeker | yes, quite handy (theoretically) in the distcheck scripts :-) |
| 13:47.39 | starseeker | seldom tries distcheck with anything but make |
| 13:47.48 | starseeker | suppose I should at least try it with ninja |
| 13:49.52 | *** join/#brlcad ih8sum3r (~ih8sum3r@122.173.119.118) | |
| 14:50.06 | starseeker | ``Erik: is Notify down? |
| 15:00.59 | *** join/#brlcad ries (~ries@D979C47E.cm-3-2d.dynamic.ziggo.nl) | |
| 15:09.12 | *** join/#brlcad sofat (~androirc@106.192.23.0) | |
| 15:20.29 | Notify | 03BRL-CAD:vasco_costa * 66091 brlcad/trunk/src/librt/primitives/rt.cl: compute ocl transparency using attenuation. |
| 15:20.39 | Notify | 03BRL-CAD:starseeker * 66092 (brlcad/branches/RELEASE/CMakeLists.txt brlcad/branches/RELEASE/misc/CMake/BRLCAD_CMakeFiles.cmake and 7 others): Minimal CMake updates for CMake >=3.0 |
| 15:20.42 | Notify | 03BRL-CAD:starseeker * 66093 (brlcad/branches/RELEASE/src/conv/fast4-g.c brlcad/branches/RELEASE/src/libged/importFg4Section.c): Fix build errors. |
| 15:21.07 | Notify | 03BRL-CAD:starseeker * 66094 (brlcad/branches/RELEASE/CMakeLists.txt brlcad/branches/RELEASE/include/brlcad_ident.h brlcad/branches/RELEASE/misc/CMake/multiconfig_path_read.cmake.in): Apply a few more changes to the build logic, working towards Windows building - the build-time path reporting was causing more problems than it was worth, removed from ident. |
| 15:21.10 | Notify | 03BRL-CAD:starseeker * 66095 (brlcad/branches/RELEASE/CMakeLists.txt brlcad/branches/RELEASE/include/config_win_cmake.h.in): Grab some of the Windows checks needed with newer MSVC |
| 15:21.53 | Notify | 03BRL-CAD:starseeker * 66096 brlcad/branches/RELEASE/src/other/stepcode/src/cleditor/STEPfile.inline.cc: Pull in MSVC build fix from r62446 |
| 15:21.57 | Notify | 03BRL-CAD:starseeker * 66097 (brlcad/branches/RELEASE/CMakeLists.txt brlcad/branches/RELEASE/include/config_win_cmake.h.in): The isinf and isnan tests don't seem to be working correctly with this merge... must need more than just these changes. Revert for now. |
| 15:22.11 | Notify | 03BRL-CAD:starseeker * 66098 brlcad/branches/RELEASE/include/config_win_cmake.h.in: Since the MSVC tests don't seem to be functioning reliably in this version and I'm not sure what other pieces are needed from trunk, take the direct approach just for the patch release. |
| 15:22.17 | ``Erik | heh, woops, the pump thread died the other day (bad encoding in an email, I need to fix the parser) and I checked one of the threads, but not the one that actually died... O:-) once it catches up, it should run normal (until it breaks again) |
| 15:22.55 | Notify | 03BRL-CAD:brlcad * 66099 brlcad/trunk/HACKING: revert back to the more strict wording regarding commit access |
| 15:23.10 | Notify | 03BRL-CAD:starseeker * 66100 brlcad/branches/RELEASE/include/bu.h: math.h may have INFINITY definition. |
| 15:23.12 | Notify | 03BRL-CAD:starseeker * 66101 (brlcad/branches/RELEASE/src/libicv/bw.c brlcad/branches/RELEASE/src/libicv/encoding.c and 5 others): Merge r57642 - HIDDEN causing problems on Windows build. |
| 15:23.14 | Notify | 03BRL-CAD:starseeker * 66102 (brlcad/branches/RELEASE/src/libicv/bw.c brlcad/branches/RELEASE/src/libicv/dpix.c and 9 others): More HIDDEN updates. |
| 15:23.16 | Notify | 03BRL-CAD:starseeker * 66103 brlcad/branches/RELEASE/include/common.h: We're CMake only now - not going to patch up config_win.h |
| 15:23.33 | Notify | 03BRL-CAD:brlcad * 66104 (brlcad/branches/RELEASE/include/db5.h brlcad/branches/RELEASE/include/magic.h and 11 others): merge initial datum support from c65612 and c65614 with a bit of conflict resolution due to header movement |
| 15:23.37 | Notify | 03BRL-CAD:brlcad * 66105 (brlcad/branches/RELEASE/NEWS brlcad/branches/RELEASE/include/conf/PATCH): stub the 7.24.4 patch release section, bump the patch revision |
| 15:23.38 | Notify | 03BRL-CAD:brlcad * 66106 (brlcad/branches/RELEASE/src/libbu/heap.c brlcad/branches/RELEASE/src/libbu/tests/bu_str_isprint.c and 8 others): compilation quellage not seen on trunk due to the different header inclusions. |
| 15:23.40 | Notify | 03BRL-CAD:brlcad * 66107 brlcad/branches/RELEASE/src/fbed/fbed.c: more quellage |
| 15:23.43 | Notify | 03BRL-CAD:brlcad * 66108 (brlcad/branches/RELEASE/NEWS brlcad/branches/RELEASE/src/fb/png-fb.c and 14 others): merge the remainder of datum work based on log activity including the following revisions: 65613,65627,65648,65649,65729,65743,65754,65763,65764,65775,65842. |
| 15:25.12 | Notify | 03BRL-CAD:vasco_costa * 66109 brlcad/trunk/src/librt/bool.c: eliminated seven gotos. thx branch elimination optimizations. |
| 15:25.34 | Notify | 03BRL-CAD:vasco_costa * 66110 brlcad/trunk/src/librt/bool.c: don't export booleval symbol. it is only called by rt_boolfinal in same file. |
| 15:26.40 | Notify | 03BRL-CAD:starseeker * 66111 (brlcad/trunk/include/rt/wdb.h brlcad/trunk/src/conv/CMakeLists.txt and 14 others): Using the BRLCAD_ENABLE_TARGETS=1 build config setting, remove last stray bits of Tcl dependency in librt and libsysv. Can now build the level one targets without needing to build Tcl. |
| 15:26.42 | Notify | 03BRL-CAD Wiki:202.164.45.212 * 9483 /wiki/User:Hiteshsofat/GSoc15/log_developmen: |
| 15:26.44 | Notify | 03BRL-CAD Wiki:Vasco.costa * 9484 /wiki/User:Vasco.costa/GSoC15/logs: /* Week 13 : 17 Aug-23 Aug */ |
| 15:26.46 | Notify | 03BRL-CAD Wiki:Vasco.costa * 9485 /wiki/User:Vasco.costa/GSoC15/logs: /* Week 13 : 17 Aug-23 Aug */ |
| 15:26.48 | Notify | 03BRL-CAD Wiki:BoqueteBRL * 0 /wiki/User:BoqueteBRL: |
| 16:12.33 | *** join/#brlcad Boquete (~piotr@bwa86.neoplus.adsl.tpnet.pl) | |
| 16:19.01 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) | |
| 17:08.27 | *** join/#brlcad ih8sum3r (~ih8sum3r@122.173.119.118) | |
| 17:44.15 | *** join/#brlcad ries (~ries@D979C47E.cm-3-2d.dynamic.ziggo.nl) | |
| 18:05.06 | Notify | 03BRL-CAD:starseeker * 66112 brlcad/trunk/src/libbrep/shape_recognition_cylinder.cpp: Hmm - this causes problems for some cylinder faces. |
| 18:10.17 | Notify | 03BRL-CAD:starseeker * 66113 brlcad/trunk/src/libbrep/CMakeLists.txt: Add some tests for brep->csg conversion. |
| 19:01.18 | *** join/#brlcad merzo (~merzo@221-12-132-95.pool.ukrtel.net) | |
| 19:40.40 | *** join/#brlcad konrado (~konro@41.205.22.54) | |
| 20:51.26 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) | |
| 21:08.56 | Notify | 03BRL-CAD:vasco_costa * 66114 brlcad/trunk/src/librt/bool.c: eliminated some gotos with equivalent code. actually makes code more readable imo. |
| 21:15.54 | *** join/#brlcad ries (~ries@D979C47E.cm-3-2d.dynamic.ziggo.nl) | |
| 21:24.00 | *** join/#brlcad ries_nicked (~ries@D979C47E.cm-3-2d.dynamic.ziggo.nl) | |
| 22:26.53 | Notify | 03BRL-CAD:vasco_costa * 66115 brlcad/trunk/src/librt/bool.c: refactor it a bit to make it look nicer. in a pinch if the debug msg is disabled all the remaining goto's to done_wave can be replaced with returns. |