| 00:15.25 | CIA-28 | BRL-CAD: 03starseeker * r48063 10/brlcad/trunk/misc/CMake/ThirdParty.cmake: oops - zigging when we need to zag. no turning things on when we have SYSTEM settings. |
| 00:21.17 | brlcad | whaaa .... when did that file get committed??? |
| 00:21.36 | brlcad | ah, r48053 |
| 00:34.59 | CIA-28 | BRL-CAD: 03starseeker * r48064 10/brlcad/trunk/ (misc/CMake/ThirdParty_TCL.cmake src/other/CMakeLists.txt): Few more tweaks for that annoying 'system only even though there's nothing installed on it that I need' case... |
| 00:38.17 | CIA-28 | BRL-CAD: 03starseeker * r48065 10/brlcad/trunk/misc/CMake/ThirdParty_TCL.cmake: No need to double warn |
| 00:52.51 | CIA-28 | BRL-CAD: 03starseeker * r48066 10/brlcad/trunk/misc/CMake/ (ThirdParty.cmake ThirdParty_TCL.cmake): Don't show disabled 3rd party opts, and make sure to show them if they become enabled. |
| 00:56.07 | CIA-28 | BRL-CAD: 03starseeker * r48067 10/brlcad/trunk/misc/CMake/xsltproc.cmake.in: Note why it's important to make the directory first when running xsltproc |
| 01:02.51 | CIA-28 | BRL-CAD: 03starseeker * r48068 10/brlcad/trunk/misc/CMake/ThirdParty.cmake: Oops - just because we've already added the subdirectory, that doesn't mean we don't need to set the executable variable for the second executable... |
| 01:04.09 | CIA-28 | BRL-CAD: 03starseeker * r48069 10/brlcad/trunk/misc/CMake/ThirdParty.cmake: in fact, the exe var setting is only conditional on the build setting. |
| 01:07.18 | CIA-28 | BRL-CAD: 03starseeker * r48070 10/brlcad/trunk/doc/docbook/books/en/BRL-CAD_Tutorial_Series-VolumeI.xml: fix xmllint error |
| 01:07.43 | starseeker | what's the LOTR line... |
| 01:07.51 | starseeker | "but that is only a few leaves in a forest" |
| 01:08.27 | starseeker | isn't feeling OCD enough to tackle 'em all right now... |
| 01:09.16 | starseeker | Looks like the on/off behavior of the CMake logic is improved considerably - will have to ask dli to give it a go when comes back |
| 02:15.43 | CIA-28 | BRL-CAD: 03starseeker * r48071 10/brlcad/trunk/ (8 files in 2 dirs): Woo hoo! autogenerate the guts of the configure rosetta script right along with the docs. |
| 02:17.05 | starseeker | "I object to doing things that computers can do." - Olin Shivers |
| 04:33.01 | CIA-28 | BRL-CAD: 03starseeker * r48072 10/brlcad/trunk/src/tclscripts/archer/BotUtility.tcl: use bu_brlcad_root to find libbu.<whatever> - wasn't working from build directory |
| 04:54.22 | starseeker | holds his breath... do I actually have it fully working now? |
| 08:08.13 | *** join/#brlcad jordisayol (~jordisayo@unaffiliated/jordisayol) | |
| 13:00.36 | starseeker | hmm http://pastebin.mozilla.org/1413600 |
| 13:01.29 | starseeker | clang 2.9 |
| 13:01.36 | starseeker | tries 3.0 |
| 13:44.55 | starseeker | /src/libbu/simd.c:33:5: error: extension used [-Werror,-pedantic,-Wlanguage-extension-token] asm volatile("cpuid":"=b"(b),"=c"(c),"=d"(d):"a"(0x1)); ^ |
| 13:45.11 | starseeker | apparently asm isn't standard ANSI C? |
| 13:54.32 | starseeker | well... it builds a working archer with the strict flag off at least |
| 14:28.32 | brlcad | looks like a valid bug it detected with the [Z] indexing past the end of the array |
| 14:30.22 | brlcad | asm isn't ansi, it's gcc |
| 14:30.35 | brlcad | could try changing it to __asm__ which gcc also supports |
| 14:32.06 | brlcad | otherwise, that logic can have another strict define added to keep clang from entering |
| 14:32.44 | brlcad | yeah, from clang manual: The parser recognizes "asm" and "typeof" as keywords in gnu* modes; the variants "__asm__" and "__typeof__" are recognized in all modes |
| 14:32.53 | brlcad | so it might not consider it an extension as __ |
| 14:33.03 | ``Erik | or break the func out to a .s file (which is probably the strict ansi permissible approach, but not exactly portable) |
| 14:47.48 | *** join/#brlcad hackrat (~packrator@c-98-209-146-133.hsd1.mi.comcast.net) | |
| 14:48.06 | starseeker | ``Erik: why is that not portable? |
| 14:48.29 | starseeker | (sure you're right, just wondering what the reasons would be) |
| 14:52.16 | ``Erik | a unix system will expect an at&t format .s file, windows will expect an intel format .asm file |
| 14:52.35 | ``Erik | linux will depend on whether you're using nasm or gas or yasm or ... |
| 14:52.51 | starseeker | ah - so embedding it in C code lets the C compiler translate it? |
| 14:53.17 | ``Erik | nah, but it lets the precompiler choose the code path |
| 14:53.53 | ``Erik | hm, I thought I put the winderz one in simd.c, guess I only did *nix |
| 14:53.58 | ``Erik | (gcc specific, even) |
| 14:55.06 | starseeker | so we'd have to conditionalize in CMake, looks like: http://www.cmake.org/Wiki/CMake/Assembler |
| 14:55.33 | starseeker | doable |
| 14:55.42 | ``Erik | and maintain parallel versions, unfortunately... I believe bullet has cross platform detection, so if that becomes a core part, we can let them handle that detail |
| 14:57.56 | starseeker | ``Erik: iirc, wasn't the simd stuff in bullet nicely contained in its own little bit of code? |
| 14:59.12 | starseeker | ah, right |
| 14:59.16 | starseeker | http://bullet.svn.sourceforge.net/viewvc/bullet/trunk/Extras/simdmathlibrary/ |
| 15:00.24 | starseeker | any reason not to just use that? |
| 15:01.30 | starseeker | bemusedly notes we could just use that instead of looking for the "m" library, assuming it actually supported all the platforms we're interested in... |
| 15:01.42 | starseeker | "a SIMDized version of the |
| 15:01.44 | starseeker | 7 C99 standard math library (libm)" |
| 15:02.04 | starseeker | "a SIMDized version of the C99 standard math library (libm)" rather |
| 15:09.10 | starseeker | ah, phooey - looks like it's only for PowerPC and Cell |
| 15:14.16 | starseeker | oh, I see - they've got it in vectormath, I think... http://code.google.com/p/bullet/source/browse/trunk/src/vectormath/ |
| 15:22.20 | starseeker | humph |
| 15:22.49 | starseeker | why can't someone just do a straight-up simd-libm under BSD license somewhere... |
| 15:23.49 | starseeker | ptview guy has one that looks close, but it's GPL3 |
| 15:26.49 | starseeker | oh, here's the bullet overview: http://bulletphysics.com/ftp/pub/test/physics/demos/Vector_Math_Library-Overview.pdf |
| 15:39.28 | starseeker | ah... looks like the benefits of simd manifest more at the higher vector level operations... could be the libm implementations for those platforms are because they needed 'em there... |
| 15:41.46 | starseeker | shakes head - ``Erik, I leave it to you :-) |
| 16:29.08 | CIA-28 | BRL-CAD: 03tbrowder2 * r48073 10/brlcad/trunk/src/tclscripts/mged/botedit.tcl: need manual dependency to define BotEditor' this fixes bug number 3392650 |
| 16:31.44 | CIA-28 | BRL-CAD: 03tbrowder2 * r48075 10/brlcad/trunk/NEWS: shameless credit for two bug fixes |
| 16:31.44 | CIA-28 | BRL-CAD: 03tbrowder2 * r48074 10/brlcad/trunk/src/tclscripts/boteditor/botTools.tcl: correct typo in menu |
| 16:56.15 | CIA-28 | BRL-CAD: 03starseeker * r48076 10/brlcad/trunk/src/other/step/src/express/expprint.c: Not really sure what's going on with printScope here... |
| 17:31.03 | brlcad | simd calls for just a few ops is utterly ridiculous |
| 17:31.25 | brlcad | you get a benefit when you can push your data into the simd pipeline AND KEEP IT THERE |
| 17:31.48 | brlcad | so a low-level simdified-libm isn't really feasible unless you overhaul all math |
| 17:34.07 | brlcad | as for the original elephant in the room, fixing that clang error is trivial and amounts to probably a 1-line tweak to the cpp logic |
| 17:36.07 | brlcad | moving the asm conditionalization from cppland to cmake wouldn't actually fix anything (and it'd be worse imho) |
| 18:23.15 | starseeker | brlcad: so Sony probably did it just to have a libm api for those platforms. Probably why bullet stuck it in extra |
| 18:28.18 | CIA-28 | BRL-CAD: 03starseeker * r48077 10/brlcad/trunk/src/libbu/simd.c: asm -> __asm__ |
| 18:37.25 | brlcad | that looks like the right fix :) |
| 18:39.01 | brlcad | starseeker: I meant feasible for our use -- a libm where your data is already in simd entities wouldn't be a bad idea |
| 18:39.38 | brlcad | heck, even something like gpm using simd under the hood would probably be a (modest) benefit (to us) |
| 18:46.28 | CIA-28 | BRL-CAD: 03n_reed * r48078 10/brlcad/trunk/src/other/perplex/scanner.re: fixed patterns that sometimes skipped quote characters, causing bad parsing of quoted text |
| 19:26.11 | CIA-28 | BRL-CAD: 03tbrowder2 * r48079 10/brlcad/trunk/regress/mged.sh: add more detailed regression tests for the mged solids and regions commands |
| 19:35.54 | CIA-28 | BRL-CAD: 03n_reed * r48080 10/brlcad/trunk/src/other/perplex/scanner.re: need to update scope flag at end of condition scope |
| 20:15.53 | CIA-28 | BRL-CAD: 03bob1961 * r48081 10/brlcad/trunk/ (3 files in 3 dirs): Added the ability to move a data object (i.e. line, arrow, polygon) or point. Previously only data points could be moved. |
| 20:59.55 | CIA-28 | BRL-CAD: 03erikgreenwald * r48082 10/brlcad/trunk/src/libgcv/test_bottess.c: add some meat |
| 21:13.22 | *** join/#brlcad Technicus (~Technicus@DSLPool-net208-2.wctc.net) | |
| 21:32.30 | CIA-28 | BRL-CAD: 03brlcad * r48083 10/brlcad/trunk/src/libbu/str.c: mark a few more of the expressions that result in logging as UNLIKELY |
| 22:15.19 | CIA-28 | BRL-CAD: 03brlcad * r48084 10/brlcad/trunk/ (4 files in 2 dirs): |
| 22:15.20 | CIA-28 | BRL-CAD: add a new bu_str_escape() function for escaping string characters with |
| 22:15.20 | CIA-28 | BRL-CAD: backslashes. supports static and dynamic memory models as well as overlapping |
| 22:15.20 | CIA-28 | BRL-CAD: or same input/output buffers. should be threadsafe and re-entrant as long as |
| 22:15.20 | CIA-28 | BRL-CAD: the inputs are. |
| 22:55.26 | CIA-28 | BRL-CAD: 03brlcad * r48085 10/brlcad/trunk/ (include/bu.h src/libbu/escape.c): |
| 22:55.27 | CIA-28 | BRL-CAD: add a new bu_str_unescape() function for removing backslash characters from |
| 22:55.27 | CIA-28 | BRL-CAD: strings. simpler to implement since the output is always less than or equal in |
| 22:55.27 | CIA-28 | BRL-CAD: size to the input. otherwise,similar to bu_str_escape() in that it supports |
| 22:55.27 | CIA-28 | BRL-CAD: static and dynamic memory models as well as overlapping or same input/output |
| 22:55.27 | CIA-28 | BRL-CAD: buffers and should be threadsafe and re-entrant as long as the inputs are. make |
| 22:55.28 | CIA-28 | BRL-CAD: sure we null-terminate. |
| 23:06.31 | CIA-28 | BRL-CAD: 03n_reed * r48086 10/brlcad/trunk/src/other/perplex/ (CMakeLists.txt scanner_template.c): lose libm dependency |
| 23:16.51 | CIA-28 | BRL-CAD: 03brlcad * r48087 10/brlcad/trunk/src/libbu/test_quote.c: doesn't actually use stdarg or stdlib |
| 23:17.12 | CIA-28 | BRL-CAD: 03tbrowder2 * r48088 10/brlcad/trunk/regress/mged.sh: document reason for new special tests |
| 23:39.23 | CIA-28 | BRL-CAD: 03brlcad * r48089 10/brlcad/trunk/src/libbu/escape.c: don't call strlen() before we know whether input is NULL |
| 23:49.20 | CIA-28 | BRL-CAD: 03brlcad * r48090 10/brlcad/trunk/src/libbu/test_escape.c: add an initial stab at a unit test for the new escape/unescape routines |
| 23:49.36 | CIA-28 | BRL-CAD: 03brlcad * r48091 10/brlcad/trunk/src/libbu/ (CMakeLists.txt Makefile.am): hook test_escape into the build |