| 00:52.31 | starseeker | hmm - bz ran out of disk space |
| 00:52.42 | starseeker | scrubs unnecessary crud... |
| 02:43.52 | *** join/#brlcad andymanak (~andy@c-24-130-199-23.hsd1.ca.comcast.net) |
| 03:51.01 | *** join/#brlcad louipc (~louipc@archlinux/fellow/louipc) |
| 05:02.07 | *** join/#brlcad andymanak (~andy@c-24-130-199-23.hsd1.ca.comcast.net) |
| 05:09.40 | *** join/#brlcad andymanak (~andy@c-24-130-199-23.hsd1.ca.comcast.net) |
| 08:11.54 | *** join/#brlcad jordisayol (~jordisayo@unaffiliated/jordisayol) |
| 15:00.56 | brlcad | starseeker: anything useful from https://github.com/mpictor/StepClassLibrary/pull/143 we could use in brl-cad? |
| 15:01.11 | brlcad | (for non-scl logic) |
| 15:14.24 | brlcad | jordisayol: possibly, it depends on the cause of the illegal instruction |
| 15:15.16 | brlcad | I wouldn't expect it be needed for a generic -O3 but if the debian build adds cpu-specific tuning flags then mtune=generic should help |
| 15:16.37 | jordisayol | hello brlcad |
| 15:18.07 | jordisayol | I just do this to override notius problem when installing deb package on a not new processor: |
| 15:18.07 | jordisayol | https://sourceforge.net/tracker/?func=detail&atid=640803&aid=3464421&group_id=105292 |
| 15:23.24 | brlcad | jordisayol: I know, I read that |
| 15:23.38 | brlcad | that didn't really identify the cause of the illegal instruction |
| 15:24.01 | brlcad | does the deb build add march or mtune flags (prior to mtune=generic) |
| 15:25.42 | jordisayol | deb build process do not add any flag directly, until now |
| 15:26.56 | brlcad | could be mistaken but the O3 flag by itself shouldn't result in platform-specific optimizations, afaik |
| 15:27.19 | brlcad | there should be some other flag causing extra optimization |
| 15:28.17 | jordisayol | but did gcc detect ans make specific for the system cpu, isn't it? |
| 15:28.36 | brlcad | not with just O3 |
| 15:28.49 | jordisayol | aha |
| 15:28.51 | brlcad | implies there is some other flag on the actual compilation line |
| 15:29.25 | jordisayol | these flags are from cmake/brlcad of from gcc itself? |
| 15:29.48 | brlcad | not from gcc, from somewhere else in the build |
| 15:30.09 | brlcad | the only way to know for sure would be to run with VERBOSE=1 and see what flags are being used |
| 15:30.36 | jordisayol | ok, I so that, wait a moment please |
| 15:32.22 | brlcad | I'd suspect somewhere something is adding an march= or mtune= flag .. |
| 15:32.37 | brlcad | maybe not, if not then we should add -mtune=generic to the cmake default |
| 15:32.52 | jordisayol | mh, i'm not quite sure but i don't think so |
| 15:33.05 | jordisayol | aha |
| 15:33.22 | jordisayol | just a moment. I'll paste the arguments here |
| 15:36.35 | jordisayol | on my system, with this cmake: |
| 15:36.35 | jordisayol | cmake -DBRLCAD_BUNDLED_LIBS=ON -DBRLCAD_ENABLE_OPTIMIZED_BUILD=ON -DBRLCAD_ENABLE_STRICT=OFF -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/usr/brlcad -DDATA_DIR=share -DMAN_DIR=share/man |
| 15:36.55 | jordisayol | I got this: |
| 15:36.55 | jordisayol | CC = /usr/bin/gcc |
| 15:36.55 | jordisayol | CXX = /usr/bin/c++ |
| 15:36.55 | jordisayol | CFLAGS = -pipe -fno-strict-aliasing -fno-common -fexceptions -msse3 -std=gnu99 -m64 -O3 -fstrength-reduce -fexpensive-optimizations -finline-functions -finline-limit=10000 -fomit-frame-pointer -pedantic -Wall -Wextra -Wundef -Wfloat-equal -Wshadow -Winline -Wno-long-long |
| 15:36.55 | jordisayol | CXXFLAGS = -pipe -fno-strict-aliasing -fno-common -fexceptions -ftemplate-depth-128 -m64 -O3 -fstrength-reduce -fexpensive-optimizations -finline-functions -finline-limit=10000 -fomit-frame-pointer -pedantic -Wall -Wextra -Wundef -Wfloat-equal -Wshadow -Winline -Wno-long-long |
| 15:36.55 | jordisayol | LDFLAGS = -m64 |
| 15:37.00 | brlcad | no no, need an actual gcc line |
| 15:37.28 | brlcad | since there could be some flag getting set elsewhere maybe |
| 15:37.36 | jordisayol | aha |
| 15:38.09 | brlcad | run "make VERBOSE=1" |
| 15:39.31 | brlcad | then look for a gcc line ( NOT in src/other ) |
| 15:41.28 | jordisayol | hmmm, like this? |
| 15:41.29 | jordisayol | /usr/bin/gcc -pipe -fno-strict-aliasing -fno-common -fexceptions -msse3 -std=gnu99 -m64 -O3 -fstrength-reduce -fexpensive-optimizations -finline-functions -finline-limit=10000 -fomit-frame-pointer -pedantic -Wall -Wextra -Wundef -Wfloat-equal -Wshadow -Winline -Wno-long-long -o CMakeFiles/buildtimestart.dir/CMakeTmp/builddelta_start.c.o -c /home/jordi/svn/b/CMakeTmp/builddelta_start.c |
| 15:56.41 | jordisayol | is it useful, or have to do something else? |
| 16:53.15 | *** join/#brlcad Stattrav (~Stattrav@61.12.114.82) |
| 16:53.16 | *** join/#brlcad Stattrav (~Stattrav@unaffiliated/stattrav) |
| 16:53.57 | brlcad | jordisayol: that's useful |
| 17:36.58 | *** join/#brlcad merzo (~merzo@170-118-132-95.pool.ukrtel.net) |
| 17:39.11 | *** join/#brlcad pawleeq (~pawleeq@212-96-188-229.cust.selfnet.cz) |
| 17:47.47 | *** join/#brlcad Stattrav (~Stattrav@unaffiliated/stattrav) |
| 18:11.06 | jordisayol | brlcad: there are not any -mtune nor -march flags. mean this that gcc compiles to a generic cpu? |
| 18:28.35 | *** join/#brlcad merzo (~merzo@238-170-132-95.pool.ukrtel.net) |
| 21:02.35 | starseeker | brlcad: possibly - the uninstall is interesting, although I'd have to test it with our build... |
| 21:02.57 | starseeker | need to see what OPTION_WITH_DEFAULT is all about |
| 21:03.59 | starseeker | there are a couple things there that might be problematic for us - in particular they're forcing the CMAKE_BUILD_TYPE... |
| 21:07.42 | starseeker | well... hmm. Guess I'd have to try it. We might be fine if we set SCL_BUILD_TYPE to our own build type first. Nothing unusual - we do more than that for png... |
| 22:15.45 | *** join/#brlcad KimK (~Kim__@209.248.147.2.nw.nuvox.net) |