**** BEGIN LOGGING AT Tue Jun 12 02:59:58 2012 Jun 12 06:28:23 so my oe does have TARGET_FPU = "hard" and my target is a omap3 .. i still can see that it uses -mfloat-abi=softfp shouldnt this then becom-mfloat-abi=hardfp ? Jun 12 06:40:58 rob_w: having fpu and having hard fpu calling convention are two things on arm Jun 12 06:42:32 rob_w: for having hard fpu calling convention you need to have callconvention-hard in TUNE_FEATURES Jun 12 06:42:34 can you enlight me a abit about those Jun 12 06:43:00 arm traditionally did not have fpu Jun 12 06:43:14 therefore default calling convention is soft Jun 12 06:43:47 ok Jun 12 06:43:51 and when neon and other coprocessors came into being gcc started using them to do some integer arithmetic smartly Jun 12 06:44:09 but calling convention still was soft Jun 12 06:44:17 so this was called softfp Jun 12 06:44:24 so will my results be faster if i now use hardfp abi ? Jun 12 06:44:50 where gcc can use FP intruction inside the function e.g. for vectorization but calling conventions will be soft Jun 12 06:45:01 rob_w: I doubt it will be fast Jun 12 06:45:18 does your application use lot of floating point ? Jun 12 06:45:37 and in such a way that it passes fp arguments to functions ? Jun 12 06:45:50 if answer is no then you will hardly see any bebefit Jun 12 06:46:11 ok Jun 12 06:48:13 reading through some docs , i get that making sure i use -mfpu=neon for my cortex8 will be the biggest optimisation ? Jun 12 06:48:22 and remember hard float ABI is not intermixable with soft-float abi Jun 12 06:48:37 yep got taht Jun 12 06:48:45 yes using -mfpu=neon -mfloat-abi=softfp Jun 12 06:49:34 but you can try to use hard float ABI if you wish in OE-Core I have made sure that both work Jun 12 06:49:51 what processor do you have Jun 12 06:50:15 TI dm3730 Jun 12 06:50:33 and some customer asked me about hardfp support .. Jun 12 06:51:02 so in that case i would need to rebuild my whole chain to hardfp using ARM_FP_ABI = "hardfp" Jun 12 06:51:29 but i wonder if that makes any sense at all and compared to the price of incompatibility to softfp binarys Jun 12 06:51:34 in OE-Core you need to add callconvention-hard to TUNE_FEATURES Jun 12 06:52:03 with classical OE I dont guarntee it at all Jun 12 06:53:19 rob_w: most of arm software was soft float friendly thus far but with hardfloat abi you can port some of application that use floating point with out penalty of calling convention Jun 12 06:53:30 so forward thinking it might be a good thing Jun 12 06:53:39 but dont expect anything immediate Jun 12 06:56:07 k Jun 12 06:56:35 i will let come down to soome performance tests , i can then rebuild .. thx alot for the clarifcations Jun 12 06:58:40 yes, please benchmark both and share results here. Jun 12 06:58:51 in my tests I did not get much Jun 12 06:59:34 unless I used something like NAS parallel benchmark which is written in fortran and uses openmp Jun 12 08:19:53 good morning Jun 12 08:33:22 good morning Jun 12 09:55:47 morning folks Jun 12 10:01:07 bluelightning: hi Jun 12 10:01:14 hi mckoan Jun 12 10:39:25 hi bluelightning , did you ever try the acx-mac80211 ? Jun 12 10:42:19 lautriv: no, I haven't Jun 12 10:42:22 hi bluelightning, all Jun 12 10:42:25 hi pb_ Jun 12 13:01:16 beware - I will use OE again ;D Jun 12 13:05:14 hrw: oh no! (LOL) Jun 12 13:05:43 hrw: welcome back ;-) Jun 12 13:07:33 can layer.conf use DEFAULT_PREFERENCE on recipes? I know that it is rather $DISTRO stuff but I have layer which needs this Jun 12 13:07:54 or should I rather use bbappend on selected recipes? Jun 12 13:17:25 hrw: hi, better to use .bbappends. See http://lists.linuxtogo.org/pipermail/bitbake-devel/2011-August/001169.html Jun 12 13:18:18 (I still have that link as not yet fully resolved...) Jun 12 13:18:23 thx Jun 12 13:20:59 awesome Jun 12 13:24:52 hrw, what will you do with it? Jun 12 13:25:02 hrw: great to hear :) welcome back! Jun 12 13:25:13 more linaro android optimizations? :p Jun 12 13:25:29 Crofton|work: can not tell yet but it will be nice thing Jun 12 13:25:56 ynezz: I love that subject ;d Jun 12 13:26:15 heh Jun 12 13:27:54 well, android layer in oe would be welcome :p Jun 12 14:07:12 WARNING: Buildfile specified, dependencies will not be handled. If this is not what you want, do not use -b / --buildfile. Jun 12 14:07:20 I love that warning ;) Jun 12 14:19:20 hrw: howdy Jun 12 14:20:25 khem: howdy Jun 12 14:23:04 hrw: back at OE hacking ? Jun 12 14:23:53 kind of Jun 12 14:25:03 can someone try to compile beecrypt(-native) with gcc-4.7? Jun 12 14:27:01 hrw: seeing issues ? Jun 12 14:28:35 khem: http://paste.debian.net/174123/ Jun 12 14:30:11 hmm Jun 12 14:32:12 khem: --disable-cplusplus for native may be other option Jun 12 14:33:30 hrw: yes thats probably good Jun 12 19:26:44 so yesterday someone mentioned a patch for kernel 3.5 regarding an alignment patch for arm, i assumed because -rc2 came out after that and had certain changes in this area, i would be ok. that wasn't the case. i found some mailings about it but nowhere a link to this patch, anyone able to tell me a location ? Jun 12 22:24:49 hello all. If i try to build meta-toolchain-qte on an x86_64 machine, will I get x86_64 binaries or i686 ones? **** ENDING LOGGING AT Wed Jun 13 02:59:58 2012