2012-04-16T00:34:01 *** arvind_khadri has joined #rtems 2012-04-16T00:44:08 *** arvind_khadri has quit IRC (Ping timeout: 240 seconds) 2012-04-16T00:46:52 *** arvind_khadri has joined #rtems 2012-04-16T00:57:31 *** arvind_khadri has quit IRC (Ping timeout: 246 seconds) 2012-04-16T01:15:19 *** arvind_khadri has joined #rtems 2012-04-16T01:53:50 *** sebhub has joined #rtems 2012-04-16T01:54:17 good morning 2012-04-16T03:10:56 *** zw__yao has joined #rtems 2012-04-16T03:11:09 *** zw__yao has quit IRC (Client Quit) 2012-04-16T04:04:48 *** QingPei has left #rtems 2012-04-16T04:05:36 *** cdcs has joined #rtems 2012-04-16T04:20:48 *** xiangfu has quit IRC (Ping timeout: 252 seconds) 2012-04-16T04:27:48 *** xiangfu has joined #rtems 2012-04-16T05:18:44 *** panzon has quit IRC (Ping timeout: 260 seconds) 2012-04-16T06:39:59 *** antgreen has joined #rtems 2012-04-16T06:46:25 *** QingPei has joined #rtems 2012-04-16T07:14:29 Hi, from where can I clone the git repo. No private repo in github 2012-04-16T07:35:01 http://git.rtems.org/ 2012-04-16T07:35:16 cdcs: thanks for the response 2012-04-16T07:35:21 12:18 < lcpfnvcy> mmorning 2012-04-16T07:35:22 12:19 < lcpfnvcy> does rtems have something similar to vmware esx 2012-04-16T07:35:22 12:36 < cdcs> Hello 2012-04-16T07:35:22 12:37 < cdcs> if i understand you question correctly: 2012-04-16T07:35:22 12:37 < cdcs> No it does not. But several hypervisors support RTEMS as a virtualized operating system. 2012-04-16T07:35:25 little late 2012-04-16T07:35:27 :( 2012-04-16T07:36:17 hm, this was for arvind_khadri 2012-04-16T07:36:42 sebhub, thanks 2012-04-16T07:37:06 and also http://wiki.rtems.org/wiki/index.php/Git 2012-04-16T08:13:35 the github repo will be up in a week or two 2012-04-16T08:13:38 i've been busy with other things 2012-04-16T08:14:15 lcpfnvcy, No problem ;) 2012-04-16T08:14:39 *** panzon has joined #rtems 2012-04-16T08:14:41 nice, how will it sync with the OAR repo? 2012-04-16T08:17:10 i wrote something a month back 2012-04-16T08:17:19 it'll be realtime once the commit is done o the master it'll push to github 2012-04-16T08:17:25 need time to set it up 2012-04-16T08:17:44 ok, so its automati 2012-04-16T08:17:45 c 2012-04-16T08:17:49 oh, of course 2012-04-16T08:18:13 i'll get it up by the end of the month hopefully 2012-04-16T08:18:23 this way gsoc students can use github over googlecode if they want 2012-04-16T08:18:37 its always good to have backups 2012-04-16T08:19:21 well, it provides a way for users to easily track the repo 2012-04-16T08:20:25 and they can publically store their repo so others can watch it similar to how we handle personal repos 2012-04-16T08:38:59 *** xiangfu has quit IRC (Quit: Leaving) 2012-04-16T08:49:36 *** cdcs has quit IRC (Ping timeout: 245 seconds) 2012-04-16T08:52:29 *** cdcs has joined #rtems 2012-04-16T09:20:07 *** gedare has joined #rtems 2012-04-16T09:33:49 good morning 2012-04-16T09:34:45 A0Sheds: i view the two projects (stm32 and beagleboard) as rather similar. Both offering support for open hardware. Beagleboard also has a quite large user community especially among hobbyists / open-source enthusiasts 2012-04-16T09:35:12 gedare: the stm32 will have millions of users 2012-04-16T09:35:34 mostly large OEMs in China will develop with it 2012-04-16T09:36:24 DrJoel understood the significance of what is going on here with replacing 20 years of closed firmware in a laptop 2012-04-16T09:36:45 i thought that is why he asked us to fid a student 2012-04-16T09:36:49 fid/find 2012-04-16T09:36:53 it is 2012-04-16T09:37:24 but i don't want to waste anymore time 2012-04-16T09:37:41 *** weiY has joined #rtems 2012-04-16T09:43:08 hi, gedare. how about the process of gsoc review? 2012-04-16T09:44:57 hey weiY: it's coming along. i think the announcements come in a week 2012-04-16T09:45:39 Both the STM32 and OMAP4 projects expect an RTEMS port to an ARM cortex M3, although different BSPs 2012-04-16T09:45:43 ok. and the mentor of the rtems is enough for all students? 2012-04-16T09:46:38 it seems that there are so many students with good proposal than past years 2012-04-16T09:46:40 and there is a OMAP gsoc project already.. 2012-04-16T09:46:44 weiY: we're a little thin but not too bad 2012-04-16T09:47:32 good news. 2012-04-16T09:49:39 omap4 is less about the bsp and more about the AMP/IPC framework and SMP on ARM 2012-04-16T09:49:54 the stm32 project is more about peripherals and getting it to work with small targets 2012-04-16T09:50:20 yup, i know. What i don't know is how much of the BSP development will overlap with STM32 2012-04-16T09:50:45 i think quite little unless they happen to share similar SOC peripherals 2012-04-16T09:51:24 i think the omap4 project have much overlap work with beagleboard 2012-04-16T09:51:40 since it should add BSP support for OMAP at first 2012-04-16T09:52:28 the beagleboard project will target Cortex A8 while the OMAP project will target the CortexM3 2012-04-16T09:54:02 now that the stm32f patch is public much of the RTEMS work is done 2012-04-16T09:54:06 therefore i don't expect much overlap 2012-04-16T09:54:22 isn't beagleboard/beaglebone using a uniproc. design? 2012-04-16T09:54:37 Google just published a working library of EC routines for the cortex M3 2012-04-16T09:54:50 link? 2012-04-16T09:54:52 we have Uwe's open stm32 library 2012-04-16T09:55:15 and we also have openEC that may be ported to arm32 2012-04-16T09:55:44 http://git.chromium.org/gitweb/?p=chromiumos/platform/ec.git;a=tree;f=chip;h=fab28b9e1ec6b544ed20d386984de3249a06bcb2;hb=HEAD 2012-04-16T09:55:56 so a lot has changed in a week 2012-04-16T09:55:58 nice 2012-04-16T09:56:01 yeah. for rtems much of that is superfluous / application level 2012-04-16T09:56:07 rather much that remains 2012-04-16T09:56:14 just glueing things together 2012-04-16T09:56:18 i wish they would have been public weeks ago vs just sitting on it :) 2012-04-16T09:56:19 which is a good bit of work 2012-04-16T09:56:21 lol yes 2012-04-16T09:56:26 maybe that is why they didn't accept you guys :/ 2012-04-16T09:56:38 80% of the project is done now 2012-04-16T09:56:39 (i doubt they have that much coordination) 2012-04-16T09:57:04 now it's just supporting more flavors of the cortex m3/m4 with RTEMS 2012-04-16T09:57:20 drivers, peripherals 2012-04-16T09:57:21 the google code is bsd which is license compatible that is nice for rtems 2012-04-16T09:57:27 yeah 2012-04-16T09:57:31 optimization for tinyRTEMS 2012-04-16T09:57:32 and maybe some tinyRTEMS work 2012-04-16T09:57:34 right. 2012-04-16T09:59:02 we still may ask for the slot so that the student can do drivers/peripherals and tinyRTEMS work 2012-04-16T09:59:09 but i wouldn't hold my breath 2012-04-16T09:59:51 we need a working laptop asap 2012-04-16T10:00:01 so this might be done by june 1 2012-04-16T10:00:23 to show at computex 2012-04-16T10:06:41 hi gedare. whether the student slots assigned to rtems was announced? 2012-04-16T10:06:51 i don't believe the # is public data. 2012-04-16T10:07:59 oh, what is the meaning of # 2012-04-16T10:09:32 *** arvind_khadri has quit IRC (Ping timeout: 272 seconds) 2012-04-16T10:15:52 # means the slot number, i guess 2012-04-16T10:18:58 yeah, i think so. i think it should be ##, haha 2012-04-16T10:40:39 *** weiY has quit IRC (Ping timeout: 250 seconds) 2012-04-16T10:47:45 *** panzon has quit IRC (Ping timeout: 252 seconds) 2012-04-16T11:08:57 *** panzon has joined #rtems 2012-04-16T11:13:40 *** panzon has quit IRC (Ping timeout: 252 seconds) 2012-04-16T11:25:53 *** panzon has joined #rtems 2012-04-16T11:29:10 *** sebhub has quit IRC (Ping timeout: 252 seconds) 2012-04-16T13:00:01 *** cdcs has quit IRC (Ping timeout: 245 seconds) 2012-04-16T13:26:01 *** QingPei has left #rtems 2012-04-16T13:34:54 *** panzon has quit IRC (Quit: Leaving) 2012-04-16T13:35:09 *** panzon has joined #rtems 2012-04-16T14:34:19 *** DrJoel has joined #rtems 2012-04-16T14:34:20 *** DrJoel has joined #rtems 2012-04-16T14:34:20 *** ChanServ sets mode: +o DrJoel 2012-04-16T14:56:47 howdy drjoel. 2012-04-16T14:57:31 hey.. been busy offline .. and then had a power supply in a PC at home die overnight. :( 2012-04-16T14:58:11 that's not fun. better than a hard drive though 2012-04-16T14:58:22 things settling at all for you yet? 2012-04-16T15:01:29 this week should be a bit better. Michele is still in a lot of pain and needs help. 2012-04-16T15:03:26 as if a dead power supply isn't bad enough.. the one I bought was DOA. 2012-04-16T15:03:36 I knew because we had a spare one here at OAR I tested with 2012-04-16T15:03:37 ouch 2012-04-16T15:04:00 they looked surprised when I returned it .. but Jeff said it has happened to him there also 2012-04-16T15:04:15 funny to think that a 3Ghz dual core with 4GB RAM could be 4 years old 2012-04-16T15:05:57 DrJoel: Hi 2012-04-16T15:06:50 DrJoel: what do you think about kqueues in RTEMS? 2012-04-16T15:07:57 verm__ suggested that ummm what did we say... 2012-04-16T15:08:59 we ended up linking them from http://wiki.rtems.org/wiki/index.php/RTEMSSystemEvents 2012-04-16T15:09:13 thing is Erlang needs either a working select, poll or kqueues 2012-04-16T15:09:14 I remember not knowing enough to have a final opinion .. but an open to anything that makes porting easier 2012-04-16T15:09:32 At the moment we have neither 2012-04-16T15:09:44 select only works with sockets 2012-04-16T15:09:53 poll + kqueues doesn't exist 2012-04-16T15:10:17 and the select implementation looks like a hack to me 2012-04-16T15:10:50 is kqueue synchronous? 2012-04-16T15:11:05 (from the userland side) 2012-04-16T15:11:13 what exactly do you mean by synchronous? 2012-04-16T15:11:40 actually it doesn't matter / make sense. either the user will sleep on the event or poll it 2012-04-16T15:12:18 yeah I was confused ... since stuff like kqueue select and poll are mostly used to make stuff more asynchronous 2012-04-16T15:12:19 peerst.. then kqueues is important.. I personally want RTEMS to be a strong platform for as many languages as possible 2012-04-16T15:12:36 does the wiki page make it clear that kqueues are important for Erlang? 2012-04-16T15:13:01 DrJoel: thing is we very probably have customer funding for porting kqueues 2012-04-16T15:13:02 nope. we were unaware of what users might exist for the kqueues immediately. 2012-04-16T15:13:17 we will adopt kqueues if it is done rihgt ;) 2012-04-16T15:13:44 dumping a mess of bsd code into the kernel to implement kqueues is "not right"... 2012-04-16T15:13:45 Sounds good 2012-04-16T15:14:14 the bsd implemenation of kqueues is pretty neat IMHO 2012-04-16T15:14:16 http://people.freebsd.org/~jlemon/papers/kqueue.pdf 2012-04-16T15:14:37 it may be OK I lack the spare capacity to investigate thoroughly right now 2012-04-16T15:15:10 if it just sits between an application and rtems then it's fine... 2012-04-16T15:15:19 if we need to hook into rtems kernel services then we have to worry more 2012-04-16T15:15:24 yep it does 2012-04-16T15:15:37 no need to go quite deep 2012-04-16T15:15:54 needs smallish support in every event source though 2012-04-16T15:16:06 that will be fine 2012-04-16T15:16:13 so basically everything that is a file descriptor 2012-04-16T15:16:29 as I said.. done carefully and with feedback from the devel community it should be ok 2012-04-16T15:16:58 additional event sources can easily be added optionally as user extensions 2012-04-16T15:17:49 the hen and egg problem is that the customer paying for it will be disappointed if we have to maintain it separately forever 2012-04-16T15:18:24 well it should be approached like the other bsd port works... 2012-04-16T15:18:34 but if the generall concept of kqueues is accepted he'll take the risk 2012-04-16T15:18:46 probably the port will be done by sebastian 2012-04-16T15:18:48 the page i linked to kind of says it all 2012-04-16T15:18:53 (sebhub) 2012-04-16T15:19:05 we had an event notification system but it is "broken"... 2012-04-16T15:19:08 we're open to better solutions 2012-04-16T15:19:19 if kqueue solves everything then great 2012-04-16T15:19:22 cool 2012-04-16T15:19:23 I have nothing against any particular implementation of kqueues... and having a user who "cares" is enough to drive requirements in my mind. As long as we end up with something that is well documented and well tested 2012-04-16T15:19:34 yup...and maintainable 2012-04-16T15:19:37 dont' forget that one :0 2012-04-16T15:20:55 :) 2012-04-16T15:21:06 any othe 'ibles? 2012-04-16T15:21:58 kqueues are really low impact: you need a pointer of storage per event source, then you write 3 small so called filter functions and put pointers to them in a struct, the kqueues infrastructure takes care of the rest 2012-04-16T15:22:24 so its really easy to add kqueue support for new filedescriptorish things 2012-04-16T15:22:36 which is part of maintainability 2012-04-16T15:22:49 :) i think it sounds good. 2012-04-16T15:23:43 if you have time sometimes read the paper I linked to, its well written 2012-04-16T15:23:59 if you lost the link its linked from the wikipedia entry of kqueue 2012-04-16T15:24:07 i'll try 2012-04-16T15:24:16 and i'll keep an eye out for development :0 2012-04-16T15:25:01 well announce it if we work out the monetary things with the customer and can start with it 2012-04-16T15:25:18 ok great. 2012-04-16T15:44:38 rtems needs kqueue/kevent 2012-04-16T15:47:14 :-) 2012-04-16T16:29:13 *** gedare has quit IRC (Quit: Leaving) 2012-04-16T17:16:43 *** DrJoel has quit IRC (Quit: Leaving) 2012-04-16T17:56:33 *** gedare has joined #rtems 2012-04-16T18:07:42 *** dr__house has quit IRC (Ping timeout: 272 seconds) 2012-04-16T18:07:43 *** mwalle has quit IRC (Ping timeout: 272 seconds) 2012-04-16T18:07:46 *** dr__house has joined #rtems 2012-04-16T18:07:47 *** dr__house has joined #rtems 2012-04-16T18:08:18 *** mwalle has joined #rtems 2012-04-16T18:43:30 gedare, do all m68k bsps build with the 1/4 m68k patch ? 2012-04-16T18:43:58 kiwichris: drjoel was testing that.. he ran into an unrelated bug during one of them though 2012-04-16T18:44:23 i'm not sure what his status was.. i built one of the bsps.. lacking the infrastructure right now to try building all of them 2012-04-16T18:44:26 Yes I also found the iess else where a few weeks go and managed to work around 2012-04-16T18:44:36 less -> issue 2012-04-16T18:45:11 i'd have to ask him to see what he managed to build. 2012-04-16T18:45:16 Understood. I cannot get gdb for the PowerPC to build on MacOS. Everything else does 2012-04-16T18:45:28 You can build on the RTEMS servers. 2012-04-16T18:45:34 i'll have to figure that out 2012-04-16T18:45:59 That patch is ok with me if it builds all m68k bsps 2012-04-16T18:46:30 Also this just shows why we need a patch tool and have get all effect code to build before committing 2012-04-16T18:47:32 yup. 2012-04-16T18:47:46 and making this change has furhter convinced me that automake must die. 2012-04-16T18:48:00 hah 2012-04-16T18:48:06 half my effort in updating this set_vector code was modifying makefile.am 2012-04-16T18:48:10 at least i could do it with sed mostly. 2012-04-16T18:49:00 Feeble, real automake hackers use ed ;) 2012-04-16T18:49:24 ... and boot their machine with dip switches 2012-04-16T18:52:53 and press the turbo button to make it go fast. 2012-04-16T18:54:27 I am now calling the makefile.am's and related m4 cruft a toxic zone. You only head in when you really need to and with maximum protection 2012-04-16T18:55:20 If you slip and break through the surface you find yourself in black quick sand 2012-04-16T18:55:28 if you're lucky. 2012-04-16T18:55:49 someday i expect my bootstrap will go into infinite recursion and fill my hard drive 2012-04-16T18:55:58 just a funny feeling i get 2012-04-16T18:56:10 What is the area on Everest where the climbs must use ladders over gaps ? 2012-04-16T18:56:46 no clue 2012-04-16T18:56:49 sounds ominous though 2012-04-16T19:02:49 Here is an interesting stat ..... 2012-04-16T19:02:50 $ cat $(find . -name Makefile.am -o -name Makefile.in -o -name \*.m4 -o -name configure -o -name \*.ac ) | wc -l 2012-04-16T19:02:50 1510827 2012-04-16T19:03:08 ack 2012-04-16T19:03:12 I suspect more lines in the autoconf/automake cruft we carry around than code 2012-04-16T19:03:22 would be an interesting study 2012-04-16T19:03:25 i wonder... 2012-04-16T19:03:54 And that is not the cache stuff 2012-04-16T19:04:01 I think 2012-04-16T19:05:54 *** Deb has joined #rtems 2012-04-16T19:06:48 well there is a slight difference between generated files and not, but only minorly so 2012-04-16T19:26:37 Generate or not they exists. Generated files can have bugs just like any other piece of code. 2012-04-16T19:26:52 The more lines the high the chance 2012-04-16T19:36:12 *** Deb has quit IRC (Quit: Leaving) 2012-04-16T19:36:28 *** xiangfu has joined #rtems 2012-04-16T19:43:55 lol 2012-04-16T19:44:03 ansic: 745153 (35.83%) 2012-04-16T19:44:04 sh: 638451 (30.70%) 2012-04-16T19:44:04 makefile: 551563 (26.52%) 2012-04-16T19:44:04 m4: 73475 (3.53%) 2012-04-16T19:44:04 asm: 38574 (1.85%) 2012-04-16T19:44:04 ada: 26261 (1.26%) 2012-04-16T19:44:06 cpp: 5248 (0.25%) 2012-04-16T19:44:08 pascal: 814 (0.04%) 2012-04-16T19:44:10 perl: 279 (0.01%) 2012-04-16T19:44:29 thats with counting Makefile.in and Makefile.am as 'makefile' 2012-04-16T19:44:53 and without the autom4te.cache 2012-04-16T19:45:46 hah 2012-04-16T19:45:47 *** xiangfu has quit IRC (Ping timeout: 260 seconds) 2012-04-16T19:45:52 after running bootstrap obviously. 2012-04-16T19:45:58 ywah 2012-04-16T19:46:02 yeah 2012-04-16T19:46:38 That piece of perl code is impossible to understand 2012-04-16T19:46:44 *** panzon_ has joined #rtems 2012-04-16T19:48:43 *** panzon has quit IRC (Ping timeout: 246 seconds) 2012-04-16T19:54:19 Totals grouped by language (dominant language first): 2012-04-16T19:54:19 ansic: 745153 (87.00%) 2012-04-16T19:54:19 asm: 38574 (4.50%) 2012-04-16T19:54:19 makefile: 30083 (3.51%) 2012-04-16T19:54:19 ada: 26261 (3.07%) 2012-04-16T19:54:20 sh: 7086 (0.83%) 2012-04-16T19:54:24 cpp: 5248 (0.61%) 2012-04-16T19:54:26 m4: 2996 (0.35%) 2012-04-16T19:54:28 pascal: 814 (0.10%) 2012-04-16T19:54:30 perl: 279 (0.03%) 2012-04-16T19:54:32 that's without bootstrapping 2012-04-16T19:54:40 the moral is that we use almost twice as much generated code as we do what is written / stored in the repo 2012-04-16T19:54:47 just for rtems building 2012-04-16T19:55:17 I suspect this and while you can argue it is generated it is silly. 2012-04-16T19:55:47 Another stat to collect would be the number of forks to build the everything. 2012-04-16T19:56:00 I think we have about 5 per file compiled 2012-04-16T19:56:22 sh + test + gcc + cc1 + asm 2012-04-16T19:56:50 Actually that means 2 shell forks and not forks 2012-04-16T19:58:10 *** xiangfu has joined #rtems 2012-04-16T20:04:50 yeah 2012-04-16T20:12:39 http://gedare-csphd.blogspot.com/2012/04/crufty-stuff.html 2012-04-16T20:12:42 there you go :) 2012-04-16T20:12:45 and im out. later 2012-04-16T20:12:49 *** gedare has quit IRC (Quit: Leaving) 2012-04-16T20:52:40 *** xiangfu has quit IRC (Ping timeout: 276 seconds) 2012-04-16T20:53:02 *** xiangfu has joined #rtems 2012-04-16T21:00:23 *** xiangfu has quit IRC (Quit: Leaving) 2012-04-16T21:00:40 *** xiangfu has joined #rtems 2012-04-16T21:36:14 *** QingPei has joined #rtems 2012-04-16T21:37:08 *** antgreen has quit IRC (Ping timeout: 245 seconds) 2012-04-16T21:47:47 *** Deb has joined #rtems 2012-04-16T21:57:08 *** xian9fu has joined #rtems 2012-04-16T21:57:40 *** xiangfu has quit IRC (Ping timeout: 276 seconds) 2012-04-16T21:57:44 *** xian9fu has joined #rtems 2012-04-16T21:58:09 *** xian9fu is now known as xiangfu 2012-04-16T22:21:43 *** xiangfu has quit IRC (Ping timeout: 276 seconds) 2012-04-16T22:33:30 *** xiangfu has joined #rtems 2012-04-16T23:03:30 *** Deb has quit IRC (Quit: Leaving) 2012-04-16T23:16:33 *** xiangfu has quit IRC (Remote host closed the connection) 2012-04-16T23:17:12 *** xiangfu has joined #rtems 2012-04-17T01:32:18 *** sebhub has joined #rtems 2012-04-17T01:57:37 *** xiangfu has quit IRC (Ping timeout: 246 seconds) 2012-04-17T02:23:55 *** panzon__ has joined #rtems 2012-04-17T02:26:31 *** panzon_ has quit IRC (Ping timeout: 260 seconds) 2012-04-17T02:28:45 *** arvind_khadri has joined #rtems 2012-04-17T02:28:46 *** arvind_khadri has joined #rtems 2012-04-17T02:28:56 *** xiangfu has joined #rtems 2012-04-17T02:30:47 *** panzon_ has joined #rtems 2012-04-17T02:33:40 *** panzon__ has quit IRC (Ping timeout: 260 seconds) 2012-04-17T02:44:30 *** panzon__ has joined #rtems 2012-04-17T02:46:16 *** panzon_ has quit IRC (Ping timeout: 246 seconds) 2012-04-17T02:46:21 *** arvind_khadri has quit IRC (Ping timeout: 260 seconds) 2012-04-17T02:50:42 *** panzon_ has joined #rtems 2012-04-17T02:52:00 *** arvind_khadri has joined #rtems 2012-04-17T02:52:01 *** arvind_khadri has joined #rtems 2012-04-17T02:53:29 *** panzon__ has quit IRC (Ping timeout: 245 seconds) 2012-04-17T03:00:02 good morning 2012-04-17T03:45:24 *** cdcs has joined #rtems 2012-04-17T03:58:25 *** panzon__ has joined #rtems 2012-04-17T04:01:08 *** panzon_ has quit IRC (Ping timeout: 272 seconds) 2012-04-17T04:04:08 *** QingPei has quit IRC (Remote host closed the connection) 2012-04-17T04:51:27 *** xiangfu has quit IRC (Remote host closed the connection) 2012-04-17T05:50:06 *** panzon__ has quit IRC (Ping timeout: 260 seconds) 2012-04-17T05:51:14 *** Hesham has joined #rtems 2012-04-17T06:52:39 *** arvind_k has joined #rtems 2012-04-17T06:53:06 *** arvind_khadri has quit IRC (Ping timeout: 260 seconds) 2012-04-17T08:09:36 *** gedare has joined #rtems 2012-04-17T08:15:37 *** panzon__ has joined #rtems 2012-04-17T08:21:32 *** leyyin has joined #rtems 2012-04-17T08:25:57 *** xiangfu has joined #rtems 2012-04-17T08:33:40 *** weiY has joined #rtems 2012-04-17T08:38:41 *** QingPei has joined #rtems 2012-04-17T09:43:55 *** Hesham has quit IRC (Quit: Leaving.) 2012-04-17T09:49:09 *** xiangfu has quit IRC (Quit: Leaving) 2012-04-17T10:24:31 *** QingPei has quit IRC (Ping timeout: 276 seconds) 2012-04-17T10:26:00 *** QingPei has joined #rtems 2012-04-17T10:29:19 *** sebhub has quit IRC (Ping timeout: 260 seconds) 2012-04-17T10:42:07 *** arvind_k has quit IRC (Ping timeout: 245 seconds) 2012-04-17T10:57:32 *** weiY has quit IRC (Ping timeout: 245 seconds) 2012-04-17T11:21:38 *** Hesham has joined #rtems 2012-04-17T11:36:28 hey gedare 2012-04-17T11:36:36 hello hesham 2012-04-17T11:37:22 is your schoolwork going well? 2012-04-17T11:37:25 about rbheap , should i consider its implementation to the high level API or just mark it as future work ? 2012-04-17T11:37:50 yeah i have finished mid-term exams :) 2012-04-17T11:37:54 nice 2012-04-17T11:38:24 thanks 2012-04-17T11:38:45 i've been thinking about that same thing 2012-04-17T11:39:02 and ? 2012-04-17T11:39:12 the problem we have to tackle is how to deal with memory management versus memory allocation 2012-04-17T11:39:30 the rbheap is nice because it shows a very clear separation between management and allocation 2012-04-17T11:39:40 i think i will stuck with memory management in general in RTEMS 2012-04-17T11:39:51 yeah 2012-04-17T11:40:24 right. the allocation shouldn't matter too much, except that if the management data structures reside together with the management structures then you cannot provide adequate protection mechanisms 2012-04-17T11:40:56 take a look at how Region Manager uses its attributes and write a brief summary 2012-04-17T11:41:36 and where it saves its management data structures ? 2012-04-17T11:41:54 it would be interesting to find out if the Region Manager can have a flexible way to specify its memory pool and whether there is space in the attribute set for specifying an allocator and whether protection migth be enabled at the Region level 2012-04-17T11:42:10 i think data structures the manage attributes of Region/Partition should be in separate area 2012-04-17T11:42:15 and whether the Region API as-is can be left alone but extended with new functions to handle these features. 2012-04-17T11:42:23 not necessarilly 2012-04-17T11:42:35 for achieving protection yes 2012-04-17T11:43:08 ok so.. 2012-04-17T11:43:13 there is a distinction to make 2012-04-17T11:43:54 protection , allocation , attributes 2012-04-17T11:43:57 between allocator management data structure and the Manager Object data structure 2012-04-17T11:44:15 exactly 2012-04-17T11:44:32 allocators can either embed their management data within the memory region or have external structures 2012-04-17T11:44:49 traditional heap (malloc) embeds management data before/after an allocated region 2012-04-17T11:45:18 and then set its attributes ? 2012-04-17T11:45:25 don't worry about attributes yet 2012-04-17T11:45:29 understand how allocation works 2012-04-17T11:45:40 if you view the memory region as a large chunk then an allocator can do one of two things 2012-04-17T11:46:04 it can set aside one chunk for management and the rest for handling allocation requests 2012-04-17T11:46:30 or it can set aside the entire chunk for handling allocation requests and keep the management data co-located with the allocated blocks 2012-04-17T11:46:39 malloc does the second approach 2012-04-17T11:46:48 rbheap implements the first approach 2012-04-17T11:47:08 partition manager implements the second approach because the partition objects reside on a linked list 2012-04-17T11:47:26 region manager uses the score Heap which implements malloc so it uses the second approach 2012-04-17T11:47:40 you cannot implement useful MMU-based protection / attributes using the second approach 2012-04-17T11:47:51 right 2012-04-17T11:48:11 so we should use first approach 2012-04-17T11:48:19 We should support both approaches 2012-04-17T11:48:37 and for the first approach we should offer the possibility of using hardware-based protection / attributes enforcement 2012-04-17T11:49:02 that's means Arenas and MMU 2012-04-17T11:49:32 probably. how to hook the allocator up to the arena is unclear 2012-04-17T11:50:39 should we think about how to implement it now ? or just take in mind that Arenas would be used along with rbheap at the design time ? 2012-04-17T11:51:35 we should consider it now 2012-04-17T11:51:54 i believe it will help us to decide where to put the Arena code 2012-04-17T11:52:06 most of which I think is written 2012-04-17T11:52:30 Then , it would be good if you provided me with any references/code that would help me 2012-04-17T11:52:38 (other than the tasking problems) 2012-04-17T11:53:12 i had read about Partition / Region Managers at online library 2012-04-17T11:53:19 I think a simple starting point is to read the code for partition and regions 2012-04-17T11:53:37 cpukit/rtems/include/rtems/rtems/region.h and part.h are the headers 2012-04-17T11:53:51 and cpukit/rtems/inline/rtems/rtems/region.inl and part.inl are their inline functions (if any) 2012-04-17T11:54:07 and various files in cpukit/rtems/src/region* or part* are their functions 2012-04-17T11:54:48 is it well documented/commented ? 2012-04-17T11:55:06