2013-09-16T00:07:35 *** monstr has joined #rtems 2013-09-16T01:42:05 *** sebhub has joined #rtems 2013-09-16T01:42:38 good morning 2013-09-16T02:03:49 *** MegaAlex|away has quit IRC (Ping timeout: 246 seconds) 2013-09-16T02:10:48 *** MegaAlex|away has joined #rtems 2013-09-16T03:25:04 *** arvind_khadri has joined #rtems 2013-09-16T05:41:52 *** MegaAlex|away has quit IRC (Ping timeout: 264 seconds) 2013-09-16T05:42:31 *** MegaAlex|away has joined #rtems 2013-09-16T07:49:02 *** gedare has joined #rtems 2013-09-16T08:21:04 *** sree has joined #rtems 2013-09-16T08:22:06 hey gedare, 2013-09-16T08:22:08 hi sree 2013-09-16T08:23:39 i just sent you an email before you hopped on here :) 2013-09-16T08:24:04 oh i'l check my mail now 2013-09-16T08:24:27 can we talk now? 2013-09-16T08:24:31 it's ok i just wanted to know if you have any progress with periodic tasks 2013-09-16T08:25:39 the tasks are not successful, i tried simplifying the spedfsched task...i am getting some spurious interrupts, i am working on that... 2013-09-16T08:25:55 i pushed in the changes for locking on to github as well 2013-09-16T08:28:35 ok i saw that, although sebhub suggested the locking is not necessary currently due to the Giant lock. 2013-09-16T08:29:04 do you know what happens when a periodic task is scheduled under your globaledf scheduler? 2013-09-16T08:29:23 the "control flow" during a period update? 2013-09-16T08:34:19 ahhh 2013-09-16T08:34:26 using "_Scheduler_simple_Insert_priority_fifo_order" is a problem for EDF. 2013-09-16T08:34:53 The deadline-driven tasks are differentiated by a set most-significant-bit 2013-09-16T08:34:54 for the period update, in the entry point i am calling the edf update, 2013-09-16T08:35:06 in the priority field 2013-09-16T08:35:20 this means you can't make direct comparison of priority values 2013-09-16T08:35:20 you have to mask off the bit 2013-09-16T08:35:52 see score/src/scheduleredfprioritycompare.c 2013-09-16T08:36:33 or rather, the non-periodic tasks get shifted like that. 2013-09-16T08:36:43 either way, the direct comparison of priority fields is no longer valid 2013-09-16T08:36:53 do you understand? 2013-09-16T08:36:58 yes, 2013-09-16T08:37:50 but can we modify priority_compare for globaledf 2013-09-16T08:38:32 that's not the point 2013-09-16T08:38:37 which data structures do you have in the scheduler? do you use the scheduled list to assign threads to processors? 2013-09-16T08:38:48 you have to provide a different version of "_Scheduler_simple_Insert_priority_fifo_order" 2013-09-16T08:39:04 yes a chain_control and rbtree_control 2013-09-16T08:39:26 rather you need to rewrite _Scheduler_globaledf_Enqueue_priority_fifo 2013-09-16T08:39:28 i meant the same thing, write a new version for scheduler_globaledf 2013-09-16T08:39:39 ok. yes, that is at least one thing that needs to be done. 2013-09-16T08:39:51 do you use the _Scheduler_SMP_Allocate_processor() function in your scheduler? 2013-09-16T08:40:16 sebhub: so far he has copied your G-FP code, and replaced the ready queue with an rbtree. 2013-09-16T08:40:46 ah, ok, this is a good start 2013-09-16T08:41:12 yes. now he has to fix some fine details related to difference between periodic and aperiodic tasks. 2013-09-16T08:41:28 the processor to thread assignment is a bit tricky, you have to avoid a thread executing on two processors at once 2013-09-16T08:41:38 this is good 2013-09-16T08:41:49 he ran most of the smptests i think. 2013-09-16T08:41:57 though the platform gives him some "spurious interrupt 7" sometimes 2013-09-16T08:42:14 (qemu-i386 with core2duo) 2013-09-16T08:42:15 what is this for an interrupt? 2013-09-16T08:42:28 what i read it is the parallel port (printer). 2013-09-16T08:42:39 hm, strange 2013-09-16T08:42:44 but i don't know if Qemu might post something on it 2013-09-16T08:42:51 you can also use the realview bsp on qemu 2013-09-16T08:43:05 cat /proc/irq/7 2013-09-16T08:43:11 looks like in an SMP system it is used for affinity 2013-09-16T08:43:23 by linux 2013-09-16T08:43:42 sorry, ls :0 2013-09-16T08:44:01 qemu runs in parallel really? 2013-09-16T08:44:57 hm, nvm. I don't know what it is fo 2013-09-16T08:44:58 for 2013-09-16T08:45:05 anyway he only gets it sometimes 2013-09-16T08:45:29 i get it for smp08 2013-09-16T08:46:12 http://www.rtems.com/pipermail/rtems-bugs/2010-October/001917.html 2013-09-16T08:46:23 looks liek this is a recurring problem 2013-09-16T08:46:37 smp08 is a good test if you remove the timeout and let it run for a couple of hours 2013-09-16T08:47:02 *** MegaAlex|away has quit IRC (Ping timeout: 240 seconds) 2013-09-16T08:47:03 ah, it could be floating-point related. 2013-09-16T08:47:45 this interrupt 7 may be an exception, not an irq. 2013-09-16T08:47:59 based on that thread from RTEMS mailing list ;) 2013-09-16T08:48:32 ok sree. I think you have to adjust some of the globaledf.c code to handle the difference with periodic tasks 2013-09-16T08:48:33 *** MegaAlex|away has joined #rtems 2013-09-16T08:50:29 i was thinking, the priority comparision we are making in enqueue_ordered is based on the chain_priority when we compare rbtree nodes as well as chain nodes. 2013-09-16T08:50:42 it would be also good to have a new test smpedf01 or so 2013-09-16T08:51:06 yeah i asked him to make one of those 2013-09-16T08:51:50 oh, where it calls (*order)(...)? 2013-09-16T08:52:05 yes when i extract highest_ready 2013-09-16T08:52:18 yeah that could be a problem too. 2013-09-16T08:52:20 if i can compare the priorities differently 2013-09-16T08:54:07 the order function should not use chain parameters 2013-09-16T08:54:20 this is a bug 2013-09-16T08:55:30 sree: I think the priority comparison may not matter, as long as the comparison is unsigned int, because the background tasks will be treated as having really large priorty values, but the comparison should still "just work".. 2013-09-16T08:55:51 (because the background tasks have their most-sig-bit set) 2013-09-16T08:55:53 hmmm. 2013-09-16T08:56:18 sree, do you have an smpedf01 available? 2013-09-16T08:56:56 or are you just modifying the spschededf directly right now? 2013-09-16T08:57:55 i m modifying spschededf 2013-09-16T08:58:13 can you make your changes into a new smp test so that I can try it out too? 2013-09-16T08:58:20 sure 2013-09-16T08:58:29 i'll add a new smp test 2013-09-16T08:58:42 ok 2013-09-16T08:59:39 where does the "order" function get set? 2013-09-16T09:00:04 oh i see it is passed in by parameter 2013-09-16T09:02:14 from enqueue 2013-09-16T09:02:20 enqueue_fifo 2013-09-16T09:04:45 ok.. 2013-09-16T09:05:09 what you need to do is to compare the regular EDF scheduler hooks with your globaledf Scheduler hooks to see what functionality is missing 2013-09-16T09:05:36 for example, enqueue should set SCHEDULER_EDF_QUEUE_STATE_YES 2013-09-16T09:06:43 sure, i'l do that. 2013-09-16T09:07:11 you might want to start with initialize.. 2013-09-16T09:07:15 and verify each change as you make it 2013-09-16T09:07:32 and i'll also change the (*order)(...) when extracting highest_ready with edf_priority_compare 2013-09-16T09:07:48 sure 2013-09-16T09:11:24 ok. make sure the comparison is used properly (e.g. < makes sense) 2013-09-16T09:11:45 since a < b for priority means a has higher priority and should be used first 2013-09-16T09:12:09 sure, i'll verify that 2013-09-16T09:12:11 there might also be a bug from taking out the (-1)*_Scheduler_Priority_compare() that I suggested earlier. 2013-09-16T09:13:02 oh, since edf has inverted priority comparisions? 2013-09-16T09:13:06 yeah 2013-09-16T09:13:07 as you mentioned previouslt 2013-09-16T09:13:19 previously* 2013-09-16T09:13:49 well, the rbtree sorts things in low-to-high, so you just have to be careful what functions you use to access it 2013-09-16T09:14:19 RBTree_First(...,RBT_LEFT) will give you the minimum 2013-09-16T09:15:25 yes, 2013-09-16T09:16:16 ok. i consider this handling of periodic tasks to be very important. 2013-09-16T09:17:28 i'll try to make all the changes we discussed now,ASAP 2013-09-16T09:17:56 good. start with a useful small test case 2013-09-16T09:18:26 for the ratemonotonic manager there is a small example in the docs 2013-09-16T09:22:00 *** arvind_khadri has quit IRC (Remote host closed the connection) 2013-09-16T09:23:19 *** arvind_khadri has joined #rtems 2013-09-16T09:35:23 *** freenix has quit IRC (Remote host closed the connection) 2013-09-16T10:34:15 *** sree has quit IRC (Ping timeout: 250 seconds) 2013-09-16T10:35:08 *** monstr has quit IRC (Ping timeout: 245 seconds) 2013-09-16T10:42:14 *** vipulnayyar has joined #rtems 2013-09-16T10:42:56 *** arvind_khadri has quit IRC (Remote host closed the connection) 2013-09-16T10:53:00 *** sebhub has quit IRC (Remote host closed the connection) 2013-09-16T11:24:04 *** arvind_khadri has joined #rtems 2013-09-16T11:25:25 *** arvind_k has joined #rtems 2013-09-16T11:28:54 *** arvind_khadri has quit IRC (Ping timeout: 264 seconds) 2013-09-16T11:30:50 *** vipulnayyar has quit IRC (Ping timeout: 245 seconds) 2013-09-16T11:34:47 *** vipulnayyar has joined #rtems 2013-09-16T11:40:50 *** vipulnayyar has quit IRC (Ping timeout: 245 seconds) 2013-09-16T12:03:21 *** arvind_k has quit IRC (Ping timeout: 245 seconds) 2013-09-16T12:47:49 *** arvind_khadri has joined #rtems 2013-09-16T14:30:01 *** arvind_khadri has quit IRC (Ping timeout: 246 seconds) 2013-09-16T15:02:30 *** arvind_khadri has joined #rtems 2013-09-16T16:17:31 *** gedare has quit IRC (Ping timeout: 245 seconds) 2013-09-16T17:36:33 *** arvind_khadri has quit IRC (Remote host closed the connection) 2013-09-16T18:56:54 *** gedare has joined #rtems 2013-09-16T19:07:50 *** gedare has quit IRC (Ping timeout: 240 seconds) 2013-09-16T20:50:01 *** freenix has joined #rtems 2013-09-16T20:54:37 hi,kiwichris_. The relocation type in libm.a is R_ARM_THM_CALL, if libm.a is loaded at high address, but it reference a symbol in rtld which is loaded at low address, RTL can not resolve this now. 2013-09-16T21:00:29 Is the libm and libc in the same directory ? 2013-09-16T21:00:42 Can you please look with -v on the command line ? 2013-09-16T21:02:55 yeah. libm.a and libc.a are both at /opt/rtems-4.11/arm-rtems4.11/lib/thumb/armv7-a 2013-09-16T21:03:08 ok 2013-09-16T21:04:00 Ok. 2013-09-16T21:04:28 What does the code look like if added to the base kernel image ? 2013-09-16T21:05:14 Now the size of rtld including the python-lib is 14M 2013-09-16T21:05:36 elf size 2013-09-16T21:06:08 nice 2013-09-16T21:06:37 Are you forcing the whole of Python to the base image ? 2013-09-16T21:07:03 I mean a single libm call 2013-09-16T21:07:55 The parser and Module and python are all included, but I #undef many macros in pyconfig.h when compiling. 2013-09-16T21:08:25 In the base image ? 2013-09-16T21:08:30 yeah the whole of Python is included in the base image 2013-09-16T21:08:53 Because I can not resolve the reloc related problem in libm.a 2013-09-16T21:09:25 if using rap, the reloc related problem should be resolved. 2013-09-16T21:09:38 so I just merge the python into rtl 2013-09-16T21:09:58 thus i do not have to care about the reloc related problem. 2013-09-16T21:10:01 Sure 2013-09-16T21:11:06 I have to leave now. I have an exam at 15 minutes later. 2013-09-16T21:12:11 As an example s_floor.c is build with '-mthumb -march=armv7-a' 2013-09-16T21:12:58 I wonder is this is a limitation with thumb code. 2013-09-16T21:13:19 What is you removed -mthumb from the BSP 2013-09-16T21:14:37 there is also '-mthumb -march=armv7-a -mfpu=neon -mfloat-abi=hard' 2013-09-16T21:14:48 which is another variant 2013-09-16T21:18:23 thumb libc also has them 2013-09-16T21:18:33 the reloc records 2013-09-16T21:28:37 freenix, have you looked at ARM's IHI0044E page 32 ? 2013-09-16T21:35:19 *** vipulnayyar has joined #rtems 2013-09-16T21:58:01 *** vipulnayyar has quit IRC (Ping timeout: 246 seconds) 2013-09-16T22:47:19 *** vipulnayyar has joined #rtems 2013-09-16T22:56:07 hi,kiwichiris_. The exam is over just now. yeah. I have read page 23 of IHI0044E. I think the point is libm.a is not compiled with `-mlong-calls`. 2013-09-16T23:01:58 Oh really. Interesting 2013-09-16T23:02:09 Same with libc 2013-09-16T23:02:21 Sebastian will be around soon. We should ask him 2013-09-16T23:03:14 Last exam ? 2013-09-16T23:07:47 it is just a job hunting related exam 2013-09-16T23:07:49 ok. 2013-09-16T23:09:26 I am not sure this should be handed in the rtems-ld side or on the target side 2013-09-16T23:09:32 handed->handled 2013-09-16T23:15:47 I mean all the object files or libs should be compiled with `long-call` support ,because we do not know the where the rap file will be loaded 2013-09-16T23:25:26 How does the standard linker handle this ? 2013-09-16T23:26:31 I have not looked into this. 2013-09-16T23:27:25 This may need digging into the source of bfd or gcc 2013-09-16T23:28:00 Can you see what it does by look at the outout ? 2013-09-16T23:28:30 Does it change the code or add something to make it work ? 2013-09-16T23:29:59 I am not sure.But the output may not give any info about this. ok. I think I should look into bfd to find this. 2013-09-16T23:31:44 Ok 2013-09-16T23:36:53 *** vipulnayyar has quit IRC (Ping timeout: 248 seconds) 2013-09-16T23:43:20 *** monstr has joined #rtems 2013-09-17T01:22:11 *** arvind_khadri has joined #rtems 2013-09-17T01:41:35 *** the9a3eedi-2 has joined #rtems 2013-09-17T01:44:38 *** the9a3eedi has quit IRC (Ping timeout: 240 seconds) 2013-09-17T02:01:46 *** sebhub has joined #rtems 2013-09-17T02:03:06 good morning 2013-09-17T02:24:42 sebhub, hi 2013-09-17T02:24:48 hi chris 2013-09-17T02:25:49 thanks for the rsb update 2013-09-17T03:50:52 yeah no problem; thanks for the send in the changes 2013-09-17T03:51:28 sebhub, freenix_ is seeing an issue with a very large RAP app 2013-09-17T03:51:57 he is building python as an app as an example of using dynamic loading 2013-09-17T03:52:39 on ARM he is seeing R_ARM_THM_CALL relocations from libm that are too far away 2013-09-17T03:52:53 he is looking at how the gnu ld handles this 2013-09-17T03:53:08 Is there a size limit with thumb code ? 2013-09-17T03:59:18 there are size limits for arm and thumb code 2013-09-17T03:59:35 the linker has to perform relaxations in this case I hope 2013-09-17T04:01:03 Ok. Peng will have to take a look at what the standard linker does 2013-09-17T04:07:30 on powerpc the gnu linker will turn a "blr addr" into a "li addr, addis addr, mtctr, bctrl" for example 2013-09-17T04:08:13 So the code offsets change ? 2013-09-17T04:08:22 It make the code size larger ? 2013-09-17T04:08:51 yes, its an iterative process 2013-09-17T04:09:01 Ouch 2013-09-17T04:09:55 search in the binutils sources for "relaxation" 2013-09-17T04:47:38 *** the9a3eedi-2 has quit IRC (Ping timeout: 245 seconds) 2013-09-17T05:18:02 *** the9a3eedi has joined #rtems 2013-09-17T05:47:13 *** MegaAlex|away has quit IRC (Ping timeout: 245 seconds) 2013-09-17T05:49:45 *** MegaAlex|away has joined #rtems 2013-09-17T06:09:02 *** the9a3eedi has quit IRC (Ping timeout: 240 seconds) 2013-09-17T06:49:04 *** marcinb has joined #rtems 2013-09-17T07:15:52 *** gedare has joined #rtems 2013-09-17T07:27:19 *** phipse has joined #rtems 2013-09-17T07:43:13 *** marcinb has quit IRC (Quit: Leaving) 2013-09-17T08:45:54 *** the9a3eedi has joined #rtems 2013-09-17T08:52:22 *** the9a3eedi has quit IRC (Read error: Connection reset by peer) 2013-09-17T09:15:08 *** freenix has quit IRC (Remote host closed the connection) 2013-09-17T09:25:40 *** phipse has quit IRC (Ping timeout: 264 seconds) 2013-09-17T09:48:10 *** phipse has joined #rtems 2013-09-17T09:55:04 *** phipse has quit IRC (Ping timeout: 264 seconds) 2013-09-17T10:15:37 *** sebhub has quit IRC (Remote host closed the connection) 2013-09-17T11:25:04 *** monstr has quit IRC (Ping timeout: 264 seconds) 2013-09-17T11:47:20 *** arvind_khadri has quit IRC (Remote host closed the connection) 2013-09-17T12:27:45 *** vipulnayyar has joined #rtems 2013-09-17T12:33:30 *** vipulnayyar has quit IRC (Quit: Leaving) 2013-09-17T12:47:33 *** dmbaturin has left #rtems ("iQuit") 2013-09-17T13:09:02 *** vipulnayyar has joined #rtems 2013-09-17T13:25:32 *** vipulnayyar has quit IRC (Quit: Leaving) 2013-09-17T13:44:54 *** phipse has joined #rtems 2013-09-17T13:55:40 *** phipse has quit IRC (Ping timeout: 264 seconds) 2013-09-17T14:56:49 verm__, about ? 2013-09-17T15:14:19 verm__, I built the latest MinGW tools on mutex and the old autoconf bug of embeddeding host tool paths in a crosscompled version of autoconf. It seems /usr/local/bin/gm4 has appeared on the machine and causing an issue. 2013-09-17T15:15:16 probably makes sense to make a jail for building 2013-09-17T15:15:40 m4 from ports is required for the new autoconf and automake 2013-09-17T15:16:16 Does this mean we lock up autoconf and remove it from society for our safety ? 2013-09-17T15:16:20 deleted 2013-09-17T15:16:45 /usr/local/bin/gm4 deleted ? 2013-09-17T15:19:15 all of it, yes 2013-09-17T15:19:37 i'll ask for another IP then i can give you a jail with root, you can do whatever you want in there 2013-09-17T15:19:45 just dedicate it to mingw building 2013-09-17T15:20:07 when we get the new hardware we can rsync it over to the new host without changing anything (other than the installed packages since new hosts will be freebsd 10) 2013-09-17T15:20:39 actually, i could just bind the jail to the local host for now 2013-09-17T15:26:28 Why not just stop RTEMS using autoconf and automake ? ;) 2013-09-17T15:26:45 Then I do not need to build and ship it 2013-09-17T15:30:24 i don't know about that, sounds like a multi-year project, would take a brave person to take on that task 2013-09-17T15:30:29 brave and handsome. 2013-09-17T15:32:38 shame, that rules us out 2013-09-17T15:33:25 agreed... 2013-09-17T15:48:31 i think there is a potion you can get. 2013-09-17T15:58:22 tried beer and wine, no difference 2013-09-17T15:59:53 i think the other people need it to make a difference 2013-09-17T22:04:19 *** gedare has quit IRC (Ping timeout: 264 seconds) 2013-09-18T00:00:01 *** monstr has joined #rtems 2013-09-18T01:32:10 *** arvind_khadri has joined #rtems 2013-09-18T01:46:12 *** sebhub has joined #rtems 2013-09-18T01:48:43 *** MegaAlex|away has quit IRC (Ping timeout: 268 seconds) 2013-09-18T01:49:24 *** MegaAlex|away has joined #rtems 2013-09-18T03:29:22 *** phipse has joined #rtems 2013-09-18T05:44:46 *** MegaAlex|away has quit IRC (Ping timeout: 256 seconds) 2013-09-18T05:45:29 *** MegaAlex|away has joined #rtems 2013-09-18T06:02:15 *** arvind_khadri has quit IRC (Ping timeout: 276 seconds) 2013-09-18T06:13:23 *** arvind_khadri has joined #rtems 2013-09-18T07:23:34 *** freenix has joined #rtems 2013-09-18T07:53:40 *** gedare has joined #rtems 2013-09-18T08:14:43 *** S_Somani has joined #rtems 2013-09-18T08:18:50 *** arvind_khadri has quit IRC (Ping timeout: 264 seconds) 2013-09-18T08:34:54 *** arvind_khadri has joined #rtems 2013-09-18T08:36:51 *** S_Somani has left #rtems 2013-09-18T08:38:46 *** freenix has quit IRC (Quit: Leaving) 2013-09-18T09:01:54 *** gedare has quit IRC (Ping timeout: 264 seconds) 2013-09-18T09:45:58 *** sebhub has quit IRC (Read error: Operation timed out) 2013-09-18T10:25:14 *** phipse has left #rtems 2013-09-18T10:35:46 *** arvind_khadri has quit IRC (Remote host closed the connection) 2013-09-18T11:01:23 *** monstr has quit IRC (Ping timeout: 245 seconds) 2013-09-18T11:15:38 *** S_Somani has joined #rtems 2013-09-18T11:22:43 *** vipulnayyar has joined #rtems 2013-09-18T11:30:00 *** vipulnayyar has quit IRC (Ping timeout: 245 seconds) 2013-09-18T11:32:32 *** jceel has joined #rtems 2013-09-18T11:32:50 *** jceel has left #rtems 2013-09-18T11:38:06 *** arvind_khadri has joined #rtems 2013-09-18T11:53:48 *** arvind_khadri has quit IRC (Remote host closed the connection) 2013-09-18T12:36:14 *** antgreen has quit IRC (Ping timeout: 240 seconds) 2013-09-18T14:30:19 *** S_Somani has left #rtems 2013-09-18T18:52:23 *** gedare has joined #rtems 2013-09-18T19:01:01 *** gedare has quit IRC (Ping timeout: 268 seconds) 2013-09-18T20:46:34 *** antgreen has joined #rtems 2013-09-18T23:05:20 *** monstr has joined #rtems 2013-09-18T23:28:00 *** freenix has joined #rtems 2013-09-19T00:50:47 *** kiwichris_ has quit IRC (Quit: This computer has gone to sleep) 2013-09-19T01:17:35 *** kiwichris has joined #rtems 2013-09-19T01:43:36 *** sebhub has joined #rtems 2013-09-19T01:44:58 good morning 2013-09-19T01:50:35 *** kiwichris has quit IRC (Read error: Connection reset by peer) 2013-09-19T02:05:20 *** kiwichris has joined #rtems 2013-09-19T02:29:53 *** freenix has quit IRC (Quit: Leaving) 2013-09-19T04:15:01 *** arvind_khadri has joined #rtems 2013-09-19T05:13:20 *** MegaAlex|away has quit IRC (Ping timeout: 245 seconds) 2013-09-19T05:14:14 *** MegaAlex|away has joined #rtems 2013-09-19T07:21:15 *** gedare has joined #rtems 2013-09-19T08:36:33 *** tictoc has joined #rtems 2013-09-19T08:37:52 *** arvind_khadri has quit IRC (Remote host closed the connection) 2013-09-19T08:39:05 Hi everyone. One question, I have a controller thread and several other threads that communicate with the controller. Apart from synchronization and mutual exclusion, does it make a difference in access times if the shared memory between the threads is on the heap or part of the controller stack? 2013-09-19T08:39:30 Memory allocation would happen during system intialization, so it's really just read / write access 2013-09-19T08:47:24 hi tictoc 2013-09-19T08:48:00 hi gedare 2013-09-19T08:48:26 i don't think the access time would differ, although accessing stack memory between tasks is risky business 2013-09-19T08:48:48 risky, but doable ;) 2013-09-19T08:50:00 In that case I will probably just create it on the heap 2013-09-19T08:50:02 rtems provides message queues which could be useful also. these are pre-allocated on the heap. 2013-09-19T08:50:10 that's what i would do. 2013-09-19T08:52:37 I will, thanks. I have some common control structures that need to be accessible by all threads but for the rest I will use message queues. 2013-09-19T08:53:11 good luck! 2013-09-19T08:53:17 or should i say, good haacking 2013-09-19T08:56:33 actually let me rephrase my earlier question. In terms of determinism of the access time, do heap memory and stack memory behave the same way? (when planning a hard real time application) 2013-09-19T08:58:28 tictoc: assuming it is on the same memory chip 2013-09-19T08:58:55 and you never call free. :) 2013-09-19T08:58:57 not e.g. stacks on in CPU memory and heap in external memeory 2013-09-19T08:59:21 other than that there is caching 2013-09-19T09:00:27 tictoc: or said differently some embedded systems have memory regions with different accesstimes 2013-09-19T09:00:45 but timing depends mostly on this 2013-09-19T09:00:51 stack and heap (prealloced no free calls) are the same 2013-09-19T09:01:04 perfect, thanks peerst and gedare : ) 2013-09-19T09:01:12 but I wouldn't use stack for sharing 2013-09-19T09:01:20 thats calling for trouble later 2013-09-19T09:01:35 heck I'd avoid sharing if possible 2013-09-19T09:09:26 howdy peerst been awhile, doing well? 2013-09-19T09:11:06 hi gedare, yeah I think so … future projects still unclear while being superbusy in the present 2013-09-19T09:11:26 but thats probably just the normal state of being self employed 2013-09-19T09:11:27 that's about the way of it when things are good :) 2013-09-19T09:11:35 how are your studies/projects 2013-09-19T09:11:43 yep.. my mom has been self-employed for > 10 yrs and is almost always like that. 2013-09-19T09:12:06 well I prefer when the financial cliff is not as close 2013-09-19T09:12:07 good.. i got my phd late last year, and am working as a postdoc trying to get funding to become a research prof 2013-09-19T09:12:10 haha yeah 2013-09-19T09:12:32 i have about 5-6 months to be successful or find a new job :) 2013-09-19T09:12:43 well, either way i'll have a new job, ha! 2013-09-19T09:12:46 need to change your nick in drgedare 2013-09-19T09:13:17 heh, no we have enough "dr"s around 2013-09-19T09:13:18 porting Erlang to RTEMS is being quite succesful 2013-09-19T09:13:23 awesome 2013-09-19T09:13:34 the Erlang community loves it 2013-09-19T09:14:00 well they love if a conference talk show some real world hardware 2013-09-19T09:14:07 that's great. do you have any examples? 2013-09-19T09:14:29 many of them have too many layers of virtualization between them and the hardware 2013-09-19T09:14:41 oh yeah sure 2013-09-19T09:14:58 examples? like projetcs running with it? 2013-09-19T09:15:28 yeah, or an easy-to-follow tutorial 2013-09-19T09:15:38 for the RTEMS community :) 2013-09-19T09:15:40 you mean for Erlang on RTEMS or on Erlang itself? 2013-09-19T09:15:48 Erlang+RTEMS 2013-09-19T09:15:55 Erlang on RTEMS is not distributed yet 2013-09-19T09:15:59 ahhh ok 2013-09-19T09:16:21 i don't have experience with Erlang, but I think the idea is interesting to use it with RTOS 2013-09-19T09:16:34 and i'm curious about its runtime system 2013-09-19T09:16:42 but i don't have time to explore it presently 2013-09-19T09:16:45 probably need to move it to RTL to make dis-entangling it with the customer project more easy 2013-09-19T09:16:53 ah 2013-09-19T09:17:10 having Erlang on small hardware on a RT OS thin layer rocks 2013-09-19T09:17:35 I acces the hardware from Erlangs port drivers directly 2013-09-19T09:18:21 nice 2013-09-19T09:18:26 sometimes with a RTEMS thread to handle IRQs 2013-09-19T09:18:39 *** phipse has joined #rtems 2013-09-19T09:18:40 but starting with a dead simple driver e.g. just shoving a bunch of bytes to a SPI 2013-09-19T09:18:49 and build all the higher levels in Erlang as prototype 2013-09-19T09:18:54 * gedare does the "bug fix" dance. 2013-09-19T09:18:57 is totally saving time 2013-09-19T09:19:06 i could see that 2013-09-19T09:19:14 sometimes i get quite sick of writing C ;) 2013-09-19T09:19:38 having something like Erlangs shell trying out the hardware not rebooting the system for 3 days in a row 2013-09-19T09:19:51 are there any parts of Erlang you need to avoid, or that cause performance overheads/bottlenecks? 2013-09-19T09:20:34 Erlang sequential performance is not very fast 2013-09-19T09:20:41 expcet a factor of 10 to C 2013-09-19T09:20:52 why the slowdown? is it interpreted? 2013-09-19T09:20:53 *but* its reactivity is superfast 2013-09-19T09:20:59 guess i could go read more about it 2013-09-19T09:21:08 its a Threaded code VM 2013-09-19T09:21:20 there is HiPE a native compiler 2013-09-19T09:21:51 ok 2013-09-19T09:21:56 but its not very well maintained and you loose the run a *.beam file everywhere (BEAM is the emulator) fun 2013-09-19T09:22:08 there is *very* intersting work on a JIT though 2013-09-19T09:22:56 but if one does not have experienced it its hard to belive how reactive Erlang systems can be 2013-09-19T09:23:00 soft-realtime 2013-09-19T09:23:07 yeha 2013-09-19T09:23:35 millions of http connections on a normal server 2013-09-19T09:23:47 this scales down very well 2013-09-19T09:24:26 often people think they need NIF (native implemented functions written in C) 2013-09-19T09:24:40 sounds like where Java was a few years ago 2013-09-19T09:24:42 but then they learn that all this C code messes with Erlangs schedulers 2013-09-19T09:25:23 and they actually get a faster complete system when they rewrite their NIF's in C 2013-09-19T09:26:17 *nah* in Erlang 2013-09-19T09:26:20