**** BEGIN LOGGING AT Wed Jun 04 02:59:58 2014 Jun 04 03:00:44 this is after "test_and_set_bit(PRU_DCF_DOWNCALL_REQ, &ppc->dc_flags)" is successful in pru_downcall Jun 04 03:14:52 panto : everything seems correct uptill that error, I can't figure out whats wrong! Jun 04 07:44:02 greetings everyone Jun 04 07:55:00 hi Jun 04 07:58:25 panto : did you see my msg? Jun 04 08:05:52 how did you made the downcall? Jun 04 08:06:02 did you follow the same steps as in pru_downcall? Jun 04 08:06:43 yep, same driver Jun 04 08:07:01 did not even modify pru_downcall Jun 04 08:08:11 looks like the downcall did not set the status bits properly Jun 04 08:08:48 so you execute a system call in PRU that said you are ready for a downcall Jun 04 08:08:59 but the ARM did not send a downcall request Jun 04 08:09:15 or it executed a downcall but you did it twice Jun 04 08:09:19 The ARM did send a request Jun 04 08:09:26 put some debugging prints and try to figure out what's going on Jun 04 08:09:35 I'm on it Jun 04 08:09:44 I'll get back in a few Jun 04 08:53:49 panto : you are right, something is firing twice. I'll figure out how and why! Jun 04 08:54:00 *where and why Jun 04 09:53:13 * karki is frustrated enough to eat his BBB. Jun 04 09:59:38 * Abhishek_ yawns Jun 04 09:59:49 morning Jun 04 10:00:47 vmayoral|pc: How's it going with Xenomai? Jun 04 10:01:22 Abhishek_: good, the timings even in userspace are quite nice Jun 04 10:01:47 but for BeaglePilot we will proabably move to vanilla, vanilla-PREEMPT or RT at the end. We will just avoid the capemgr Jun 04 10:02:21 vmayoral|pc: Do you have some numbers> Jun 04 10:02:58 I am curious about interrupt latencies Jun 04 10:04:13 i will share them soon :) Jun 04 10:04:19 sure Jun 04 10:07:06 vmayoral: Avoiding the capemgr means building the overlays right into the device tree, correct? Jun 04 10:07:44 vvu_: pong Jun 04 10:08:22 yes, Anuj is working on a DT directly Jun 04 10:08:37 meanwhile we are working with some overlays on vanilla Jun 04 10:08:47 anyhow i have to share that even the results in vanilla are pretty nice Jun 04 11:14:03 panto : figured out the problem, in "pru_handle_syscall", in the switch case statement, the control is jumping to PRU_SC_DOWNCALL_READY both the times. the second time it should actually jump to "PRU_SC_DOWNCALL_DONE", which it is not. Jun 04 11:15:26 odd, cause I have used 254 in "R14" the first time and 255 the next time. yet same switch is getting trigged. Jun 04 11:15:47 Anyway, I'll carry out some more debugging to figure out whats happening. Jun 04 12:00:34 panto : the scno value is 254 both the times, why? The downcall handler on the PRU is invoked ( I checked by setting GPIO pin values). but while signaling to the ARM that downcall is done, 254 is read again ( though I LDI r14 with 255) Jun 04 12:00:54 panto : I'm *really* close. just can't figure this out. Jun 04 12:01:05 what is your PRU code doing? Jun 04 12:01:24 this is read from the PRU when it is stopped Jun 04 12:03:58 PRU code is the same as your code. Jun 04 12:04:06 I have changed only the handler Jun 04 12:07:41 panto : I'm sure I'm tracing this correctly (I also drew a flowchart for the flow of control during downcalls so as to be clear), but I just don't get why/where the register values are going wrong..... Jun 04 12:08:10 something is obviously different Jun 04 12:09:30 here is the hacky version of the asm part : https://gist.github.com/deepakkarki/329ab90de34f187c3e0e Jun 04 12:14:09 karki, err, are you advancing the program counter when taking the exception? Jun 04 12:14:21 looks like you just resume execution and hitting halt again Jun 04 12:14:57 the driver takes care of that.... right? Jun 04 12:16:05 yes, in theory Jun 04 12:16:05 panto : the handler is getting executed correctly. Jun 04 12:16:18 anyway, lunch Jun 04 12:16:27 ok :) Jun 04 12:40:07 hi jkridner Jun 04 12:40:21 gm Abhishek_ Jun 04 13:17:17 hey jkridner Jun 04 13:17:25 have time for a quick talk? Jun 04 14:14:04 hi DiegoTc Jun 04 14:14:11 I have a few minutes now Jun 04 14:19:26 hey jkridner I send you a PM Jun 04 14:19:31 did you read it? Jun 04 14:19:37 or I do copy paste? Jun 04 14:19:42 I read it. Jun 04 14:20:55 Want to talk a little? o we wait for the meeting about it? Jun 04 14:21:18 tell me what is on your mind. Jun 04 14:21:25 Do you have any updates now to show? Jun 04 14:21:38 did VoltVisionSteve get you any design help? Jun 04 14:22:26 After the email Steve send me, I haven't talk with him. Jun 04 14:23:36 I have the back part of the code, (I'm sending it by parameters "hard code" this will be replace of course with what it will be on the text fields) Jun 04 14:24:35 So that's nothing to show for the status Jun 04 14:25:47 by "back part" do you mean writing TO gist.github.com? Jun 04 14:25:55 yeap Jun 04 14:26:20 well, I like to see code as much if not more than HTML pages. :-) Jun 04 14:27:16 Cool, when I arrive home I will upload the branch and share it. Jun 04 14:27:29 DiegoTc: I see some traffic from Steve today. Jun 04 14:29:07 Yes, I will communicate with the professor Steve recommend, besides they speak spanish it will be easier Jun 04 14:29:25 jkridner: what do you think about my recomendation: (It's fine for you mentors to inform you daily or every two days my updates and receive your feedback) Jun 04 14:36:37 yes, more communication the better. Jun 04 14:36:48 from both sides. sorry I am difficult to reach at times. Jun 04 14:52:11 ds2: A little digging with the logs and I found out that /proc/kmsg doesn't work but /dev/xconsole does. Jun 04 14:55:00 almost meeting time? Jun 04 14:55:21 one hour to go, I guess Jun 04 14:56:28 k Jun 04 15:13:39 jkridner, we have a lunch at work today so I likely won't be able to make the meeting, sorry, but I'll scroll back through when I'm back to my pc Jun 04 15:21:22 Abhishek_: Here you have a summary of the latency tests https://groups.google.com/forum/#!topic/beagleboard-gsoc/BISQ3fJj3P4. You can follow the link for more ;) Jun 04 15:22:46 Xenomai = 630ms ?? Jun 04 15:22:58 yes Jun 04 15:22:58 that's unexpected max latency Jun 04 15:23:06 cyclictests are runned in userspace Jun 04 15:23:15 not in the separate xenomai kernel Jun 04 15:23:47 guess that the overlays in userspace are notable Jun 04 15:24:00 still, note that the Avg is below the RT_PREEMPT Jun 04 15:24:25 i runned the tests several times because i didn't expect the vanilla kernels to perform so good Jun 04 15:25:22 well sure you did bust the "realtime" myth it seems for your application ;) Jun 04 15:26:38 i'm happy about the results but i wish i could have more time to test it throughfully Jun 04 15:26:50 anyhow it seems the data is good enough to through my quad to the air Jun 04 15:26:56 we'll see ;) Jun 04 15:28:05 ;) Jun 04 15:30:17 is the vanilla 3.8.13 kernel "tickless"? Jun 04 15:32:22 yeap Jun 04 15:33:19 https://github.com/RobertCNelson/bb-kernel.git Jun 04 15:45:37 all the 'vanilla' is making me hungry :p Jun 04 15:48:56 av500: pong Jun 04 15:49:02 karki: get a chocolate then :P Jun 04 15:49:09 meeting now right ? Jun 04 15:49:17 almost time Jun 04 15:49:31 couple of min more..... Jun 04 15:52:20 hi disdi Jun 04 15:52:26 hi karki Jun 04 15:52:34 hows it going? :) Jun 04 15:52:48 its kind of going Jun 04 15:52:56 isn't is always ;) Jun 04 15:53:07 yes :) Jun 04 15:53:22 do you graduate this year? Jun 04 15:53:37 post graduate in september Jun 04 15:57:51 vvu_: yeah, just a couple more minutes. Jun 04 15:58:47 hi cdsteinkuehler Jun 04 15:59:06 Status report posted Jun 04 16:00:14 Just reading your status report, looks like you're making good progress! Jun 04 16:00:17 kinda offtopic question: is anyone familiar with TCL around here ? Jun 04 16:01:05 TCL? Jun 04 16:01:21 tool command line Jun 04 16:01:27 some scripting language Jun 04 16:01:52 I have passing familiarity with TCL, it 's the scripting language used by my FPGA tools, but I'm no expert. Jun 04 16:01:55 Tcl/Tk, don't know much about it other than my use of OpenOCD which uses it for config Jun 04 16:02:40 cdsteinkuehler: did some oo things with it ? i need to set up some classes and stuff like this Jun 04 16:02:54 but anyway meeting now, i will see what is happening Jun 04 16:03:26 * karki is not feeling well. Jun 04 16:03:34 That's beyond my usage of Tclk/Tk. I generally have to dig out a language reference whenver I need to do something. Jun 04 16:04:01 * Abhishek_ would be doing with Node.JS in a few weeks from now Jun 04 16:04:09 *doing the same Jun 04 16:07:08 gm all Jun 04 16:07:11 gm Jun 04 16:07:14 hello ! Jun 04 16:07:18 hi Jun 04 16:07:48 good morning Jun 04 16:07:49 hello Jun 04 16:08:18 ACK Jun 04 16:09:18 ! Jun 04 16:09:18 REQ Jun 04 16:11:24 I think some students are struggling right now keeping up with the high expectations around GSoC. Jun 04 16:12:17 I'd like to brainstorm any ideas to give some trainings or similar items people feel may help jumpstart some of the coding. Jun 04 16:13:35 most projects are going well, but it is also possible to feel like things are mostly ok without getting things clean and shareable. Jun 04 16:14:32 does the idea of hosting some trainings sound helpful, or are the topics too diverse for that to be useful. Jun 04 16:14:46 I'm thinking on the topic of 'git' and perhaps 'kernel'. Jun 04 16:14:58 jkridner : +1 (kernel) Jun 04 16:15:12 though I'm done with it. Jun 04 16:15:21 but I'd still like it :) Jun 04 16:15:22 i think mostly kernel Jun 04 16:15:30 not 'git'? Jun 04 16:15:36 kernel! that should be fun too Jun 04 16:15:40 * jkridner worries about clean commit histories, etc. Jun 04 16:15:57 producing pull requests to lists, cleaning up patches after-the-fact, etc. Jun 04 16:16:13 ok 'git' too :) Jun 04 16:16:58 maybe start with how to send a patch. Jun 04 16:17:13 jkridner, training sounds good but the short span of GSoC may be too short for that Jun 04 16:17:39 although I was not fully clear and probably am still not what's the balance between training versus doing something useful in the GSoC program Jun 04 16:17:45 I agree... it is something that would have been better a month ago. :( Jun 04 16:17:56 jkridner: yup, a little bit on the practical nuances which help a lot though have to be dug up running through references. Jun 04 16:17:59 now we are well into the coding time and the time is very short. Jun 04 16:18:28 other brainstorms? Jun 04 16:19:07 how long would the training take? Jun 04 16:19:09 but need more pateince form mentors too.... some of us are too new to kernel development specially to the embedded domain(ARM) Jun 04 16:19:19 Abhishek1: how goes the libsigrok world? Jun 04 16:19:23 additional meeting times at more convenient hours covering specific topics and of shorter duration each? Jun 04 16:19:28 i think every project should have in their git repo TODO which is updated daily with problems and what is needed Jun 04 16:19:33 training wont be needed since most of the mentors make up for it Jun 04 16:19:38 so people do not get stuck and just ask during meetings Jun 04 16:19:44 vvu: that is a good idea Jun 04 16:20:12 and we can sync them from time to time and help with what we can Jun 04 16:20:14 vvu: having a clear TODO list is definitely required. I'm fine with it being in their git repo. Jun 04 16:20:36 vvu: getting help on the TODO items is what has me most concerned. don't want them simply sitting in the repo and not getting solved. Jun 04 16:20:45 mranostay: libsigrok driver is currently still boilerplate. Once the kernel work is frozen it would get through Jun 04 16:20:49 yes, but it must exist 1st :) Jun 04 16:20:54 vvu: +1 Jun 04 16:21:29 github issues are nice for todo items because they're less hidden away than a text file Jun 04 16:21:31 like this year is not so much activity on the IRC channel Jun 04 16:21:32 Abhishek_: are you using a character device to pass traffic to userspace? Jun 04 16:21:36 or yes github issues Jun 04 16:21:46 alexanderhiam: and no conflicts to fix :P Jun 04 16:21:56 hey disdi sorry to say this, but atleast for me as a mentor- the "too new to kernel development" point is not at all true. I have emails from you stating that you had worked on kernel previously. Atleast for me, the expectations were set in advance. Sorry if I being an ass and not so polite, but I'm just saying it the way it is. Jun 04 16:22:40 mranostay: hehe Jun 04 16:22:57 jkridner: I'm fine with trainings. Think it's a great chance. We can all learn from the more experience users :) Jun 04 16:22:59 aren't we like 2 or so weeks from midterms? Jun 04 16:23:02 mranostay: currently working on it as an alternative to shared memory [which will be there by default]. Jun 04 16:23:03 vvu: yeah, getting people into the IRC channel more often really seems needed. Jun 04 16:23:06 As for the ARM point, the MMC/SD spec and DMA are pretty old-time and not very embedded'ish. We're just dealing with drivers here Jun 04 16:23:15 joel: no offence but my point is that we also dont want to let you down but it might take a while to get through a new peice of code or a new stack Jun 04 16:23:33 may be some of us havnt worked on it Jun 04 16:23:41 we are 3 weeks from midterms: https://www.google-melange.com/gsoc/events/google/gsoc2014 Jun 04 16:23:54 for the kernel, there is really no "embedded" specifics. it is a kernel. it is the same kernel as anything else Jun 04 16:24:26 ds2: except for linus hates us :P Jun 04 16:24:39 mranostay: What I currently have in my head is that read would just read 64K blocks from shared memory, that's all. Jun 04 16:24:41 ds2, joel_: I'm just thinking that we could use some of the TI training materials or something else to try to give a few more examples that might fill in holes. Jun 04 16:24:41 my point i that some havbe worked on x86 enviroment and some on ARM,,,,,a lot of things cahnge when it comes to ARM Jun 04 16:24:42 mranostay: it started like that, it will be like that. live with it :D Jun 04 16:24:45 thats it Jun 04 16:25:07 shooting for some time next week. Jun 04 16:25:21 disdi: on the device driver side, there isn't that much of a difference IF you used the proper APIs on x86 Jun 04 16:25:26 it isn't a silver bullet... hard, round-the-clock work is needed to make progress. Jun 04 16:25:38 now if you did *(int *)RandomISAaddress, then you are SOL Jun 04 16:25:43 with good testing, clear patch sets, good reviews, etc. Jun 04 16:25:53 ds2: sorry but you aint getting my point Jun 04 16:26:08 we should not start now to go away from the scope of this meeting Jun 04 16:26:13 ds2: I am saying not every one has worked on same piece of code Jun 04 16:26:15 disdi: try not to be defensive... Jun 04 16:26:16 vvu : +1 Jun 04 16:26:28 disdi, even I'm not getting your point. There is NOTHING you're doing that's ARM specific at this point Jun 04 16:26:36 disdi: it is hard. Jun 04 16:26:59 vmayoral|pc: how's life writing flight ware? Jun 04 16:27:01 maybe not everyone has sympathy for how hard different tasks are to learn and perform. Jun 04 16:27:06 jkrinder: thanks ... I have nothing else to say Jun 04 16:27:11 ds2: :) Jun 04 16:27:20 ds2: dying to let it fly Jun 04 16:27:29 vmayoral|pc: things seem a bit quiet after the move to the main list Jun 04 16:27:30 ds2: maybe next week i'll get encouraged Jun 04 16:27:38 'k Jun 04 16:27:44 I spent a bit of time this week with the kernel internals, I think a little bit of training could have saved me a little time this week though, but there is progress Jun 04 16:28:01 vmayoral|pc: if you still need kernel drivers, I found some for the MPU6050 Jun 04 16:28:13 jkridner, there is some training material, Im not sure what disdi needs though, I can look up some if I can know what's the gap Jun 04 16:28:21 joel_ : I guess disdi is trying to say students have worked with different parts of the kernel. working with a untouched part is always an uphill task. Jun 04 16:28:30 disdi, what exact kernel training are you in need off? Jun 04 16:28:34 ds2: Thanks :). Jimmy, one of the contributors coded the MPU6K and is testing it now https://github.com/BeaglePilot/BeagleBone-Black-MPU6000-Driver Jun 04 16:28:59 ds2: still i appreciate a link to the MPU6050, probably can refer Jimmy to that one Jun 04 16:29:51 perhaps students can shout out some other general topics they could use some help with? Jun 04 16:29:51 jkridner : I hacked through pantos driver. I'm progressing at a satisfactory speed. Jun 04 16:30:00 joel: I have already been through the hard part.....now I do not require training.... I want to concentrate and meet my project goals Jun 04 16:30:09 ds2: to be honest It's quite exciting. I'm working a lot though. The hardware part is specially tricky (the scope became my bestie) Jun 04 16:30:13 karki: PRU and kernel talking nicely using virtio? Jun 04 16:30:32 I though have an opinion now that the best way to try something new is to take it hands on, that gives more insight than just reading tons of documentation and then coding up something. Jun 04 16:30:33 jkridner : no virtio, shared memory (upcalls and down calls) Jun 04 16:30:35 Time for me to go...email if there are any issues I can help with. Jun 04 16:30:39 karki, I agree. but I disagree that we are doing any rocket science type of work. and I don't think its that hard to pick up something new, the technique matters. I also don't think we are dealing with something very complex here. Jun 04 16:30:54 thanks cd... gone Jun 04 16:31:04 vmayoral|pc: https://github.com/xcstacy/kernel-N8000/blob/master/drivers/input/mpu6050_input.c Jun 04 16:31:12 all I heard was some noise about ARM vs x86, I never heard anything about picking up something new from disdi Jun 04 16:31:23 vmayoral|pc: just if out the sensor_register stuff and you'll be fine Jun 04 16:32:22 joel: sorry if you missed it.... my point was that some people are working on a newer piece of code for themselves atleast so make take time to catch up Jun 04 16:32:24 Abhishek_: yep, hands on is a good way to learn... worse case you do a rev 2 with improvements based on what you did learn Jun 04 16:32:45 joel_ : It took me about 4-5 days to get familiar with panto 's code...... but then things are going fast now! So it's the initial period which is well, slow! Jun 04 16:33:16 5-days is quite a good initial period, we are in week 3 now I believe. Jun 04 16:33:18 jkridner : status updates about the BBB rev C? Jun 04 16:33:26 ds2: I can relate to that ^^ Jun 04 16:33:33 karki: not yet shipped. probably close to the end of the project. :( Jun 04 16:33:38 and add another 4 weeks before GSoC for that Jun 04 16:33:59 ds2: appreciated! Jun 04 16:34:34 vmayoral|pc: feel free to ping me on kernel stuff (still discouraging user land things ;)) Jun 04 16:34:41 jkridner : okay :( Jun 04 16:34:41 feedback done on brainstorming for how to help students speed up coding efforts and resolve doubts? Jun 04 16:35:05 my understanding is that there is still lots of hardware issues, but none are blocking, correct? Jun 04 16:35:15 ds2: :) Jun 04 16:35:30 For me no. Have to finish the platform first :) Jun 04 16:35:33 ds2: what opinion do you have on the results i posted? http://erlerobot.com/blog/beaglepilot-cyclictests-different-kernels/ Jun 04 16:36:13 anyone in favor of additional meeting times (shorter, less critical) as a larger group to help keep discussion about blocking issues occurring more often? Jun 04 16:36:33 jkridner: what about a hangout? Jun 04 16:36:36 jkridner: for me is ok and i think it would be better Jun 04 16:36:52 rseethamraju: I'm guessing your other hardware might be a while as well, let's just move ahead with SPI and I can do some testing here as needed Jun 04 16:36:53 we will know each other via web cam at least Jun 04 16:36:58 DiegoTc: this is kinda tricky from time to time *think so* Jun 04 16:37:18 DiegoTc: Adding a hangout sounds good. Need to pick a different time of day. Perhaps later in the US evening on Monday? Jun 04 16:37:37 vvu: probably a Saturday or Sunday like "Friends - coworkers" Jun 04 16:37:47 might not be as reliable as this. Jun 04 16:37:59 vmayoral|pc, just wondering. for your purposes isn't rt_prio sufficient? Jun 04 16:38:02 Sunday, maybe. I don't like Saturday as it interferes with social time. Jun 04 16:38:03 yep that would work for me, during day i'm from morning until like now at work :) Jun 04 16:38:14 vmayoral|pc: cyclictests do not do I/O to flash, right? Jun 04 16:38:16 alexanderhiam I have an spi sensor from clg I can do that Jun 04 16:38:16 I meant Linux real-time scheduling priorities which don't need an RT kernel Jun 04 16:38:51 jkridner: why not keep it on IRC? Jun 04 16:39:00 ds2: no i think it doesn't, and yeap it might be enough Jun 04 16:39:10 last year IRC has proven to be really good Jun 04 16:39:19 ds2: the main problem here is that i'm lacking of time. I would love to have a couple more weeks just to dedicate to this matter Jun 04 16:39:23 i would encourage more interaction here Jun 04 16:39:26 rseethamraju: ok great Jun 04 16:39:32 vmayoral|pc: beware if things get flushed out to flash at bad times Jun 04 16:39:39 av500m, _av500_ not around ? Jun 04 16:39:49 vmayoral|pc: to be fair, this problem has been plaguing people for as long as I can remember Jun 04 16:39:50 ds2: meant to be a supplement. I'd be happy with also meeting at a different time on IRC, in addition to this one.... just need to keep it short and get into the issues quickly (and let people go not involved in the issues). Jun 04 16:40:19 av500 has been on travel to Asia, last I heard. Jun 04 16:41:03 vvu: bringing problems here all the time at any time of day or night really helped last year. Jun 04 16:41:04 * vmayoral|pc nods Jun 04 16:41:19 * jkridner is just looking for ways to try to encourage that tmore. Jun 04 16:41:25 * Abhishek_ nods Jun 04 16:41:42 yes, please bring problems here Jun 04 16:41:49 jkridner: are there any news regarding the Google Meeting and the chance of assisting as an ambassor from BeagleBoard? Jun 04 16:42:25 * jkridner reminds vmayoral|pc, disdi, karki, DiegoTc, Abhishek_, rseethamraju to use nick names to ping people Jun 04 16:42:40 sure Jun 04 16:42:49 ok Jun 04 16:43:02 ok Jun 04 16:43:10 having names of people who might know some answers helps them realize they should look at the traffic. Jun 04 16:43:20 just a ping without a question is bad. Jun 04 16:43:37 you need to get the question in there and leave the window open a couple hours looking for a response. Jun 04 16:44:00 also try if you guys can to set up headless irssi to keep you open all the time Jun 04 16:44:02 good to ping on IRC even you've sent an e-mail. DiegoTc has certainly learned that about me. :-) Jun 04 16:44:29 i remember how annoying was when somebody pinged me and i was not online and neede to check logs Jun 04 16:44:50 if you are not sure about who, a nameless ping is still better then sitting there not doing anything Jun 04 16:45:03 you can also try #beagle for general questions if there's no one in here Jun 04 16:45:05 You learn something new daily Jun 04 16:45:09 alexanderhiam: +1 Jun 04 16:45:29 I use IRCCloud and (try to) keep it always online through my phone. Jun 04 16:45:30 and always feel free to bring in other experts. Jun 04 16:46:31 k, I'll end the official meeting and allow the unofficial meeting to continue to solve "real" issues. Jun 04 16:46:34 * jkridner sounds gavel. Jun 04 16:46:55 DiegoTc: let me know when you commit the stuff you've made... guess being careful not to push secure credentials. Jun 04 16:47:35 ok. I will work on that tonight. Thanks for your help Jun 04 16:47:37 karki: I can't wait to play with libsigrok on Beagle! Jun 04 16:47:46 jkrinder: thank you for your time and suppport Jun 04 16:47:53 er, Abhishek_ Jun 04 16:48:00 hehe Jun 04 16:48:05 er, me keeps getting all the projects confused. :-) Jun 04 16:48:55 karki: will you start getting an interpreter up soon? Jun 04 16:48:59 panto : the problem is the the 2.0 compiler ABI, so I was jumping back to the start of the sc_downcall rather than continuing, I fixed it (now scno takes value 255 the second time), but another small problem remains... Jun 04 16:49:12 jkridner : I should have one by tonite :) Jun 04 16:49:13 karki: I'm still confused why you are using memory mapped I/O instead of virtio still. Jun 04 16:49:58 jkridner : panto 's advice! mmap is faster and cleaner Jun 04 16:50:06 karki: k. Jun 04 16:50:16 karki what is ur github url for the project ? Jun 04 16:50:22 karki: but, don't you want a virtual serial like interface for the interpreter? Jun 04 16:50:49 karki, panto: do you have thoughts about when to meet with Chris and the students at Tufts? Jun 04 16:51:19 jkridner, probably by next weekend [I know deadlines can be messed up sometimes but this is a best estimate]. Jun 04 16:51:25 jkridner : I have mentioned meeting chris at a part of my TODO for this week! (in the google group) Jun 04 16:51:26 err, you mean on irc? Jun 04 16:51:36 tufts? isn't that a geologic formation thingie? Jun 04 16:51:59 yeah, the virtio serial port seemed like a nice easy way to be able to have bindings in other userspace languages Jun 04 16:52:19 alexanderhiam jkridner : sysfs! Jun 04 16:52:45 oh right, now I remeber you were figuring that out Jun 04 16:52:50 ds2: Tufts University in Boston. Jun 04 16:53:32 karki: so, are you going to have a simple sysfs interface for messages soon? Jun 04 16:53:37 alexanderhiam jkridner I'm creating binary sysfs attributes that can be mmap'd into user space , for fast, clean, direct and light interaction b/w userspace and PRU Jun 04 16:53:39 peeps i'm out! have fun! Jun 04 16:53:49 bye vvu Jun 04 16:53:53 bye vvu Jun 04 16:54:08 jkridner : I already have it :) I'm able to control pins from userspace Jun 04 16:54:09 karki: I like the sound of that Jun 04 16:54:15 karki: good for doing character terminated serial data streams? Jun 04 16:55:15 jkridner : I didn't quite get you. Jun 04 16:55:24 where would I be using that? Jun 04 16:55:47 oh Jun 04 16:55:48 karki: when streaming script commands to the interpreter. Jun 04 16:56:41 jkridner : I'll be writing the commands in the shared area and issuing a down call telling the interpreter that the instructions are ready to be read Jun 04 16:56:52 all right then, i'll take my leave now. Ciao ;)! Jun 04 16:57:07 karki: one instruction at a time, or the whole script? Jun 04 16:57:19 alexanderhiam : whole script Jun 04 16:57:27 much faster this way Jun 04 16:57:30 ok Jun 04 16:57:45 disdi: if you got kernel questions, please ask them on IRC... there are quite a few of us that can help there Jun 04 16:58:02 karki: so who writes the script into shared memory? Jun 04 16:58:36 karki: k, that is a bit different than how it was originally envisioned, but might be OK. what happens if you only get a partial script at a time? do you copy it elsewhere to build a script? Jun 04 16:58:41 alexanderhiam : the userspace PRU library. Jun 04 16:58:52 I find it weird though when I find out that virtio is used by virtualizers like KVM to get gigabit thorughput on virtual devices yet here mmaps are faster. Jun 04 17:00:02 how could virtio be faster than mmap in any case? mmap is the most simple and basic thing. only single write and single read Jun 04 17:00:25 depends on what kind of mapping you need Jun 04 17:00:37 isn't virtio just a ring buffer on top of shared memory? Jun 04 17:00:38 because with virtio you don't have to do data movement Jun 04 17:01:06 don't you pay a price with the tlbs/caches in virtio? Jun 04 17:01:10 jkridner : me and panto discussed about this a week and a half ago. I'll make a wiki, hopefully it will answer your questions Jun 04 17:01:25 k. Jun 04 17:01:26 ds2, still beats copying data Jun 04 17:01:39 gotta go Bye! Jun 04 17:01:46 bye rseethamraju Jun 04 17:01:57 panto: are you liking the resulting architecture flexibility? Jun 04 17:02:19 jkridner, it is going to the right direction Jun 04 17:02:21 panto: is it good enough you'd ever consider using the resulting interpreter, or just some aspect of the framework? Jun 04 17:02:59 well, potentially you don't even need an interpreter Jun 04 17:03:14 you could even compile/jit PRU code and just execute it Jun 04 17:03:49 I like the way it might be a general digital I/O description format Jun 04 17:03:56 but we still have ways to go Jun 04 17:04:17 well, I'd hope the framework would make it easier to create some higher level functions and just call them when needed, rather than sending down entire code chunks. Jun 04 17:04:39 panto: k Jun 04 17:04:43 jkridner : Don't worry it will :) Jun 04 17:05:02 JIT? JVM/PRU project for next year? ;) Jun 04 17:05:16 I like that newbs can start sending simple commands, yet it is something you'd want to extend as a "pro". Jun 04 17:05:23 JVM? lol! Jun 04 17:05:34 but JIT is certainly possible Jun 04 17:05:38 ds2: espruino? ;-) Jun 04 17:05:59 put down the crack pipe please :) Jun 04 17:06:02 lol Jun 04 17:06:18 you can't even come close to java with so much limited instruction code space Jun 04 17:06:33 let me find that link Jun 04 17:07:34 panto: http://dmitry.gr/index.php?r=05.Projects&proj=12.%20uJ%20-%20a%20micro%20JVM Jun 04 17:07:47 done already...need to port to PRU :D Jun 04 17:08:21 oh wow Jun 04 17:08:26 ds2, still crack pipe territory :) Jun 04 17:08:57 this guy is the crazy one who ran linux on an 8Bit MCu Jun 04 17:09:03 Java is so 2000. Today is all about JavaScript. ;-) Jun 04 17:09:07 hahah Jun 04 17:09:11 The way I would like the PRU is to have a full set of I/O libraries (serial, parallel I/O, transparent data capture [what I am currently trying to do with BeagleLogic], even framebuffer or cams) with the PRU instead of just libprussdrv or remoteproc with kernel and appropriate userspace infrastructure with a port to say Python or Node.JS . That would be way Jun 04 17:09:11 better Jun 04 17:09:24 guys, guys Jun 04 17:09:37 swift! Jun 04 17:09:44 have you guys had a look at micro python? Jun 04 17:09:54 oh no, invasion of AWWDC Jun 04 17:10:13 karki: http://embeddedpython.org/index.php?title=The_Owl_Embedded_Python_System Jun 04 17:10:55 Ah, looks like micropython picks up from here and slaps it on the STM32 Jun 04 17:11:15 I could use my Python library to generate code to run in a Python interpreter on the PRU! It's perfect! Jun 04 17:11:18 jkridner : http://micropython.org/ Jun 04 17:11:23 panto: I appreciate your timely comment Jun 04 17:11:47 alexanderhiam : that was my job right ? Jun 04 17:12:02 karki: yeah, but I know people working on the http://embeddedpython.org project. :) Jun 04 17:12:03 :P Jun 04 17:13:11 * jkridner must head out Jun 04 17:13:27 I don't get why not just have a BotSpeak --> PRU assembly generator, and call the assembler and load on the fly :p Jun 04 17:13:42 karki: why not? :P Jun 04 17:13:45 we discussed that a couple of days ago Jun 04 17:13:55 thanks jkridner Jun 04 17:14:16 I seriously don't get why :p (though there are small advantages using an interpreter ) Jun 04 17:14:21 * Abhishek_ waves to jkridner Jun 04 17:15:56 the next big thing would be to port some sort of JS to microcontrollers. I like the idea of being able to program my STM32 boards with that, though the BBB should be enough though ;) Jun 04 17:17:14 Abhishek_, if you want to burn power while doing nothing just disconnect the BBB and put a nice big resistor between power and ground :) Jun 04 17:18:33 Better to replace my laptop with a bigger one though ;) Jun 04 17:18:50 a bigger resistor Jun 04 17:20:38