**** BEGIN LOGGING AT Wed May 28 02:59:59 2014
May 28 03:17:02 <karki> panto : from where did you get "pruss_dt_ids" here : https://github.com/beagleboard/linux/blob/3.8/drivers/remoteproc/pru_rproc.c#L2808 ?
May 28 04:27:15 <karki> panto : I got a basic platform driver ready. doesn't do anything as of now. but successfully binds to the platform code and probe gets called when I load the DTS for pru speak.
May 28 05:18:44 <karki> panto : so the platform driver is working perfectly, to sysfs now?
May 28 05:19:18 <karki> I'll try bin files and try to implement mmap for them.
May 28 05:44:58 <ds2> \
May 28 05:57:06 <Abhishek_> morning
May 28 06:00:36 <Abhishek_> av500: I still don't understand why calling the function directly leads to a segfault, whereas calling it through a pointer doesn't.
May 28 06:03:22 <karki> Abhishek_ : can I have a look at the code
May 28 06:03:24 <karki> ?
May 28 06:06:40 <Abhishek_> see yesterday's log
May 28 06:10:50 <_av500_> Abhishek_: if it corrupts the stack, that might make a difference
May 28 06:11:00 <_av500_> Abhishek_: pastebin with code?
May 28 06:11:27 <Abhishek_> _av500_: https://gist.github.com/abhishek-kakkar/e2a4a0aefd7d1e7fc15d
May 28 06:13:52 <_av500_> Abhishek_: can you comment the actual vst ops
May 28 06:13:56 <_av500_> o you only read
May 28 06:14:17 <_av500_> I think this is nothing aobut memcpy but you mangle the stack somehow when just caling the function
May 28 06:14:21 <_av500_> softfp vs hard
May 28 06:17:13 <Abhishek_> _av500_: Commented the vst's. still segfaults
May 28 06:19:38 <_av500_> see
May 28 06:19:51 <_av500_> so you need to check what changed in the ABI
May 28 06:19:56 <_av500_> and rewrite that part
May 28 06:39:29 <ds2> Abhishek_: yes but "write" depends on how the hardware is setup... you want a 128bit burst which may not necessarily mean you "write" 128bits from a SW stand point
May 28 06:39:53 <ds2> the other problem with the DDR is there are refresh cycles and burst patterns along with latency issues if you write things in the wrong order
May 28 06:40:22 <Abhishek_> ds2: Currently I write in 256bit bursts (32 bytes)
May 28 06:40:38 <ds2> from the ARM or PRU?
May 28 06:40:47 <Abhishek_> from the PRU into DDR
May 28 06:41:01 <ds2> writting to DDR is dangerous
May 28 06:41:25 <ds2> you can run into refresh
May 28 06:41:51 <Abhishek_> In my tests it didn't seem to, even I overwrote the same DDR area 500 times.
May 28 06:42:01 <ds2> depends on memory timing
May 28 06:43:07 <ds2> you can get very varied timing with the DDR since L3 traffic is also a factor
May 28 06:43:26 <Abhishek_> ds2: In this case, the memcpy on the ARM core becomes very slow
May 28 06:43:59 <Abhishek_> ds2: have you read Sunday's email?
May 28 06:44:07 <ds2> yes
May 28 06:44:19 <ds2> I was on a slow link
May 28 06:44:41 <ds2> do things in kernel
May 28 06:45:18 <ds2> or you canget cache/paging issues
May 28 06:49:29 <ds2> the DDR resource is being slammed by the ARM (instruction/data), the video output, and the PRU
May 28 06:49:32 <ds2> so of course things get slow
May 28 06:49:48 <ds2> and this is all coming through the L3...on the ARM, the caches matter a lot
May 28 06:50:17 <ds2> once you get it into the kernel, playing with the cache settings may have effects
May 28 06:50:22 <ds2> but in userspace, you are kind of doomed
May 28 06:50:34 <Abhishek_> IIRC the UIO reserved area isn't cached.
May 28 06:50:55 <ds2> but does it allow for stuff like buffering, etc?
May 28 06:51:05 <ds2> i think the terms are something like posted reads, etc
May 28 06:51:23 <Abhishek_> I am not aware of that
May 28 06:51:27 <ds2> another thing to try is to remove all video support and see if things go faster
May 28 06:51:49 <Abhishek_> that is already done: The input pins are the same HDMI pins
May 28 06:51:51 <ds2> the ARM core's interface to the outside world can be buffered in various ways for preformance reasons
May 28 06:51:54 <ds2> no no
May 28 06:52:14 <ds2> if the video driver is running, the hw may still be slamming the bus for framebuffer access even though it never leaves the chip
May 28 06:52:20 <ds2> i.e. it is still eatting memory bw
May 28 06:52:41 <ds2> video support here means software, not HW
May 28 06:52:58 <Abhishek_> hmm, you mean disable usability through VNC?
May 28 06:53:17 <ds2> vnc shouldn't care about it
May 28 06:53:50 <Abhishek_> k
May 28 06:54:14 <ds2> ypu want /dev/fb0 gone
May 28 06:55:37 <Abhishek_> ds2: There's no /dev/fb0 on my BBB system
May 28 06:55:46 <ds2> ok
May 28 06:59:32 <Abhishek_> ds2 mranostay: To build kernel drivers, there would be rejuggling some aspects of work till midterm (week 5). The original plan was to have the core ready by week 2-2.5, so the LA core and back-end server work will have to be merged till the target of midterm.
May 28 07:00:44 <ds2> how fatal is the current memory bw limitation you are running into?
May 28 07:01:36 <Abhishek_> even at 25 MHz (83ms interrupts), the memcpy to remove data from the PRU shared area takes 140ms
May 28 07:02:24 <ds2> hmmm I am very curious if you burst to the SRAM (it is small), if things would appear faster... this is more to show there is a DDR bottle neck then anything else
May 28 07:02:49 <Abhishek_> ds2: It isn't a DDR bottleneck, the interrupt slamming rate is more of a bottleneck
May 28 07:03:04 <ds2> oh..
May 28 07:03:11 <ds2> then you definitely need kernel side
May 28 07:03:21 <ds2> interrupts in userland is painful at best
May 28 07:03:25 <ds2> hmmm
May 28 07:04:38 <ds2> if mranostay is good with that, I can along the sched change
May 28 07:05:46 <Abhishek_> ds2: Smaller memory, memcpy is fast but the interrupts come in faster. More memory, interrupts come in slower but memcpy limits the rate then
May 28 07:06:30 <Abhishek_> I experimented with transfer sizes yesterday and noticed this
May 28 07:06:32 <ds2> interrupts to userland never really worked
May 28 07:06:43 <ds2> hence my strong prefernce for proper kernel drivers
May 28 07:07:09 <panto> ds2, +1
May 28 07:07:25 <panto> don't have more time to expand on that today unfortunately
May 28 07:07:36 <Abhishek_> panto: I was about to ping you and tell your scatter-gather point right now :)
May 28 07:09:47 <panto> heh, I'm sure ds2 gets the point :)
May 28 07:13:48 <ds2> :D
May 28 07:13:55 <ds2> userland == nothing works
May 28 07:16:26 <Abhishek_> ds2: a SBBO to SRAM or DDR from the PRU, in small bursts, takes 1 extra cycle per word
May 28 07:16:39 <ds2> depends on L3 traffic
May 28 07:17:29 <Abhishek_> yep, mine was an idle BBB so it wasn't an issue
May 28 07:19:25 <Abhishek_> the new PRU C Compiler is also making things a little difficult
May 28 07:21:08 <ds2> forget about C
May 28 07:21:11 <ds2> do it all in assembly
May 28 07:21:18 <ds2> you are running things pretty tight anyways
May 28 07:21:53 <Abhishek_> yep, that's not an issue for me. the vrings and stuff in the remoteproc implementation concerns me
May 28 07:22:39 <Abhishek_> I need to understand more on how upcalls and downcalls work with the new implementation.
May 28 07:36:09 * Abhishek_ is thinking what kind of device driver should the PRU buffer expose
May 28 07:38:49 <ds2> a simple character device might be enough
May 28 07:38:59 <ds2> and just a rule that all reads must be the buffer size
May 28 07:39:13 <ds2> keep like 2-3 PRU sized buffers in the main memory
May 28 07:39:17 <_av500_> if reading less, throw away the rest
May 28 07:40:13 <Abhishek_> when I consulted in #sigrok about this, a zero copy approach was highlighted
May 28 07:40:40 <panto> ds2, as long as you don't copy
May 28 07:40:49 <panto> you will need a mmap interface
May 28 07:40:59 <panto> and then you'll need cache coherency on top of that
May 28 07:41:15 <panto> got to shut up now, got work to do :)
May 28 07:41:24 <ds2> :)
May 28 07:41:37 <ds2> even if you copy, it'll still be better then userland crap
May 28 07:41:45 <panto> agreed there
May 28 07:41:49 <ds2> <-- hates doing things in userland
May 28 07:41:58 <_av500_> ds2: kernel irc daemon?
May 28 07:42:30 <Abhishek_> before coming into eLinux, I was working with STM32 and ChibiOS
May 28 07:43:10 <ds2> sure
May 28 07:56:20 <Abhishek_> How do the ethernet drivers handle data coming in at such high speeds? IIRC AM335x does support gigabit Ethernet but there's only a 10/100 PHY
May 28 07:56:57 <panto> skbs
May 28 07:57:16 <Abhishek_> Currently that's the closest analogue I'm thinking of, the bandwidth is quite similar too [50 MHz @ 8 bits = 400mbps]
May 28 08:03:24 <Abhishek_> skbs looks interesting, I wonder if the LA can directly emit TCP packets that can be streamed over the network.
May 28 08:03:37 <panto> wut, no!
May 28 08:04:19 <panto> got to be afk now, but try to keep it sane :)
May 28 08:04:38 <Abhishek_> k
May 28 08:11:46 <ds2> dma descriptor
May 28 08:31:59 <Abhishek_> panto:  were you talking about the use of dma_map_sg() ?
May 28 08:32:18 <panto> it's one of the components
May 28 08:33:30 <Abhishek_> k
May 28 12:15:35 <jkridner> hi _av500_
May 28 12:17:34 <karki_> jkridner : do I give my week 2 status as a separate post, or could I just reply to my Week 1 update?
May 28 12:18:23 <jkridner> replying to previous post is nice to save people getting a newly started thread. thanks for being considerate.
May 28 12:18:53 <Abhishek_> morning jkridner
May 28 12:19:26 <jkridner> hi Abhishek_
May 28 12:19:38 * jkridner has some conflicts today.
May 28 12:19:49 * jkridner is looking to see if av500 can run our meeting.
May 28 12:20:13 * jkridner also needs to spend some time catching up with DeigoTc.
May 28 12:21:39 <Abhishek_> jkridner: The issue is the same as mranostay's, his code needs to be tested by me now [see status report], I'm hitting the limits of the UIO PRU driver
May 28 12:22:43 <Abhishek_> Though my code will not be necessarily written in C, it would be in ASM only
May 28 13:06:24 <karki_> hmm...
May 28 13:07:14 <karki_> it's nice to see how this channel fills up on wednesday ;)
May 28 13:07:34 <Abhishek_> the RNDIS link manages ~82 Mbps [iperf]
May 28 13:08:04 <jkridner> Abhishek_: k
May 28 13:10:51 <vvu> Abhishek_: i'm curious on USB3.0 how much will go
May 28 13:11:23 <Abhishek_> the BBB doesn't seem to have one
May 28 13:12:06 <vvu> yeah but still would be nice to see some specs
May 28 13:13:02 <Abhishek_> yeah, it should clock more than 1 Gbps [only estimate].
May 28 13:13:30 * jkridner wonders if it is commute time for _av500_
May 28 13:13:52 <Abhishek_> jkridner: Would USB 3.0 come up in further iterations of BeagleBoards?
May 28 13:19:47 <jkridner> yes, most likely.
May 28 13:20:19 <jkridner> GigE would probably happen first given that AM3358 already has a GigE controller.
May 28 13:20:34 * vvu subscribed to get a TRE dev edition
May 28 13:20:54 <jkridner> but USB3 is a highly requested feature
May 28 13:21:08 <vmayoral|pc> jkridner: good morning, just answered your mail about where the discussion is going on
May 28 13:25:54 <jkridner> thanks vmayoral|pc
May 28 13:28:59 * karki_ wonders how vvu subscribed for the TRE......
May 28 13:29:14 * karki_ wants a TRE too.....
May 28 13:29:49 <vvu> karki_: http://store.arduino.cc/index.php?main_page=product_info&cPath=11&products_id=534
May 28 13:31:28 <karki_> vvu : the price :'D
May 28 13:31:56 <Abhishek_> I'd buy a wandboard instead
May 28 13:32:13 <DiegoTc> whoa jkrinder left
May 28 13:32:18 * karki_ wonders if there is a way to get his hands on a complementary dev edition!
May 28 13:32:35 <karki_> maybe jkridner can help....
May 28 13:40:43 <karki_> I wanted to buy a wandboard, bought a BBB instead a few months back. not a bad investment at all ;)
May 28 13:52:54 <DiegoTc> Hi VoltVisionSteve
May 28 13:52:58 <DiegoTc> just read your email
May 28 13:53:02 <DiegoTc> thanks
May 28 13:54:03 <VoltVisionSteve> DiegoTc, yes.  How is it going?
May 28 13:54:58 <DiegoTc> personal life good(got some problems but resolve). Project got some issues. Here is the report of this week; http://diegoturcios.blogspot.com/2014/05/bone101-status-week-2.html
May 28 14:14:31 <VoltVisionSteve> DiegoTc, cool. Looks good.  By the way, do you have a Wii controller available to you?
May 28 14:17:26 <DiegoTc> yes, well my girl friend has
May 28 14:17:32 <DiegoTc> so I do have also :p
May 28 14:17:51 <Abhishek_> :D
May 28 14:22:41 <DiegoTc> VoltVisionSteve: have in mind a cool projetc?
May 28 14:26:32 <VoltVisionSteve> DiegoTc, I have been using a Wii controller for live music performance lately where I can change songs/lights interactively.  I can make a video of it this week.  I am working with someone now to port it to BBB using a cheap Bluetooth dongle.
May 28 14:27:51 <VoltVisionSteve> ...my favorite is being able to push the "1 button" and rotate to change the volume...then let go of the "1button".  It responds to relative changes...pretty cool.
May 28 14:29:10 <VoltVisionSteve> I could put this up as one of the Bluetooth tutorials.
May 28 14:47:00 <Abhishek_> jkridner: Where did the Flot Demo go?
May 28 14:49:01 <DiegoTc> cool VoltVisionSteve, waiting to see the video
May 28 15:08:24 <karki_> panto : did you fill in the hardware requirements form against my name?
May 28 15:30:43 <praveendath92> vvu: ping
May 28 15:47:58 <karki_> meeting in a few minutes?
May 28 15:48:07 <jkridner> yes, 10.
May 28 15:48:31 * jkridner has limited connectivity at the moment.
May 28 15:48:39 <Abhishek_> jkridner: Where did the flot demo in Bone101 go? I am not able to find it
May 28 15:49:05 <jkridner> Abhishek_: should be on http://jsfiddle.net/user/jkridner/fiddles/
May 28 15:49:15 <Abhishek_> ok, will check.
May 28 15:49:23 <disdi> hi all
May 28 15:50:41 <jkridner> http://jsfiddle.net/jkridner/rC4AD/
May 28 15:50:44 <jkridner> hi disdi
May 28 15:51:05 <Abhishek_> thanks jkridner
May 28 15:51:17 <disdi> jkrinder: hi
May 28 15:52:02 <disdi> jkrinder: I guess joel wont be present today since he is travelling.
May 28 15:52:28 <disdi> jkrinder: we will be having our weekly  meeting this friday
May 28 15:53:23 <jkridner> disdi: thanks.
May 28 15:53:47 <jkridner> disdi: glad to hear you got your board.
May 28 15:54:13 <DiegoTc> hi jkridner, sorry for dissapearing this week, got some problems
May 28 15:54:42 <jkridner> DiegoTc: hope nothing too serious. starting to work out the github and flickr apis now?
May 28 15:55:11 <disdi> jkrinder: yes.... it would not have been possible without help from cathy and others
May 28 15:55:21 <disdi> jkrinder: including you
May 28 15:55:38 <DiegoTc> jkridner: Github yes. Flickr wanting to confirm if we're going to use it
May 28 15:56:14 <disdi> I am maintaing my wiki and source repo at github
May 28 15:56:17 <jkridner> I think it is most practical. haven't heard what objections might be, other than needing an additional account.
May 28 15:57:22 <disdi> I havent found anything easier to raise pull requests , commits and review for your code
May 28 15:57:43 <jkridner> DiegoTc: don't know if Google Plus is any better for an image repo?
May 28 15:57:52 <vmayoral|pc> hi everyone
May 28 15:58:13 <DiegoTc> i don't know the amot of space
May 28 15:58:26 <praveendath92> Hello vmayoral|pc
May 28 15:58:37 <DiegoTc> I will compare and make a survey, let the community choose. What do you think?
May 28 15:58:49 <jkridner> DiegoTc: just found https://github.com/hecticjeff/gist-img
May 28 15:58:51 <vmayoral|pc> praveendath92: how is it going? any big issues?
May 28 15:58:53 <Abhishek_> hello everyone
May 28 15:59:04 <vmayoral|pc> greetings Abhishek_
May 28 15:59:27 <jkridner> DiegoTc: I suspect it doesn't work anymore.
May 28 16:00:06 <praveendath92> vmayoral|pc: Nothing big as of now but I must admit it's going slow currently.
May 28 16:00:33 <praveendath92> I have a trip next week and had to do some work related to that also so.
May 28 16:00:34 <DiegoTc> jkridner: and only 50k size
May 28 16:00:57 <jkridner> DiegoTc: right. :(
May 28 16:01:07 <praveendath92> vmayoral|pc: You seem to make very good progress.
May 28 16:01:10 <jkridner> converting the image to a DataURI seems to also be an approach.
May 28 16:01:25 <vmayoral|pc> praveendath92: nah, feeling the same as you. Slower than I'd like
May 28 16:01:44 <jkridner> k, seems to be time, so let me call attendance and move forward with our weekly meeting. :-)
May 28 16:01:46 <praveendath92> Seriously?
May 28 16:02:04 <praveendath92> I mean, I take motivation from your github activity actually.
May 28 16:02:08 <DiegoTc> I try it and fail. (Probably didn't try enough)
May 28 16:02:33 <vmayoral|pc> praveendath92: haha glad to hear that. Last year i realized that it takes a bit to get comfortable with your project
May 28 16:02:33 <jkridner> hi DiegoTc, praveendath92, vmayoral|pc, disdi, Abhishek_, karki_.
May 28 16:02:46 <vmayoral|pc> praveendath92: then your efficiency boosts up :)
May 28 16:02:53 <disdi> jkrinder: hi
May 28 16:03:15 <karki_> jkridner : Hi
May 28 16:03:18 <praveendath92> Yeah. Hopefully :)
May 28 16:03:23 <praveendath92> Hello jkridner
May 28 16:03:25 <vmayoral|pc> jkridner: hi
May 28 16:03:54 <DiegoTc> Hi
May 28 16:04:05 <Abhishek_> hi
May 28 16:04:09 <VoltVisionSteve> hello!
May 28 16:04:37 <alexanderhiam> ahoy all
May 28 16:04:39 <karki_> jkridner : all is going well on my side, I've started coding the driver. just wanted to know whether you have my entry in the h/w requirements list?
May 28 16:04:46 <karki_> alexanderhiam : Hi :D
May 28 16:05:01 <alexanderhiam> hello karki_
May 28 16:05:19 <VoltVisionSteve> hello rseethamraju !
May 28 16:05:29 <rseethamraju> Hello!
May 28 16:05:34 <Abhishek_> hello rseethamraju
May 28 16:05:39 <rseethamraju> how late am I>
May 28 16:05:41 <rseethamraju> ?
May 28 16:05:49 <rseethamraju> hey everyone!!
May 28 16:05:49 <praveendath92> Not so late really.
May 28 16:05:55 <rseethamraju> ok
May 28 16:05:56 <praveendath92> Just the hello's so far.
May 28 16:05:58 <Abhishek_> just about to begin
May 28 16:06:00 <karki_> alexanderhiam : finally figures the event to interrupt mapping. ( It's in the dev tree and driver ;) ) I had to figure it out since I'm coding the driver now :P
May 28 16:06:27 <alexanderhiam> oh awesome
May 28 16:06:36 <karki_> busy week?
May 28 16:06:44 <Abhishek_> karki_: Even the firmware loading is through DT :)
May 28 16:06:48 <karki_> didn't see you very often?
May 28 16:07:01 <karki_> Abhishek_ : I know.
May 28 16:07:11 <alexanderhiam> karki_:  me?
May 28 16:07:31 <karki_> Abhishek_ : but still have to figure out how the driver takes care of the loading :P
May 28 16:07:37 <karki_> alexanderhiam : yes
May 28 16:08:15 <alexanderhiam> yeah, very busy week!
May 28 16:08:50 <karki_> Abhishek_ : help me figure the firm ware loader bit if you are already done. (will save me my effort ;) )
May 28 16:09:20 <karki_> rseethamraju : when do your exams start?
May 28 16:09:22 <Abhishek_> karki_: It parses the ELF, from what I have read.
May 28 16:09:32 <rseethamraju> next wed
May 28 16:09:38 <rseethamraju> =(
May 28 16:09:43 <praveendath92> That's sad.
May 28 16:09:54 <praveendath92> Who else still have exams?
May 28 16:09:59 <jkridner> anyone seen rseethamraju?
May 28 16:10:00 <karki_> Abhishek_ : Too many static DS all around for my liking.....
May 28 16:10:05 <jkridner> grrr... did I mention my connection was bad today?
May 28 16:10:07 * jkridner will try to keep things brief...
May 28 16:10:12 <rseethamraju> ya hey
May 28 16:10:33 * karki_ feels for rseethamraju :(
May 28 16:11:01 <karki_> rseethamraju : don't worry too much about it, it'll be gone sooner than you think
May 28 16:11:02 <rseethamraju> jkridner: updating the status thing. Sorry I forgot
May 28 16:11:05 <ds2> hmm
May 28 16:11:07 <Abhishek_> rseethamraju: are you done with the prep or one night before?
May 28 16:11:19 <jkridner> hi panto, av500, alexanderhiam, tridge and mranostay and any other mentors around.
May 28 16:11:21 <jkridner> hi ds2
May 28 16:11:26 <rseethamraju> karki_: one night b4
May 28 16:11:39 <rseethamraju> started yday so
May 28 16:11:46 <jkridner> rseethamraju: yes, thanks for sending weekly status to the beagleboard-gsoc list.
May 28 16:11:46 <alexanderhiam> hey jkridner
May 28 16:12:04 <karki_> panto is busy today.
May 28 16:12:33 * jkridner has a very unreliable connection, but wants to make sure no one has immediate blocking needs for hardware, etc. that I should be solving. seems disdi was finally able to get a board.
May 28 16:13:06 <praveendath92> jkridner: I received mine too. Long back actually.
May 28 16:13:33 <jkridner> disdi: any update on making a video to intro your project to complete the "collection"?
May 28 16:13:33 <disdi> jkrinder: yes ... i got my board
May 28 16:13:45 <karki_> jkridner : all okay from my side :) but want to know if you have my entry in the h/w list! did panto enter the details?
May 28 16:13:51 <Abhishek_> jkridner: I understand that my board is delayed, but just let me know whenever it ships.
May 28 16:13:57 <disdi> yes I am just finishing it....was finishing my setup for BBB today
May 28 16:14:14 <disdi> would be up with the video in a day or two
May 28 16:14:42 <jkridner> karki_: no, I don't see the details.
May 28 16:15:02 <karki_> jkridner : hmmm...... can someone else enter it for me?
May 28 16:15:20 <karki_> panto : seems to be *very* busy today
May 28 16:16:07 <jkridner> ds2, alexanderhiam: do either of you want to take that on to enter the info? I think we have waited a while now and should jump in.
May 28 16:16:08 <karki_> jkridner : don't know when he'll be free, so could someone else fill it up? AFAIK I only need BBB rev C
May 28 16:16:28 <alexanderhiam> sure I caan do that
May 28 16:16:38 <karki_> alexanderhiam : Thanks :)
May 28 16:16:40 <jkridner> karki_: you can send info directly to me if alexanderhiam or ds2 can't input it sooner.
May 28 16:16:44 <jkridner> thanks alexanderhiam
May 28 16:17:09 * jkridner confirms alexanderhiam is on the editors list.
May 28 16:17:51 <jkridner> karki_: Chris Rogers is interested in having a weekly chat with you regarding the project. I think it'd be helpful.
May 28 16:17:54 <rseethamraju> I got my board too today but it got delevered to college so I’ll go pick it up tomorrow
May 28 16:18:14 <alexanderhiam> karki_: email me the list?
May 28 16:18:22 <karki_> jkridner : sure
May 28 16:18:33 <karki_> alexanderhiam : what details do you need?
May 28 16:19:00 <Abhishek_> ds2: did you hear from mranostay?
May 28 16:19:38 <ds2> Abhishek_: sorry, nope. but I was sleeping just 20 mins ago :D
May 28 16:19:49 <ds2> he might be too, we are both in the same timezone
May 28 16:19:58 * jkridner saw PRU C compiler issue (address reach?) from mranostay and karki_
May 28 16:20:41 <alexanderhiam> karki: everything :P address, hw needs, shirt size...
May 28 16:20:44 <jkridner> er, Abhishek_, not karki_
May 28 16:21:03 <jkridner> Abhishek_: that isn't a blocker, is it? just a maximum capture-size issue?
May 28 16:21:19 <disdi> ds2: same here, IST
May 28 16:21:24 <karki_> thought so, I don't remember any C compiler complaints.....
May 28 16:21:36 <jkridner> praveendath92: is this error code -110 a blocker issue?
May 28 16:21:49 <jkridner> yeah, sorry karki_... was just thinking PRU, but wrong project.
May 28 16:22:09 <ds2> what project is praveendath92?
May 28 16:22:18 <Abhishek_> jkridner: The C compiler is not a blocker for me, but the maximum capture size is a blocker for continuous capture
May 28 16:22:20 <ds2> -110 sounds familar in a bad way
May 28 16:22:29 <jkridner> the issue was with BeagleLogic being able to address over 8M or something like that.
May 28 16:22:41 <Abhishek_> jkridner: yep
May 28 16:22:44 <karki_> alexanderhiam : sure. (I don't really know what all you'd need! but as of now address, shirt size, h/w )
May 28 16:22:50 <jkridner> ds2: praveendath92 is working on BARD (Android remote display)
May 28 16:23:15 <praveendath92> jkridner: Not sure.
May 28 16:23:28 <praveendath92> I couldn't find a documentation of the error codes
May 28 16:23:33 * karki_ particularly likes the BARD project, will be hacking on it sometime after the GSoC.
May 28 16:23:47 <praveendath92> I was referring https://www.kernel.org/doc/htmldocs/usb/API-usb-control-msg.html
May 28 16:23:54 <Abhishek_> jkridner: Might have to go into kernel to unblock the 8 M barrier
May 28 16:24:35 <Abhishek_> was discussing with ds2 this morning
May 28 16:24:36 <jkridner> vmayoral: i see discussions here (https://groups.google.com/forum/#!topic/beaglepilot/7DKcdm0AEPo), but I'd want to get the linux-omap community involved in such an issue and escalate to TI if no quick resolution.
May 28 16:25:13 <ds2> praveendath92: you can look at the other side's source
May 28 16:25:18 <vmayoral> jkridner: np
May 28 16:25:30 <vmayoral> jkridner: the dicussion continues here https://groups.google.com/forum/#!topic/drones-discuss/7DKcdm0AEPo
May 28 16:25:35 <ds2> for a general info, look at /usr/include/asm-generic/errno.h
May 28 16:25:57 <praveendath92> ds2: Thanks.
May 28 16:26:00 <ds2> Abhishey_: heh... from my point of view, that was last night :D
May 28 16:26:09 <vmayoral> jkridner: i will now forward the message to  linux-omap@vger.kernel.org.
May 28 16:26:24 <Abhishek_> ds2: timezone blues :D
May 28 16:26:27 <karki_> disdi : do you know anil pugalia?
May 28 16:26:51 <vmayoral> jkridner: I'm hoping to save some time by the end of this coming weeks to look at it again in more detail. But vanilla and Xenomai are working fine so it won't be a problem
May 28 16:27:00 <vmayoral> coming week*
May 28 16:27:05 <disdi> karki_: yes I have been in his contact for 3 years or so
May 28 16:27:13 <jkridner> VoltVisionSteve: did anything come from trying to get a designer?
May 28 16:27:33 <karki_> disdi : explains your interest in drivers ;)
May 28 16:27:56 <jkridner> vmayoral: having a close-to-mainline approach that includes real-time would really be nice though. vanilla have enough predictable performance?
May 28 16:28:29 <jkridner> vmayoral: be sure to give good background, ie., "how-to-ask-good-questions"
May 28 16:28:57 <disdi> karki_: hahaha....to be honest its the contrary....actually my interest in drivers explains my acquintance to him
May 28 16:29:00 <VoltVisionSteve> jkridner, yes.  The Guatemala professor is picking the best student to help Diego now....also David Tran just responded as this meeting started
May 28 16:29:04 <jkridner> vmayoral: http://catb.org/~esr/faqs/smart-questions.html is our standard guide for asking questions on most mailing lists, including linux-omap. :-)
May 28 16:29:23 <karki_> disdi : do you think anil would be game at mentoring BB.org for the next GSoC?
May 28 16:29:30 <DiegoTc> jkridner & VoltVisionSteve I got in contact with some teachers at college, they will give me a hand with the design. Besides David (the guy you email jkridner jkridner ) answer the email. I will take into consideration his opinions specially with github.
May 28 16:29:55 <disdi> karki_: he is a busy man....dont know
May 28 16:30:08 <DiegoTc> I'm going to wait for the Guatemala professor answer
May 28 16:30:17 <disdi> karki_: ofcourse he is best suited for it
May 28 16:30:23 <karki_> disdi : all the mentors are busy ;) I'll ask him though!
May 28 16:30:24 <vmayoral> jkridner: @tridge mentioned that vanilla seemed fine when ardupilot was getting the whole processor for itself but once you launch other (consuming) processes there're misses
May 28 16:30:26 <Abhishek_> jkridner: Apart from max sample size, the interrupt latencies are also cropping up, and UIO package does not seem good at it.
May 28 16:31:02 <karki_> disdi : Do you keep meeting him? (I missed my monthly hackathon with him! )
May 28 16:31:06 <vmayoral> jkridner: i will do so thanks for the guidelines. Anyhow, this week i will present some RT results comparing at least Xenomai and Vanilla.
May 28 16:31:34 <disdi> karki_: I met hima month ago when I was in Banglore
May 28 16:31:45 <jkridner> Abhishek_: going to make the move to RemoteProc like karki_?
May 28 16:32:02 <ds2> Please refrain from doing userland crap
May 28 16:32:35 <Abhishek_> it seems so, I will have to discuss with mranostay regarding this
May 28 16:32:46 <ds2> Abhishek_: I believe our last discussion was waiting for a vote from mranostay on a possible schedule reshuffle to make it proper kernel code
May 28 16:32:56 <Abhishek_> yep
May 28 16:33:15 <Abhishek_> waiting for mranostay's response
May 28 16:33:37 <jkridner> k.
May 28 16:34:30 <jkridner> ds2, Abhishek_: please do try to resolve direction quickly to keep coding time productive.
May 28 16:35:13 <jkridner> well, I tried to bring up anything I thought might be blocking for students/mentors....
May 28 16:35:23 <jkridner> please follow-up if you have any potential blockers.
May 28 16:36:02 * jkridner sounds the end-of-meeting gavel to hopefully leave everyone to a productive rest-of-week.
May 28 16:36:30 * karki_ leaves the meeting.
May 28 16:36:44 <ds2> wow... teh gavel knock him off the air ;)
May 28 16:37:16 * vmayoral is going to get some sun ;)
May 28 16:37:21 <vmayoral> keep in touch everyone!
May 28 16:37:36 <Abhishek_> vmayoral: looking forward to your Xenomai test reports.
May 28 16:38:16 <praveendath92> ds2: Sorry. Got disconnected. Thanks for the link.
May 28 16:38:29 <praveendath92> ds2: Is errno.h error codes relevant in all contexts of kernel C?
May 28 16:38:46 <praveendath92> I mean with any error codes that are resulted from the system or built-in functions?
May 28 16:38:59 <ds2> Abhishek_: should be more responsive this week...was traveling...will have one more blackout window toward the end of June
May 28 16:39:21 <ds2> praveendath92: that is the convention...most drivers will follow that
May 28 16:40:07 <praveendath92> ds2: So, it's possible that the usb.h functions followed it too right?
May 28 16:40:33 <Abhishek_> hey cdsteinkuehler
May 28 16:40:51 <cdsteinkuehler> Sorry I'm late, I had a meeting ITRW
May 28 16:41:10 <ds2> usb definitely does it
May 28 16:41:21 <ds2> but the -110, IIRC is related to a URB handling thing
May 28 16:41:26 <praveendath92> Yeah. Just saw the source code.
May 28 16:41:52 <praveendath92> IIRC?
May 28 16:42:04 <Abhishek_> cdsteinkuehler: jkridner was facing connectivity problems, he finished it early today
May 28 16:42:07 <mranostay> Abhishek_: so bottleneck solved?
May 28 16:42:29 <Abhishek_> mranostay: still bottlenecked @25 MHz
May 28 16:43:05 <Abhishek_> mranostay: Status update at the GSoC mailing list
May 28 16:43:37 <ds2> mraonstay: short summary - IMO, he needs to do stuff in K-space...that might change schedule. need your input.
May 28 16:44:50 <Abhishek_> cdsteinkuehler: Would have to move to kernel space, memcpy does not perform well enough to remove data
May 28 16:46:41 <cdsteinkuehler> For performance on the ARM, it will be critical that you execute burst transactions for both read and write at the SoC interconnect level.  Otherwise, you'll be chewing up a *LOT* of time in bus transactions.  This is a question for an ARM assembly language expert, which I am not.
May 28 16:47:19 <ds2> at the moment, that is largely irrevelant for userspace crap
May 28 16:47:46 <Abhishek_> cdsteinkuehler: are 256-bit bursts (32-byte) bursts okay for the PRU side (given there's a 128bit OCP L3 for the DDR?)
May 28 16:49:14 <cdsteinkuehler> Ideally you will want to match the SDRAM burst length, and/or the cache line length on the ARM core.
May 28 16:49:35 <cdsteinkuehler> I do not know the size of either of these off-hand
May 28 16:49:36 <Abhishek_> cdsteinkuehler: Following your advice on printf, I ran my tests on buffer sizes smaller than 4 MB, something like 32K , or 64K, but then found out that memcpy was still lagging
May 28 16:52:37 <Abhishek_> a wild idea: Storing the first 8K samples in the PRU0 SRAM, then writing to DDR and raising an interrupt, writing 8 K to PRU1 SRAM , then writing the rest samples to DDR, and then raising an interrupt. This way there's a short window in which DDR memory is free and can be moved around while PRU writes to SRAM.
May 28 16:53:52 <ds2> write up a quick and dirty kernel driver then re-evaluate this
May 28 16:57:19 <mranostay> yeah kernel space is probably needed
May 28 16:57:33 <mranostay> what you using for a test source for signals?
May 28 16:58:28 <Abhishek_> mranostay: I currently do not have a test source but I am reading 0x3C on PRU1_R31_0 to PRU1_R31_7, which is the setting for boot resistors
May 28 16:59:02 <Abhishek_> mranostay: I have a audio player (I2S), I2C and SDIO bus that I can use to test it
May 28 16:59:27 <Abhishek_> an external STM32 circuit
May 28 17:00:25 <ds2> power it from the BBB 3.3V rail
May 28 17:00:52 <Abhishek_> I power both the circuits from different USB ports on the same laptop, does that count?
May 28 17:01:11 <ds2> no
May 28 17:01:12 <Abhishek_> okay, I can always connect their grounds together
May 28 17:01:24 <ds2> problem is the am335x does not like its pins driven before it is powered up
May 28 17:01:42 <ds2> by powering it from the BBB's 3.3V rail, you have some implicit power sequencing
May 28 17:01:54 <Abhishek_> yep, I am aware of the powerup issue
May 28 17:02:07 <ds2> k
May 28 17:02:37 <Abhishek_> I remember when I tested the old firmware and binary, I got an 8 MB dump, but didn't really fire up sigrok and study the signals
May 28 17:02:51 <Abhishek_> Old = the reference example which I talked about
May 28 17:03:04 <ds2> *nod*
May 28 17:04:45 <Abhishek_> mranostay: What kind of a device should the LA expose, how good will your ioctl stuff work with this?
May 28 17:05:29 <Abhishek_> mranostay: panto was talking about using scatter-gather and mmap.
May 28 17:06:20 <Abhishek_> ds2: Are you in a different time zone wrt jkridner, or the same time zone?
May 28 17:13:54 <ds2> different
May 28 17:14:10 <ds2> he's usually 2 hours ahead of me
May 28 17:15:57 <Abhishek_> It must be 1015 there?
May 28 17:16:30 <ds2> 1015 here
May 28 17:16:43 <ds2> if he's in his office, it'd be 1215
May 28 17:18:36 <Abhishek_> we have our meetings @12pm Eastern time,
May 28 17:19:24 <ds2> but he's not on eastern
May 28 17:19:30 <Abhishek_> ah
May 28 17:19:31 <ds2> he's central, I am pacific
May 28 17:19:59 <Abhishek_> I had always assumed that you, jkridner and mranostay had the same timezones
May 28 17:35:47 <Abhishek_> ds2: If I rebuild a 3.8 kernel from source (Koen's kernel), will the current kernel modules (3.8 by rcn) be compatible with it, or I have to reinstall the modules too?
May 28 18:00:13 <ds2> reinstall to be sfe
May 28 18:00:20 <ds2> or just don't use modules
May 28 18:01:54 <Abhishek_> there's something strange going on. my compiled kernel doesn't seem to start the gadget interface, so no RNDIS or Mass storage device is coming up
May 28 18:03:17 <ds2> u using angstrom or debian userland?
May 28 18:03:27 <Abhishek_> debian
May 28 18:03:52 <Abhishek_> the latest system image by rcn - 15th May
May 28 18:07:26 <Abhishek_> I had faced this once before when I tried to generate a new initrd for my existing setup after which I upgraded to the latest kernel
May 28 18:07:35 <Abhishek_> *latest image
May 28 18:30:27 <ds2> oh... RCN has a hack for that
May 28 18:30:35 <ds2> found it once, but forgot about it promptly
May 28 18:31:16 <Abhishek_> do I have to enable it at compile time?
May 28 18:34:18 <Abhishek_> mranostay: Do you have any specific inputs on how the kernel driver for the LA should be, or which interfaces are to be used?
May 28 18:37:32 <ds2> no, it plays games with the modules
May 28 18:37:55 <Abhishek_> yep I see a g_multi module loaded
May 28 18:38:52 <ds2> then what's the problem again?
May 28 18:39:40 <Abhishek_> ds2: With another system image, the offending image is in microSD, I am running off eMMC for digging into it
May 28 18:43:38 <Abhishek_> ds2: Where to look for his hack?
May 28 18:48:10 <ds2> ah
May 28 18:48:19 <ds2> either in /opt or /boot, IIRC
May 28 18:48:30 <ds2> he had a script that picks the USB mode
May 28 18:50:16 <Abhishek_> ds2, found such a script @/opt/scripts/boot/
May 28 18:50:31 <ds2> that sounds about right
May 28 18:51:18 <Abhishek_> but I guess this script does run at boot time, correct?
May 28 18:56:24 <Abhishek_> ah, that means I have to install new kernel modules matching my new kernel version; modprobe is unable to find them
May 28 18:58:18 <ds2> yep
May 28 18:58:36 <ds2> or just compile it in with that as the default
May 28 18:58:45 <ds2> going to get lunch. BBL
May 28 18:58:55 <Abhishek_> sure
May 28 19:49:15 <Abhishek_> ds2: Installed modules but didn't quite help
**** ENDING LOGGING AT Thu May 29 03:00:00 2014