IRC logs for #openrisc Monday, 2014-09-22

--- Log opened Mon Sep 22 00:00:04 2014
poke53282stekern: At the moment I have the following patches to allow ptrace to work.
poke53282the question is if PT_ORIG_GPR11 and PT_SYSCALLNO  should be offered to the user00:38
poke53282the other question how to handle the delayed ins.00:39
poke53282in the kernel or do the user space has to handle it?00:39
poke53282the pc is reduced by 4 if l.trap is in a delayed slot.00:40
poke53282for example ltrace wouldn't find the breakpoint in this case.00:40
poke53282the correction can be applied everywhere. In the kernel or userspace. But which position is correct?00:41
poke53282The problem is that if we correct it in kernel space than we need another table entry. The pc to give to the user space in ptrace and the real position where to return. Of course then setting the pc via ptrace in user space is then also difficult.00:43
poke53282this is the way ltrace finds the breakpoint.00:48
poke53282and this will fail if the trap is a delayed instruction00:49
poke53282stekern: Do you use the  SND_SOC_GENERIC_DMAENGINE_PCM ?04:53
stekernpoke53282: yes, but I'm still in the progress of trying to figure out how everything is put together04:53
poke53282does this work without a dma driver.04:53
stekernI don't think it does04:53
poke53282soc/sh/fsi.c don't use it.04:54
stekernI tried to avoid writing a dma driver and just do everything in a simple cpu-dai driver, but I gave up on that and am now writing a dma driver for the wb_streamer04:54
poke53282Ok :)04:55
stekernbut from my standpoint, I tried to do it in a way so that it'd be easy to switch over to a real dma driver later, so if you avoid that, it's maybe more feasible04:58
poke53282I will try. I tried to do it as simple as possible two or three registers. , pointer, buffer_size, period and so on.05:00
poke53282do you the abbreviative dai means?05:02
poke53282Haven't figured out yet05:02
poke53282dai = ??05:08
poke53282digital audio interface05:09
poke53282Hmm, "pcm-audio 98000000.sound: Missing dma channel for stream :005:23
stekernpoke53282: that's where I'm at too ;)05:23
poke53282guess, this is is such an error05:23
poke53282Ok, then I use the implementation from soc/sh/fsi.c. It Is not that big.05:24
poke53282without dma05:24
stekernhmm, doesn't that still use the dmaengine?05:31
poke53282yes, but it supports also pio05:34
poke53282I know that this works, at least it is accepted and I get a and device.05:34
poke53282dma is only an option05:34
poke53282they use snd_soc_register_platform instead of devm_snd_dmaengine_pcm_register05:39
poke53282devm_snd_dmaengine_pcm_register calls snd_soc_register_platform internal05:40
poke53282and they need around 7 additional functions for this05:42
poke53282like open, ioctl, free, ....05:42
poke53282the real /dev interface if I see it right05:43
poke53282the function are pretty small05:45
poke53282now it works05:48
stekernah, right05:58
poke53282these are the driver functions executed. Now I have to fill them with something.06:26
poke53282And this is, how it currently looks like06:38
poke53282more or less a dummy driver06:40
olofkAnyone here who is at least a bit handy with basic signal processing theory?07:15
stekernI got scared off signal processing theory by a terrible teacher at the university07:16
olofkI'm trying to generate a stream if I/Q samples and look at the generated waveform. I'm just not sure what I should expect to see07:16
olofkstekern: I stopped listening after my teacher told us about Q-kanalen (swedish joke)07:17
stekernI think that was the only exam I had to do twice07:17
stekernwhich is a bit of a pity, signal processing is kinda fun. at least the little I've done with my synth endeavours have been.07:18
olofkYeah, I thought it was fun when I started learning about it, but haven't done a shit with it since then07:19
stekernthe teacher we had looked like a bum, to the extent that we actually thought that he *was* a bum that had wandered off into our lecture the first time we saw him07:21
olofkThat's probably a good reason why I should never start teaching :)07:22
stekernthat wouldn't make him a bad teacher of course, but he was07:22
mithroHeyo, does anyone know who "owns" ?07:22
olofkmithro: Yep. blueCmd set it up07:22
mithrobtw should give you access to most of the opencores infrastructure while is down07:23
mithroblueCmd: ping?07:23
olofkmithro: Yeah, I took a look at it and was pleased to see that the OpenRISC wiki was there as well07:24
olofkmithro: Are you with ORSoC?07:24
mithroolofk: nope07:25
mithroolofk: just a concerned data packrat :P07:25
olofkJust a knight in shining armour? :)07:25
olofkI think we should set some time aside at orconf to discuss how to deal with a potential loss of opencores07:26
mithroolofk: I personally think you should just move everything to github :P07:26
mithrobut I've never contributed to opencores at all07:27
olofkmithro: All new development and projects that we update frequently are on github already07:27
olofkmithro: We still have bugzilla and a wiki on OpenCores that won't be as smooth to move however07:28
olofkThe wiki is messy, we know that, but until someone comes along and moves all the relevant documentation elsewhere, it's the best we got07:29
mithroolofk: I assume you don't have any access to the wiki server?07:29
olofkmithro: Nope. I used to work for the company who owns opencores, but I quit some time ago07:30
olofkAnd honestly, if we were to move it, I would rather pick out the relevant information07:31
mithroolofk: got a link to the wiki?07:31
olofkBut it looks like is back up now07:31
mithroolofk: looks like they paid their bill :P07:34
stekernto be fair, as soon as they've been aware of the problem, it's been handled fairly quickly07:37
stekernso far at least07:37
mithrostekern: well, it did expire 4 days ago according to whois07:38
stekernwell, friday 23:59, so basically a couple of hours of office time07:41
stekernI bet no-one there care during weekends...07:41
olofkYeah, that's true07:42
olofkLoneTech: Hi07:45
olofkLoneTech: Is the DNS issue sorted out now?07:45
LoneTechI don't know (would guess not for a few hours still), but the ball should be rolling. turns out debugging third party issues is harder on sundays.07:46
olofkLoneTech: Great!08:00
olofkLoneTech: Are you coming to orconf btw?08:03
LoneTechnot sure yet. I'd like to08:03
olofkHope you can make it08:04
olofkVHDL must still be the worst language on the planet. I've been doing VHDL work full time for about six years now, but as soon as I'm away from it for a few weeks I forgot basic syntax08:38
LoneTechI used to carry a little reference book that included nicely where everything went08:39
olofkThat's pretty handy08:39
LoneTech - originally published as a supplement for a swedish vhdl course, iirc08:41
LoneTechremove -book for original, that was formatted for printing08:41
sb0__olofk, try migen ;-)08:43
olofksb0__: I can't. FuseSoC doesn't support migen :P09:15
olofkI'm starting to get the hang of this I/Q nonsense09:44
blueCmdmithro: hello!09:55
olofkHmm.. Quartus doesn't accept $sin, even if it's just for preinitializing an array in an initial block. Anyone has ideas for workaround?10:15
olofkI'm currently using a python script to build the table, but I'm not very happy with that solution10:16
LoneTechI tend to use that solution too10:16
olofkMaybe I should just use a cordic instead. Eventually I want to control the frequency anyway10:17
sb0__olofk, there is no fundamental problem with integrating migen cores in a non-migen design10:18
sb0__though I do recommend that you switch to misoc, too10:18
LoneTechwhat makes me really unhappy is when such tables make it into a spec like SHA without explanation of how they are generated (they are explained there, but it happens sometimes)10:27
sb0__olofk, you can use python's math.sin() from migen, no problem. we have a cordic core for you, too.10:33
sb0__both will work with ise, vivado, and quartus.10:36
mithroblueCmd: do you "maintain" the opencores github repo?11:23
AnupHi All, any one working on OR1K synthesis on ML605 or Virtex-7 here11:45
-!- Anup is now known as Guest8554511:45
Guest85545i need some details on synthesis11:45
Guest85545OR1K on ML605 or ZYNQ11:46
Guest85545anybody there11:48
olofkmithro: It's just a backup. We're having a hard time finding maintainers for most of the cores? Interested in adopting any of them? :)12:07
mithroolofk / blueCmd: do you sync anything from opencores into that repo at all?12:12
olofkmithro: Sorry. No idea12:31
blueCmdmithro: nope12:33
blueCmdmithro: due to the state of opencores I wasn't able to feasibly automate it12:34
blueCmdmithro: if the website was 1) able to do checkouts like a proper SVN site and 2) did not contain CRC errors with what I assume is on-disk corruption, it would be a whole other matter12:35
poke53282olofk: Looks like I am coming to orconf16:13
poke53282Have to book the flight in the next two days16:13
olofkpoke53282: Good to hear! Don't forget to register too16:26
olofksimoncook: Why on earth did you get three tickets to orconf? Are you planning to sell the other two? :)16:27
simoncookMyself, Jeremy and I have one other colleague coming along whose been working on LLVM for OpenRISC16:30
olofksimoncook: Cool16:32
olofkThe more, the merrier16:32
simoncookYou may actually have it it down as four of me, that I'll need to fix :)16:32
olofksimoncook: Hmm... I don't know. Too many cooks.. you know :)16:33
olofkstekern: You're not registered either.16:35
olofkIt's not like we're running out of tickets, but still16:36
stekernolofk: there's a 8 minute time limit on the registration, I needed some practice before I succeded to fill it out within that16:45
olofkI know the feeling. There's 12 minute time limit on the swedish train systems. I have about 20% success rate on that16:46
olofkMake sure to register this week. There's a 37% discount until sunday16:50
poke53282olofk: registered16:55
poke53282and I would like to give a presentation.16:56
olofkpoke53282: Awesome. Send me a mail once you have a preliminary title. Hoping to have the schedule up soon16:56
olofk25 minutes + 5 minutes discussion. Will that work?16:57
poke53282Hmm, "Hard core introduction to quantum physics for computer freaks" =)16:58
poke53282Yes, that works16:58
olofkI like the sound of that :)16:58
poke53282"Nature and people of Vancouver - why the city is on the third place of the most livable cities of the world and what to do when you see a bear"17:01
poke53282No, it will be about jor1k of course.17:05
poke5328225 minutes + 5 minutes if Ok. I will send you the title when I have it.17:11
poke53282when I know it17:11
poke53282Does it harm if I come with a Windows laptop? Or with an iPad?17:17
olofkpoke53282: You will be forced to install Limux as soon as you enter the city18:04
poke53282Ahh yeah, Munich is a Linux city. I remember the news.18:05
poke53282olofk: Approx. How many people will be there?18:23
olofkpoke53282: Just over 20 people atm18:23
olofk(including simoncook and his three clones)18:23
poke53282"Writing an OpenRISC emulator in Javascript – fun and pain"18:26
olofkNo good title. That won't help me at all when I see a bear18:30
poke53282Yes, jor1k doesn't have sound support yet. So you can't annoy the bear with beepy and choppy sound. But I will add a slide.18:32
poke53282But you can annoy humans pretty easily with it, so that they leave the room.18:35
olofkThat's good.18:41
poke53282I hope that 25 minutes will be enough.19:40
olofkpoke53282: No worries. We can give you more time if you want that19:43
--- Log closed Tue Sep 23 00:00:06 2014

Generated by 2.15.2 by Marius Gedminas - find it at!