IRC log of #cubox of Tue 03 Dec 2013. All times are in CET < Back to index

09:38 dv rabeeh: you guys super busy with shipping out the preorders? :)
09:53 MikeSeth I sure hope they are
09:53 MikeSet 09:53 * MikeSeth ponders volunteering to help.. ship.
09:54 MikeSeth I really wanna go visit solidrun offices
10:32 dv they need these amazon drones :)
10:37 MikeSeth our flight authority would probably be unwelcoming
10:37 MikeSeth and by unwelcoming I do mean shoot it down
10:40 jnettle 10:40 * jnettlet expects the same to happen in the US, only by gangstas and rednecks
10:41 MikeSeth I'd take those over Hezbollah any day, frankly
10:42 jnettlet I don't know. I am sure Hezbollah have better aim.
10:42 MikeSeth and worse ideas..
10:43 jnettlet dv, are you seeing a performance drop with the Vivante driver under the 3.5.7 kernel?
10:43 dv I didnt try 3.5.7 so far
10:44 dv neither 3.10 (did you mean that one perhaps?)
10:44 dv I'm currently busy finishing some tasks for the beta release of my plugins
10:46 jnettlet I know the problem exists in 3.10 but not in 3.0.35, just wondering if it existed in 3.5.7 without having to screw around with getting another kernel I could care less about running.
10:46 dv hmm
10:46 dv well I was told by otavio not to use 3.5.7
10:46 dv or at least the freescale fork of 3.5.7
10:47 dv I think the reason was that it was in alpha state and got abandoned at some point, since the focus was shifted towards 3.10
10:48 jnettlet probably shifted when it was announced at the next linaro LTS kernel.
10:48 jnettlet well Linaro of gkh
10:49 dv and/or the pengutronix efforts
10:49 d 10:49 * dv finds all of this so confusing
10:49 dv pengutronix, boundarydevices, linaro..
10:56 hste_ dv: and the freescale versions of 3.0.35 ......1.0.0. 1.1.0 4.0.0 4.1.0 ....
10:59 jnettlet dv, best as I can tell pengutronix hasn't released much on their own recently. It is floating around somewhere in pengutronix land.
11:04 dv yeah, the versioning is also weird
11:04 dv could be worse, though. vivante has pretty much no versioning except for v2 and v4 :)
11:26 jnettlet dv, well their main versioning problem is that they let each chip maker to build their own userland. So really they don't control released versions.
13:39 jnettlet MarcusVinter, welcome back. saw your message the other day.
13:40 MarcusVinter Oh. Thanks jnettlet. I appriciate it.
13:40 jnettlet MarcusVinter, what do you plan on hacking on if you don't mind me asking?
13:42 MarcusVinter Mainly just use it for similar purposes as a thin client. I'm waiting for the latest cubox and the C1 is the closest chip-wise. What about yourself?
13:44 dv_ jnettlet and _rmk_ are our kernel gurus
13:44 jnettlet Mostly u-boot and kernel stuff
13:45 dv_ thanks to them, we might see a nice working 3.10 kernel with VPU and GPU support
13:45 jnettlet well it works, it is just slow as crap right now.
13:46 jnettlet my list of things that could be broken is dwindling so I hope to have that fixed soonish.
13:46 _rmk_ btw, Mark Brown now says that my kirkwood asoc patches look good, but there's still a sticking point: Liam's patch to fix DPCM which is incomplete and has no real changelog description
13:47 jnettlet _rmk_, is that the one he hacked up at the conference?
13:47 MarcusVinter Oh great! It's nice to meet you both. Can't wait for GPU support.
13:47 _rmk_ jnettlet: it's one Liam's had since April and hadn't posted
13:48 jnettlet oh _rmk_, dv_, I may have made some headway into getting some Vivante contacts. We are sorting out NDA discussions.
13:48 dv_ jnettlet: can you also involve wumpus in this?
13:48 dv_ because that sounds awesome
13:48 jnettlet dv_, it is NDA so wumpus wants nothing to do with it.
13:48 dv_ hm. right.
13:48 jnettlet can't taint the etnaviv driver
13:48 _rmk_ jnettlet: great, I wonder if that means we can start feeding back fixes and improvements into their codebase
13:49 jnettlet _rmk_, that is what I am hoping.
13:49 dv_ but that also means you cant tell us anything you learn
13:49 jnettlet and get a unified driver release
13:49 dv_ but at least yes, the fixes can be fed back to them
13:49 jnettlet dv_, I already can't as I have seen their v4 client code under an old NDA
13:49 jnettlet that is why I am a good candidate to push the relationship.
13:50 jnettlet I can't work on kernel and OSS userspace, just can't work on the "hal" layer
13:52 jnettle 13:52 * jnettlet also final got a beta firmware to hopefully fix my powerline ethernet adapters so they don't randomly lose their connection. Another 6 month process.
13:55 rabeeh MarcusVinter: there is ubuntu with accelerated gpu support if you want
13:55 rabeeh ubuntu 13.04 that was packed by jas-hacks_ and hste on the channel
13:57 wumpus indeed, NDAd stuff wouldn't really help me much, and I'm not interested in the closed source driver anyway
13:58 MarcusVinter rabeeh: Yes please. Where might I find this?
14:02 rabeeh please PM me with your email.
14:02 MarcusVinter Thanks.
14:36 _rmk 14:36 * _rmk_ now has a Lindy spdif to rca converter, replacing the cheapo chinese one - and it seems better.
14:36 _rmk_ it's about the size of the cubox too :)
14:39 MarcusVinter Can I ask why you went for the spdif on the cubox anyway?
14:40 MarcusVinter Its quite an unusual connector appose to 3.5mm
14:47 _rmk_ well, spdif gives the possibility to do more than 2 channels
14:48 _rmk_ such as squirting out the raw mpeg or ac3 stream into something that can do all the 5.1 etc surround
14:48 MarcusVinter Ahhh I see. Makes sense then.
14:50 jnettlet and you can don't have the sound out quality compromised by price constraints you may have trying to get a high quality DAC onto the board.
14:51 _rmk_ the down-side is you need either a hifi which can accept optical in or a converter. there's some cheap chinese 5.1 converters (spdif to 6 phonos) which, although they "work" they're not all that good (they don't decode the spdif stream synchronously, which causes noticable artifacts such as pitch bending)
14:51 _rmk_ the other converters like Lindy are just a DAC, so they don't accept AC3/MPEG streams
14:52 _rmk_ but at least they decode synchronously
14:53 _rmk_ the cheap chinese one is fine for rock/pop music where you don't have long continuous notes - but really bad with classical music.
14:53 MarcusVinter Ahh I see. We have a hifi with spdif in the office and a spare cable about. I'll give it a shot later today. I don't think the cable fits quite right into the cubox though so I may have to bore the hole out a bit.
14:54 _rmk_ yea, I had to do that too - you can get some optical cables which do fit :)
14:56 _rmk_ my main motivation for getting this box is more for my new panasonic tv, since it also gives optical out and that's about the only way to get stuff out of it and fed into an amp... and being a modern tv, they've gone back to poor quality tinny sound
14:57 MarcusVinter Yeah, I've noticed as TV have got thinner, the
14:57 _rmk_ previously they improved it so it was actually quite reasonable including the lower end - proper integrated speaker enclosures etc - now they seem to have done away with all that even on the more expensive models in favour of making them slim line
14:57 MarcusVinter They have been putting lower quality speakers in*
14:57 MarcusVinter Yeah
14:57 MarcusVinter It sucks :(
14:57 _rmk_ and now you have to also buy a "sound bar" or find some other solution
14:58 MarcusVinter Luckily my TV has a 3.5mm that I aux into an old amp to some huge pioneer speakers heh
15:00 MarcusVinter It wouldnt supprise me if the manufacturers purposely reduced the quality over time to sell sound bars
15:01 _rmk_ I think you're right there
15:02 _rmk_ the other reason to using the spdif output from the cubox/carrier1/cubox-i is that you can get audio out of them without needing the tv on :)
15:04 MarcusVinter Oh yeah, good point. I guess i'm starting to see the benifits of spdif now.
15:06 _rmk_ esp. important as last christmas I had something like two days of music playing off the cubox
15:07 MarcusVinter Did SolidRun manage to get around the HDMI audio licensing?
15:07 MarcusVinter Oh nice. Did it get hot?
15:08 _rmk_ obviously this was the marvell dove version - it gets warm during normal running but not excessively so.
15:09 MarcusVinter I like the idea of using the adhesive thermal pad to conduct from the CPU to the metal housing. I assume the idea is that the heat will then dissipate out through the plastic?
15:09 _rmk_ I've also had that playing back 4 or 5 hours worth of material almost continuously without problem
15:09 _rmk_ well, black radiates better than white :)
15:11 MarcusVinter Thats true. When the next cubox is released I might overclock this and use it as a heater.
15:11 MarcusVinter Jokes
15:11 jnettlet _rmk_, what is your mainline kernel setting the SLUB's cache line size to? I just realized with device-tree my 3.10 kernel is setting it to 64 which wrong should be 32
15:12 jnettlet which is wrong
15:12 jnettlet wow I am having some major typing problems today.
15:12 _rmk_ some armv7's have 64-byte cache lines
15:13 _rmk_ so we do use 64-byte alignment on armv7
15:13 _rmk_ that shouldn't cause a slowdown since 64-byte alignment just results in a little bit of memory wastage
15:16 _rmk_ if anyone looks for a SPDIF to RCA box, stay away from the "HD Audio Rush" stuff - the Lindy box so far seems to be streets ahead even though it won't do anything but PCM.
15:18 _rmk_ I got the Lindy for ?40 from Amazon
15:19 _rmk_ probably cheaper from ebay but can be more problematical if you have to return
15:20 MarcusVinter The lindy box stuff looks nicer in your house anyway.
15:20 MarcusVinter Could mount it
15:33 rabeeh cortex a15 uses 64byte
15:34 _rmk_ some a9s too
15:34 _rmk_ and a8s
15:36 _rmk_ and that means in a multiplatform kernel, we have no option but to use 64 byte alignment for cacheline aligning
16:00 MarcusVinter Anyone tried working with Windows EC 7 or anything like that?
20:59 jnettlet excellent ubuntu still has a known unpatched bug for ARM on their xorg-server package. Nice work Ubuntu.
21:02 hste jnettlet: what sort of bug?
21:03 jnettlet hste, nasty crasher in the scheduler for ARM and PPC
21:04 jnettlet I think I am done with my Ubuntu image testing. They just can't keep it together at this point.
21:04 jnettlet The bug has been diagnosed and fixed for almost 6 months now.
21:05 hste we still have to compile our own to use vivante
21:07 jnettlet compile our own xserver? nah should just need to compile the driver.
21:07 _rmk_ have you stuffed it into their bug reporting system or have you just given up with them? :)
21:08 hste have to compile to use dri for vivante
21:08 jnettlet I was just thinking about that. But then I looked at my account and not a single bug I have filed has been answered.
21:09 _rmk_ oh, the standard problem I run into with bug reporting
21:09 _rmk_ which has now resulted in me never reporting bugs to any distro ever
21:09 jnettlet actually that isn't true. one bug has been closed as will not fix, or fixed in next release.
21:11 jas-hacks_ jnettlet: which ubuntu release you testing?
21:11 jnettlet jas-hacks_, the latest 13.11
21:12 jas-hacks_ interesting ... I'm starting on a 13.10
21:13 jas-hacks_ which kernel you at?
21:14 jnettlet jas-hacks_, it doesn't happen with the 3.0.35 kernels. It only occurs with 3.10 kernels
21:14 jas-hacks_ how stable is 3.10?
21:15 jnettlet I am still patching but reasonably stable
21:16 jnettlet jas-hacks_, did you have a 3.5.7 built kernel that you were using?
21:18 jas-hacks_ Currently I've stayed with 3.0.35 as it seems stable but old. Next task is 3.10
21:19 jas-hacks_ problem is that FSL don't give away much on the state of the kernels
21:20 jnettlet jas-hacks_, well I have good chunk of patches I am still working through. Mostly just hacks for stability at this point with TODO's of what the real problem is.
21:20 jas-hacks_ Are you just porting C1 board into 3.10 or more?
21:22 jnettlet C1 porting is done. I am fixing Vivante stuff and trying to sort out an ever increasingly complex userspace set of drivers.
21:23 jnettlet torn between wanting a stable set of drivers now, and not wanting to deal with this binary mess any longer.
21:26 jas-hacks_ Does X11 run with Vivante?
21:26 jnettlet not right now that is what I am bitching about :-P
21:27 jas-hacks_ random crashes/locks ?
21:27 jnettlet The problem is the more bugs I fix in the kernel driver the more userspace bugs it exposes.
21:27 jnettlet crashes in userspace. I have fixed all the locks in the kernel
21:27 _rmk_ jnettlet: you could always go back to the v2 version :)
21:28 jnettlet _rmk_, except the GLESv2 implementation in v2 is horrendous
21:29 jas-hacks_ So it works 3.0.35?
21:30 jas-hacks_ works with 3.0.35
21:31 jnettlet works better.
21:38 purch third usb-ttl cable and finally 3.3V!
21:40 xraxor anyone received a cubox yet?
21:41 jnettlet oh this is interesting. I just read further into Freescale's EULA. Vivante's license is actually listed under appendix A and it seems to be less restrictive.
21:41 jnettlet we may be able to use the Freescale libs for the Marvell chips as well.
21:41 jnettlet http://fpaste.org/58770/86103310/
21:47 purch where do you keep the latest kernels and images for C1?
21:50 _rmk_ jnettlet: so... the question becomes, is the vivante userspace stuff not deemed to be "Freescale proprietary software"...
21:50 jnettlet _rmk_, that is what I am trying to discern.
21:51 jnettlet Since it is in the Appendix with OSS and Microsoft software then I would have to assume no.
21:52 jnettlet Also Freescale's "only use on our systems" legalese isn't legal for OSS licensed software.
21:53 jas-hacks_ Are freescale lib/blobs not restricted to the SOC by some chip id check?
21:55 jnettlet nah they know better than that. They are shipping most of this to OSS developers so it would be hacked to work anywhere without blinking an eye.
21:57 jnettlet _rmk_, Section 2. Licenses
21:57 jnettlet 2.1 Separate license grants and rights to Third Party Software,
21:57 jnettlet if different from those granted in this Section 2, are as identified
21:57 jnettlet on Appendix A.
21:57 jnettlet so Vivante is explicitly excluded from the section regarding Freescale licensed technologies
21:59 jnettle 21:59 * jnettlet may package them and see if I get served with a take down notice.
22:01 _rmk_ so... this is not how to do a website upgrade: announce it the day before, saying it'll be N hours. after N hours say it'll be another two hours. then say it'll be back by X pm. then say it'll be just two more hours. then say they'll get an update from the engineers. then say it'll be a few more hours. then say it'll be tomorrow.
22:04 jnettlet :-) I guess they haven't heard of the fallback contingency planning method.
22:04 hste I thought such information was just sending a 404 :)
22:05 jnettlet 8.2 Choice of Law. This Agreement will be governed by,
22:05 jnettlet construed, and enforced in accordance with the laws of the State of Texas.
22:06 jnettlet so Freescale is bringing all offenders to justice in Texas!
22:06 jnettlet This is also important people. 8.11 Unauthorized Use. The Licensed Software is not intended or
22:06 jnettlet authorized for use in anti-personnel landmines, and you agree that it
22:06 jnettlet will not be used for this purpose.
22:06 hste yes justice is simpler in texas :)
22:07 jnettlet yeah, in that the corporation usually wins
22:07 jnettlet we can not turn our carrier-1 boards into anti-personnel landmines. Other landmines are okay just as long as they aren't anti-personnel.
22:11 jnettle 22:11 * jnettlet kicks off a new xorg build to fix this stupid bug.