On Mon, Jun 16, 2014 at 3:27 PM, Philippe Coval < [email protected]> wrote:
> hi, > > I've been investigating about ARM support for tizen common > > (we demoed it at TdcSf14, were any of you there ? > if not watch this : > https://www.youtube.com/watch?v=6JDy9uUqH4Y > ) > This is great. Do you have other artifacts? Is there a central place we're keeping ARM related links? I know the wiki has some stuff, and I assume that is central, but I want to be sure. > > On 06/16/2014 02:59 PM, Jeremiah Foster wrote: > >> On Mon, Jun 16, 2014 at 11:29 AM, Dominig ar Foll < >> [email protected] <mailto:[email protected]>> wrote >> >> >> Le 16/06/2014 09:28, Jeremiah Foster a écrit : >> > >> > That list doesn't seem to get much traffic -- June is half over and >> >> > Sorry the correct list to be used is : >> https://lists.tizen.org/listinfo/dev >> >> >> Okay, thanks! >> > > I also noted some ongoing works in progress on this page : > > https://wiki.tizen.org/wiki/ARM > > Feel free to use this page as entry point and edit it at will > > (we can split it into several pages later if it gets messy) Excellent, this seems like an efficient way of working on ARM in Tizen. > > >> > > The source used to build the ARM repo in Common are all public. >> > >> > There doesn't seem to be an agreed upon approach, or if there >> is, its >> > unclear. Let's take the Renesas request for >> > example: https://lists.tizen.org/pipermail/ivi/2014-June/ >> 002636.html > > > >> Will answer them ... better late than never > >> >> So far we have no refrence ARM platform for IVI. The proposal is >> to add >> Renesas as our initial ARM platform and that is what we are working to >> acheive with them. >> >> this sounds a good idea > > > > We will add more build option if needed when other reference platform >> commit to work for IVI. >> >> >> I think it needs to be the other way around. Tizen IVI has to be shown to >> work on production-quality ARM silicon before silicon vendors will commit >> resources. Currently there is nothing in Tizen IVI that can't be found >> elsewhere with significantly better support for ARM. Debian, for example, >> has a working ARM v8 software stack, Linaro creates LTSI kernels, etc. >> > > I guess tizen aims to do the same check the plans from some samsung team : > > http://www.slideshare.net/MyungJooHam/tdc-mjham-armkernelintizen3noani That looks very interesting indeed, shame about the quality of the video. > > > >> > If Tizen had a small team that tested Tizen Common against an ARM >> > kernel, then they could at least guide on suitable approaches. >> > Currently, one would have to; >> >> We are a couple and there are also some community efforts > > > > >> > 1. Identify a kernel that fits Tizen common and your hardware >> >> since hardware is odroid we'll use that linux-3.10 branch maintained by > exynos team This seems reasonable. I note though that the merge window is closed for 3.15, so 3.10 is likely a little bit behind. I wonder what the criteria are for choosing a kernel? It has to be a little more rigorous than "because its there." I think that its a really good start that its done by the exynos team, but are they going to work in the open, accept patches, help with graphics acceleration, support newer kernel features. All the good stuff that comes with real open source collaboration. > > 2. Compile that kernel, with toolchain, to test on the hardware >> >> > done using armv7l EABI > So, in this terminology, armv7l is what debian calls "armel", i.e. softfloat? > need some time and/or support to make it work on the odroid u2 > that's where I am now Its pretty awesome you're doing this. And I hope to be able to get a board and do some testing too. I don't mean to sound negative, but how do we expect to bring in other silicon vendors? Exynos is hardly representative of the automotive industry. The current incumbents are companies like Freescale, nvidia, TI, Renesas, and STMicro. Don't you think it better to adopt the process that these companies, along with GENIVI, uses? That is to say use Open Embedded -- Yocto to create recipes for Tizen that can be used against _any_ silicon vendor recipe? > 3. Build Tizen common from source in a build system suitable and >> > widely used for ARM hardware. >> >> this is done because the armv7l rootfs > worked fine on sunxi boards Hmm, okay. Should Tizen be listed here then: http://linux-sunxi.org/Bootable_OS_images > > That is exaclty what we have agreed to do using the Odroid reference >> platform. We are wainting for HW to be delivered to enable the >> Auto test >> on Common for ARM. >> >> To what I know yes, smoke testing > > > >> While I'm sure the Odroid platform is awesome, >> > it has MALI GPU as in other samsung exynos hardware > > > > does it have automotive connectivity and peripherals? Of course, we want >> something easy to buy and inexpensive so that a whole lot of people can buy >> one, >> > > the question was asked in list .. can you suggest some Beagle Bone Black Wandaboard Renesas has a lower cost offering. > > > > but even there Odroid seems a bit of strange choice, wouldn't the Beagle >> Bone Black be better? >> > > do we have OMAP / PVR GPU support ? Do you have MALI support? > > > >> > Just identifying a kernel is a non-trivial step -- are you going to go >> > with a LTSI kernel (hardware in automotive needs long-term support)? >> > > probably yes see upper link > > > > Are you going to use Linaro's LTS kernel? >> > > Usure , I have some pple to contact to see if we could get good support > > > It may have features you >> > need. Are those kernels going to bring in the needed features for >> > Tizen common? >> > no idea yet > > > Has someone identified the needed functionality in the >> > kernel that the Tizen common userland will need so that one doesn't >> > spend a lot of time wondering why certain features don't work? >> > > smack and gpu seems the big features AFAIK > > > That is an other topic which deserve a thread by itself. >> >> :-) >> >> As Tizen we do >> not enforce a specific version of the Kernel but rather the provision >> for specific feature (e.g. Smack and SystemD boot). >> >> >> This begs the question: who's in charge of architecting Tizen Common? And >> how strong is the preference for SMACK? What about projects like TOMOYO: >> http://en.wikipedia.org/wiki/TOMOYO_Linux How would one go about >> removing SMACK in favor of TOMOYO and would it still be "Tizen"? >> > Question for architects but I bet smack is going to stay :) > > >> People maintaining >> the Kernel of each reference platform have the duty to provide a >> Kernel >> with the needed features. >> > > seems a bit early now > > > Getting it based on LTSi is certainly an >> interresting approach but not a requirement. >> > >> > There seems to be a lot missing before a suitable public port to ARM >> > can be done repeatedly in a manner consistent with production >> systems. >> > Is there a goal to add this type of support? >> >> > For now we want to make common supporting ARM > I have some idea in mind , like automatization of publishing an armv7l > rootfs > and then users will do the adaption using their own kernel and bootloaders > > > > Or is the ARM port of >> > Tizen something that the community will fund and build in their >> spare >> > time? >> >> > Yes there is a community project, the name is tizen-sunxi from a bulgarian > friend > he did adapt an beagle ivi image to some allwinner boards > same hardware used in the demo I setup and mentioned earlier... > so It will follow Tizen Common ... > > I think this project is relevant because this is this only project > community project > that use wildly available openhardware and able to be rebuilt from scratch > to attract hackers this is a good point. > > > > > Today we are strongly limited by the lack of publicly available >> reference platforms. >> > yes reference platforms are lacking > > > > On Common Odroid should fix the issue by end of >> June. >> > > yes I hope to have some result to share soon > > > > For IVI, it will be up to the reference platform providers to fix >> the issue of availability. >> >> true > > > >> I suspect the lack of available hardware to be a bottleneck, especially >> when a key deliverable for success is an ADK or SDK. Isn't it much faster >> if we use something like the BBB? >> > > BBB could be an option with or without GPU support ? > With. I think that the binaries are released under the GPL. So you don't get source, but you can redistribute. > > what are your thoughts on allwinner / sunxi ? > I know people who say it is an awesome chip with excellent price/performance qualities. I also know that they try to be pretty open and seem to get open source. I don't think it is representative of the automotive industry though and I don't know how interested they are in supporting Tizen. We should find out. :-) > > Regards see you on dev mailing list > Thanks for all your answers! Regards, Jeremiah > > -- > mailto:[email protected] -- gpg:0x467094BC > xmpp:[email protected] > https://dockr.eurogiciel.fr/blogs/embedded/author/pcl/ > . > > -- Jeremiah C. Foster GENIVI COMMUNITY MANAGER Pelagicore AB Ekelundsgatan 4, 6tr, SE-411 18 Gothenburg, Sweden M: +46 (0)73 093 0506 [email protected]
_______________________________________________ IVI mailing list [email protected] https://lists.tizen.org/listinfo/ivi
