On Sun, Mar 23, 2014 at 6:44 AM, Peter Robinson <pbrobinson@xxxxxxxxx> wrote: >> I don't have an issue with ARM (or PPC) builds of the workstation, but >> I don't think we should decide to make them officially supported platforms >> before we feel very certain there is a viable community and ecosystem around >> them to make the product workable medium to long term on those platforms. >> This means of cause the basic lithmus test of having the shell 'work' on a specific >> piece of hardware, but also there needs to be a viable roadmap for that hardware >> going forward. I mean I don't want a situation where we declare ARM supported >> because someone got a build working on a specific dev board, only to have the >> manufacturer of that devboard switch GPU provider in the next iteration and leave >> us without a working open driver. a potential issue for the SoC's which use 3p gpu (everyone but qcom and nvidia). ofc, the flip side is some SoC vendor could switch from (for example) IMG to vivante.. Not sure what that means from a planning standpoint. But I don't think we should let that block us. It just might mean that way say (for example) that "XYZ imx6 board is supported" rather than "freescale is supported". A specific SoC part # is not going to have multiple variants with different GPU. It would be a different part #. > Believe me you are not alone in that regard, it's a discussion the ARM > people have on a regular basis. We've already had one vendor and > another SoC go from hero to zero in a short period of time :-) > >> Rob Clark is doing stellar work on Freedreno and the new Broadcom source code release >> is good news in this regard, but I think I personally need to feel that a >> officially supported ARM platform needs to be something we can believe will >> continue to exist and not a one shot 'the stars aligned for us' situation. > > Personally I'm not sure either of those are of much value. The QCom > devices are primarily used in phones which aren't really targets for > Fedora ARM. There's currently one dev board I'm aware of and it's not > widely available and it's not currently anywhere on our roadmap when > it comes to the kernel. I think upstream kernel on ifc6410 in time for f21 is actually doable. I have very basic stuff (serial console) on upstream kernel since last Dec. Getting display and gpu going has a lot of dependencies, but the clk stuff is there already, and we aren't that far off on the rest. Once linaro qcom landing team is ramped up, things should only be moving faster. Personally, I'd like to see both etnaviv and freedreno in f21.. etnaviv/imx6 is in better shape on general kernel, but missing drm and x11 support (but that is in-progress). Freedreno/snapdragon is in better shape in userspace and has drm/kms driver for gpu and display, but rest of kernel is further behind (but still I think within reach for f21) Just fyi: current qcom boards: + ifc6410 - has been shipping for a while now ($149) + ifc6412 - not shipping yet ($99) + bstem - not shipping widely yet ($??) + dragonboard - shipping for a while ($if_you_have_to_ask..) (and, well, I expect we will see more in the future with qcom as a linaro member now.. they were a pretty late on the whole community-hacker-board scene but are catching up) > The Broadcom one would possibly be of interest if there was a usable > driver and HW we could support of which there is neither at the > moment. they now know how to drive the hw.. (big props to bcom for docs!).. but there is a lot of work before we can have an upstream driver, due to some of the memory protection challenges w/ the r-pi architecture. And also the armv6 thing. But we should definitely keep an eye on bcom for their future SoC's.. > The two platforms that are of interest to me and I think will provide > us value in the short to medium term is SoCs with the Vivante GPU of > which there is an open driver that is very close to supporting mesa. > These cover the i.MX6 devices of which we already will support well > over a dozen discrete devices in the F-21 timeframe, the ARM based > OLPC XO laptops, and a bunch of differing Marvell SoC based devices. > The other platform is the Tegra K1 devices which will be supported by > nouveau and are a being announced in decent devices like netbooks, > likely chromebooks and all in one style desktops. given the timeframe for availability of the jetson board, tegra K1 should also be a very real possibility. If not f21 then most certainly f22. BR, -R >> There is also a question of what kind of hardware we want to support here, >> for instance if someone made ARM based laptops or desktops that seems like an >> obvious target, but officially supporting something like the RasperryPi or PandaBoard >> seems maybe something of an overkill. A homebrew devboard seems like it can be >> 'supported' well by just having an unofficial build for it. > > There's already a bunch of mini desktops based on the aforementioned > i.MX6 and there's an interesting array of devices that are not > dissimilar to a large tablet on a stand all in one style devices but > with keyboard and mouse attached, of course there's various netbooks > esq devices too. Fedora ARM isn't interested in phones and similar as > I just don't believe they would provide a decent user experience with > the workstation product. > > Peter > -- > desktop mailing list > desktop@xxxxxxxxxxxxxxxxxxxxxxx > https://admin.fedoraproject.org/mailman/listinfo/desktop -- desktop mailing list desktop@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/desktop
