[Freedreno] [GSoC-15] Enabling Freedreno on Android

[Emil Velikov]

On 19 March 2015 at 18:54, Daniel Thompson <daniel.thompson at linaro.org> wrote:
> On 18/03/15 22:28, Nicolas Dechesne wrote:
>>
>> On Wed, Mar 18, 2015 at 1:02 PM, Rob Clark <robdclark at gmail.com> wrote:
>>>>
>>>> So, I can have CM integration as a stretch goal (will need to decide a
>>>> suitable platform) possibly with their help, depending on how complex
>>>> the upstreaming part turns out. Does this sound feasible?
>>>
>>>
>>> Linaro also does android builds for some platforms.. none of the
>>> snapdragon platforms yet, but I guess they should have most of the
>>> infrastructure, so they might be another good candidate.
>>
>>
>> speaking about Linaro... Dan (cc'd) from Linaro has started to put
>> things together, namely:
>> * the Linaro AOSP builds
>> * the Linaro kernel 'integration' branch for IFC6410 (the branch Rob
>> has mentioned here)
>>
>> and the current status is
>> * mainlin'ish kernel (3.19)
>> * boot from eMMC
>> * framebuffer (HDMI)
>>
>> overall it should be a good candidate to start with the graphics
>> pieces since it has all the recent DRM features (atomic modeset and
>> updates).
>
Hi Dan, glad to see you :-)

>
> I agree. It should allow a GSoC student to get straight into the actual
> graphics work without having to spend much time nursing though initial
> builds.
>
> Anyhow I'm still doing the final checks but I think this is more or less
> ready to go. Fairly detailed README can be found here:
> http://git.linaro.org/people/daniel.thompson/device/linaro/ifc6410.git
>
>From a quick look it looks quite good. One might need to find and
update the correct pieces for freedreno to be picked up (egl.cfg is
one example), although I guess that is not the prime goal of the repo.

>
>> I am personnally interested to get AOSP + freedreno support (and
>> possibly some multimedia as well..), so i can try to help/support, if
>> time permits..
>
>
> Regarding AOSP versus LCR: I have billed this an LCR port but I think the
> LCR releases are structured as AOSP supplied by .repo/manifests augmented by
> a Linaro-specific plus-pack in .repo/local_manifests. I can double check
> this if anyone needs this explicitly confirmed.
>
> Anyhow I think this means the differences between LCR and AOSP are
> enumerated here:
> https://git.linaro.org/people/daniel.thompson/platform/manifest.git/blob/1569c9b8f527fefdf0cab4227e558b460e5590d7:/linaro.xml
>
Is it normal for Android distributions (not sure if it's the right
word here), to clone locally most/all the things they use ? What would
be the merits and/or benefits behind doing so ?

>From your POV what should upstream do in order to encourage Android
distros and/or companies to be more open and contribute ?

I can see that you guys have some trivial (and some not so trivial)
patches on top of upstream projects - drm, waffle, piglit that never
made it to the mailing lists. I realise that you guys are busy,
although I was under the impression that one of Linaro's goals is to
get features/fixes upstream, yet there are some one-line patches for
waffle that have been there for 5+ months. Honestly this makes me
quite sad :'-(

On the similar note, all the patches (or equivalent) from Benjamin's
libdrm have been upstream for quite some time. Perhaps one can
consider using the upstream repo ? The alternative libdrm in the above
manifest list to a TI's version. Seems like they prefer hand picking
patches rather than sending their work to be up-streamed. That seems
rather counter-productive imho, although it would explain why they
have been so quiet, since Rob is no longer with them.

I realise that you have no take in most (all) of this, but I would
appreciate your honest input how we can, at least attempt to, resolve
this.

Thanks
Emil

P.S. Sorry for the wall of text.