[Gta04-owner] Which distri
H. Nikolaus Schaller
hns at goldelico.com
Tue Sep 29 13:12:41 CEST 2015
Am 29.09.2015 um 12:31 schrieb Radek Polak <psonek2 at seznam.cz>:
> On Tuesday 29 of September 2015 11:48:52 H. Nikolaus Schaller wrote:
> > Well, for me it is the reverse thing: I have spent so much time to improve
> > hardware and keep the kernel following mainline but have no nice userspace
> > which is useable for it (well, we have one: Replicant - but it is currently
> > stuck on 3.12 kernel).
> > So looks like a deadlock :)
> But you have different goal.
Why do you think so? Just because we have not reached it yet, you assume that we don't have the same goal?
Yes, originally the GTA04 was only a project to provide urgently demanded hardware, but after the hardware became available, software goals got priority.
> For me goal is to have phone - the thing that you use for calling your family/friends.
I have the same. That is what it was developed for.
> I really tried GTA04, but it was always discharged when i forgot it for more then 12 hours. And if you have small children then it's not funny - working phone can save a life...
> GTA04 is comparable to car that has max speed 15 km/h - nice for test ride, but you cant take your family and go to the city with it. And correct me if i am wrong but there was zero progress for years in the power management area.
Because it simply wasn't possible to work on power management yet (or more people would have done).
Each time I wanted to spend some hours and tried to look into an issue it was changed upstream faster than I could understand and fix it. So I couldn't even boot my test scripts.
And since device tree was introduced, we are still fixing missing things (which might also have influence on power consumption).
Now you could question the strategy to follow mainline at all. We could have stuck with 3.12 (which supports all features of the device). But to optimize power consumption, you also have to fix things that already have been fixed upstream in 3.13 or later. So you constantly end up in backporting good improvements by the kernel community - which reduces time to work on those specific things that they don't care about. And at some point the dependencies have grown to a level that you can't even backport a fix.
It turned out that it is much less double-work just following mainstream and automatically participate in their improvements.
But as said we have not reached the inner shell of working on power management subtleties and still have to make the mainline kernel working. A simple example: we have no display backlight control.
To improve speed of the car, the steering wheel, the doors, the driver seat and the brakes must work first. Before you dare to test how to make it faster...
So it is more like a car which has been demonstrated that it can run 150 km/h - but the motor control software (who did say v w?) doesn't permit.
> Only person who did progress in this area was Neil.
I wouldn't even say progress since 3.7. The only kernel with good power management is 3.7. Afaik there is no kernel useable for daily phone (because important features are missing) and at the same time has better power management than 3.7. That is indeed our problem.
And the question is why only Neil worked on it? Maybe because he has the deepest knowledge and experience with Kernel development (he is maintainer of the RAID subsystem).
> Honestly I would much more expect that from you as HW guy.
Yes, you can expect everything, but we can only deliver what we are experienced in. Optimizing Linux kernels for power consumption isn't my prcatical core experience. I know how to develop hardware, compile some kernel, install it, fix some bugs, and know the theory of power management. But thats it. Implementation into the kernel is tough.
Well, we have gained a lot of experience by the gta04-kernel project, but it is mainly about writing device trees for the devices and fixing minor bugs in drivers.
The only thing I can do beyond doing something myself is to invite to participate actively in the gta04-kernel project , discuss ideas, solution options, findings, measurement results, contribute patches, work on the known issues, write documentation etc. I.e. everything which makes a community project out of it.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Gta04-owner