Not new to folding - question about f@h Android client

[Hameed]

What is the timeline for updating the f@h Android client to have the following functionality?

• Flexibility to allow setting start and end time
• Removal of the requirement to have the battery of my smartphone fully charged
• Allow me to contribute even when I'm connected to Wifi
• Allow me to specify my user name, team and passkey

Thanks.

[7im]

Hello Hameed, welcome to the folding support forum.

This is welcomed feedback, but not really a support question this community can answer. Feedback and questions for the fah developers on this client may be sent to these locations:

Send email to: apps@sonymobile.com
or
https://www.reddit.com/r/foldingathome/ Post feedback and questions
or
https://play.google.com/store/apps/deta ... ting&hl=en Write a Review

[gwildperson]

Those are good questions, but you'll need to ask Sony or at https://www.reddit.com/r/foldingathome/

I wonder if it will run on the new Samsung hardware. (They will be using a custom CPU).

[Hameed]

Thanks. Will do

[eric2kwok]

My Samsung Galaxy C5 Pro I tried to install folding@home android but doesn't work. If I installed to Samsung Galaxy Note 3 with no complaint. I feel quite strange.................

[Joe_H]

My Samsung Galaxy C5 Pro I tried to install folding@home android but doesn't work. If I installed to Samsung Galaxy Note 3 with no complaint. I feel quite strange.................

If you read the newer topics there is information about newer phones. Basically, the closer a phone is in hardware to an existing Sony model, the more likely the Android client will work. When Sony ended its development and open sourced the code they had not made changes to support processors like in your C5 with a mix of low and high power processor cores. Changes might be made in the future to support this, but I have not seen any announcements.

[bruce]

I've heard that Sony has withdrawn their support. That probably means that somebody will need to rewrite the client from OpenSource repository. (That probably would have been a good idea some time ago.)

Do I hear any volunteers.

[ChristianVirtual]

I have two questions: how serious are the result taken as now in open source the integrity need to be checked only in the backend and second without integration into regular points system (as BOINC on Android) I don’t see much benefit.

[bruce]

1) BOINC found one solution to that problem: They don't believe any result unless it's validated by running it at least twice. That wastes at least 50% of the resources spent on the project, so I don't like that solution, but it could be enforced on an Open Source client. Hopefully there will be more efficient ways of validating the results.

FAHCore_21 does run periodic validations so that if at designated intervals, the partial results can be aborted if they've already gone astray. That can save a fraction of the processing of an errant WU but probably wouldn't help if an evil person is intent on submitting false results.

2) If I were (re-)writing the Android client, I'd certainly enhance it's ability to work on a wider range of phones. I'd also add the code to generate the appropriate points that could be integrated into the existing credits. The PG would be responsible for accepting only those oints from WUs that pass whatever validation methodology is adopted.per question 1.

That's just IMHO, not anything official.

[JimboPalmer]

(As always, I am just a Donor, and have no inside information)

One solution that the current cores use is to take Open Source GOMACS and wrap it in a proprietary core and only accept the proprietary core. That avoids false flag cores that claim to be doing more then they do.

I have not looked at the current Android code but I would guess that it has the following limits: all CPUs must be the same model running at the same speed, and I question if it is using GPUs at all. (ARM GPUs have very few shaders, 1 to 32, so will be too weak to fold, I think)

An all new mechanism to find the fastest CPUs and use all the CPUs that match that specification would address some of the current limits. It is possible some phones could also do useful work using the weaker CPUs in a different WU. (The issues with sharing computation with CPUs of differing speeds is HARD)

The ARM CPU architecture that causes grief is called big.LITTLE and comes in 3 flavors:
a) we have x low power draw CPUs, if we need heavy computing, we use x high power CPUs instead. Android F@H would need to switch to high power mode, then stay there until the WU is done. The low power CPUs and high power CPUS never run at the same time. This is the easiest.
b) each lower power CPU has a high power CPU that can run instead. In this scenario we need to force each Pair into high power mode and keep all the pairs in high power mode for the entire WU. This is harder, as we have x pairs we need to manage as opposed to one switch above.
c There are some low power CPUs and some high power CPUs and we can run all of them at the same time if we want. This will be hard, we need to find all the high power CPUs and run one WU on them and find all the low power CPUs and run a different WU on them. WUs will not finish at the same time. The system will be fighting you to swap all the F@H jobs to a high power CPU, we may have to admit we just can't use the lower power CPUs even though they are available.

https://en.wikipedia.org/wiki/ARM_big.LITTLE
https://en.wikipedia.org/wiki/Mali_(GPU)