Folding on Android

[bruce]

I've explained the serial simulation problem previously in terms of a relay race. (You can search the forum.)

Bottom line: A running team cannot compete in a 20 mile relay race by letting 100 walkers cover the first 0.2 miles. Nobody ever reaches the end because nobody can start from the first 0.2 milepost until the first walker reaches that point. Many slow computers cannot make up for missing faster computers because the simulation, itself, is serial and you have to know how step N ends before anybody can start on step N+1.

Those 10 runners that can complete a mile in maybe 6 or 7 minutes are worth a lot more to the 20 mile relay race than any number of walkers. They can get the baton to the finish line in a couple of hours whereas none of the walkers will complete the course, no matter how long you give them. For FAH, speed is more important that simply completing a small bit of a WU.

[poppakap]

Ok, good, because that wasn't my intention, it was (and is) to stir the pot.

I guess the short answer to that question is, 'not so much.' And so, I do have a lot to study up on. I am here merely out of burning curiosity. I have a very deep desire to understand the code aspect of this. It is absolutely fascinating, and yet is the science part that also drives me to learn more and more.

On the other hand, I do have a couple decades of heavy coding experience under my belt. And that is the part of me that's trying to push the envelope, if you will. I understand that there are many people in this field that are much more intelligent than I am and I'm fine with that, even at the risk of being laughed at.

If there is one thing that I have learned time and time again, it's that you have to question everything, like, I never say, 'no we can't do this and that.' There's always a way... it just hasn't been developed yet. And I promise you, as sure as I am breathing, there's a way to have devices all over the planet folding in harmony... even toasters if we wanted to.

And I'm going to make another statement that will be sure to infuriate many of those involved, and cause great commotion and ruckus, a rain of you-dont-know-wth-youre-talking-about replies; If the only answer is more hardware, then the software design needs another examination. There. I said it, that's right. Now everyone unleash their wrath!

[7im]

Almost all of the FAHCores run based on code from Gromacs.org and requires SSE based code optimizations. Diving in head first over there (open source) would help explain a lot about folding.

[poppakap]

Hi bruce, I was looking at it as parallel race, in which all 100 walkers completed at their own speed.
If not, then why should I contribute anything? If I can only put out 2k per day, why bother at all when someone else can do 1 million?

[P5-133XL]

Because everything helps.

[poppakap]

Thanks man, I was starting to feel a bit inadequate there for a moment.
I'm curious how small a work unit can be, and if perhaps there were a way to assign the tiny ones to something like an android (in theory) and let the pc's and mass StorageForum.nets do the heavy lifting.
It does seem as though I get the smaller wu's < 5mb so I'm thinking there is already a sized-based assignment already built into the logic. I'm also curious as to the factors that determine the size, and what the data returned consists of. But, I'll do more research.
Gracias

[poppakap]

I think I understand the factors now... disregard that part!

[bruce]

The science of Folding@home has various needs. A very important goal is to create very long trajectories in a short period of elapsed time. That means fast computers (faster than the best of today's technology allows) running continuously, so my example of a 20 mile relay race might actually be more like a cross-continent relay race than only 20 miles. To meet that objective, the speed of each leg of the journey is more important than the number of runners.

Nevertheless, that's not the only goal. There are other important goals that can be achieved in parallel, but there, too, speed is important.

More than ten years ago, FAH started with the computers of that day and eventually settled on a minimum platform of (more or less) a Pentium 4 or an Athlon. Hardware has made great strides since then, but there's still work that can be done by those computers (if anybody still has any).

Over the years, most of the development work has focused on newer/faster platforms which can be helpful on projects like a cross-continent relay race. Recently, however, there have been new development efforts that have focused on untapped resources nearer the low-end. My example of 100 walkers that can only manage 0.2 miles or maybe 1.0 miles can contribute useful work as long as the work assigned can be matched up to the effectivity of the hardware. FAH has not done a very good job at utilizing those walkers but I predict they soon will.

The NaCl client is a relatively new development. The current NaCl client only runs on Windows/Linux/MacOS inside of the Chrome browser. The WUs being assigned can be completed by those PCs in a hour or so rather than requiring continuous processing over several days, which is a boon to people who don't choose to fold 24x7. Stanford also has other developments in the pipeline but there's no prediction when they might surface or if they'll fit your objectives. Also, I'm not sure where the Android platform or the iPhone/iPad fits in processing power (compared to a low-end desktop/laptop PC) or whether there might be a plan to develop clients for either the CPU or the GPU on other operating systems.

Stanford rarely pre-announces their plans. I suggest you sign up for Tuesday's webinar and get the official word rather than my naive predictions about the future.

[panthro1971-2]

Just wondering, because I will get an Android mobile.....Would be possible?

[P5-133XL]

Not yet.

While there is a pNaCl version, they do not have ARM support.

[panthro1971-2]

Can not be done accesing the webpage only? Not installing the client per se

[7im]

No, the code still needs specific processor instructions to run folding@home. The ARM chip doesn't have those cpu instructions, and the fah client hasn't been ported to ARM yet.

[panthro1971-2]

Oh ok.....Maybe in a future update then. Or maybe never..... .....

[PantherX]

From this blog post (https://folding.stanford.edu/home/addin ... e-browser/):

Next steps. Finally, as many will probably notice, this opens the door to naturally moving Folding@home from the browser on a computer to a browser on a phone or tablet. We are actively investigating those possibilities.

[bruce]

The most recent information I've seen is that the simultaneously keeping a guest application (like a FAHCore) in the sandbox (to prevent malware apps) is going to be very difficult in Android. "The interprocess communication model is extremely different on Android and since NaCl must reside in it's own separate process... there are lot of tricky pieces which will need to be implemented in completely different way on Android."

This was 9 months go and I'm not aware of any recent progress from the Developers at Chrome.
https://groups.google.com/forum/#!topic ... EZP458cVQY

If anybody has more recent news, please post it.