FAH on A7 chip with Metal and Swift

[TonyStewart14]

I just watched the Apple WWDC conference and they introduced two major innovations for their mobile platforms. The first, Metal, is a replacement for OpenGL and can render graphics up to 10x faster. The second, Swift, is a new programming language that replaces Objective-C and requires less code and is thus more efficient. I think this could allow FAH to finally run on mobile devices such as iPad Air and iPhone 5S. As these devices and the A7 chip increase in popularity, this could be the biggest step forward for FAH since the PS3 client.

Here are some links on it. Any thoughts on this are welcome.

http://appadvice.com/appnn/2014/06/appl ... ics-to-ios

http://appadvice.com/appnn/2014/06/appl ... lled-swift

[Jesse_V]

Metal/OpenGL isn't applicable to F@h. OpenCL is the library that FahCore 17 uses, not OpenGL. Rendering graphics for games and doing general-purpose work on the GPU are quite different.

It's nice to see them introducing a replacement for Objective-C. I've never really liked that language, but I think it's a good time for them to introduce something new.

I doubt F@h is going to run on Swift for the time being, since it would require rewriting the entire client or core in Swift, which is likely a major undertaking. So I'm excited to see this progress in Apple's world and I think it will help them, but I don't see any impact for F@h.

[7im]

Graphics/Gaming is not Folding, so Metal does nothing for FAH.

Swift may have potential. But porting to a new language is a lot of work (for unknown benefit) when the current code already runs so well.

It still all boils down to the folding performance of the chip in the device, and how many hours a day it can contribute, regardless of the code it runs on. It may have potential, but every new code announcement is not a magic bullet.

As in the Android thread, and the other alternate platforms threads, PG watches all of these developments with interest.

To see the future of fah, consider watching the webinar mentioned in the blog.

[P5-133XL]

The picking of a language for folding has very little to do with coding efficiency. Much more important are the libraries and processors it supports as well as cross-platform support i.e. can you use the same code for Linux, Windows, OSX, as well as potentially Android, and IOS then does it also support OpenCl in all those platforms too. If the programmers have to change even small things because of the platform it vastly out weighs coding efficiency.

[TonyStewart14]

Thanks all for your replies; they all clarify it a lot. I think it'll be interesting to see how it plays out for games and app development, and maybe even folding several years down the line.

[bruce]

One more fact that hasn't been mentioned:

OpenCL is a non-proprietary language which is currently supported on NVidia GPUs and AMD GPUs and can potentially be supported by Intel and Apple. That means they only have to write the FahCore code once to support two or 3 or 4 different platforms. Swift is a proprietary language which probably will only run on Apple products. CUDA is a proprietary language which only runs on Nvidia GPUs. Supporting either of those languages means porting the same analysis to two or three languages. While it's possible that the extra development costs plus the extra maintenance costs to support multiple cores that are functionally equivalent is a worthwhile expense, it's unlikely, even if there's s (small ?) increase in efficiency. It's more likely that the money saved by supporting a single FahCore will be better spent on some other aspect of FAH.