Would it be possible to build the application for generic ARM Linux platforms like arm-unknown-linux-gnueabi (without hardware support for floating-point, like ARMv5) and arm-unknown-linux-gnueabihf (with support for floating-point, like ARMv6)?

Though ARM is not a high-performance processor by today's standards, it may be as fast as the typical PC of a few years ago and comparable to a current Intel Atom.

Other projects like Enigma, OProject, QCN, Radioactive, Yoyo and WUProp already provide an application for such platforms.

I've helped out other projects getting the applications built and tested, as can be seen here. Please, let me know if I can help.

TIA

---

Would it be possible to build the application for generic ARM Linux platforms like arm-unknown-linux-gnueabi (without hardware support for floating-point, like ARMv5) and arm-unknown-linux-gnueabihf (with support for floating-point, like ARMv6)?

Though ARM is not a high-performance processor by today's standards, it may be as fast as the typical PC of a few years ago and comparable to a current Intel Atom.

Other projects like Enigma, OProject, QCN, Radioactive, Yoyo and WUProp already provide an application for such platforms.

I've helped out other projects getting the applications built and tested, as can be seen here. Please, let me know if I can help.

TIA

My understanding is that there are currently at least two theoretical problems with running Rosetta on ARM, and possibly more fundamental practical problems:

First, as you've already mentioned, ARM CPUs are comparatively weak, especially at FPU calculations (not sure of different relative throughputs for the different cores, but the Raspberry PI team say their CPU is comparable to a Pentium II 350MHz, so something like 0.2% to 1% the FPU performance of an i7-2700k)

Second, they tend not to have enough RAM. Some more modern kit might be ok on that front - my phone has 2GB with 4 cores which would appear to be enough if comparable to x86, but the vast majority don't have that much.

Probably more importantly though, I believe Rosetta is over 1 million lines of code, so I wouldn't be surprised if it's the biggest code-base of the BOINC projects, so it has been stated numerous times in the past (mainly for GPUs) that it's not a trivial task.

I'm not saying it's not possible, or that it's not worth doing; just that it doesn't appear that it would be anything like a quick and easy port if the lab were to release the code for that.

I would think a good target for expansion would be the consoles though as they're x86.

---

Would it be possible to build the application for generic ARM Linux platforms like arm-unknown-linux-gnueabi (without hardware support for floating-point, like ARMv5) and arm-unknown-linux-gnueabihf (with support for floating-point, like ARMv6)?

Though ARM is not a high-performance processor by today's standards, it may be as fast as the typical PC of a few years ago and comparable to a current Intel Atom.

Other projects like Enigma, OProject, QCN, Radioactive, Yoyo and WUProp already provide an application for such platforms.

I've helped out other projects getting the applications built and tested, as can be seen here. Please, let me know if I can help.

TIA

Have you been to WCG yet? They are an IBM run project and have several smaller projects they run under their umbrella. Some are short in duration and not too complicated so could be ideal if enough people were to crunch. Rosetta has been resistant to much change for a long time, not that they never will, but they say 'things are working just fine for us' when people ask about new things.

GPU and standard CPU computing will soon be a thing of the past. There is a company call Adapteva that will, if things stay on schedule, be releasing a 1024 core FP chip in 2014. Although the chip does not have much memory like a gpu/cpu, its cost is much less and the cost to operate it is significantly less.

For example, Nvidia released a Titan GPU video card. This card costs $1K. It has 2688 cuda cores and a tflop rating of about 4.5Tflps. It needs a 600w power supply to run it as I understand. On the other hand, once Adapteva releases their chips, you could buy them way, way, way cheaper and they will consume much less power. In fact, the power savings will allow you to afford more chips to add crunching power as you go.

Now, these chips need something to feed data to them. They are parallel processors. Adapteva has gotten rid of all of the overhead instructions and made their chips just for pure computation. A computer could download a Work Unit (WU) and load each parallel processor with a specific task. Once finished, it reports back to the computer for another assignment. You could conceivably have 500,000 cores siting next to your computer that will do some serious number crunching.

Now, I understand Rosetta may be a large program, however, like any serial programed cpu/gpu, they don't run every piece of code at the same time. They do it in a step by step fashion. This type of programing can always be reprogramed to work in parallel for the most part.

Rosetta needs to take a serious look into the near future and see where processing power is headed. Get ahead or be left behind, especially with us WU junkies.

Brian

Rosetta needs to take a serious look into the near future and see where processing power is headed. Get ahead or be left behind, especially with us WU junkies.

I think this will be a piece of future: HSA Foundation

---

GPU and standard CPU computing will soon be a thing of the past. There is a company call Adapteva that will, if things stay on schedule, be releasing a 1024 core FP chip in 2014. Although the chip does not have much memory like a gpu/cpu, its cost is much less and the cost to operate it is significantly less.

For example, Nvidia released a Titan GPU video card. This card costs $1K. It has 2688 cuda cores and a tflop rating of about 4.5Tflps. It needs a 600w power supply to run it as I understand. On the other hand, once Adapteva releases their chips, you could buy them way, way, way cheaper and they will consume much less power. In fact, the power savings will allow you to afford more chips to add crunching power as you go.

Now, these chips need something to feed data to them. They are parallel processors. Adapteva has gotten rid of all of the overhead instructions and made their chips just for pure computation. A computer could download a Work Unit (WU) and load each parallel processor with a specific task. Once finished, it reports back to the computer for another assignment. You could conceivably have 500,000 cores siting next to your computer that will do some serious number crunching.

Now, I understand Rosetta may be a large program, however, like any serial programed cpu/gpu, they don't run every piece of code at the same time. They do it in a step by step fashion. This type of programing can always be reprogramed to work in parallel for the most part.

Rosetta needs to take a serious look into the near future and see where processing power is headed. Get ahead or be left behind, especially with us WU junkies.

Brian

Heterogeneous System Architecture (HSA), maintained by HSA Foundation, is a system architecture that allows accelerators, for instance, graphics processor, to be a first-class component as CPU in the system. HSA-compliant accelerators eases the programming of heterogeneous applications in various aspects, including being ISA agnostic for both CPUs and accelerators, supporting high-level programming languages, having the ability to access pagable system memory, and maintaining cache coherency for system memory with CPUs. HSA is widely used in System-on-Chip devices, such as tablets, smartphones, and other mobile devices. HSA allows programs to use the graphics processor for floating point calculations without separate memory or scheduling.

That does not mean, that the processor would have a huge number of cores, but the processor and graphics card are able to share the same resources. The same resources = processor is able to take advantage of the graphics card FPU units in the calculation, if the executable program that supports this.

My 2 cents on this issue is that both Boinc and Rosetta@home need to fold in both Arm and Atom devices into the supported mix. These devices are going to proliferate much more widely and quickly than than big performance platforms in the coming decade.

I've always enjoyed supporting science and research this way - but as we enter our 'later years' I can see us down sizing. Which means no space for the odd tower system churning away. I'd much rather 3 or 4 mobile devices - or Raspberry Pi like things - taking up less than say 6 to 8 square inches of space and using less power. And have churning away 365 24/7.

Its gotta happen ......

Ok, this post is not strictly on ARM device, but from HCC (Help Conquer Cancer - WCG project) forum, an admin post:

To be honest, the combined production of the GPU's surprised us as well. We had anticipated that with the launch of GPU we would double the daily output. instead it increased by a factor of 16-17 (before GPU we were completing ~60k/results per day with 1 image per job, we are now >500k/results per day with 2 images per job

I know rosetta is a big platform of tools, codes, projects, but i think that a little sub-project with a very specific kind of code and a specific protein target, may run on arm/gpu/fpga/etc with a great benefit.
OpenCL may be the first step.

---

Ok, this post is not strictly on ARM device, but from HCC (Help Conquer Cancer - WCG project) forum, an admin post:

To be honest, the combined production of the GPU's surprised us as well. We had anticipated that with the launch of GPU we would double the daily output. instead it increased by a factor of 16-17 (before GPU we were completing ~60k/results per day with 1 image per job, we are now >500k/results per day with 2 images per job

I know rosetta is a big platform of tools, codes, projects, but i think that a little sub-project with a very specific kind of code and a specific protein target, may run on arm/gpu/fpga/etc with a great benefit.
OpenCL may be the first step.

Rosetta has said in the past that their way of getting the result did not lend itself to porting it to a gpu, your idea of 'a different kind of unit' may get them thinking though. The idea would have to percolate thru and may take awhile to go up thru all the dusty halls and preconceived notions and then to come back down again to the programmers and finally to us users. Rosetta does not appear to be a nimble, quick change, pivot and move forward type of place. They do things 'the Rosetta way' and their 'always have and always will' type of thinking will have to change for this to happen. But I DO think time is on your side and it will happen, eventually.

Rosetta does not appear to be a nimble, quick change, pivot and move forward type of place. They do things 'the Rosetta way' and their 'always have and always will' type of thinking will have to change for this to happen. But I DO think time is on your side and it will happen, eventually.

From Ralph@Home forum, ad admin:

That said, one of the researchers in the lab was/is working on a "skunkworks"-type project regarding incorporating the use of GPUs into Rosetta - though I hesitate to mention this because it's very uncertain if anything will actually come of it, or if it does, when. Even if there are parts of Rosetta that are GPU-ized, there's no guarantee that it would be those parts which would make a significant difference in the typical runs performed on Rosetta@Home

But this answer was september 2011, i don't know the progress...

---