Parallella Community

distributed.net has source code for their RC5 and OGR projects here:

Just had a look at Einstein@home on my Cubieboard2 and Raspberry Pi. The executable einsteinbinary_BRP4_1.06_arm-unknown-linux-gnueabihf is 7694642 bytes i.e. 7.33 MBytes so is not going to run on a Parallela Epiphany Core with just a recompile.

When running top shows it with 95228 Virtual and 89m Res

As has been said previously maybe enigma@home is the best boinc candidate.

@keithsloan52: thanks for posting that! I was also thinking about trying Einstein and Milkyway when I'll buy the cluster kitt
So has anyone run E@home or MW@home on a board or a cluster?

BOINC is installed on my board and is currently crunching away on the 2 ARM cores.

I would be delighted to do any alpha/beta testing for any project. Just ask.

That's awesome!!! Fantastic to see that you didn't give up on the board and got it up and running.

Really curious to hear thoughts from hewsmike and everyone else regarding how to best move forward with BOINC. How can we make Parallella really effective for BOINC?

Is next task an FFT interface? If so, what does it look like. Is anybody working on a scalable FFT library already or should we make it our responsibility at Adapteva to provide one.

Inputs appreciated.

Andreas

Over the last 6 weeks I've got my self a Raspberry Pi, and have been playing around getting Boinc compiled on it, as in Boinc 7.2.42 the latest non-development version,
have my Ubuntu Laptop and the Pi running my Boinc 7.2.42 compilations, the Pi runs Einstein, and Albert at home's apps O.K, been trying to compile a Stock Seti app,
and the Lunatics optimised variant for Arm v6 without much success, Jackyman38 had the same issues as me with the seti source, and has submitted a fix
to the Debian Seti source enabling armhf to be built again (I also asked for a change):

http://packages.qa.debian.org/b/boinc-app-seti.html

https://launchpad.net/ubuntu/+source/boinc-app-seti

Jackyman has managed to get an app running on lubuntu armv7-a, but not on the Pi yet (there is a Daniel Carrion setiathome app, but it is really slow, something like ~600 hours a workunit), for comparison my Nexus 7 running one of the Android setiathome apps is taking around 30 hours per workunit, (not exactly a fair comparison, Arm v6 hf without neon, but with vfp at 700MHz, against Arm v7 with Neon, vfp and vfpv3 at 1200MHz), hopefully there'll be more progress getting seti running on different arm devices over the coming months, and Parallella once Joe, Raistmer, Richard and myself get our Parallellas.

Claggy

Would it help if you had remote access to a physical board?

Hi

I'm currently stress-testing my brand new Parallella board with Einstein@Home, it runs out-of-the-box with BOINC installed from the package manager.

So far, so good. Performance wise, Einstein@Home on the Parallella's ARM CPU seems to run faster by a factor of ca 1.4 (per core) compared to the single core Raspberry Pi (overclocked), so considering the dual core , the Parallella's ARM CPU is a bit less than 3 times more productive than a Raspberry Pi wrt. Einstein@Home. It's also comparable in performance to mid-range Android smart phones, btw. , which are at a disadvantage because Android uses the 'soft float' ABI.

Given this moderate performance, I'm puzzled why the Parallella's ARM CPU+FPGA needs the amount of cooling that it does. The Raspberry PI's SoC package of CPU, GPU and memory, all packaged in one stack of chips, doesn't even need a heat sink, and the same holds for comparable SoCs in smart phones.

HBE