BOINC projects with open source code?

[notzed]

I've used this following lib in the past and it was quick - and easy to use and port. It was a while ago but at the time it was certainly very fast on NEON:

https://github.com/anthonix/ffts

edit: looks like it's not very good once the data outgrows the cache, so looks like it needs some work. in-cache though it's quick.
edit2: oops i was timing the verification code, no, it's fast.

It has a code generator so could be modified to create epiphany code directly although code-size might be an issue (and it's not trivial). This sprang out of another project which might be equally applicable to epiphany though:

https://github.com/anthonix/ffts-fpga

An fpga implementation.

This next library(-suite) may have some useful building blocks. Since the epiphany isn't SIMD some of the more exotic libraries shouldn't be necessary for good performance.

http://www.kurims.kyoto-u.ac.jp/~ooura/fft.html

I've done some poking around from fundamentals but nothing i'm coming up with is very fast so far. I hit some hardware/firmware issues which killed the mood a bit.

FWIW ... Although the C compiler does a pretty good job of the arithmetic and scheduling it's not very good at the memory accesses and indexing so i've been playing with some assembly as well although haven't got to the point of timing anything to see if it really matters.

[Bikeman]

Hi

sorry for beeing silent on the beer bet for so long!

I had seriously underestimated the effort needed to program the Pi's GPU, so I must accept defeat since I did not come up with a working PI implementation in time. OTOH I haven't seen a Parallella version for the bet's task either, so while I clearly owe Adapteva a create of beer, the question that the bet set out to answer is yet unsolved.

It's a shame, I'm actually quite close to Adapteva's HQ right now (@ Woburn, MA), and could almost deliver the beer in person, but I'm here on a conference and not (entirely) for fun and I don't think I can make it to Lexington while I'm here. But be assured you'll get that beer

Cheers
HB

[aolofsson]

HB,

Hey, you are basically 5 minutes away from my house. Catch me here or at @adapteva if you want to grab a beer later tonight!

Apologies for dropping the ball on this. For fun, I just ran FFTW on the parallella and got (7.7seconds) running ONLY on the ARM (N=12582912). Same program ram in 0.233sec on my i7. Clearly something is weird with the package or I screwed something up in my test code?). As a side note, now we are also finally starting work on a scalable FFT (I won't be doing this work myself..but I am excited about the approach.)

Andreas

[Bikeman]

Hi Andreas,

Thanks for the invitation, but I couldn't make it (still jet-lagged by 6 hrs).

Runtime on the ARM alone seems to be in he same order of magnitude as for the PI. The Parallela's ARM supports NEON so it it faster even at the same clock rate.

>As a side note, now we are also finally starting work on a scalable FFT (I won't be doing this work myself..but I am excited about the approach.)

Intreresting indeed. FFT is just too important an algorithm to be ignored . So I suggest to make the bet an open ended race. Whenever you have something ready, let me know, and we can compare that to the best I have ready for the PI at that time.

I might be in the Boston area again in about two weeks.

Cheers
HB

[Glen]

New BOINC version 7.4 supports Parallella

[gregb]

[aolofsson]

Yes, great news indeed. Of course, someone still needs to tune OpenCL code for Epiphany..
Andreas

[emerzon]

So, are there any projects that already make use of the Epyphany's cores?

I tried a couple Boinc projects, but they seem only to use the ARM cores.

[keithsloan52]

[Bikeman]

Hi Keith!

Yeah as we discussed in other places, there is still a lot of work remaining to get this going and there might be some problems with the numerical accuracy of the Raspi GPU FFT lib.

Cheers
HB