Parallella Community

Hi all,
I am new to this forum. I saw this forum and Parallela arround 2 week ago. I don't know much about the platform. BUT i have seen the specs and I expect this can be a good platform for heavily numeric crunching application.
One of my near future projects demand such thing( OpenCV based ) second in the line is a setting of an "animation studio" for local movie house using Blender.

I think parallela will be much better option than putting a cluster of PCs for HPC.

Any feedback or thought is welcome See I am a novice to this platform. any pointer will help.

One more thing to do.

I recently find out that the people at povray have taken notice of Parallella, I'll have to try it if they port it.

A while back I wrote a feed forward, back propagation neural network simulator.

I've been watching the Multicore Programming Primer from MIT on iTunes U and if I'd know the basics of parallel programming I'd have done it very differently. They were using the IBM Cell processor but a lot of the material is general in nature. One tip was that starting up a core by creating a new thread is expensive. Therefore you should design it to run for as long as possible before exiting. For those who know about this type of neural network, I created a thread for value passing along every connection and the same in reverse on the training pass i.e. multiplying two numbers together.

A better approach would be to to farm out the weights on the connections between nodes to each core and feed in the values to each core so that it can multiply all of the output values and then pass the result onto another core to generate the output value of the next node for processing on the next core. Perhaps... I have not really thought through the whole process yet.

opps, wrong forum

I'm extremely interested in the DSP applications, but being RISC cores vs dedicated DSP cores, even with the two float intruction per cycle, I'm very curious about this. The setup seems tailored to video/audio, but the limited ram would require some playing around.

Can't wait to try something new.

Hello All,

Right now, I am doing my master thesis on power-aware scheduleing for many-core real-time systems. I have decided to use parallella board to implement my scheduler (hoping that I get the board in time).
I have been thinking to use the Epiphany in it's bare metal mode and program it directly using C, because the scheduler needs to have full control on the hardware and it should know which core is running what application to be able to meet the application's deadlines.
On the other hand, Parallella is able to support OpenCL. I am wondering if I could use OpenCL as my framework to implement the scheduler. Will I have the required control on the hardware?
I would appreciate any information and guidelines on this subject.

Thank you,

Hi all!

My vote goes for XBMC media center and a real desktop pc alternative with a low power consumption profile.

cheers!