Is parallella based on the original shared-memory IPC?

[fmotta]

Unfortunately, I have not found any documented on parallella yet.
I have worked with advanced multi-core systems that provide hardware queues for inter core communications (IPC) and they allow for scalability closer to the 1:1 levels. With the traditional shared memory IPC the non-parallelizable portion of the overall application is increased and generally one is lucky if they can achieve 0.65:1 scalability.

Does the parallella provide any solution for this problem or will we only get ~60% value of each additional core running?

Thanks

[aolofsson]

We have demonstrated 1:1 scaling for examples like matrix multiplication and filters, but the actual speed up achieved will depend heavily on the application and the skill of the programmer. Documentation can be found here:

http://www.adapteva.com/support/docs/e3 ... nce-manual

[fmotta]

[parallel-future]

[fmotta]

[parallel-future]

[fmotta]

[jim.uazu]

[fmotta]

[jim.uazu]