Parallella Community

How much is network latency per byte in epiphany chip?
I use e-get-global-address to find other core array, but ot seems that has a lot latency. How much should it be?
which points should consider when using noc in order to being efficient?

Thanks

As I read page 22 of epiphany_arch_ref.pdf file, it says about cMesh:
"The cMesh network connects a mesh node to all four of its neighbors and has a maximum bidirectional
throughput of 8 bytes/cycle in each of the four routing directions. At an operating frequency of 1GHz, the cMesh network has a total throughput of more than 0.5 Terabit/sec."
but in paper titled "Programming the Adapteva Epiphany 64-core Network-on-chip Coprocessor " published in 2014 the following diagram is presented:

communication_rate.JPG (40.14 KiB) Viewed 10535 times

based on this report for message size below 50 byte the bandwidth is less than 90MB/s.

I want to know is there any way to make it more efficient, because I have an algorithm that need high communication rate between cores in 8 stage.

thanks

Also is there anyone who knows how memcpy function work and how we can make it more efficient. For example if this function copy data byte by byte force it copy data eight byte by eight byte, that can increase communication rate.