It looks like you're new here. If you want to get involved, click one of these buttons!Sign In
It looks like you're new here. If you want to get involved, click one of these buttons!
Based on my understanding, the NCS has two level of memory: One is the global memory (4Gb LPDDR) and local buffer/cache inside the SHAVE processing units.
Since mvNCProfile can report the bandwidth number, I'm wondering this number get from the real data transmission from global memory to Myriad-2 SOC, or just the ''volume of data for a layer/layer processing time''
Also, it would be good if anyone can provide a dataflow chart showing what's the dataflow between global memory, on-chip buffer, and processing engine, etc.
Thanks for your help~