Multi-core, mutli-thread, multi-processors, lower power, more performance, shorter response time, higher throughput. These are all the sales pitch and potential of these new technology. Be it Nvidia with a large number of small cores, or the Monarch with it is local embedded memory or the Intel architectures or the Snapdragon, multi-cores are the new approach.
This blog hopes to provide information on how to best architect systems using multi-core. This includes techniques to partition code into concurrent operations, select the right hardware platform configurations and mapping of the software threads on to individual cores.
Multi-cores being new require new metrics. For multi-application spaces with small data, old adage of instructions per cycle and instructions per watt are sufficient. For data-intensive applications, can it be video frames per second or effective memory throughput or …..
If there are any subjects of interest, feel free to contact us.