Introduction to Multicore Systems On-Chip
Abderazek Ben Abdallah · Atlantis ambient and pervasive intelligence · 2013
Semiconductor companies and processor architects have historically invested time and money in micro-architectural and performance enhancements. Many of these efforts such as deep pipelining, increased large cache size, and sophisticated dynamic Instruction Level Parallelism (ILP) extraction exhibit diminishing returns due to increased area and power consumption. When considering the limitations associated with voltage supply scaling, threshold scaling, and clock frequency scaling, along with the above design complexity, architects were already looking for an alternative to the single core approach. Multicore was therefore the natural next revolution in staying on the ever increasing performance driven curve. But, was it really the good timing to switch from uni-processor approach to the more complex parallel structure of multiprocessor/multicore platforms? The direct answer from major hardware companies was very clear: yes; it is time for revolution and not for evolution! This decision was fueled by the shift that started from around 2004 when market leaders in the production of general purpose computer systems and embedded devices started offering an increasing number of multicore chips, contributing to an unprecedented paradigm that has led to what is know today as multicore revolution. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.