an hint to the next generation computational architectures

Computational architectures are loosing their traditional structure. Nowadays there is no CPUs with a single core, a CPU has definetly become a computing system. The more technology advances the less computational structures (in its elementary devices and their interconnections) can be complex. An interesting approach to HW automatic synthesis at nano-metric scale are regular fabrics that could represent, in the next future, the right way to build complex architectures by exploiting regular interconnection among homogeneous logic blocks . Such an approach is very similar to the concept at the base of programmable logic.

FPGA are very interesting by the research point of view and well exploited in some critical applications, above all in the telecom domain. A post completely dedicated to the FPGA will be written in the next future.I personally think that every computational architectures will be equipped with a programmable core in the future, because this is the only way to "respect" the Moore law and to build faster and faster processors.
Since "yesterday", ... :) ... in my opinion the revolution has already begun( see terascale,cell processor, IAAL) ... , processor performance gain was provided by component miniaturization which granted a shorter gate delay. Wire delay, instead, has always remained at the same level causing great difficulties in communication and synchronization (by a consistent clock signal on the chip surface). The physical problems due to high resolution photolitography will by described in next posts ...
On the other hand, today, the performance improvement is reached by evolution of architectural paradigm: multi-cores, DSP array, and finally reconfigurable array.