Nowadays the FPGA based microporcessor systems are very popular and widely used,
because of the developement of the technology it is possible to create multi-core or many-
core system in one chip. The computing performance of these systems is highly depends
on the communication method between the CPU cores, therefore in my thesis I studied
the scaling of systems characteristic features which were implemented in FPGAs.
I describe the features of the frequently used multi-core systems and the communication
possibilities between them. I introduce the conditions of the efficient data exchange based
on the graph theory, with them it is possible to evalute the different communication
topology’s efficiency.
To create a processor network I introduce a self developed communicaton modul, called
Router. With the help of this hardware modul it is possible to easily build up a
communication channel between even more CPUs. So I could create multi-core and many-
core systems in simple way. In these systems I measured the maximum working frequency,
occupied slices, power consumption and the runtime of the translator’s different processes
depending on the number of the processors.
The measurement results are given in table and data chart too, for the better
understanding trendlines were fitted to charts. The description of the results follow the
interpretation and the deduction of consequences, then the comparison with the other
topologies.
The measurements were made for ring, star, 2D mesh and torus topology many-core
systems.