In spite of the fact that the first graphics cards came on the scene in the middle of the 80’s and hardware acceleration of computer algorithms is a longstanding idea, this field of researches is still current nowadays. Because of CPU frequency was near the bound of physic at the turn of the millennium, it became evident that the only way to further output increase can be achieved by architectural development.
For special problems already excellent hardware acceleration was found out (for example grapics cards and signal processing CPUs), that’s why my dissertation is based on these common hardware accelerations on the evidence of technical literature. After a short introduction, I grouped and showed the already appeared ideas, which are the basis of this topic, and I summarised the advantages and disadvantages of the mentioned groups.
Next I sketched an idea, which is able to increase output significantly without altering (parallelizing) the present-day algorithms, moreover it is invisible for software developers.
I made measurements on a self-assembled experimental system to verify the ability of it in practice. I examined the difference between the output increase in case of a merely software based implementation and hardware acceleration via three different algorithms.