Optimal multi-core processors for fast discrete cosine transform

This paper explores optimal multi-core processors for FDCT (fast discrete cosine transform), which is a kernel of JPEG image compression, by evaluating the effect of different numbers of processors and memories for a fixed image size in terms of area efficiency and energy efficiency. The different numbers of processors and memories have significant impacts on the overall area and energy efficiencies of the computational multi-core processors. Experimental results using six different processing element (PE) configurations indicated that PEs=4,096 and PEs=1,024 yield the highest area efficiency and energy efficiency, respectively.