Synergistic Approximation of Computation and Memory Subsystems for Error-Resilient Applications

Approximate computing is a new design paradigm that exploits the <italic>intrinsic error resilience</italic> exhibited by emerging applications to significantly improve their energy efficiency and performance. Prior work in this domain has proposed approximation techniques targeting either the computational subsystem or the memory subsystem. For the first time, this letter proposes a methodology to perform synergistic approximations across the computation and memory subsystems together that results in a significant improvement in energy consumption compared to the case when the subsystems are approximated independently. We implemented our proposed methodology using an Altera Stratix IV GX FPGA based Terasic TR4-230 development board containing a 1GB DDR3 DRAM module, which executes three error-resilient benchmarks. Experimental results demonstrate energy improvements in the range of <inline-formula> <tex-math notation="LaTeX">$2.05\boldsymbol {\times }$ </tex-math></inline-formula><monospace>-<inline-formula> <tex-math notation="LaTeX">$4.45\boldsymbol {\times }$ </tex-math></inline-formula></monospace> for minimal loss in application quality (<monospace><1%</monospace>). Compared to individual approximations, our technique achieves an additional <inline-formula> <tex-math notation="LaTeX">$1.6\boldsymbol {\times }$ </tex-math></inline-formula><monospace>-<inline-formula> <tex-math notation="LaTeX">$2.35\boldsymbol {\times }$ </tex-math></inline-formula></monospace> energy savings for the same quality specifications.