PROJECT TITLE :

Approximate Computing: Making Mobile Systems More Efficient

ABSTRACT:

Approximate systems will reclaim energy that's currently lost to the "correctness tax" imposed by ancient safety margins designed to forestall worst-case scenarios. Researchers at the University of Washington have co-designed programming language extensions, a compiler, and a hardware co-processor to support approximate acceleration. Their end-to-end system includes two building blocks. 1st, a new programmer-guided compiler framework transforms programs to use approximation during a controlled method. An Approximate C Compiler for Energy and Performance Tradeoffs (Settle for) uses programmer annotations, static analysis, and dynamic profiling to find parts of a program that are amenable to approximation. Second, the compiler targets a system on a chip (SoC) augmented with a co-processor that may efficiently evaluate coarse regions of approximate code. A Systolic Neural Network Accelerator in Programmable logic (Snnap) may be a hardware accelerator prototype which will efficiently evaluate approximate regions of code in an exceedingly general-purpose program.