Architecture support for disciplined approximate programming

Abstract

Disciplined approximate programming lets programmers declare which parts of a program can be computed approximately and consequently at a lower energy cost. The compiler proves statically that all approximate computation is properly isolated from precise computation. The hardware is then free to selectively apply approximate storage and approximate computation with no need to perform dynamic correctness checks.In this paper, we propose an efficient mapping of disciplined approximate programming onto hardware. We describe an ISA extension that provides approximate operations and storage, which give the hardware freedom to save energy at the cost of accuracy. We then propose Truffle, a microarchitecture design that efficiently supports the ISA extensions. The basis of our design is dual-voltage operation, with a high voltage for precise operations and a low voltage for approximate operations. The key aspect of the microarchitecture is its dependence on the instruction stream to determine when to use the low voltage. We evaluate the power savings potential of in-order and out-of-order Truffle configurations and explore the resulting quality of service degradation. We evaluate several applications and demonstrate energy savings up to 43%.

Load Value Approximation

“…Programming and ISA support have been proposed for identifying approximate computations [2,13]. For example, the EnerJ framework allows Java programmers to declare data as either precise or approximate [35].…”

“…Architectural techniques include novel accelerator designs [14,41]. Work in programming languages include programming frameworks to identify approximate data [5,35], ISA extensions to leverage approximation [13] and compiler support [34]. We focus on a microarchitectural application of approximate computing; our technique has low overhead and does not require radical or complex changes to the processor pipeline leading to a low barrier to entry.…”

Hierarchical Synthesis of Approximate Multiplier Design for Field-programmable Gate Arrays (FPGA)-CSRmesh System

“…And [9] had introduced approximate speculative adders used in the third step of a multiplier. Recently IBM [17], Intel [16], Microsoft [2], [1], [5] and a lot of companies and research groups [20], [3], [15], [18], [26] proposed some effective results on approximate computing. However, most of them are limited to the software applications.…”

A variability-aware OpenMP environment for efficient execution of accuracy-configurable computation on shared-FPU processor clusters

“…To ensure practical use, approximate computation must be controllable at the granularity of instructions [15] given the difficulties in identifying large blocks of 'error-tolerant' instructions. This requires interleaved execution of approximate and precise computations.…”

“…Green [19] also trades off quality of service for improvements in energy consumption, while providing statistical quality of service guarantees. Truffle [15], a dual-voltage microarchitecture design, supports mapping of disciplined approximate EnerJ programs through ISA extensions. It applies a high voltage for precise operations and a low voltage for approximate operations.…”