Retargetable code optimization with SIMD instructions

Abstract

Retargetable C compilers are nowadays widely used to quickly obtain compiler support for new embedded processors and to perform early processor architecture exploration. One frequent concern about retargetable compilers, though, is their lack of machine-specific code optimization techniques in order to achieve highest code quality. While this problem is partially inherent to the retargetable compilation approach, it can be circumvented by designing flexible, configurable code optimization techniques that apply to a certain range of target architectures. This paper focuses on target machines with SIMD instruction support which is widespread in embedded processors for multimedia applications. We present an efficient and quickly retargetable SIMD code optimization technique that is integrated into an industrial retargetable C compiler. Experimental results for the Philips Trimedia processor demonstrate that the proposed technique applies to real-life target machines and that it produces code quality improvements close to the theoretical limit.

Instruction Selection

“…For example, there exist a number of methods-many of which are based on polyhedral transformations-for making efficient use of such instructions (see for example [10,39,180,206,213,214,323,342]). Common to most is that they rely on aggressive loop unrolling and are thus restricted to handling only these instructions.…”

Ingredients of Adaptability: A Survey of Reconfigurable Processors

“…To take advantage of the spatial and temporal parallelism of an application, SIMD instructions [26] and fused custom instructions [27] are used. Such instructions basically reduce the decoding overhead and accomplish more operations in one go.…”

A SIMD optimization framework for retargetable compilers

“…Therefore, our approach emphasizes efficient utilization of SIMD instructions and compiler retargetability at the same time. Our SIMD optimization comprises a loop-vectorizer and a unroll-and-pack based technique (an earlier version of the latter is described in Hohenauer et al [2006]), which are both driven by the same SIMD specification. The retargeting formalism is fully integrated into the compiler backend specification.…”