Modern processors incorporate SIMD instructions to improve the performance of multimedia applications. Vectorizing compilers are therefore sought to efficiently generate SIMD instructions. With the existence of different families of SIMD instruction sets, the task of compiler writers is more complex. Moreover virtual machines, such as JVMs, are currently widely used for increasing the portability of programs across different platforms; performing SIMDization on these virtual machines would further require 'fast' compilation. This paper selects an efficient retargetable compilation technique, based on tree-pattern matching, which generates efficient SIMD code on static compilers, and studies its utility on the Jikes RVM. The paper extends BURS system used in Jikes optimizing compiler accordingly, and adds new rules for manipulating subword data for the IA-32 architecture. Initial experimental results show an overall speedup at runtime despite dynamic compilation overheads.

International audienceEmbedded systems such as mobile devices are currently ubiquitous. The performance potential of these devices is rapidly improving by incorporating multi-core and GPU technologies, and is rapidly catching up with the workstation platforms. Nevertheless, the heterogeneity of the underlying hardware as well as the low-power constraints severely limit performance portability. In this paper we consider the case of leveraging JIT compilers to provide portable parallelization while hiding the corresponding expensive runtime analysis. We propose a novel lightweight JIT framework that exploits the device idle time and the large storage space generally available on these devices. The framework performs 'incremental' analysis while the processor is idle (such as during charging time), and exploits the storage space to cache intermediate analysis results. Such approach requires reengineering existing complex optimization analysis methods. For this paper, we focus on the traditional loop parallelization analysis, and implement a working prototype into the LLVM framework, integrating a lightweight dynamic profiling method to identify hotspots. Initial results demonstrate the low overhead of our method for parallelizing simple loops on an embedded GPU