Energy consumption analysis of embedded programs necessitates the analysis of low-level program representations. This is challenging because the gap between the high-level program structure and the low-level energy models needs to be bridged. Here, we describe techniques for recreating the structure of low-level programs and transforming these into Horn clauses in order to make use of the CiaoPP resource analysis framework. Our analysis framework, which makes use of an energy model we produce for the underlying hardware, characterizes the energy consumption of the program, and returns energy formulae parametrised by the size of the input data. We have performed an initial experimental assessment and obtained encouraging results when comparing the statically inferred formulae to direct energy measurements from the hardware running a set of benchmarks. Static energy estimation has applications in program optimization and enables more energy-awareness in software development.

This article examines a hardware multithreaded microprocessor and discusses the impact such an architecture has on existing software energy modeling techniques. A framework is constructed for analyzing the energy behavior of the XMOS XS1-L multithreaded processor and a variation on existing software energy models is proposed, based on analysis of collected energy data. It is shown that by combining execution statistics with sufficient data on the processor's thread activity and instruction execution costs, a multithreaded software energy model used with Instruction Set Simulation can yield an average error margin of less than 7%. ACM Reference Format:Steve Kerrison and Kerstin Eder. 2015. Energy modeling of software for a hardware multithreaded embedded microprocessor.

Energy models can be constructed by characterizing the energy consumed by executing each instruction in a processor's instruction set. This can be used to determine how much energy is required to execute a sequence of assembly instructions, without the need to instrument or measure hardware. However, statically analyzing low-level program structures is hard, and the gap between the high-level program structure and the low-level energy models needs to be bridged. We have developed techniques for performing a static analysis on the intermediate compiler representations of a program. Specifically, we target LLVM IR, a representation used by modern compilers, including Clang. Using these techniques we can automatically infer an estimate of the energy consumed when running a function under different platforms, using different compilers. One of the challenges in doing so is that of determining an energy cost of executing LLVM IR program segments, for which we have developed two different approaches. When this information is used in conjunction with our analysis, we are able to infer energy formulae that characterize the energy consumption for a particular program. This approach can be applied to any languages targeting the LLVM toolchain, including C and XC or architectures such as ARM Cortex-M or XMOS xCORE, with a focus towards embedded platforms. Our techniques are validated on these platforms by comparing the static analysis results to the physical measurements taken from the hardware. Static energy consumption estimation enables energy-aware software development, without requiring hardware knowledge

General rightsThis document is made available in accordance with publisher policies. Please cite only the published version using the reference above. Abstract. The static estimation of the energy consumed by program executions is an important challenge, which has applications in program optimization and verification, and is instrumental in energy-aware software development. Our objective is to estimate such energy consumption in the form of functions on the input data sizes of programs. We have developed a tool for experimentation with static analysis which infers such energy functions at two levels, the instruction set architecture (ISA) and the intermediate code (LLVM IR) levels, and reflects it upwards to the higher source code level. This required the development of a translation from LLVM IR to an intermediate representation and its integration with existing components, a translation from ISA to the same representation, a resource analyzer, an ISA-level energy model, and a mapping from this model to LLVM IR. The approach has been applied to programs written in the XC language running on XCore architectures, but is general enough to be applied to other languages. Experimental results show that our LLVM IR level analysis is reasonably accurate (less than 6.4% average error vs. hardware measurements) and more powerful than analysis at the ISA level. This paper provides insights into the trade-off of precision versus analyzability at these levels.

Energy transparency is a concept that makes a program's energy consumption visible, from hardware up to software, through the different system layers. Such transparency can enable energy optimizations at each layer and between layers, and help both programmers and operating systems make energy-aware decisions. In this paper, we focus on deeply embedded devices, typically used for Internet of Things (IoT) applications, and demonstrate how to enable energy transparency through existing Static Resource Analysis (SRA) techniques and a new target-agnostic profiling technique, without hardware energy measurements. Our novel mapping technique enables software energy consumption estimations at a higher level than the Instruction Set Architecture (ISA), namely the LLVM Intermediate Representation (IR) level, and therefore introduces energy transparency directly to the LLVM optimizer. We apply our energy estimation techniques to a comprehensive set of benchmarks, including single- and also multi-threaded embedded programs from two commonly used concurrency patterns, task farms and pipelines. Using SRA, our LLVM IR results demonstrate a high accuracy with a deviation in the range of 1% from the ISA SRA. Our profiling technique captures the actual energy consumption at the LLVM IR level with an average error of 3%.Comment: 33 pages, 7 figures. arXiv admin note: substantial text overlap with arXiv:1510.0709