Energy-Aware Memory Allocation Framework for Embedded Data-Intensive Signal Processing Applications
| Autor: | Florin BALASA, Ilie I. LUICAN, Hongwei ZHU, Doru V. NASUI |
|---|---|
| Rok vydání: | 2009 |
| Předmět: |
Flat memory model
business.industry Computer science Applied Mathematics Real-time computing Semiconductor memory Computer Graphics and Computer-Aided Design Memory map Extended memory Memory address Memory management Computer engineering Signal Processing Computer data storage Interleaved memory Electrical and Electronic Engineering business |
| Zdroj: | IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences. :3160-3168 |
| ISSN: | 1745-1337 0916-8508 |
| DOI: | 10.1587/transfun.e92.a.3160 |
| Popis: | The storage requirements in data-intensive signal processing systems (video and image processing, artificial vision, medical imaging, real-time 3-D rendering, advanced audio and speech coding, etc.) have an important impact on both the system performance and the essential design parameters—the overall power consumption and chip area. Therefore, the optimization of the memory subsystem is an important step in the early stages of the system-level exploration. This work proposes an energy-aware memory allocation methodology for embedded data-intensive signal processing applications. Starting from the high-level behavioral specification of a given application, the framework proposed in the thesis performs the assignment of the multidimensional signals to the memory layers—the on-chip scratch-pad memories (SPM) and the off-chip main memory—the goal being the reduction of the dynamic energy consumption in the hierarchical memory subsystem. Based on the assignment results, the framework subsequently performs the mapping of signals into both memory layers such that the overall amount of data storage be reduced. The multidimensional signal processing applications are algorithmically specified in high-level programming languages, the main data structures being multidimensional arrays. The major contributions of the Ph.D. thesis are (1) a formal model (operating with multidimensional polyhedra and linearly bounded lattices) for the reduction of the dynamic energy consumption in the hierarchical memory subsystem of embedded data-intensive signal processing applications, and (2) a novel methodology (using inter-array memory sharing) for mapping the multidimensional arrays from the high-level algorithmic specifications into the physical memory. This software system yields a complete allocation solution: the exact storage amount on each memory layer, the mapping functions that determine the exact locations for any array element (scalar signal) in the specification, and an estimation of the dynamic energy consumption in the memory subsystem. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|