SoD: Enabling Design Strategies for Single Chip Heterogeneous Multiprocessors

[0438948] paul, JoAnn M. CMUSoD：单片异构多处理器的使能设计策略在未来十年内，将有可能在单个芯片上提供足够的晶体管，用于一百个32位RISC处理器。这些新兴的、软件密集型的单芯片异构多处理器（SCHM）计算机系统的设计排除了经验分析或具体综合方法的可能性。目前还没有针对软硬件混合单芯片多处理系统的设计科学。战略计算机系统组织-需要一个创造性的过程，不能自动化，因此这种设计师的创造力必须促进在系统层面。这项研究的完整愿景是发展所需的科学，使未来的设计系统能够摆脱当前CAD工具的具体和综合思维模式，成为一种特征和发明范式，这种范式仍然与当前较低层次的集成电路设计方法流程有关。我们研究背后的另一个主要主题是，计算机系统设计者从趋势分析开始，在应用程序集的主要特性的预期变化和执行它们的系统的主要架构特性的物理能力之间寻找未来的交叉点。由此，设计策略被发明出来，这些策略可以成为整个可编程系统类的基本组织主题。

Abstract0438948Paul, JoAnn M. CMUSoD: Enabling Design Strategies for Single Chip Heterogeneous MultiprocessorsWithin the next ten years, it will be possible to provide enough transistors on a single chip for on the order of a hundred 32-bit RISC processors. The design of these emergent, software-intensive single chip heterogeneous multiprocessor (SCHM) computer systems precludes the possibility of seat-of-the-pants analysis or specify-and-synthesize methodologies. No science of design has been developed for mixed hardware/software single chip multiprocessing systems. Strategic computer system organization - requires a creative proces that can not be automateds,and thus this designer creativity must be facilitated at the system level. The full vision of this research is to develop the science needed to enable future design systems to break away from the specify-and-synthesize mind set of current CAD tools to become a characterize-and-invent paradigm that still has ties into the current lower level integrated circuit design methodology flows. Another major theme behind our research is that computer system designers start from trend analyses, looking for future cross-over points between anticipated changes in the major features of applications sets and the physical capabilities of the major architectural features of the systems that carry them out. From these, design strategies are invented which can become basic organizing themes for entire classes of programmable systems.