首页> 外文学位 >Architectural and compiler techniques for energy reduction in high-performance microprocessors.
【24h】

Architectural and compiler techniques for energy reduction in high-performance microprocessors.

机译:用于降低高性能微处理器能耗的体系结构和编译器技术。

获取原文
获取原文并翻译 | 示例

摘要

The microprocessor industry has started viewing power, along with area and performance, as a decisive design factor in today's microprocessors. The increasing cost of packaging and cooling systems poses stringent requirements on the maximum allowable power dissipation. Most of the research in recent years has focused on the circuit, gate, and register-transfer (RT) levels of the design. In this research, we focus on the software running on a microprocessor and we view the program as a power consumer. Our work concentrates on the role of the compiler in the construction of “power-efficient” code, and especially its interaction with the hardware so that unnecessary processor activity is saved. We propose techniques that use extra hardware features and compiler-driven code transformations that specifically target activity reduction in certain parts of the CPU which are known to be large power and energy consumers.; Design for low power/energy at this level of abstraction entails larger energy gains than in the lower stages of the design hierarchy in which the design team has already made the most important design commitments. The role of the compiler in generating code which exploits the processor organization is also fundamental in energy minimization. Hence, we propose a hardware/software co-design paradigm, and we show what code transformations are necessary by the compiler so that “wasted” power in a modern microprocessor can be trimmed.; More specifically, we propose a technique that uses an additional mini cache located between the instruction cache (I-Cache) and the CPU core; the mini cache buffers instructions that are nested within loops and are continuously fetched from the I-Cache. This mechanism can create very substantial energy savings, since the I-Cache unit is one of the main power consumers in most of today's high-performance microprocessors. Results are reported for the SPEC95 benchmarks in the R-4400 processor which implements the MIPS2 instruction set architecture.
机译:微处理器行业已经开始将功率以及面积和性能视为当今微处理器的决定性设计因素。包装和冷却系统成本的增加对最大允许功耗提出了严格的要求。近年来,大多数研究都集中在设计的电路,门和寄存器传输(RT)级别上。在这项研究中,我们专注于在微处理器上运行的软件,并且我们将该程序视为功耗者。我们的工作集中在编译器在“省电”代码的构造中的作用,尤其是其与硬件的交互,从而节省了不必要的处理器活动。我们提出了使用额外的硬件功能和编译器驱动的代码转换的技术,这些转换专门针对减少CPU某些部分中的活动,这些部分是已知的大功率和能量消耗者。与设计团队已经做出最重要的设计承诺的设计层次结构的较低阶段相比,在此抽象级别进行低功耗/能量设计需要更大的能量获取。编译器在生成利用处理器组织的代码方面的作用也是最小化能量的基础。因此,我们提出了一种硬件/软件协同设计范例,并说明了编译器需要进行哪些代码转换,以便可以削减现代微处理器中的“浪费”功率。更具体地说,我们提出一种使用位于指令高速缓存(I-Cache)和CPU内核之间的附加小型高速缓存的技术。迷你高速缓存缓冲嵌套在循环中并从I-Cache连续获取的指令。这种机制可以节省大量能源,因为I-Cache单元是当今大多数高性能微处理器的主要功率消耗者之一。报告了实现MIPS2指令集体系结构的R-4400处理器中SPEC95基准测试的结果。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号