首页> 外国专利> PROCESSOR CONFIGURED TO GENERATE LOOKAHEAD RESULTS FROM COLLAPSED MOVES, COMPARES, AND SIMPLE ARITHMETIC INSTRUCTIONS

PROCESSOR CONFIGURED TO GENERATE LOOKAHEAD RESULTS FROM COLLAPSED MOVES, COMPARES, AND SIMPLE ARITHMETIC INSTRUCTIONS

机译:处理器经过配置,可以从折叠的移动,比较和简单的算术指令中生成“笨拙的结果”

摘要

A processor includes a lookahead address/result calculation unit which is configured to receive operand information (either the operand or a tag identifying the instruction which will produce the operand value) corresponding to the source operands of one or more instructions. If the operands are available, lookahead address/result calculation unit may generate either a lookahead address for a memory operand of the instruction or a lookahead result corresponding to a functional instruction operation of the instruction. The lookahead address may be provided to a load/store unit for early initiation of a memory operation corresponding to the instruction. The lookahead result may be provided to a speculative operand source (e.g. a future file) for updating therein. A lookahaead state for a register may thereby be provided early in the pipeline. Subsequent instructions may receive the lookahead state and use the lookahead state to generate additional lookahead state early. On the other hand, the subsequent instructions may receive the lookahead state and hence may be prepared for execution upon dispatch to an instruction window (as opposed to waiting in the instruction window for execution of the prior instruction). In one embodiment, the processor also includes an operand collapse unit confgured to collapse the lookahead results into subsequent, concurrently decoded instructions (intraline dependencies). Additionally, the operand collapse unit may be configured to collapse a compare instruction into a subsequent branch instruction which depends upon the result of the compare.
机译:处理器包括超前地址/结果计算单元,其被配置为接收与一个或多个指令的源操作数相对应的操作数信息(操作数或标识将产生操作数值的指令的标签)。如果操作数可用,则超前地址/结果计算单元可生成指令的存储器操作数的超前地址或与指令的功能指令操作相对应的超前结果。可以将超前地址提供给加载/存储单元,以提早启动与该指令相对应的存储器操作。可以将前瞻结果提供给推测操作数源(例如,将来的文件)以在其中进行更新。从而可以在流水线的早期提供用于寄存器的lookahaead状态。后续指令可能会接收超前状态,并使用超前状态尽早生成其他超前状态。另一方面,后续指令可以接收超前状态,因此可以在调度到指令窗口时准备执行(与在指令窗口中等待执行先前指令相反)。在一个实施例中,处理器还包括操作数折叠单元,该单元被配置为将超前结果折叠为后续并发解码的指令(行内依赖性)。另外,操作数折叠单元可以被配置为根据比较的结果将比较指令折叠为随后的分支指令。

著录项

  • 公开/公告号EP1031074B1

    专利类型

  • 公开/公告日2006-06-28

    原文格式PDF

  • 申请/专利权人 ADVANCED MICRO DEVICES INC.;

    申请/专利号EP19980953695

  • 发明设计人 WITT DAVID B.;

    申请日1998-10-19

  • 分类号G06F9/38;

  • 国家 EP

  • 入库时间 2022-08-21 21:32:00

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

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