第四章指令并行-软件方面.pptVIP

Chapter 4 (2) Instruction-Level Parallelism Software Approaches Computer Architecture ----A Quantitative Approach Lecture for ILP:Software approaches(软件方法) Basic Compiler Technique for Exposing ILP Loop unrolling（基本的发现ILP的编译技术是循环展开） Static Branch Prediction(静态分支预测) Static multiple Issue: VLIW（静态多指令发射VLIW） Advanced Compilor Support for Exposing and Exploiting ILP(对发现和开发ILP的高级编译器支持) Software pipelining(软件流水) Global Code scheduling（全局代码调度） Hardware Support for Exposing More Parallelism at compile time(对编译时开发ILP的硬件支持) Conditional or Predicated(断言的) instructions(条件指令或预测指令) Compiler speculation with hardware support(在硬件支持下的编译器投机技术) FP Loop: Where are the Hazards? Loop: LD F0,0(R1) ;F0=vector element ADDD F4,F0,F2 ;add scalar from F2 SD 0(R1),F4 ;store result SUBI R1,R1,8 ;decrement pointer 8B (DW) BNEZ R1,Loop ;branch R1!=zero NOP ;delayed branch slot Reducing stalls from schedulling in BB and delayed branch Unroll Loop Four Times (straightforward way) Rewrite loop to minimize stalls? Unrolled Loop That Minimizes Stalls What assumptions made when moved code? OK to move store past SUBI even though changes register OK to move loads before stores: get right data? When is it safe for compiler to do such changes? Using Loop unrolling and scheduling with static Multiple Issue Static Branch Prediction静态分支预测 Static branch predictors are used in processors when branch behavior is expected highly predictable at compile time. (静态分支预测一般用于分支行为在编译器时就具有很高有可预测性的情形) Several different methods Always predict a branch as taken or untaken (总是预测转移成功或不成功) Predict on the basis of branch direction(基于转移方向的预测) Backward-going branch to be taken,(向后预测为成功) Forward-going branch to be not taken.(向前预测为不成功) Profile-based Prediction(基于以往概要信息(含多方面的行为)的预测) Static Multiple issue: VLIW(静态多发射：VLIW) VLIW: Very Long Instruction Word(超长指令字) Each “instruction” has explicit coding for multiple operations(每条“指令”都显式地包括多个操作) In EPIC, grouping called a “packet” In Transmeta, groupi