EE 4720 Lecture Notes

Generated from file lslifr.dvi.

This page was generated using an imperfect translator. Text may be poorly positioned and mathematics will be barely readable and some characters may be omitted. Illustrations will not show at all. If possible, view the PostScript or PDF versions of these notes.


fr-1                               Fall 1999 Final Exam Review                                     *
 *     fr-1



    When / Where


       Alternate: Tuesday, 7 December, 15:00-17:00 (3 pm to 5 pm) CST, Room EE 117.


        Original: Monday, 6 December, 10:00-12:00 CST, Room EE 145 (Here)



    Conditions


       Open Book, Open Notes


       Bring any amount of books, notes, and other material : : :

       : : :as long as they can be carried 100 meters without stopping.


       No communication devices.



    Format


       Two or three problems.


       One set of short answers.



fr-1                           EE 4720 Lecture Transparency. Formatted 10:37, 3 December 1999 from *
 *lslifr.                           fr-1

fr-2 * * fr-2 Resources Study recommendations: href = "http://www.ee.lsu.edu/ee4720/exam.html" Solved tests and homework:. href = "http://www.ee.lsu.edu/ee4720/prev.html" Study Recommendations Solve_ Old Problems Memorizing solutions is not the same as solving. Following and understanding solutions is not the same as solving. Use the solutions for brief hints and to check your own solutions. fr-2 EE 4720 Lecture Transparency. Formatted 10:37, 3 December 1999 from * *lslifr. fr-2
fr-3 * * fr-3 Topics Introductory Material ISA v. Implementation. Technological Factors: Transistor speed and quantity, memory speed and size. Different factors influencing ISA and implementation. Design principles: Amdahl's law, locality. CPU Performance Equation Benchmark types. Instruction Set Design Data Types: What to include, what to leave out. Basic integer and floating point Packed types: BCD, integer, saturating integer. Size choices. fr-3 EE 4720 Lecture Transparency. Formatted 10:37, 3 December 1999 from * *lslifr. fr-3
fr-4 * * fr-4 Memory and Register Organization: Why ss 32 registers? Stack and accumulator architectures. Memory/Memory, Register/Memory. Addressing Modes: What they do, which ones to include. Register, Immediate, Direct, Register Deferred (Register Indirect), Displacement, In- dexed, Memory Indirect, Autoincrement, Autodecrement, Scaled. fr-4 EE 4720 Lecture Transparency. Formatted 10:37, 3 December 1999 from * *lslifr. fr-4
fr-5 * * fr-5 Control Transfer Instructions: Types, when to use. Branch, Jump, Jump & Link, Call, Return Format of displacements in instruction. Specification of condition: condition code registers, integer registers, loop counter. Delayed and predicated instructions; prediction hints. Instruction Coding. Fixed-length, variable-length, and bundled instructions. Splitting of opcode field (as in DLX type-R instructions). ISA Classifications: RISC, CISC, VLIW, Stack, Accumulator Synthetic Instructions fr-5 EE 4720 Lecture Transparency. Formatted 10:37, 3 December 1999 from * *lslifr. fr-5
fr-6 * * fr-6 DLX Classification: RISC Goals: ISA should allow simple, high-speed implementation. Instruction types. Know how to read and write DLX programs. Chapter-3 (Static Scheduled) DLX Implementations Unpipelined Implementation Pipelined Implementations Basic (3-cycle branch delay). One-cycle branch delay. Bypassed. fr-6 EE 4720 Lecture Transparency. Formatted 10:37, 3 December 1999 from * *lslifr. fr-6
fr-7 * * fr-7 Hazard Definitions For a Given Pipelined Implementations Show pipeline execution diagrams. Show register contents at any cycle. Determine control hardware. Determine CPI. fr-7 EE 4720 Lecture Transparency. Formatted 10:37, 3 December 1999 from * *lslifr. fr-7
fr-8 * * fr-8 Interrupts, Exceptions, Traps Difference between interrupt, exception, trap. Causes of exceptions, role of handler. Privileged Mode. Pipeline activity leading to execution of handler. Vectored traps (using trap table). Precise exceptions, achieving with floating-point operations. fr-8 EE 4720 Lecture Transparency. Formatted 10:37, 3 December 1999 from * *lslifr. fr-8
fr-9 * * fr-9 Long Latency Operations Types of operations. (Floating point and maybe load.) Degree of pipelining: Initiation interval, latency. Detecting functional unit structural hazards. Detecting WB structural hazards: reservation register. Detecting and handling RAW hazards: ID-stage v. pre-WB stall. Handling WAW hazards. Instruction-Level Parallelism Definition, analysis. Dependency definitions. Basic block definition. fr-9 EE 4720 Lecture Transparency. Formatted 10:37, 3 December 1999 from * *lslifr. fr-9
fr-10 * * fr-10 Loop Unrolling Dynamic Scheduling Reservation stations. Register Renaming How renaming allows out-of-order execution. Where registers get renamed. Role of register map. Register Names: Reorder buffer entry number. Reservation station number. Physical register number. fr-10 EE 4720 Lecture Transparency. Formatted 10:37, 3 December 1999 from* * lslifr. fr-10
fr-11 * * fr-11 Reorder Buffer Normal Use Issue: placement of instructions in reorder buffer. Completion: updating reorder buffer entry Commitment (retirement): removal from reorder buffer entry. Recovery Goal: Undo execution starting at some instruction. Reasons: exception, branch misprediction. Steps when register map backed up. Steps when register map not backed up. fr-11 EE 4720 Lecture Transparency. Formatted 10:37, 3 December 1999 from* * lslifr. fr-11
fr-12 * * fr-12 Load/Store Unit Store/load ordering rules. Load/store bypassing. Multiple Issue Superscalar Duplication of Resources. For n instructions / cylce: Fetch, decode, rename, writeback, commit. For < n instructions: load/store, floating-point units. Added Complexity Instruction fetch inefficiency. Rules for fetching a new group. fr-12 EE 4720 Lecture Transparency. Formatted 10:37, 3 December 1999 from* * lslifr. fr-12
fr-13 * * fr-13 VLIW Difference with superscalar: instruction bundling. Dependence information in bundles. Branch Prediction One-level branch prediction. Two-level (gshare, gselect) branch-prediction. Branch target prediction. Branch folding. Memory Connection of memory devices. Set-Associative Caches. fr-13 EE 4720 Lecture Transparency. Formatted 10:37, 3 December 1999 from* * lslifr. fr-13

ECE Home Page 4720 Home Page Up
David M. Koppelman - koppel@ee.lsu.edu
Modified 3 Dec 1999 11:41 (17:41 UTC)