chapter 2 instruction set principles and examples l.
Download
Skip this Video
Loading SlideShow in 5 Seconds..
Chapter 2. Instruction Set Principles and Examples PowerPoint Presentation
Download Presentation
Chapter 2. Instruction Set Principles and Examples

Loading in 2 Seconds...

play fullscreen
1 / 43

Chapter 2. Instruction Set Principles and Examples - PowerPoint PPT Presentation


  • 315 Views
  • Uploaded on

Rung-Bin Lin. Chapter 2. Instruction Set Principles and Examples. Topics Present a taxonomy of instruction set alternatives Analyze some instruction set measurements Discuss instruction set architecture not aim at desktops or servers: DSPs and Media processors

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about 'Chapter 2. Instruction Set Principles and Examples' - oren


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
chapter 2 instruction set principles and examples

Rung-Bin Lin

Chapter 2. Instruction Set Principles and Examples
  • Topics
    • Present a taxonomy of instruction set alternatives
    • Analyze some instruction set measurements
    • Discuss instruction set architecture not aim at desktops or servers: DSPs and Media processors
    • Address the issue of languages and compilers
    • Overview the MIPS
classification of instruction set architectures
Classification of Instruction Set Architectures
  • The type of internal storage in the CPU is the most basic differentiation.
    • The major choices are a stack, an accumulator, or a set of registers. Operands may be explicit or implicit.
      • Stack architecture : Early machines
      • Accumulator architecture : Early machines
      • General purpose register (GPR) architecture : machines after 1980.
code sequence for c a b
Code Sequence for C=A+B

Stack Accumulator Register-memory Register-register

Push A Load A Load R1, A Load R1, A

Push B Add B Add R1, B Load R2, B

Add Store C Store C Add R3, R1, R2

Pop C Store C, R3

memory access for gpr machines
Memory Access for GPR Machines
  • Two ways to access explicit operands
    • First loaded into temporary storage
    • Accessed directly from memory
  • Memory access for register machine
    • Register-memory architecture: one can access memory as part of any instruction.
    • Register-register or load-store architecture: memory access only by load or store instructions.
  • Memory access for memory-memory architecture
  • Reasons for emergence of general-purpose register (GPR) machines
    • Registers are faster than memory
    • Registers are easily used by a compiler and used more effectively.
      • Example: (A*B)-(C*D)-(E*F) for stack machine? for GPR machine?
    • Registers can be used to hold variables: Reduce memory traffic, improve code density, speed up program.
types of gpr machines
Types of GPR Machines
  • Two major instruction set characteristics divide GPR architectures:
      • Whether an ALU instruction has two or three operands.
      • How many of the operands may be memory address in ALU instructions?
      • Example (fig. 2.3 possible combinations)

Number of memory Maximum number of Examples

addresses operands allowed

0 3 SPARC, MIPS, PA, PowerPC, Alpha

1 2 Intel 80X86, Motorola 68000

2 2 VAX

3 3 VAX

advantages disadvantages of gpr machines
Advantages & Disadvantages of GPR Machines

TypeAdvantagesDisadvantages

Register Simple, fixed inst. length, Poor code density

(3,0) similar clocks to execute

Register Easy to encode, Operands destroyed,

(1,2) good code density CPI varies

Memory-Memory Most compact Inst. Length and CPI

(2,2) or (3,3) varies greatly

memory addressing
Memory Addressing
  • How is a memory address interpreted?
    • Byte addressed: Provide access for bytes, half words, words, and double words (64 bits)
    • Conventions for ordering the bytes within a word:
      • Little Endian: put byte whose address xxxx00 at LSB position.

Word address Data

0 3 2 1 0

4 7 6 5 4

      • Big Endian: Put byte whose address xxxx00 at MSB position.

Word address Data

0 0 1 2 3

4 4 5 6 7

address alignment
Address Alignment
  • Access to objects larger than a byte must be aligned.
  • An access to an object of size S bytes at byte address A is aligned if A mod S =0.
    • Fig. 2.5 aligned and misaligned access

Object Aligned at Misaligned at

addressed byte offsets byte offsets

------------------------------------------------------------

Byte 0,1,2,3,4,5,6,7 Never

Half word 0,2,4,6 1,3,5,7

Word 0,4 1,2,3,5,6,7,

Double word 0 1,2,3,4,5,6,7

    • A misaligned memory access will take multiple aligned memory references
addressing mode
Addressing Mode
  • How architectures specify the address of an object they will access?
    • In a GPR, an addressing mode can specify
      • a constant,
      • a register,
      • a location in memory (used to compute effective address).
    • Immediate or literals are usually considered as memory addressing mode.
    • Addressing modes that depend on the program counter is called PC-relative addressing.
    • Addressing modes can significantly reduce instruction counts, but may add to the complexity of building a machine and increase the average CPI.
addressing modes for desktops and servers
Addressing Modes for Desktops and Servers

Register ADD R4, R3

Immediate ADD R4, #3

Displacement ADD R4, 100(R1)

Register Indirect ADD R4, (R1)

Indexed ADD R3, (R1+R2)

Direct (Absolute) ADD R1, (1001)

Memory Indirect ADD R1, @(R3)

Autoincrement ADD R1, (R2)+

Autodecrement ADD R1, -(R2)

Scaled ADD R1, 100(R2)[R3]

addressing mode for dsps
Addressing Mode for DSPs
  • 95% of DSP addressings are same as the Addressing modes for desktops and servers
  • Circular (modulo) addressing mode
    • Starting address register and ending address register
  • Bit-reverse addressing mode
    • Reverse the bit order in an address register
      • 6(1100)  3(0011)
      • Used for Fast Fourier Transform(FFT)
  • Bit-reverse and circular takes only 5% of DSP addressings
type and size of operands
Type and Size of Operands
  • How is the type of an operand designated?
    • Encoding it in the OpCode
    • Annotated with tags
  • For desktops(servers) and DSPs
operand types for desktops and servers
Operand Types for Desktops and Servers
  • Character, half word, word, single-precision floating point, double-precision floating point.
  • Distribution of data accesses by size
operands for media and signal processing
Operands for Media and Signal Processing
  • Vertex for graphic operations
    • A vertex has four components
      • X-coordinate, Y-coordinate, Z-coordinate, and W-coordinate (to help with color or hidden surface)
  • Pixel for imaging Processing
    • A pixel typically has 32 bits which is divided into four 8-bit channels.
      • R(red), G(green), B(blue), and A (denote the transparency of the surface or the pixel)
  • Fixed-point operand type (in addition to floating point)
  • Data widths
    • 32 bits, 24 bits and 16 bits are common (fig. 2.13)
operations in the instruction set
Operations in the Instruction Set
  • Desktops and servers
    • ALU, Data transfer, control, system, floating point, decimal, string, graphics.
    • SIMD instructions for medial and signal processing (fig. 2.17)
    • Ten most used 80x86instructions (takes 96% of the total inst.)
      • Load(22%), conditional branch(20%), compare(16%), store(12%), …. (fig. 2.16)
  • DSPs (beside the above operations in desktops)
    • MAC (multiply and accumulate)
    • Saturating arithmetic
      • If the result is too large to be represented, it is set to the largest representable number.
instructions for control flow
Instructions for Control Flow
  • Types and their frequency
destination address of a control flow instructions
Destination Address of a Control Flow Instructions
  • Specified explicitly except procedure return.
    • The most common way to specify the destination is to use PC-relative addressing. Its advantages:
      • Require fewer bits
      • Form re-loadable code
  • Procedure return and indirect jump require a destination address be specified dynamically.
    • The target address may be as simple as naming a register.
    • Permit any addressing mode to be used to supply the target address.
use of indirect jump
Use of Indirect Jump
  • The register indirect jump can be useful in
    • case or switch statements
    • dynamically shared library
    • virtual functions or methods in objected oriented languages
    • high-order functions or function pointer in C or C++
how branch condition is tested
How Branch Condition Is Tested?
  • Condition code (CC)
    • 80x86, ARM, Power PC, SPARC, SuperH
    • Test special bits set by ALU
    • Condition is set for free
    • CC is extra state and constrains the ordering of inst.
  • Condition register
    • Alpha, MIPS
    • Test arbitrary register with the result of a comparison
    • Simple
    • Use up a register
  • Compare and branch
    • Compare is a part of branch
    • PA-RISC, VAX
    • One instruction for doing branch
    • Too much work
procedure invocation options
Procedure Invocation Options
  • Procedure calls and returns involve saving of machine states
    • Program counter must be saved for returning to the caller
    • Some other registers automatically saved by hardware (Old architecture)
    • Registers are saved by executing the code generated by the compiler
      • Caller saving
      • Callee saving
encoding an instruction set
Encoding an Instruction Set
  • How to encode the addressing mode with the operations?
    • Depend on the number of operands per instruction and the number of addressing modes per operand.
      • If the number of operation and addressing mode combinations is large, a separate address specifier is needed for each operand (ex. VAX machine).
      • If it is small, the addressing mode can be encoded as part of the OpCode (Load-store machine, one memory operand and one or two addressing mode).
    • The architects must balance several competing forces when encoding the instruction set.
      • The desire to have as many registers and addressing modes as possible.
      • The impact of the size of the register and addressing mode fields on the average instruction size and on the average program size.
      • A desire to have instructions encoded into lengths that will be easy to handle in the implementation (i.e., multiples of bytes).
an example of the variable encoding
An Example of The Variable Encoding

ADD EAX, 1000(EBX)

    • Indicates a 32-bit integer ADD instruction with two operands.
    • Opcode takes 1 byte
    • Needs a 1-byte address specifier to indicate addressing mode and the register being used.
    • The displacement needs four bytes.
    • Total length: 1+1+4=6 bytes
  • 80x86 instruction length
    • 1 to 17 bytes
code size reduction
Code Size Reduction
  • Use of hybrid format (40% reduction)
    • ARM series processors
      • 32-bit ARM instruction set
      • 16-bit Thumb instruction set
    • MIPS (40% reduction in code size)
      • 32-bit MIPS
      • 16-bit MIPS
  • Use of code compression
    • IBM PowerPC decodes the compressed code when it is first fetched from main memory
the role of compilers
The Role of Compilers
  • Crossed coupled with the instruction set architecture
  • It greatly affects the code size and execution speed of a program
  • Goal of a compiler writer
    • Correctness
    • Speed of compiled code
phase ordering problem
Phase-ordering Problem
  • Compilers make assumptions about the ability of later steps to deal with certain problems. It limits the optimization effectiveness of compilation.
  • Examples
    • Choose which procedure calls to expand inline before they know the exact size of the procedure.
    • Global common subexpression elimination assumes the value will be allocated with a temporary register
optimization of compilation
Optimization of Compilation
  • High-level optimization
    • Processor independent
  • Local optimization
    • Within a basic block
  • Global optimization
    • Across branches
  • Register allocation
    • Machine dependent
    • Can be solved by Graph Coloring
  • Other processor-dependent optimization
how the architect can help compiler writer
How the Architect Can Help Compiler Writer
  • Instruction set properties help compiler writers
    • Regularity: Operations, data types, and addressing modes should be orthogonal (independent).
    • Provide primitives, not solutions: Don’t try to match a language construct.
    • Simplify the trade-offs among alternatives: One of the most difficult instances of complex trade-offs occurs in a register-memory architecture in deciding how many times a variable should be referenced before it is cheaper to load it into a register.
    • Provide instructions that bind the quantities known at compiled time as constants.
the mips architecture
The MIPS Architecture
  • The MIPS architecture emphasizes
    • A simple load-store instruction set.
    • Design for pipelining efficiency, including a fixed instruction set encoding.
    • Efficiency as a compiler target.
  • Registers for MIPS64
    • 32 64-bit general purpose registers, named R0, R1, …, R31. The value of R0 is always 0.
    • 32 single(double)-precision registers, named F0, F1, …, F31.
  • Data types for DLX
    • 8-bit byte, 16-bit half word, 32-bit word for integer.
    • 32-bit single-precision, 64-bit double-precision for floating point.
addressing modes for mips64
Addressing Modes for MIPS64
  • Immediate and displacement, both with 16-bit fields.
  • Register deferred can be obtained by placing 0 in the 16-bit displacement field. Ex. LD R1, 0(R1).
  • Absolute addressing can be obtained by using register R0 as the base register. Ex. LW R1, xx(R0).
  • Byte addressable with a 64-bit address.
  • A mode bit to select Big-Endian or little-Endian mode
  • A load-store architecture.
mips operations
MIPS Operations
  • Load and stores (fig. 2.28 on page 133)
  • ALU operations (fig. 2.29 on page 134)
  • Branches and jumps (fig. 2.30 on page 135)
    • Jump
      • target address: 26-bit offset + PC+4 or a register containing the whole 32-bit address.
    • Jump and Link
      • Target address: same as jump.
      • Return address: PC+4
    • Conditional Branch
      • Target address: 16-bit offset + PC+4
    • Branch condition is specified in the instruction, which may test the register source of zero or non-zero.
  • Floating-point operations
    • Add, subtract, multiply and divided.
    • MOV.S and MOV.D copy a single-precision (MOVF) or double-precision floating-point register to another register of the same type.
concluding remarks
Concluding Remarks
  • Changes of the instruction set architecture in 1990s
    • 32-bit address ----> 64-bit address
    • Optimization of conditional branches via conditional execution
    • Optimization of cache performance via prefetch
    • Support for multimedia
    • Faster floating-point operations
  • Trends in next decade
    • Long instruction word.
    • Increased conditional execution.
    • Blending of general-purpose and DSP architectures
    • 80x86 emulation