The Instruction Set Architecture

							The Instruction Set Architecture
Hardware – Software boundary

 Java Program        C Program        Ada Program

     Compiler       Compiler       Compiler

           Instruction Set Architecture

                    Microcode

                   Hardware
Data types
 The ISA defines the data types native to a
  particular machine. Some examples of numeric
  data types that may or may not be present on the
  machine:
   –   Signed integer
   –   Unsigned integer
   –   Floating point
   –   Binary Coded Decimal (BCD) integers
   –   Packed numeric representations
 Java does not support unsigned, BCD, or packed
  integer representations.
Instruction Format
 The ISA defines the instruction format(s) for the
  machine
   – Fixed or variable length instructions
   – 0, 1, 2, or 3 operand instructions
   – Addressing modes
 The Java Virtual Machine
   – 9 instruction formats
   – Variable length
   – Little in the way of addressing modes; immediate,
     indexed
Examples
 See handout of IBM System 360 ISA
   – 5 instruction formats
   – A long list of instructions
   – Condition codes
 This is a CISC (Complex Instruction Set
  Computer) Architecture
Addressing modes
 Why more than one type?
  – Short is good, but not sufficient for all purposes
  – Immediate address – the data is in the same memory
    word as the instruction
  – Register – the data is in one of the registers, memory
    locations on the processor chip itself
  – Absolute – specify the exact memory location in use
  – Indexed – Data is specified by a base and displacement
  – Indirect – address specified contains an address, rather
    than data
ISA and Microarchitecture
 So, you have this wonderful idea for improving
  performance of your processor
   – Will all the existing software still run?
   – Will the same operating system work?
 The microarchitecture separates the actual
  hardware from the ISA.
   – Do the hardware improvements.
   – Modify the microcode to match the existing ISA to the
     new hardware. All existing applications continue to
     run. New applications can take advantage of new
     features.
Language Levels
  Higher level images of the computer:
  Spreadsheets, Word processors, databases –
  closer to the problem
   C, Java, etc.: designed for representing
   a problem solution; independent of the machine
   ISA: instruction set, addressing modes of
   a specific line of computers

     Microarchitecture: bit settings that
     activate specific data flows

        Hardware: understands voltages
Access to the hardware
 Programs that can manipulate the hardware
  can be dangerous.
 Hardware is protected by providing several
  classes of access rights.
 Operating system has complete access
 User programs are limited.
 Users can access the hardware by invoking
  operating system functions.
Why many high level languages?
 Notice the position of the high level
  languages in our chart
  – They exist to make it easier to express the
    problem
  – Different application areas have different needs
    in terms of expressing problems.
 Ex. APL – single instruction to multiply two
  matrices
Compiler
 A compiler is a program
 Its input is the text of a program in a
  particular language, say c
 The compiler translates the c code into the
  instruction set understood by the computer
 The translation is not a line by line
  transliteration. Optimization.
Instructions and Data
 Two distinct kinds of things
  – The machine understands a finite subset of
    possible bit patterns as instructions.
  – Data can take any bit pattern.
 Usually stored separately. Data accessed
  when an instruction calls for it.
Black boxes
 To reduce complexity, we sometimes show
  only the input and output of an operation
  without the details of how it gets done. We
  call it a black box, because the content of
  the box is hidden.
 a, b        Turing machine that adds        a+b


 We can combine the boxes and use them any way we want.
Turing Machines
 Mathematical model of a machine
 Turing was interested in understanding what could
  be computed and what could not be computed.
 The general purpose computers of today are
  implementations of the Turing model
 That’s why all computers can do the same things,
  given enough time and memory – because they are
  all Turing machines.