CSC: 345 Computer Architecture

1
CSC 345 Computer Architecture

• Jane Huang
• Lecture 6
• Instruction Sets
• Addressing Modes

2
http//fizbin.eecs.lehigh.edu/mschulte/ece401-01/
lect/my-lec03-p2.pdf
3
Instruction Set

• The operation of the CPU is determined by the
  instructions it executes ie. Its Instruction Set
• An instruction must contain all the information
  needed in the instruction cycle.

4
Elements of an Instruction

• Operation Code Specifying the operation to be
  performed. For example ADD, JUMP, STORE
• Source operand reference.
• Result operand reference.
• Next instruction reference.
• An instruction is represented as a series of bits
  but as this is not very understandable by a
  user, we can also use symbolic notation (ie ADD
  instead of 01011101 etc).

4 bits
6 bits
6 bits
Opcode
OperandReference
OperandReference
16 bits
5
Instruction Types

• Consider a basic instruction
• x x yLocation 512 stores x, location 513
  stores y
• With one register (AC)Load AC from Memory
  location 512,Add to AC from Memory location
  513Store AC back to Memory location 512
• Therefore a single high level instruction might
  require multiple instructions at the machine
  level.
• Machine instructions needed for
• Data processing (Arithmetic, logic)
• Data storage (Memory instructions)
• Data movement (I/O instructions)
• Control (Test and branch instructions)

6
http//fizbin.eecs.lehigh.edu/mschulte/ece401-01/
lect/my-lec03-p2.pdf
7
Number of Instructions

• Maximum number of addresses that could be needed
  4.
• 2 source operands
• 1 result operand
• Address of next instruction

8
Stacks

• A stack is a pushdown list.
• Access is always from the top ? LIFO list.
• Basic stack operations are PUSH (to put on the
  stack) and POP (to get off the stack)
• Simple calculation a ab
• Push aPush bAdd
• If stack is accessible to programmers need for
  explicit stack oriented instructions.
• Stack related addresses often stored in registers
  for Stack pointer, Stack base, Stack limit.
  (Optional TopElement, SecondElement)

9
Expression Evaluation

• Normally mathematical formulas are evaluated
  using infix notation in which a binary operation
  appears between two operands (eg ab)
• In reverse polish (postfix) notation the operator
  follows its operands
• a b ? ab
• a (b c) ? abc
• (ab) c ? abc
• Postfix expressions can be easily evaluated using
  a stack.
• Scan from left to right.
• If the element is a variable, push it onto the
  stack.
• If the element is an operator, pop the top two
  items, perform the operation, and push the result
  back onto the stack.
• For example abc

10
Converting from infix ? postfix

• Dijkstra defined an algorithm that uses a stack
  to convert from infix to postfix.
• Scan infix expression from left to right
  performing the following steps
• Examine the next element in the input.
• If it is an operand, output it.
• If it is an opening parenthesis, push it onto
  the stack.
• If it is an operator then
• If the top of the stack is an opening
  parenthesis, push the operator.
• If it has higher priority than the top of the
  stack, push the operator.
• Else pop operation from stack to output and
  repeat step 4.
• Example A (B C) ( D E) / F

Edsger Wybe Dijkstra    1930-2002
11
Instruction Set Design

• Many fundamental issues related to instruction
  sets is still in dispute
• Operation repertoireHow many and which
  operations to provide, how complex the operations
  should be.
• Data typesWhat types of data should be included.
• Instruction formatInstruction length, number of
  addresses, size of fields, etc.
• RegistersHow many registers should be referenced
  by instructions.
• AddressingThe mode by which addresses should be
  specified.

12
Types of Operands

• Major categories of data include
• Addresses
• Numbers
• Characters
• Logical data

Numbers

• Three common types of numerical data
• Integer or fixed point
• Floating point
• Decimal

13
Integer Representation

• Binary number system can represent all numbers
  using 0 and 1 digits, minus sign (-) and a
  decimal point.
• - 1101.01012 -13.312510
• Converting from decimal to binary Example
  355
• 355/2 177 remainder 1 ?
• 177/2 88 remainder 1 ? 88/2 44
  remainder 0 ? 44/2 22 remainder 0 ?
  22/2 11 remainder 0 ?
• 11/2 5 remainder 1 ?5/2 2 remainder
  1 ?2/2 1 ?

111011001100011100011110001111100011
14
Integer Representation

• Converting a decimal fraction to binary0.8110
• 0.81 X 2 1.620.62 X 2 1.240.24 X 2
  0.480.48 X 2 0.960.96 X 2 1.920.92 X 2
  1.84
• Computer storage has no minus signs and periods!
• Sign-magnitude representationLeftmost big holds
  the sign bit (1 negative, 0 positive)
• Example 310 0112 -310 1112
• Drawbacks include
• Two representations for zero (000,100)
• Addition and subtraction need to consider the
  sign of the number and their relative magnitude.

111110110011001110011
15

• Twos Complement Representation
• Also uses the most significant bit as a sign
  bit.Differs from signed magnitude in how the
  other bits are interpreted.
• To convert from binary ( 0011010) to 2s
  complement.
• Take the boolean complement of each bit of the
  corresponding positive number
• 1100101
• Add 1 to the resulting bit pattern viewed as an
  unsigned integer. 1100101 1
  1100110
• Quick method starting at least significant bit
  leave all bits the same until after the first
  1. From that point on, flip the bits.

16
Twos Complement Arithmetic

• Add 7 4 11
• 7 00111, 4 00100
• 00111 00100 01011

• Subtract 7 - 4 3
• 7 00111, -4 11100
• 00111 11100 00011

17
Floating-Point Representation

• Fixed point notation supports the representation
  of a range of positive and negative numbers
  centered around 0.
• Range is limited by the number of bits.
• In decimal notation, we used scientific notation.
• 976,000,000,000,000 can be represented as 9.76 X
  1014
• 0.0000000000000976 ? 9.76 X 10-14
• Binary numbers can also be represented in the
  form
• S X B E

23 bits
8 bits
Sign of signific-and
Biased exponent
Significand
1.101001 X 210100 0 10010011
101000100000000000000000 1.638125 X
220 -1.101001 X 210100 1 10010011
101000100000000000000000 1.638125 X 220
1.101001 X 2-10100 0 01101011
101000100000000000000000 1.638125 X
220 -1.101001 X 2-10100 1 01101011
101000100000000000000000 1.638125 X 220
18

• Decimal Numbers
• For applications that have comparatively little
  computation but lots of I/O, it is sometimes
  better to keep numbers in decimal format.
• Packed Decimal1834 ? 0001 1010 0011 0100
• Characters
• ASCII
• EBCDIC
• Logical Data
• An n-bit data unit is considered as n 1-bit items
• Array of boolean or binary data items.

19
Examples of IBM S/390 Data Transfer Operations L
Load (32 bits) Mem ? RegLH Load
halfword (16 bits) Mem ? RegLR Load (32
bits) Reg ? Reg

• Types of Operations
• Data transfer
• Arithmetic
• Logical
• Conversion
• I/O
• System Control
• Transfer of Control
• Data Transfer
• Move, Store, Load, Set, Clear, Push, Pop
• Location of source and destination operands
• Location could be memory, register, stack.
• Length of data to transfer
• Addressing mode (to be discussed later)
• Tradeoffs - Where to indicate the type of
  location (opcode spec or operand?)

20

• Arithmetic
• Basic arithmetic operations of add, subtract,
  multiply, divide.
• Orthogonality means these should be provided for
  all supported number types. Sometimes certain
  operations only provided for fixed point numbers.
• Arithmetic occurs in the ALU.
• Data transfer operations may be required to
  position operands and to deliver output.
• Logical
• Operations for manipulating individual bits of a
  word (bit twiddling)
• Masking registers (When did we see this???)
• NOT, AND, OR, XOR.
• How could we use XOR to test for equality of two
  numbers?
• Logical shift (used to isolate fields within a
  word).

21

• Logical left/right shift (0 shifted in on)Useful
  for isolating single fields within a word.
• Arithmetic shift treats word as signed integer
  and does not shift the bit sign. In 2s
  complement a shift to the right ? division by 2,
  shift to the left ? multiplication by 2.
• Rotate (cyclic shift) brings each bit in turn
  into rightmost bit, for testing.

22

• Conversion
• Instructions that change the format or operate on
  the format of data.
• Example convert from decimal to binary (ie
  packed decimal to signed integer, or EBCDIC to
  ASCII)
• System Control
• Can only be executed from a privileged state
• Used by the operating system.
• Example reading or altering a control register
  ( stack base etc)

23

• Transfer of Control
• Many instructions change the sequence of
  instruction execution
• Operation of these instruction causes the PC to
  be updated with a new address.
• Program loops
• Decision making
• Procedure calls
• Operation of these instruction causes the PC to
  be updated with a new address
• Most common transfer-of-control operations
  include branch, skip, procedure call.
• Branch Instructions One operand is the address
  of the next instruction to be executed.BRE R1,
  R2, X (Branch to X if R1 is equal to R2.

24
Status register bits
Status register enables quick comparison for
branching.

• Bit C (Carry) set to 1 if the end carry C8 is 1,
  it is cleared if C8 0.
• Bit S (sign) set to 1 if highest order bit F7 is
  1 (visa versa for 0)
• Bit Z (zero) set to 1 if output of ALU contains
  all 0s.
• Bit V (overflow) set to 1 if exclusive-OR of last
  two carries is equal to 1 (ie condition for
  overflow)

25

• Procedure Call Instructions
• Procedures are self-contained computer program
  that can be called (invoked) from anywhere in the
  larger program.
• Allows same code to be used multiple times.
  (economy and modularity)
• Involves two basic instructions
• A call instruction with a branch to the
  procedure.
• Return instruction (also a branch)
• Note
• A procedure can be called from more than one
  location
• One procedure may call another procedure
  (nesting)
• Each procedure call is matched by a return to the
  called program.
• Where should the return address be stored for
  CALL X?
• Register (Reg ? PC ?, PC ? X) - where ?
  instruction length.
• Start of called procedure (X ? PC ?, PC ? X
  1)
• (Both of these approaches prevent re-entrant
  procedures)
• Top of stack (more general and powerful
  approach)Return address placed on the stack.

26
http//fizbin.eecs.lehigh.edu/mschulte/ece401-01/
lect/my-lec03-p2.pdf
27
Instruction SetsChapter 11

• The address field in a typical instruction format
  are relatively small.
• In order to reference a large range of locations
  in main memory a variety of addressing
  techniques are employed.
• Each one involves a trade-off between
• Address range and/or address flexibility
• Number of memory references needed to access the
  data
• Complexity of the address calculation.
• Common addressing techniques
• Immediate
• Direct
• Indirect
• Register
• Register Indirect
• Displacement
• Stack

28
Immediate Addressing

• Operand present in the instruction.
• OPERAND A
• No memory reference other than instruction fetch
  is needed.
• Limited operand magnitude.

29
Direct Addressing

• Address field contains the effective address of
  the operand.
• EA A
• Common in early computers
• Not common on todays computers.
• Requires only one memory reference and no
  special calculation.
• Provides only a limited address space of 2K
  where k the length of the address field.

30
Indirect Addressing

• The address field refers to the the address of a
  word in memory.
• That word contains a full-length address of the
  operand.
• EA (A) where (A) means the contents of A
• Address space now increasesto 2N, where N the
  word length.
• BUT the number of different effective addresses
  is limited to 2K.
• This works well in a virtual memory environment
  in which alleffective address locations can
  beconfined to page 0 of a process.

31
Register Addressing

• Similar to direct addressing.
• The address field refers to a register rather
  than main memory address.
• EA R
• Small address field 3-5 bits.
• No memory references required.
• BUT address space is limited.
• Limited number of registers only worthwhile if
  the operandwill be used repeatedly.
• Interesting register coloring problem.

32
Register Indirect Addressing

• Analogous to indirect addressing.
• EA (R)
• Address field refers to a word in memory
  containingan address.

33
Displacement Addressing

• Very powerful addressing mode.
• Combines capabilities of directaddressing and
  register indirectaddressing.
• EA A (R)
• Instruction must have two address fields, at
  leastone of which is explicit A.
• 3 common uses
• Relative addressing in which the PC is the
  implicitly referenced register. Therefore EA
  displacement relative to the address of the
  instruction.
• Base-Register Addressing referenced register
  contains a memory address, address field contains
  a displacement.
• Indexing Address field references a main memory
  address referenced register contains a positive
  displacement from that address.

34
Instruction Formats

• Defines the layout of bits in an instruction, in
  terms of its constituent parts.
• Must include an opcode and explicitly or
  explicitly 0 or more operands.
• Each explicit operand is referenced using an
  addressing mode.

Instruction Length

• Critical decision impacted by memory size, memory
  organization, bus structure, CPU complexity, and
  CPU speed.
• Major trade off powerful instruction repertoire
  vs. need to save space.
• Programmers want more opcodes, more operands,
  more addressing modes, and greater address range.
  (shorter programs, easier to write)
• Instruction length should be equal to
  memory-transfer length (bus) or one should be a
  multiple of the other.
• Should be a multiple of character length (8
  bits), and fixed point numbers.

35
Allocation of bits

• How should bits be allocated within the
  instruction?
• Major trade-off between number of opcodes and the
  power of the addressing capability.
• Primary issues include
• Number of addressing modesAre these modes
  implicit or explicit?
• Number of operandsFewer addresses ? longer
  more awkward programs.
• Register vs. memoryIf certain operations operate
  on registers rather than memory.Number of
  registers
• Number of register setsGeneral purpose registers
  vs. specialized set in which the set to be used
  is implicit in the operation.
• Address rangeRange of memory to address,
  addressing modes supported.
• Address granularityWord or byte?

36
http//fizbin.eecs.lehigh.edu/mschulte/ece401-01/
lect/my-lec03-p2.pdf
37
http//fizbin.eecs.lehigh.edu/mschulte/ece401-01/
lect/my-lec03-p2.pdf
38
http//fizbin.eecs.lehigh.edu/mschulte/ece401-01/
lect/my-lec03-p2.pdf
39
Instruction Set Design(Homework 6 will
partially build on this)

• Design an Instruction Set for a small
  programmable embedded device that is used to
  control the temperature of a vat used for
  manufacturing chemicals. The temperature must
  follow a specific pattern during the production
  process. The device has a single temperature
  sensor.
• Consider
• Instruction size
• Instruction format
• Addressing modes you will support
• Opcodes you will support