Controllers & Instructions
Module 14
|
2,4,5 |
||
Raising the Level of Abstraction: Assembly Instructions
- These
are concise, machine readable representations of register transfer
operations
- Hardware
implementation can translate these instructions into microinstructions
(control signals)
- Examples
- add R1, R2, R3
- lw R1, R28
- shift R2, R3, R4
- Instruction Types
- Memory instructions
- Register-register instructions
- Logical instructions
- Arithmetic instructions
Instruction Formats
- Instruction
formats define how assembly instructions are encoded as binary patterns
- Instructions
are encoded in 32-bits organized into several fields
- Some fields are defined by design, e.g., opcode
- Some fields depend on the context, e.g.,
registers being used
- Format
depends on what we need to describe
- all formats specify an code for the operation
being performed: the opcode
- R-format:
instructions control the operation of any of the functional units
and memory
- we must specify the source and destination
registers
- I-format:
one of the operands is now an immediate rather than a register
- we must specify destination/source register and
an immediate value
- Sources
and destinations for instruction data are referred to as operands
- Register operands
- Immediate operands
- Assembly
instructions are easier for us to write rather than microcode. How do
these encoded instructions control the datapath?
Instructions in the Datapath
- The
encoded instruction is placed in a special register: the instruction
register (IR)
- Addition
of the instruction register adds two more modules to the datapath
- Sign extension unit
- 16-bit immediate must be signed extended to
32-bits for the building blocks
- Next instruction logic + instruction memory
- The
datapath supports 23 basic operations
- addition and subtraction
- three types of shift operations: logical,
arithmetic, and rotate
- 16 Boolean operations
- load/store from/to memory
- Instructions
can be realized with these 23 basic operations
- Addition and subtraction
- Multiplication by powers of 2 can be implemented
with shifts
- Instructions where one operand is an immediate
rather than a register operand
- Loading an immediate value into a register by
adding to 0
- Negation can be implemented by subtracting from
0
- Clear a register by performing EX-OR on its
contents with itself
Next Instruction Logic
- Instructions
in a program are stored in successive words in instruction memory
- Instructions
are fetched and executed one at a time starting from the first instruction
- Instruction
fetch logic
- The address of an instruction is provided by a
special register: the program counter (PC)
- Instruction is fetched and placed in the instruction
register
- The program counter is incremented by 4
(remember memory is byte addressed) to be ready to fetch the next
instruction
- A
new instruction is fetched every clock cycle: how fast can the clock be?
Control
- The
encoded instruction is used to drive the datapath control signals
- Some instruction fields directly control
datapath components, e.g., register fields
- Some instruction fields are interpreted, e.g.,
opcode field
- Controller
- Truth table defines the control signals for each
instruction opcode
- This truth table can be implemented in a combinational logic module such as a PLA or in a sequential module such as a ROM
Instruction Set Architecture (ISA)
- The
ISA includes
- The set of registers
- 32 registers each 32 bits
- registers identified as R1, R2, ... or
$1, $2, ....
- A range of memory addresses
- Operations that can be performed on data stored
in registers and/or memory
- Two groups of instructions
- move data between memory and registers: lw and sw
- operate on data in registers: add, sub, and, or, sa, sl, and any
logical operation
- Instructions that operate on registers may also
have immediate variants, e.g., addi
and sli.
- The
ISA serves as the interface to the compiler
- Some
example programs
- Some
example commercial ISAs
Questions and comments to Sudhakar Yalamanchili