Transcript Overview of basics

Computer Architecture
Lecture Notes
Spring 2005
Dr. Michael P. Frank
Competency Area 3:
Programming and Coding Methods
Instructions
• Main goals of this chapter:
— To be able to derive binary MIPS instruction code from
assembler code
— To be able to derive assembler code from C-code
representations
• We’ll be working with the MIPS instruction set architecture
— similar to other architectures developed since the 1980's
— used by NEC, Nintendo, Silicon Graphics, Sony
— MIPS instruction set architecture will be introduced in a stepby-step approach. By the end of the chapter, you should
have a good understanding of the design rules, and be able
to analyze MIPS ISA.
Instructions
• Instructions  Language of the Machine
• Instruction Set  “its vocabulary”
• More primitive than higher level languages
e.g., no sophisticated control flow for branches, loops
• Very restrictive
e.g., MIPS Arithmetic Instructions
* Common design goal among computer designers:
maximize performance and minimize cost, reduce design time
Instructions
• Four design principles will be introduced in this chapter which
are important in instruction set architecture design:
• Design Principle 1: Simplicity favors regularity.
• Design Principle 2: Smaller is faster.
• Design Principle 3: Good design demands good
compromise.
• Design Principle 4: Make the common case fast!
MIPS Arithmetic
• In MIPS assembly language, all instructions have 3
operands only
— Destination operand and 2 source operands.
• Instructions can perform only one operation at a
time.
— However, pipelined and superscalar implementations may
execute multiple instructions simultaneously.
• Operand order is fixed (destination first):
Example:
C code:
A = B + C
MIPS code:
add A, B, C
• A “#” symbol marks the start of a comment,
— Comments are ignored by the compiler.
MIPS Arithmetic
• MIPS architecture philosophy is to keep the hardware simple,
since complex instructions require more physical hardware
resources to implement (space, time, energy costs).
— This constraint has been becoming less important as transistors
shrink.
• ***Design Principle 1: “Simplicity favors regularity.”
However, simplicity in the ISA design can lead to a larger size
for compiled code.
C code:
f = (g+h) – (i+j);
MIPS code:
add
add
sub
t0, g, h
t1, i, j
f, t0, t1
MIPS Arithmetic
• Symbolic variable notation is used in previous
examples.
—However in MIPS architecture, only registers can
be used as operands.
• Registers are 32 bits  “word ” length
• There are 32 registers available (for integer
arithmetic) in the MIPS architecture.
—The registers (almost) all behave the same.
– Simple, regular program design & HW implementation.
MIPS Arithmetic
MIPS register conventions:
NAME
$zero
Register Number
$r0
Usage
Hardwired to the constant value 0
$v0 - $v1
$r2 - $r3
Subroutine results and expression evaluation
$a0 - $a3
$r4 - $r7
Arguments (parameters) to subroutines
$t0 - $t7
$r8 - $r15
Temporary registers (caller saves)
$s0 - $s7
$r16 - $r23
Saved registers (callee saves)
$t8 - $t9
$r24 - $r25
More temporary registers (caller saves)
$gp
$r28
Global pointer (e.g. to static data area)
$sp
$r29
Stack pointer (stack grows downwards)
$fp
$r30
Frame pointer (to local variables on stack)
$ra
$r31
Return address for subroutine calls
Note: Register $r1 is reserved for use by the assembler and
Registers $r26-$r27 are reserved for the operating system.
MIPS Arithmetic
*** Design Principle 2: smaller is faster.
• Why?
— Having a large number of registers will increase clock cycle
time (longer wires, more RC delay)
— Recall that having a smaller clock cycle time will improve
performance!
• Effective use of the principle is key to computer
performance ;
— Computer designer must balance the programmer’s desire for
more registers with the need for a minimal clock cycle time.
• Programmers also should worry about this…
— A program that uses less memory will often run faster.
– Less cache contention, virtual memory not needed.
Registers versus Memory
• Arithmetic instructions’ operands must be registers,
— No arithmetic instructions operate directly on memory contents
— Only 32 registers are provided.
• Revisit earlier example:
— C Code:
— Modified MIPS code:
add $t0,$s1,$s2
add $t1,$s3,$s4
sub $s0,$t0,$t1
f = (g+h) –(i+j);
#Register $t0 contains g+h
#Register $t1 contains i+j
#Reg. $s0 gets $t0-$t1=(g+h)(i+j);
• What if we have more than 32 variables in our program?
— Must transfer values to and from main memory to work with them.
— We can access single variables, arrays, and other data structures
– located on the stack,
– in statically allocated memory,
– or on a dynamically-allocated heap.
Registers versus Memory
• How can a computer represent and manipulate large
data structures, such as arrays?
• Recall processor contains only small amount of data
in registers, but memory can contain millions (even
billions) of data elements.
• Data transfer instructions (load/store) allow the CPU
to transfer data between registers and memory.
• To access a word in memory, the instruction must
specify a memory address (location).
Memory Organization
registers
Processor
3
2
100
10
1
0
address
114
1
data
Data transfer
Memory
• Memory can be viewed as a large, 1-dimensional array.
• A memory address serves as an index into the array.
• “Byte addressing” means that there is a unique index
for each individual byte (8 bits) of memory.
Recall from last time…
• The instruction set architecture (ISA) of a machine can
be thought of as the hardware’s “user interface.”
— Where by “users” here we mean software engineers, such as
compiler writers and assembly language programmers.
• We are studying the MIPS instruction set architecture;
— MIPS is a reduced instruction set computer (RISC), which
allows for simplified hardware.
• Recall, design principle 1:
“Simplicity favors regularity ”.
• MIPS has 32 registers, each 32 bits long.
— As opposed to hundreds of registers in some architectures.
• Design Principle #2: “Smaller is faster”
— Hence improved performance through reduced
clock cycle time.
Memory Organization
• Recall, that a list of many data elements are stored in an array.
— To access these elements (i.e. memory locations) we use data transfer
instructions: load and store.
• The data transfer instruction that moves data from memory to a
register is called load.
— Think “Load the data into the CPU for processing.”
– Like “Load the dishes into the dishwasher for cleaning.”
— In MIPS the actual instruction is “lw” for load word.
– Other load instructions transfer data of different sizes.
• To transfer data from registers to memory, use store.
— Think “Store the data back in memory after processing.”
– Like “Store the dishes back in the cabinet after washing.”
— MIPS: Use “sw” for “store word.”
• You can think of memory as essentially a large 1-dimensional array,
with the address acting as an index into that array.
Memory Organization
• Example: The address of the third word in the following array
is 8 and the value of Memory[8]=10.
12
100
8
10
4
114
0
1
address data
Memory
* This is an example of byte addressing in which the index refers to a byte of
memory. Since words are 32 bits long, the memory address increments by 4
so that words will always start at addresses that are a multiple of 4. This
requirement is known as alignment restriction; it can help to speed up data
transfers.
Memory Organization
Consider the following example:
Assume that A is an array of 100 words and the compiler has associated registers
$s1 and $s2 with the variables x and y. Also assume that the starting address, or
base address is contained in register $s3. Determine the MIPS instructions
associated with the following C statement:
x = y + A[8];
// adds 8th element in array A to y and stores result in x
Solution:
Before we can perform any arithmetic operations, we must first transfer the data
contained in A[8] to a temporary register.
lw
$t0, 32($s3)
# $s3 contains the base address of array and
# 32 is the offset address of the 8th element
add
$s1, $s2, $t0
# performs addition
Memory Organization
Note that machines can use different “endian-ness” conventions to
order bytes within a word.
Starts with
the “little”
end!
Little Endian
Byte #
0
1
a
a+1
(LSB)
Address:
Starts with
the “big”
end!
2
3
(MSB)
a+2 a+3
-DECStation 3100 Machines
-Intel 80x86 family
3
(MSB)
a
Big Endian
Byte #
2
1
a+1 a+2
0
(LSB)
a+3
- Sun SPARC
- Machintosh (PPC)
- MIPS
From Gulliver’s Travels: “Gulliver finds out that there is a law, proclaimed by the grandfather
of the present ruler, requiring all citizens of Lilliput to break their eggs only at the little ends.
Of course, all those citizens who broke their eggs at the big ends were angered by the
proclamation. Civil war broke out between the Little-Endians and the Big-Endians, resulting
in the Big-Endians taking refuge on a nearby island, the kingdom of Blefuscu.”
Memory Organization
• If a machine uses byte-addressing with 8-bit
addresses, how many different byte locations can be
accessed?
2  256
8
Memory locations with addresses ranging from 0 to 255
• If 32-bit-long byte addresses are used, how many
different aligned 32-bit word locations can be
accessed?
— (Do as an in-class exercise.)
(Hint: Memory locations increment by 4=22.)
Memory Organization Example
Example (using load and store)
Assume that A is an array of 100 words and the compiler has
associated registers $s1 with the variable x. Also assume that
the base address of the array is in register $s2. Determine the
MIPS instructions associated with the following C statement:
A[12] = x + A[8];
Solution:
lw
add
sw
$t0, 32($s2)
$t0, $s1, $t0
$t0, 48($s2)
NOTES:
(1) Store word instruction has destination last as last element.
(2) Remember arithmetic operands are registers only, not memory!
Memory Organization Example
Example (using variable array index)
Assume that A is an array of 100 elements and the base is in $s3. Also
assume that the compiler associates g, h, and i with $s1, $s2, and $s4.
Determine the MIPS instructions associated with the following C statement:
g = h + A[i];
Solution:
We need to know that address of the A[i] before we can load it into a
temporary register. Recall, that to access an element in memory we must
multiply it by 4 to account for byte addressing. To accomplish this we perform
the following sequence of operations:
4i 
add $t1, $s4, $s4
add $t1, $t1, $t1
add $t1, $t1, $s3
lw $t0, 0($t1)
add $s1, $s2, $t0
First, i + i = 2i then 2i + 2i = 4i
#
#
#
#
temp register holds 2i
temp register holds 4i
$t1 holds address of A[i]
loads A[i] into temp register $t0
Recap…
• MIPS
— loading words but addressing bytes
— arithmetic on registers only
• Instruction
add $s1, $s2, $s3
sub $s1, $s2, $s3
lw $s1, 100($s2)
sw $s1, 100($s2)
Meaning
$s1 = $s2 + $s3
$s1 = $s2 – $s3
$s1 = Memory[$s2+100]
Memory[$s2+100] = $s1
• In programs containing more variables than registers, the compiler
tries to keep the most frequently used variables in registers and the
rest in memory. This process is known as spilling. Why is this
important to system performance?
Machine Language
• Both numbers (data) and instructions are stored in computer
hardware as high and low electronic signals (e.g. binary
signals).
• MIPS Assembly Instructions are converted into machine
language using a sequence of 1’s and 0’s (a.k.a machine code.)
• MIPS Instruction Format:
— Composed of different segments called fields
— Instructions are exactly 32 bits long
— Same size as a data word
op
rs
rt
rd
shamt
funct
Machine Language
• MIPS Instruction Format:
op
rs
rt
rd
shamt
funct
• Op:
Opcode – the basic operation of the
instruction
• Rs:
First register source operand
• Rt: Second register source operand
• Rd: register destination operand
• Shamt: Shift amount (explained in Chapter 4)
• Funct: Function – selects the specific variant of the
operation in the opcode
Machine Language
• Design Principle #3: Good design demand good
compromise!
• Thus, all MIPS instructions have the same length (32 bits) but
different formats are used:
(1) R-Type (for Register)
Bit allocation for R-type format:
op: 6bits
rs: 5bits
rt: 5bits
rd: 5bits
shamt:
5bits
(2) I-Type (for data transfer type functions)
Bit allocation for I-type format:
op: 6bits
rs: 5bits
rt: 5bits
Address: 16 bits
funct: 6bits
Machine Language
• Examples:
(i) add $t0, $s1, $s2
In decimal representation:
0
17
18
8
0
32
10010
01000
00000
100000
In binary representation:
000000
10001
(ii) lw $t0, 32($s3)
In decimal representation:
35
19
8
Binary representation: (do on own)…
Where’s the compromise?
32
Machine Language
• Appendix A (on the CD-ROM), pages A-50 through
A-81, gives format for assembly language instructions
in the third edition.
• Example:
For the given C statement, determine its MIPS
assembly code, as well as, its corresponding machine
code. Assume that the base address for A is contained
in $s2.
A[100] = x + A[50];
(do on own)
Stored-Program Concept
• Since instructions are represented as numbers, programs can be
stored in memory to be read just like data.
• This idea leads to the stored-program concept, which allows
a computer to execute different programs that are stored in
memory.
Memory
Accounting
Program
Processor
Editor
Program
C Compiler
Code
Payroll
Account
memory for data, programs,
compilers, editors, etc.
Book
Text
Instructions for Decision-making
• Decision making instructions:
— Computers have the ability to make decisions.
— The “next” instruction to be executed depends on the
outcome of the decision.
— Many programming languages use the if statement and/or
the goto statement to represent decision-making.
• MIPS language uses conditional branch instructions
(branch if equal, branch if not equal):
(1) bne register1, register2, L1
- go to statement labeled L1 if value in register 1 does
not equal value in register 2
(2) beq register1, register2, L1
- go to statement labeled L1 if value in register 1 equals
value in register 2
Control Flow Examples
Example:
For the given C statement, assume that variables f through j
correspond to registers $s0 through $s4. What is the compiled
MIPS code?
L1:
if (i == j) goto L1;
f = g + h;
f = f - i;
Solution:
We’ll need a branch if equal (beq) statement to correspond to
the ‘if’ command:
beq $s3, $s4, L1
add $s0, $s1, $s2
# skipped if i==j
Now we just have to identify the code for the label L1. Consider, if the
conditional branch is true, then the add instruction is skipped. How do
we specify the label such that the last instruction is always executed?
Control Flow Examples
• In stored-program computers, instructions are stored
in memory,
— thus they are identified using memory addresses.
• L1 will correspond to the address of the subtract
instruction.
L1: sub $s0, $s0, $s3
• Complete solution:
C Code:
L1:
if (i == j) go to L1;
f = g + h;
f = f - i;
L1:
beq $s3, $s4, L1
add $s0, $s1, $s2
sub $s0, $s0, $s3
MIPS assembly:
Control Flow Examples
Example:
For the given C statement, assume that variables f
through j correspond to registers $s0 through $s4. What
is the compiled MIPS code?
if
(i == j)
f = g + h;
else
f = g - h;
MIPS Code:
bne
add
j
else: sub
exit:
$s3, $s4, else
$s0, $s1, $s2
exit
$s0, $s1, $s2
Unconditional Branch is
used; the machine always
takes (follows) this branch.
MIPS uses “j” for “jump”
to distinguish it from
conditional branches.
Unconditional Branches
• We can use MIPS unconditional branch instructions to
implement if-else statements, as well as for and while
loops using the format:
j
label
• Example:
if (i!=j) i≠j
h=i+j;
else
h=i-j;
i=j
beq
add
j
Lab1: sub
Lab2: ...
$s4, $s5, Lab1
$s3, $s4, $s5
Lab2
$s3, $s4, $s5
i=j
Unconditional Branches
Example (while loop):
Write the MIPS assembly code for the following C code segment.
Assume that i, j, k correspond to $s3, $s4, $s5 and the base of the
array SAVE is contained in $s6.
while (save [i] = = k)
i = i + j;
Solution
loop:
exit:
add $t1, $s3, $s3
add $t1, $t1, $t1
add $t1, $t1, $s6
lw $t0, 0($t1)
bne $t0, $s5, exit
add $s3, $s3, $s4
j loop
Control Flow Examples
• We looked at beq, bne, and j what about other
conditional statements?
SLT: Set-on-less-than
Compares two registers, if first register is less than
second then it sets destination to 1, otherwise
destination register contains 0.
* We use this for case/switch statements.
2
Summary
• Arrays, byte addressing, alignment restriction
• Machine language conventions
(big and little endian)
• Instruction format types (R-type and I-type)
• Stored program concept
• Decision-making instructions, conditional and
unconditional branches
• Next time: case/switch statements,
supporting procedures in computer hardware,
arrays versus pointers, etc…
2
Recall from last time…
•
•
•
•
•
Memory Organization
Machine Language
Instruction Formats
Decision-making Instructions; Control Flow
Conditional Branches
Today…
• Case/Switch statements using branches
• Procedures
• MIPS addressing modes
• Arrays versus pointers
Case/Switch Statements
• Case/Switch statements are used in many programming
languages to allow the user to select one of many choices.
• It can be implemented as a sequence of if-then-else
statements.
• We can also use a jump address table to encode
alternatives. The program will index the table and them
jump to the appropriate instruction sequence.
• MIPS uses a jump register (jr) instruction to identify
the proper address of the jump table.
Case/Switch Statements
Example
Assume the variables f through k correspond to $s0 through $s5
and register $t2 contains 4. What is the associated MIPS code for
the following switch statement written in C?
switch (k) {
case 0: f = i + j; break:
case 1: f = g + h; break:
case 2: f = g - h; break:
case 3: f = i - j; break:
}
/* k = 0 */
/* k = 1 */
/* k = 2 */
/* k = 3 */
k is an index that contains the address of the instruction to be
executed.
Case/Switch Statements
Example cont…
k should equal 0, 1, 2, or 3 to enter the jump address table, if it doesn’t then it
should exit the switch command:
slt $t3, $s5, $zero
# test if k < 0
bne $t3, $zero, Exit
# Exit if k <0
slt $t3, $s5, $t2
# test if k > 4
beq $t3, $zero, Exit
# Exit if k 
4
Convert k to a byte address:
add $t1, $s5, $s5
add $t1, $t1, $t1
# $t1 = 4k  offset
Assume that 4 sequential words in memory, starting at an address contained in
$t4, have addresses corresponding to the labels L0, L1, L2, and L3. Then we
load the proper jump address as
add $t1, $t1, $t4
# $t4  base address
lw $t0, 0($t1)
# $t0 contains address of instr
jr $t0
# jump to address in reg $t0
Case/Switch Statements
Example cont…
Next define the labels L0, L1, L2, and L3:
L0: add $s0, $s3, $s4
j Exit
L1: add $s0, $s1, $s2
j Exit
L2: sub $s0, $s1, $s2
j Exit
L3: sub $s0, $s3, $s4
Exit:
Case/Switch Statements
Complete Example
C Code:
switch (k) {
case 0:
case 1:
case 2:
case 3:
f = i + j; break:
f = g + h; break:
f = g - h; break:
f = i - j; break:
/* k = 0 */
/* k = 1 */
/* k = 2 */
/* k = 3 */
}
MIPS Assembly Code:
L0:
L1:
L2:
L3:
Exit:
slt $t3, $s5, $zero
bne $t3, $zero, Exit
slt $t3, $s5, $t2
beq $t3, $zero, Exit
add $t1, $s5, $s5
add $t1, $t1, $t1
add $t1, $t1, $t4
lw $t0, 0($t1)
jr $t0
add $s0, $s3, $s4
j Exit
add $s0, $s1, $s2
j Exit
sub $s0, $s1, $s2
j Exit
sub $s0, $s3, $s4
# test if k < 0
# Exit if k <0
# test if k < 4
# Exit if k > 4
# $t1 = 4k  offset
# $t4  base address
# $t0 contains address of instr
# jump to address in reg $t0
# define instructions for Case 1
# define instructions for Case 2
# define instructions for Case 3
# define instructions for Case 4
# End of statement
Supporting Procedures
•
A procedure is a tool that is used to structure programs to
make them easier to understand and to reuse.
•
There are 6 steps to be followed when executing a procedure:
1. Place parameters in a place where procedure can
access them;
2. Transfer control to the procedure;
3. Acquire the storage resources needed for the
procedure;
4. Perform the desired task;
5. Place the result in an accessible place;
6. Return control to the point of origin.
Supporting Procedures
•
MIPS allocates special registers for supporting procedures and
procedure calling:
$a0 - $a3  argument registers to pass parameters
$v0 - $v1  value registers to return values
$ra  return address register to return to origin
•
Also, an instruction just for procedures is used jump-and-link
(jal) instruction; it jumps to an address and saves the address of
the following instruction in the $ra register.
jal ProcedureAddr
•
The “link” portion stores the return address in $ra:
Return addr = PC + 4
Program Counter
Procedures
Note:
- Procedure calls preserve registers $s0 - $s7 (saved registers) and erase
values stored in temporary registers $t0 - $t7.
- Nested procedures are also possible. All of the registers that are needed in
the caller program are pushed onto the stack.
Example:
Determine the MIPS assembly code for the following C code:
int
leaf_example(int g, int h, int i, int j)
{
int f;
f = (g + h) – (i + j);
return f;
}
Procedures
Recall MIPS Code for the statement:
f = (g + h) – (i + j);
•
add $t0, $s1, $s2
# register $t0 contains g+h
add $t1, $s3, $s4
# register $t1 contains i+j
sub $s0, $t0, $t1
# register $s0 contains $t0-$t1= (g+h) –(i+j);
We’re going to create a subroutine around this operation. Since we’re passing
arguments, we must use argument registers $a0 - $a3 for the variables g through j.
We’ll use $s0 for variable f.
add $t0, $a0, $a1 # register $t0 contains g+h
add $t1, $a2, $a3 # register $t1 contains i+j
sub $s0, $t0, $t1
# register $s0 contains $t0-$t1= (g+h) –(i+j);
Procedures
-
Recall we’re using a LIFO structure/stack so we must store the values contained in
the $t0, $t1, $s0 registers initially:
sub $sp, $sp, 12
sw $t1, 8($sp)
sw $t0, 4($sp)
sw $s0, 0($sp)
Next we can load our code for our operation:
add $t0, $a0, $a1
add $t1, $a2, $a3
sub $s0, $t0, $t1
The return value for f is copied to the return value register:
add $v0, $s0, $zero
# returns f ($v0 = $s0 +0)
Restore old values in registers that we saved initially:
lw $t1, 8($sp)
lw $t0, 4($sp)
lw $s0, 0($sp)
add $sp, $sp, 12
Procedures
Finally we use a jump register instruction to go to the return address:
jr $ra
BEFORE Procedure Call
AFTER Procedure Call
DURING Procedure Call
$sp
$sp
Contents of $t1
Contents of $t0
$sp
Contents of $s0
- The Stack Pointer always points to the “top” of the stack or the last word in the stack.
- “Push”ing registers onto stack ensures that the stack above $sp is preserved.
Procedures
Putting it all together…
C Code:
int
leaf_example(int g, int h, int i, int j)
{
int f;
f = (g + h) – (i + j);
return f;
}
MIPS Assembly
sub
sw
sw
sw
add
add
sub
add
lw
lw
lw
add
jr
$sp, $sp, 12
$t1, 8($sp)
$t0, 4($sp)
$s0, 0($sp)
$t0, $a0, $a1
$t1, $a2, $a3
$s0, $t0, $t1
$v0, $s0, $zero
$t1, 8($sp)
$t0, 4($sp)
$s0, 0($sp)
$sp, $sp, 12
$ra
Note:
- Since $t0 and $t1 are temporary registers and are not typically
preserved during a procedure call, we can drop the 2 stores and
2 load commands. What is the updated code? (on own)
Procedures
-
The stack is used to store contents of registers as well as to store local variables that
are local to the procedure.
-
The segment of the stack that contains the procedure’s saved registers and local
variables is called the register frame or activation record.
-
A frame pointer ($fp) points to the first word of the frame of a procedure. It can be
used as a stable base register within a procedure to access local memory references.
Its use is optional.
BEFORE Procedure Call
$fp
DURING Procedure Call
$fp
Saved Arg regs
Saved rtn addr
Saves Sve regs
$sp
$sp
Local arrays
and data
structures
AFTER Procedure Call
$fp
$sp
Representing Text
-
We process numbers, as well as, text using the American Standard Code
for Information Interchange (ASCII) character representation.
-
Recall that ASCII characters are represented using 8 bits = 1 byte.
-
MIPS instructions allows us to move bytes from words using load byte (lb)
–loads a byte from memory and places it in rightmost 8 bits of a register;
and store byte (sb) – takes a rightmost byte from register and places it into
memory.
-
We can copy a byte with the following sequence:
lb $t0, 0($sp)
sb $t0, 0($gp)
# Read byte from source
# Write byte to a destination
Example
String Copy Procedure Example:
MIPS Assembly
C Code:
strcpy:
void strcpy(char x[ ], char y[ ])
{
int i;
i = 0;
while ((x[i] = y[i]) != 0) /* copy and test byte */
i = i + 1;
}
L1:
Assume base addresses for x and y are
found in $a0 and $a1 and i is in $s0.
L2:
Note also that x and y are arrays of
characters so there is no need to multiply
by 4 to obtain the address.
sub
$sp, $sp, 4
sw
$s0, 0($sp)
add
$s0,$zero,$zero
add $t1, $a1, $s0
lb
$t2, 0($t1)
add $t3, $a0, $s0
sb $t2, 0($t3)
beq $t2, $zero, L2
addi $s0, $s0, 1
j L1
lw $s0, 0($sp)
add $sp, $sp, 4
jr $ra
Constants
• Small constants are used quite frequently (50% of operands)
e.g.
A = A + 5;
B = B + 1;
C = C - 18;
• Solutions? Why not?
- put 'typical constants' in memory and load them (takes time!).
- create hard-wired registers (like $zero) for constants like one.
- Encode constant in instruction (I-type formats)
- Some “immediate” MIPS Instructions (addi, slti, andi, ori, lui).
This leads to design principle #4:
Make the common case fast!!
Procedures
•
In summary:
- Call a procedure by first putting parameters in $a0-$a3
- Use jal to jump to procedure
- Perform calculations within procedure
- Place results in $v0-$v1
- Return control to caller program by using jr $ra
•
If we need more registers to hold parameters we can use
spilling to accomplish this.
•
The ideal structure for spilling registers is called a stack (lastin-first-out queue). A stack pointer ($sp) is used to index the
most recently allocated address on the stack.
•
•
Data placed onto stack  “Push”
Data removed from stack  “Pop”
Larger Constants
•
•
We'd like to be able to load a 32 bit constant into a register
Must use two instructions, new "load upper immediate" instruction
lui $t0, 1010101010101010
filled with zeros
1010101010101010
•
0000000000000000
Then must get the lower order bits right, i.e.,
ori $t0, $t0, 1010101010101010
1010101010101010
0000000000000000
0000000000000000
1010101010101010
1010101010101010
1010101010101010
ori
$t0
Larger Constants
•
Example:
Determine the sequence of MIPS instructions for the following C segment
x[10] = x[11] + c;
Assuming that c is contained in $t0 and that array x has a base address of
(4,000,000)10
First load base address into a register:
(4,000,000)10  (003D0900)16
 (0000 0000 0011 1101 0000 1001 0000 0000) 2
Solution:
lui
ori
lw
add
sw
$t1, $t1, 0000 0000 0011 1101
$t1, $t1, 0000 1001 0000 0000
$t2, 44($t1)
$t2, $t2, $t0
$t2, 40($t1)
# load upper 16 bits
# load lower 16 bits using OR imm
# load element x[11] into $t2
# sum x[11] and c; put result in $t2
# store it back into memory
On own: Write MIPS assembly that loads 32-bit word into register $t5:
0000 0000 0011 1101 0000 1001 0000 0000
Assembly Language vs Machine Language
• Assembly provides convenient symbolic representation specific to a
particular architecture
• This is much easier than writing down sequences of binary
numbers (machine code) which is the communication mechanism
of a machine.
• Assembly can also provide 'pseudoinstructions‘ (instructions that
are not actually implemented in hardware but make assembly
coding easier for the programmer).
— e.g., “move $t0, $t1” exists only in assembly
— would be implemented using “add $t0,$t1,$zero” in MIPS
• However, when considering performance you should count real
instructions that will be implemented in hardware.
Recall…
• Procedure Example
• Character representations
• Constants and Immediate
This time…
• Addressing in branches and jumps
• MIPS addressing modes
• Arrays versus pointers
• Examples of other architectures
Addressing in Branches and Jumps
• MIPS jump instructions have the simplest addressing:
j 10000
J-type Format:
# go to location 10000
2
2500
op
(6 bits)
location
(26 bits)
• Conditional Branches Instructions:
bne
$s0, $s1, Exit
Left-shifted by 2
(multiplied by 4)
before use
# goto exit if $s0  $s1
I-type format:
op: 6bits
rs: 5bits
rt: 5bits
Address: 16 bits
Addressing in Branches and Jumps
• Most conditional branches are local meaning they
tend to branch to nearby locations (principle of
locality). Examples would be if-else statements. In
this case,
Program Counter (PC) = register + branch address
(PC-relative addressing)
• For jump and jump-and-link instructions which
execute procedures, far away branching is more
common. These instructions use j-type format.
• Consider the following:
beq
$s0, $s1, Label1
Replace this expression with a sequence of instructions that
allows greater branching distance.
Replace with 
bne
j
L2:
$s0, $s1, L2
L1
Addressing in Branches and Jumps
• Example
Assume that the following while loop is placed starting at
memory location 80000, what is the MIPS machine code for this
segment?
MIPS Machine Code
loop: add
$t1, $s3, $s3 80000 0
19
19
9
0
0
9
9
9
0
add
$t1, $t1, $t1
80004
80008
0
9
22
9
0
add
$t1, $t1, $s6
35
80012
9
8
0
lw
$t0, 0($t1)
8
21
8
bne
$t0, $s5, Exit 80016 5
19
20
19
0
add
$s3, $s3, $s4 80020 0
80024
2
20000
j
loop:
80028
...
Exit:
32
32
32
32
Note: bne instruction adds 8 bytes to the following instruction
which corresponds to 80020 + 8 = 80028 (addr of exit)
(PC-relative Addressing)
In summary:
MIPS operands
Name
32 registers
Example
Comments
$s0-$s7, $t0-$t9, $zero, Fast locations for data. In MIPS, data must be in registers to perform
$a0-$a3, $v0-$v1, $gp,
arithmetic. MIPS register $zero always equals 0. Register $at is
$fp, $sp, $ra, $at
reserved for the assembler to handle large constants.
Memory[0],
Accessed only by data transfer instructions. MIPS uses byte addresses, so
30
2 memory Memory[4], ...,
sequential words differ by 4. Memory holds data structures, such as arrays,
words
and spilled registers, such as those saved on procedure calls.
Memory[4294967292]
add
MIPS assembly language
Example
Meaning
add $s1, $s2, $s3
$s1 = $s2 + $s3
Three operands; data in registers
subtract
sub $s1, $s2, $s3
$s1 = $s2 - $s3
Three operands; data in registers
$s1 = $s2 + 100
$s1 = Memory[$s2 + 100]
Memory[$s2 + 100] = $s1
$s1 = Memory[$s2 + 100]
Memory[$s2 + 100] = $s1
Used to add constants
Category
Arithmetic
Instruction
addi $s1, $s2, 100
lw $s1, 100($s2)
sw $s1, 100($s2)
store word
lb $s1, 100($s2)
load byte
sb $s1, 100($s2)
store byte
load upper immediate lui $s1, 100
add immediate
load word
Data transfer
Conditional
branch
Unconditional jump
$s1 = 100 * 2
16
Comments
Word from memory to register
Word from register to memory
Byte from memory to register
Byte from register to memory
Loads constant in upper 16 bits
branch on equal
beq
$s1, $s2, 25
if ($s1 == $s2) go to
PC + 4 + 100
Equal test; PC-relative branch
branch on not equal
bne
$s1, $s2, 25
if ($s1 != $s2) go to
PC + 4 + 100
Not equal test; PC-relative
set on less than
slt
$s1, $s2, $s3
if ($s2 < $s3) $s1 = 1;
else $s1 = 0
Compare less than; for beq, bne
set less than
immediate
slti
jump
j
jr
jal
jump register
jump and link
$s1, $s2, 100 if ($s2 < 100) $s1 = 1;
Compare less than constant
else $s1 = 0
2500
$ra
2500
Jump to target address
go to 10000
For switch, procedure return
go to $ra
$ra = PC + 4; go to 10000 For procedure call
MIPS Addressing Modes

Different modes of addressing are used to implement different
types of instructions. MIPS addressing modes are as follows:
•
Immediate addressing:
operand is a constant within the instruction itself (e.g. ‘addi’)
•
Register addressing:
operand is a register (e.g. ‘add’)
•
Base or displacement addressing:
operand is at the memory location whose address is the sum of a
register
and a constant in the instruction (e.g. ‘lw’)
•
PC-relative addressing:
address is the sum of of the PC and the constant in the instruction
(e.g.
branches)
•
Pseudodirect addressing:
jump address is the 26bits of the instruction concatenated with the
upper bits of the PC
1. Immediate addressing
op
rs
rt
Immediate
2. Register addressing
op
rs
rt
rd
...
funct
Registers
Register
3. Base addressing
op
rs
rt
Memor y
Address
+
Register
Byte
Halfword
4. PC-relative addressing
op
rs
rt
Memor y
Address
PC (+4)
+
Word
5. Pseudodirect addressing
op
Address
PC
Memor y
<<2
Word
Word
Decoding Machine Code
Example
What is the assembly language corresponding to this machine
code?
0000 0000 1010 1111 1000 0000 0010 0000
First consider the opcode field:
op:
000000  several different arithmetic codes
Next look at function code field:
func:
100000  this corresponds to the ‘add’ instruction
Now reformat machine instruction using R-type format:
(op)
binary
decimal
(rs)
(rt)
(rd)
(shamt) (func)
000000 00101 01111 10000 00000 100000
0
5
15
16
0
32
The MIPS assembly instruction is
add
$s0, $a1, $t7
Arrays versus Pointers
Example
What is the assembly language for the following C procedures:
Array Version
Pointer Version
clear1(int array[ ],int size)
{
int i;
for (i=0; i<size; i=i+1)
array[i] = 0;
}
clear2(int *array,int size)
{
int *p;
for (p=&array[0]; p<&array[size]; p=p+1)
*p = 0;
}
Arrays versus Pointers
Example cont.
Array Version of Clear:
move
Loop1: add
add
add
sw
addi
slt
bne
$t0, $zero
$t1, $t0, $t0
$t1, $t1, $t1
$t2, $a0, $t1
$zero, 0($t2)
$t0, $t0, 1
$t3, $t0, $a1
$t3, $zero, loop1
Pointer Version of Clear:
move
add
add
add
Loop2: sw
addi
slt
bne
$t0, $a0
$t1, $a1, $a1
$t1, $t1, $t1
$t2, $a0, $t1
$zero, 0($t0)
$t0, $t0, 4
$t3, $t0, $t2
$t3, $zero, loop2
Alternate Architectures
• Design alternative:
—provide more powerful operations and flexibility in
designs
—goal is to reduce number of instructions executed
—danger is a slower cycle time and/or a higher CPI
• Example architectures: PowerPC and Intel
80x86
• Refer to your textbook for more information
on these architectures.
Summary
• Instruction complexity is only one
variable
—lower instruction count vs. higher CPI / lower
clock rate
• Design Principles:
—simplicity favors regularity
—smaller is faster
—good design demands compromise
—make the common case fast
• Instruction set architecture
—a very important abstraction