CS 11 C track: lecture 8
n Last week: hash tables, C preprocessor
n This week:
n Other integral types: short, long, unsigned
n bitwise operators
n switch
n "fun" assignment: virtual machine
Integral types (1)
n Usually use int to represent integers
n But many other integral (integer-like) types
exist:
n short
n long
n char
n unsigned int
n unsigned short
n unsigned long
n unsigned char
Integral types (2)
n Two basic things that can vary:
n unsigned vs. signed (default)
n length: char, short, int, long
n Note that char is an integral type
n can always treat char as an 8-bit integer
n Two basic questions:
n Why use unsigned types?
n When should we use shorter/longer integral types?
Integral types (2)
n Why use unsigned types?
n may be used for something that can't be negative
n e.g. a length
n gives you 2x the range due to last bit
n may want to use it as an array of bits
n so sign is irrelevant
n C has lots of bitwise operators
Integral types (3)
n When should we use shorter/longer integral
types?
n to save space when we know range is limited
n when we know the exact number of bits we need
n char always 8 bits
n short usually 16 bits
n int usually 32 bits (but sometimes 64)
n long usually 32 bits (but sometimes 64)
n guaranteed: length(char) < length(short)
<= length (int) <= length(long)
Integral types (4)
n unsigned by itself means unsigned int
n Similarly it's legal to say
n short int
n unsigned short int
n long int
n unsigned long int
n but usually we shorten by leaving off the int
Bitwise operators (1)
n You don't need to know this for this lab!
n But a well-rounded C programmer should know
this anyway...
n There are several "bitwise operators" that do
logical operations on integral types bit-by-bit
n OR ( | ) (note difference from logical or: ||)
n AND ( & ) (note difference from logical and: &&)
n XOR ( ^ )
n NOT ( ~ ) (note difference from logical not: !)
Bitwise operators (2)
n bitwise OR (|) and AND (&) work bit-by-
bit
n 01110001 | 10101010 = ?
n 11111011
n 01110001 & 10101010 = ?
n 00100000
n NOTE: They don't do short-circuit evaluation
like logical OR (||) and AND (&&) do
n because that wouldn't make sense
Bitwise operators (3)
n bitwise XOR (^) also works bit-by-bit
n 01110001 ^ 10101010 = ?
n 11011011
n Bit is set if one of the operand's bits is 1
and the other is 0 (not both 1s or both
0s)
Bitwise operators (4)
n bitwise NOT (~) also works bit-by-bit
n ~10101010 = ?
n 01010101 (duh)
n Substitute 0 for 1 and 1 for 0
Bitwise operators (5)
n Two other bitwise operators:
n bitwise left shift ( << )
n bitwise right shift ( >> )
n 00001111 << 2 = ?
n 00111100
n 00111100 >> 2 = ?
n 00001111
n Can use to multiply/divide by powers of 2
switch (1)
n Minor language feature: switch
n Used to choose from multiple integer-valued
possibilities
n Cleaner than a series of if/else if/else
statements
switch (2)
n Common coding pattern:
void do_stuff(int i) {
if (i == 0) {
printf("zero\n");
} else if (i == 1) {
printf("one\n");
} else {
printf("something else\n");
}
}
switch (3)
void do_stuff(int i) {
switch (i) {
case 0:
printf("zero\n");
break;
case 1:
printf("one\n");
break;
default:
printf("something else\n");
break;
}
}
switch (3)
n switch statements more convenient than
if/else if/else for many integer-valued
cases
n but not as general -- can only be used on integral
types (int, char, etc.)
n Lab 8 code contains one switch statement that
you don't have to write
n but you should understand it anyway
switch (4)
switch (i) {
case 0: /* Start here if i == 0 */
printf("zero\n");
break; /* Exit switch here. */
... /* other cases: 1, 2, 42 etc. */
default: /* if no case matches i */
printf("no match\n");
break;
}
switch (5) -- fallthrough
switch (i) {
case 0: /* Start here if i == 0 */
printf("zero\n");
/* oops, forgot the break */
case 1: /* "fall through" from case 0 */
printf("one\n");
break;
}
n Now, if i is 0 then prints "zero" and also "one"!
n Sometimes this is desired, but usually just a bug
Lab 8: Virtual machine (1)
n Where have you heard the term "virtual
machine" before?
n Java virtual machine
n A "virtual microprocessor"
n You define simple instructions for a
mythical computer's assembly language
n Program interprets them
Virtual machine (2)
n Our virtual machine is very simple
n Only data type will be int
n All instructions will act on ints
n Instructions include
n arithmetic
n control flow
n memory access
n printing
Virtual machine (3)
n First need to define data structures for
our virtual microprocessor:
n instruction memory to hold instructions of
program
n registers to hold temporary results of
computations
n stack to hold results that are being operated
on directly
Virtual machine (4)
n Instruction memory contains 216 locations
n = 65536
n Each location is a single byte (unsigned char)
n How many bits do we need to represent all
possible locations in instruction memory?
n 16
n Can use an unsigned short for this
n Called the "instruction pointer" or IP
n Don't confuse with C's pointers! Not the same
thing!
n It's just an index into the instruction memory
Virtual machine (5)
n 16 registers (temporary storage locations)
n How many bits do we need to represent all
possible locations in registers?
n 4
n Can use an unsigned char for this
n Registers are just an array of 16 ints
Virtual machine (6)
n Stack which is 256 deep
n How many bits do we need to represent
all possible locations in stack?
n 8
n Can use an unsigned char for this
n called the "stack pointer" or SP
n also not a pointer in the C sense, just an
index
n Stack is just an array of 256 ints
Push and pop (1)
n Stack has two operations: push and pop
n push puts a new value onto the stack
n pop removes a value from the stack
n Have to adjust stack pointer (SP) after push
and pop
n Stack pointer "points to" first UNUSED element
of stack
n starts at zero for empty stack
n Top filled element in stack is "top of stack"
(TOS)
Push and pop (2)
Stack starts off
empty;
SP points to first
unused location
SP
stack
Push and pop (3)
push 10 onto
stack
SP
TOS 10
stack
Push and pop (4)
push 20 onto
SP
stack
TOS 20
10
stack
Push and pop (5)
pop stack;
20 still there,
SP 20 but will be
overwritten next
push
TOS 10
stack
Push and pop (6)
push 30 onto
SP
stack;
TOS 30 old value (20)
gets overwritten
10
stack
Push and pop (7)
pop twice;
stack is now
30 "empty" again
SP 10
stack
VM instruction set (1)
n VM instructions are often called
"bytecode"
n because they fit into a byte (8 bits)
n represented as an unsigned char
n Our VM has 14 different instructions
n some take operands (some number of bytes)
n some don't
VM instruction set (2)
n Instructions:
n NOP (0x00) – does nothing ("No OPeration")
n PUSH (0x01) – PUSH <n> pushes the integer
<n> onto the stack
n POP (0x02) – removes the top element on the
stack
n LOAD (0x03) – LOAD <r> pushes contents of
register <r> to the top of the stack
n STORE (0x04) – STORE <r> pops top of stack
and puts contents into register <r>
Load (1)
42 ....
0 1 ....
registers
load 0
SP
stack
Load (2)
42 ....
0 1 ....
registers
load 0;
SP
42 pushed onto
stack
TOS 42
stack
Store (1)
42 ....
0 1 ....
registers
store 1
SP
42
stack
Store (2)
42 42 ....
0 1 ....
registers
store 1;
topmost element
of stack copied
SP 42 into register 1;
stack popped
stack
VM instruction set (3)
n Control flow instructions:
n JMP (0x05) – JMP <i> sets the instruction
pointer (IP) to <i> ("jump")
n JZ (0x06) – JZ <i> sets IP to <i> only if the
top value on the stack (TOS) is zero; also
pops stack ("jump if zero")
n JNZ (0x07) – JNZ <i> sets IP to <i> only if
the TOS is not zero; also pops stack ("jump
if nonzero")
VM instruction set (4)
n Arithmetic instructions:
n ADD (0x08) – pops the top two entries in the
stack, adds them, pushes result back
n SUB (0x09) – pops the top two entries in the
stack, subtracts them, pushes result back
n Watch order! Should be S2 – S1 on TOS
n MUL (0x0a) and DIV (0x0b) defined similarly
Sub (1)
SP
sub
TOS 31
(before)
42
stack
Sub (2)
sub
SP 31
(after)
TOS 11
stack
VM instruction set (5)
n Other instructions:
n PRINT (0x0c) – prints the TOS to stdout and
pop TOS
n STOP (0x0d) – terminates the virtual program
Example program (1)
n Program to generate factorial of 10 (10!)
n Which means...?
n 10 x 9 x 8 x 7 x 6 x 5 x 4 x 3 x 2 x 1
n = 3628800
n But we'll write a program in our virtual
machine's language
Example program (2)
n Register 0 will contain the count
n Register 1 will contain the running total
n Register 0 will start off at 10
n each step, will decrease by 1
n Register 1 will start off at 1
n each step, will be multiplied by register 0 contents
n Continue until register 0 has 0
n result is in register 1
Example program (3)
/* Initialize the registers. */
push 10
store 0
push 1 /* Initialize result. */
store 1
/* continued on next slide... */
push 10
....
0 1 ....
registers
SP
TOS 10
stack
store 0
10 ....
0 1 ....
registers
SP 10
stack
push 1
10 ....
0 1 ....
registers
SP
TOS 1
stack
store 1
10 1 ....
0 1 ....
registers
SP 1
stack
Example program (4)
/* Put counter value on stack.
* If it's 0, we're done; register 1
* contains the final value. */
1 load 0 /* Load current count. */
jz 2 /* if 0, jump to 2 */
/* 1,2 are "labels"; represent the
* location of instructions which are
targets of jmp, jz, jnz operations. */
Example program (5)
/* result = result * count */
load 1
load 0
mul
store 1
/* count = count - 1 */
load 0
push 1
sub
store 0
load 1
10 1 ....
0 1 ....
registers
SP
TOS 1
stack
load 0
10 1 ....
0 1 ....
registers
SP
TOS 10
stack
mul
10 1 ....
0 1 ....
registers
SP 10
TOS 10
stack
store 1
10 10 ....
0 1 ....
registers
10
SP 10
stack
load 0
10 10 ....
0 1 ....
registers
SP 10
TOS 10
stack
push 1
10 10 ....
0 1 ....
registers
SP
TOS 1
10
stack
sub
10 10 ....
0 1 ....
registers
SP 1
TOS 9
stack
store 0
9 10 ....
0 1 ....
registers
etc. ...
1
SP 9
stack
Repeating...
n Registers start off as 10, 1
n Then become 9, 10
n 8, 10*9
n 7, 10*9*8
n ...
n 0, 10!
n ... and we're done.
Example program (6)
/* Go back and loop until done. */
jmp 1
/* When we get here, we're done. */
2 load 1
print
stop
/* End of program. */
Lab 8
n Program is given to you
n You need to write the byte-code interpreter
n Most of code is supplied; have to fill in the guts
of the instruction-processing code
n Looks complicated but actually is pretty easy
n Watch out for error checking e.g.
n popping an empty stack
n pushing to a full stack
n accessing non-existent register or instruction
Lab 8 -- error checking
n One subtlety with stack pushes
n If stack pointer is at 255, and you push onto
stack, what is the new stack pointer value?
n 0
n (256 is too large for an unsigned char)
n But this is clearly incorrect
n How to detect "stack overflow"?
n Solution: don't allow overflow!
n If stack pointer is 255, a push is invalid
Finally...
n Hope you enjoyed the course!
n If so, consider taking
n other CS 11 tracks
n (C++, Java, advanced C++/Java)
n CS 11 project track
n CS 24
n CS 2 for larger-scale software projects
n CS 3 for larger-scale software projects in C
