Scribd
Upload
49 views

CS109A Notes For Lecture 1/31/96 Measuring The Running Time of Programs Example

The document discusses measuring the running time of programs. It introduces analyzing the running time by induction on program structures and their nesting depth. Simple statements like assignments take constant time O(1), while complex statements built from simple ones using loops, branches, and blocks can be analyzed inductively. As an example, a binary conversion function runs in O(log n) time, where n is the input size. A nested double loop like insertion sort has an overall running time of O(n2), even though the inner loop iterates a varying number of times.

Uploaded by

savio77
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
49 views

CS109A Notes For Lecture 1/31/96 Measuring The Running Time of Programs Example

The document discusses measuring the running time of programs. It introduces analyzing the running time by induction on program structures and their nesting depth. Simple statements like assignments take constant time O(1), while complex statements built from simple ones using loops, branches, and blocks can be analyzed inductively. As an example, a binary conversion function runs in O(log n) time, where n is the input size. A nested double loop like insertion sort has an overall running time of O(n2), even though the inner loop iterates a varying number of times.

Uploaded by

savio77
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

CS109A Notes for Lecture 1/31/96

Measuring the Running Time of Programs

Fix a measure of the \size" n of the data to which


a program is being applied.
Example: For integer arguments, the value is
often a good size measure. For strings: the length.
Compute a big-oh upper bound on the running time of a program by induction on the
complexity of program structures, i.e., the
depth to which structures are nested.
Try to make the bound simple and tight.
In the following, we assume there are no function calls in the program except for I/O operations.
Basis: Simple statements contain no statements
nested within them. In C:
1. Assignment statements.
2. Goto's, including break, continue, return.
3. Input/Output using function calls like printf
or getchar().
Fundamental assumption: Application of an
operator takes a constant amount of time.
Write \some constant" as O(1).
Operators include arithmetic, comparison, logical.
ML has some exceptions: concatenation
of strings or lists.
Thus, in C every simple statement takes O(1)
time.
Induction: Complex statements built from simple statements by recursive application of:
1. Loop formers: for-, while-, repeat-.
2. Branching statements: if else-, if-, case-.
3. Block formers: { }.
1

Structure Trees

node = complex statement; its children are


the constituent statements.

Example:
(1)
(2)
(3)
(4)
(5)

main() {
int i;
scanf("%d", &i);
while(i>0) {
putchar('0' + i%2);
i /= 2;
}
putchar('\n');
}

Block
1{5
1

While
2{4
Block
3{4
3

Details of the Induction


Blocks : Running time bound = sum of the
bounds of the constituents.
Use summation law to drop from the sum
any term that is big-oh of another term.
Conditionals : Bound = O(1) + larger of
bounds for the if- and else- parts.
O(1) is for cost of the test | usually neglectable.
Loops : Bound is usually the maximum number of times around the loop the bound on
the time to execute the loop body.

But we must include O(1) for the increment and test each time around the loop.
The possibility that the loop is executed
0 times must be considered. Then, O(1)
for the initialization and rst test is the
total cost.
Example: Consider binary-conversion function.
Size of data = i.
Lines 1, 3, 4, 5 each O(1) by the basis.
Block of 3{4 is O(1) + O(1) = O(1).
While of 2{4 iterates at most log2 i times.
Bound on body times number of iterations =
O(1) log2 i = O(log i).
Block of 1{5 is O(1) + O(log i) + O(1) =
O(log i).
i.e., it takes O(log i) time to convert i to binary by this function.

Triangular Double Loops

Sometimes we need to \give up" trying to tighten


the upper bound on running time. Example: an
inner loop iterates di erent numbers of times.
Example: Insertion sort : After i iterations, the
rst i elements of an array are sorted. At iteration
i + 1 we move the (i + 1)st element forward until
it meets an element smaller than it.
(1)
(2)
(3)
(4)
(5)

void isort(int A ], int n) {


int i,j;
for(i=1; i<n; i++) {
j = i;
while(j>0 && A j-1]>A j]) {
swap(j-1, j); /* exchange A j-1] with A j] */
j--;
}
}
}

for
1{5
block
2{5
while
3{5

block
4{5
4

Input \size" = n = length of array A.


Lines 2, 4, 5 are O(1); note swap is short for
3 assignment statements.
Block 4{5 is O(1) + O(1) = O(1).
While-loop 3{5 iterates at most i times, for
j = i down to j = 1.
But it may terminate earlier if A j 1]
A j ] ever holds.
Thus, i O(1) = O(i) is an upper bound
on lines 3{5.
Block 2{5 takes time O(1) + O(i) = O(i).
For-loop 1{5 iterates n 1 times. The body
takes O(i) time.
But i changes within the loop and makes
no sense outside the loop, so we cannot
say the for-loop takes O(ni) time.
But n i, so O(n) is an upper bound on
the while-loop of 3{5.
Then, the upper bound on the for-loop is
n O(n) = O(n2 ).
Nothing lost. If we summed the times for
each
P 1iteration of the for-loop we would get
1)=2 = O(n2 ).
=1 O(i) = O n(n
n
i

You might also like