Performing matrix operations and matrix operands.

Create a calculator for performing matrix operations that exploits the (expected)
sparseness of its matrix operands. An 𝑛 × 𝑛 square matrix is said to be sparse if the number of non-zero
entries (here abbreviated NNZ) is small compared to the total number of entries 𝑛
2
. The result will be a C
program capable of performing fast matrix operations, even on very large matrices, provided they are
sparse.
Given 𝑛 × 𝑛 matrices A and B, their product 𝐶 = 𝐴 ⋅ 𝐵 is the 𝑛 × 𝑛 matrix whose 𝑖𝑗th
entry is given by
𝐶𝑖𝑗 = ∑ 𝐴𝑖𝑘
𝑛
𝑘=1 𝐵𝑘𝑗.
Thus the element in the i
th row and j
th column of C is the vector dot product of the i
th row of A with the j
th
column of B. If we consider addition and multiplication of real numbers to be our basic operations, then
the above formula can be computed in time Θ(𝑛
3
), which is impractical for matrix sizes n of more than a
few thousand. If it so happens that A and B are sparse, then a great many of these arithmetic operations
involve adding or multiplying by zero, hence are unnecessary.
The sum S, and difference D, of A and B are the 𝑛 × 𝑛 matrices having 𝑖𝑗th
entries:
𝑆𝑖𝑗 = 𝐴𝑖𝑗 + 𝐵𝑖𝑗 and 𝐷𝑖𝑗 = 𝐴𝑖𝑗 − 𝐵𝑖𝑗
The scalar product of a real number x with A is denoted 𝑥𝐴, and has 𝑖𝑗th
entry (𝑥𝐴)𝑖𝑗 = 𝑥 ⋅ 𝐴𝑖𝑗. The
transpose of A, denoted 𝐴
𝑇
, is the matrix whose 𝑖𝑗
𝑡ℎ
entry is the 𝑗𝑖
𝑡ℎ
entry of A: (𝐴
𝑇
)𝑖𝑗 = 𝐴𝑗𝑖. In other
words, the rows of A are the columns of 𝐴
𝑇
, and the columns of A are the rows of 𝐴
𝑇
. Each of these
operations can be computed in time Θ(𝑛
2
), and just as for multiplication, their cost can be improved upon
significantly when A and B are sparse.
As one would expect, the cost of a matrix operation depends heavily on the choice of data structure used
to represent the matrix operands. There are several ways to represent a square matrix with real entries. The
standard approach is to use a 2-dimensional 𝑛 × 𝑛 array of doubles. The advantage of this representation
is that all of the above matrix operations have a straight-forward implementation using nested loops. This
project will use a very different representation however. Here you will represent a matrix as a 1-dimensional
array of Lists. Each List will represent one row of the Matrix, but only the non-zero entries will be stored.
Therefore List elements must store, not just the matrix entries, but the column index in which those entries
reside. For example, the matrix below would have the following representation as an array of Lists.
𝑀 = [
1.0 0.0 2.0
3.0 0.0 0.0
0.0 4.0 5.0
] Array of Lists: [
1: (1, 1.0) (3, 2.0)
2: (1, 3.0)
3: (2, 4.0) (3, 5.0)
This method obviously results in a substantial space savings when the Matrix is sparse. In addition, the
standard matrix operations defined above can be performed more efficiently on sparse matrices. As you
will see though, the matrix operations are much more difficult to implement using this representation. The
trade-off then, is a gain in space and time efficiency for sparse matrices, at the expense of more complicated
algorithms for performing standard matrix operations. Designing these algorithms in terms of our List ADT
operations will constitute the majority of the work you do on this assignment.
It will be necessary to make some minor changes to your List ADT from pa1. First you must convert your
ADT from a List of ints to a List of generic pointers. This entails changing certain field types, declaration
statements, method parameters, and return types from int to void*. The objects referred to by these List
elements will be defined in the Matrix ADT specified below. Second, it will be necessary to eliminate the
List operations equals() and copy(). The problem with these functions is that the List no longer knows what
it is a list of, and therefore can only perform "shallow" versions of these operations, i.e. compare or copy
pointers, not what they point to. For the same reason, function printList() is no longer required (but you
may wish to include it for diagnostic purposes only). The (optional) function cat() may also be included.
All other List operations from pa1 will be retained.
A "typical loop" in a client of List (as previously illustrated on page 3 of the pa1 project description) might
now appear as
type x; // Here "type" is whatever data type the void
// pointers in List are pointing to, known by
// the client but not by List.
moveFront(L);
while( index(L)>=0 ){
// Get current element, cast as the right kind of
// pointer, follow the pointer, assign its value to x.
x = *(type)get(L); // do something with x, then moveNext(L); } Observe that the equals(), copy() and printList() operations can still be performed, but they must now be done from the client level. Be sure to test this modified List ADT before you try to implement the Matrix ADT defined below. A (very weak) test of the modified List ADT called ListClient.c will be posted under the examples section of the class webpage. File Formats The top level client module for this project will be called Sparse.c. It will take two command line arguments giving the names of the input and output files, respectively. The input file will begin with a single line containing three integers n, a and b, separated by spaces. The second line will be blank, and the following a lines will specify the non-zero entries of an 𝑛 × 𝑛 matrix A. Each of these lines will contain a space separated list of three numbers: two integers and a double, giving the row, column, and value of the corresponding matrix entry. After another blank line, will follow b lines specifying the non-zero entries of an 𝑛 × 𝑛 matrix B. For example, the two matrices 𝐴 = [ 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 ] and 𝐵 = [ 1.0 0.0 1.0 0.0 0.0 0.0 1.0 1.0 1.0 ] are encoded by the following input file: 3 9 5 1 1 1.0 1 2 2.0 1 3 3.0 2 1 4.0 2 2 5.0 2 3 6.0 3 1 7.0 3 2 8.0 3 3 9.0 1 1 1.0 1 3 1.0 3 1 1.0 3 2 1.0 3 3 1.0 Your program will read an input file as above, initialize and build the Array of Lists representation of the matrices A and B, then calculate and print the following matrices to the output file: 𝐴, 𝐵, (1.5)𝐴, 𝐴 + 𝐵, 𝐴 + 𝐴, 𝐵 − 𝐴, 𝐴 − 𝐴, 𝐴 𝑇 , 𝐴𝐵 and 𝐵 2 . The output file format is illustrated by the following example, which corresponds to the above input file. A has 9 non-zero entries: 1: (1, 1.0) (2, 2.0) (3, 3.0) 2: (1, 4.0) (2, 5.0) (3, 6.0) 3: (1, 7.0) (2, 8.0) (3, 9.0) B has 5 non-zero entries: 1: (1, 1.0) (3, 1.0) 3: (1, 1.0) (2, 1.0) (3, 1.0) (1.5)A =
1: (1, 1.5) (2, 3.0) (3, 4.5)
2: (1, 6.0) (2, 7.5) (3, 9.0)
3: (1, 10.5) (2, 12.0) (3, 13.5)
A+B =
1: (1, 2.0) (2, 2.0) (3, 4.0)
2: (1, 4.0) (2, 5.0) (3, 6.0)
3: (1, 8.0) (2, 9.0) (3, 10.0)
A+A =
1: (1, 2.0) (2, 4.0) (3, 6.0)
2: (1, 8.0) (2, 10.0) (3, 12.0)
3: (1, 14.0) (2, 16.0) (3, 18.0)
B-A =
1: (2, -2.0) (3, -2.0)
2: (1, -4.0) (2, -5.0) (3, -6.0)
3: (1, -6.0) (2, -7.0) (3, -8.0)
A-A =
Transpose(A) =
1: (1, 1.0) (2, 4.0) (3, 7.0)
2: (1, 2.0) (2, 5.0) (3, 8.0)
3: (1, 3.0) (2, 6.0) (3, 9.0)
AB = 1: (1, 4.0) (2, 3.0) (3, 4.0) 2: (1, 10.0) (2, 6.0) (3, 10.0) 3: (1, 16.0) (2, 9.0) (3, 16.0) BB =
1: (1, 2.0) (2, 1.0) (3, 2.0)
3: (1, 2.0) (2, 1.0) (3, 2.0)
Notice that the rows are to be printed in column sorted order, and zero rows are skipped altogether. A zero
matrix will cause no output to be printed, as seen by the matrix 𝐴 − 𝐴 above. Note that, unlike the output
file, the input file may give matrix entries in any order.
Matrix ADT Specifications
In addition to the main program Sparse.c and the altered List.c from pa1, you will implement a Matrix ADT
in a file called Matrix.c. This ADT will contain a private inner struct (similar to Node in your List ADT)
encapsulating the column and value information corresponding to a matrix entry. You can give this inner
struct any name you wish, but I will refer to it here as Entry. Thus Entry is a pointer to a struct called
EntryObj that has two fields storing types int and double respectively. Your Matrix ADT will represent a
matrix as an array of Lists of Entries. Since Entry is itself a pointer, it can be assigned to a field of type
void*. It is required that each List of Entries be maintained in column sorted order. Your Matrix ADT
will export the following operations.