CS 2734 Computer Organization II

Laboratory 12 [4/21; 4/23; 4/28] Sparc Floating Point Operations

This laboratory is designed to help you review Sparc assembly language and to introduce you to the floating point operations.

Prior to the laboratory, you should do the following:

Make a new directory for this laboratory.

Write the Sparc assembly language function vectornorm:

       double vectornorm(double a[], int n)
       {
           int i;
           double sum;
  
           sum = 0.0;
           for (i = 0; i < n; i++) 
               sum = sum + a[i]*a[i];
           sum = sqrt(sum);
           return sum;
       }

Write a C main program to call vectornorm. This program should take two command line arguments:
- The name of the input file for the values for the array a.
- The number of elements in the array a.
Check for the appropriate number of command line arguments and print out a usage message if incorrect.
Allocate space for the array a using malloc. Be sure to trap on error and exit if malloc fails.
Write a Makefile that will compile these source files, first into separate object files and then into an executable mycalnorm. Use the -g compilation option.

For Checkoff:

You will be asked to run your program on a test file.
You will be asked to single-step through your function and display the contents of the floating point registers.
Hand in prtext copies of your source files.
For checkoff you will asked to run the script in
```
    /home/sdykes/pub/c12
```
You enter the executable filename either as an argument or through an interactive prompt.

Implementation Notes:

One of the test cases will have n = 0. You still have to set f0 and f1 in this case or the function will return NaN.
Floating pt registers are all initialized to -NaN to test for uninitialized usage. Consequently, if you try to zero an uninitialized register by subtraction, as in
```
	fsubd	%f0, %f0, %f0
```
the result is %f0 = -NaN.
Both the book and the compiler initialize a memory location to 0 and load from it (with ldd). This means you have to be careful of alignment for text and data segments, e.g.
```
        .data
        .align  8
zero:
        .word   0x0                        ! (double) 0.0
        .word   0x0
```
(You can also use the .double directive to declare initialized static variables containing doubles.
You can't move data between integer and floating point registers.

Revision Date: 4/18/97