Implementing myprintf()

The goal of this assignment is to implement and use leaf functions, recursive functions, string to decimal conversions, decimal to string conversions, variable length argument lists, and C linkage in SPARC assembly code.

In this assignment, you will write a modified version of the printf() function in SPARC assembly code which can be called from a C program.

The ANSI C prototype for the myprintf() function is as follows:

    void myprintf(const char *format, ...);

The first argument is a required format string. It is followed by zero or more optional arguments. The number of arguments needed is determined by the contents of the format string. The output of the function is sent to the standard output (file descriptor 1) using the write service request and the trap instruction.

The format contains two kinds of items: plain characters and conversion specifications. Plain characters are simply copied to the standard output. Conversion specifications begin with a %, have an optional unsigned decimal integer, and end with a conversion operation character from the set: u, s, r, and %.

%u interprets the next argument in the parameter list as a 32-bit unsigned integer and prints its decimal representation. In this conversion specification there can be no optional integer between the % and the u.

%s interprets the next argument in the parameter list as the address (a pointer) of a null-terminated character string (type const char *) which is printed. If the address is 0 (i.e., NULL), the behavior is as if a pointer to the string "NULL" were supplied. There can be an optional unsigned decimal integer between the % and the s which is used as the maximum field width. For example, if the value is 20, then no more than 20 characters of the string will be printed (less if the string is less than 20 characters). If the optional integer is missing, it is as if the maximum possible value of 4294967295 (2³²-1) were specified. If the optional integer is larger than this maximum, it is as if the maximum were specified. (Implementation note: Build this integer digit by digit until either: (1) all the digits have been processed or (2) you have determined that the number is larger than the maximum value. In the later case you do not need to actually determine what the number is.)

%r causes a random number generated by a call to the C standard library function rand() to be printed using the %u conversion specification. This does not consume an argument from the argument list. There can be no optional integer between the % and the r.

%% causes a single percent sign to be printed. This does not consume an argument from the argument list. There can be no optional integer between the % characters.

Anything not listed above is not a conversion specification and is printed verbatim. See the example below.

The format argument is the address of the format string, and the format string is terminated by a null character (ascii code 0).

If a NULL pointer is passed as the first parameter, myprintf() should behave as if a pointer to the string "NULL\n" were given as the format string.

Example

Given the program:

extern void myprintf(char *, ...);

int main(void)
{
   myprintf("Simple Test\n");
   myprintf("Simple Test with ignored arguments\n", 1, 2, 3);
   myprintf(NULL);
   myprintf("%10s\n", "Hello, World");
   myprintf("%00000000000000002s\n", NULL);
   myprintf("%4294967297s\n", "Hello, Again");
   myprintf("%s%r%s%r%s%r\n", "rand()=", " rand()=", " rand()=");
   myprintf("0=%u  4294967295=%u\n", 0, 4294967295UL);
   myprintf("%u%% is perfect\n", 100);
   myprintf("%100xs\n", "Not Used");
   myprintf("%1u %1r %1%\n");
   myprintf("%s", "Go" "od");
   myprintf("%s\n", "bye");

   return 0;
}

The output is:

Simple Test
Simple Test with ignored arguments
NULL
Hello, Wor
NU
Hello, Again
rand()=16838 rand()=5758 rand()=10113
0=0  4294967295=4294967295
100% is perfect
%100xs
%1u %1r %1%
Goodbye

Ground Rules

• The only external functions you may use are .udiv and rand().

• The only service requests you may use are "write" and "exit". Only use the "exit" service request if the "write" service request fails. For purposes of this assignment, assume that the "write" service request fails if the "C" bit is set after the trap or the number of characters actually written did not match the number of characters requested to be written. Use the "SYMBOL=value" construct in your program to give appropriate symbolic names to EXIT, WRITE, and STDOUT

• Do not use the %rn names for the registers. Instead, use the %gn, %in, %ln, and %on names. You may not use the %Y register or the floating point registers.

• You are required to implement and use a recursive function which outputs a 32-bit unsigned integer in decimal.

• You are required to implement and use a leaf function which multiplies a 32-bit unsigned integer by 10.

• Begin your file with a comment block which includes your name. Precede each function with a block of comments describing the function. The comments will include: how the function is called (what goes in each register), what the function does, the nature of the algorithm, and what (if anything) is returned.

• Attempt to fill each delay slot with instructions other than nop. Note that this is not always possible.

• You must submit a single, plain SPARC assembly language file called myprintf.s which must assemble without warnings or errors using lcc -c -o myprintf.o myprintf.s. The name of the function must be myprintf().

• You may not use a high-level language compiler such as gcc or lcc to generate the assembly code you submit.

• You will lose points if you deviate from the assignment. Use the above list as a final checklist for your program.

Suggestion

My solution used eight helper functions in addition to the myprintf() function. I suggest you first implement and test this program using C. Write drivers which will test the helper functions. Now, one at a time, create an equivalent SPARC routine for each C function. Start with a simple leaf function (such as the one which multiplies by 10). Link it with your working program so that everything is in C except for this low-level function. Test it carefully. Continue by translating and testing each function until at last myprintf() is complete.

Grading

• 25% - Following the above ground rules
• 25% - Style
• 50% - Passing our test cases

Submission

Submit your file electronically with the command:

/u/cs217/bin/submit 6 myprintf.s

Due: submitted by 11:59pm, Monday, April 12, 1999.