Optimization (ITI8510)

Real-Time Operating Systems and Systems Programming

Basic Linux command-line and utilities for accessing it Standard Streams Stack Exploration Exploring Debugger Signals Threads Processes File Attributes Reading Binary Files LAB: Hufmann Encoder Network Sockets Optimization (mmap() helps here)

[ @ ]

Optimization of a program (Also available in Estonian)

Programm

Write a program which will read a file and outputs its statistics.

• For a "standard" input file with random content, please run:

dd if=/dev/urandom of=random_file count=1048576 bs=1

• The output of the program should be 256 human-readable lines of ASCII numbers, which represent the content of the file. The first line should contain the number of occurences of byte with a value of 0x0 (or binary zero) in the file and the last line the number of occurences of byte with value 0xff (or 256 in decimal).
• Measure the running time of the program with:

time programname

Hints

• Does the loop unrolling trick work?
• Perhaps it is possible to use parallel programming or threads?
• Does compilation optimization with -O work?
• Can you replace printf with something faster that you write yourself?
• Perhaps the keyword inline can be used.
• (Sadly mmap for file reading fails in dijkstra) Can you speed up the reading of the file with mmap()? (Note -- mmap not yet explained in the lecture)

Example code for base program

#include <stdlib.h>
#include <stdio.h>

long int stats[256];

int main() {
        int i, c;
       FILE * f; 

        f = fopen("random_file", "r");
        if(f == NULL) {
                perror("slowstat");
                exit(EXIT_FAILURE);
       } 

        while((c = fgetc(f)) != EOF) {
                stats[c]++;
       } 

        for(i = 0; i < 255; i++) {
                printf("%ld\n", stats[i]);
        }
        exit(EXIT_SUCCESS);
}