/*
 * Josh Holtrop
 * 2008-10-01
 * Grand Valley State University
 * CS677
 * Programming Assignment #2
 */

#include <iostream>
#include <fstream>
#include <vector>
#include <pthread.h>
using namespace std;

#define eq(x, y) ( ( (x) == (y) ) || ( (x) == '?' ) || ( (y) == '?' ) )

void usage(char * prog);
void * calcSimMatrixThread(void * arg);
bool readFile(char * fileName, vector<char> & v);

pthread_barrier_t barrier;
int num_threads;
int * matrix;
vector<char> * s;
vector<char> * t;

/* Print basic usage information */
void usage(char * prog)
{
    cout << "Usage: " << prog << " [-n <num_threads>] <file1> <file2>" << endl;
    exit(42);
}

/*
 * taskAllocate() will divide a set of total_tasks tasks into
 * total_workers groups, as evenly as possible
 * Parameters:
 *   total_tasks   : IN  : the total number of tasks to divide up
 *   total_workers : IN  : the total number of workers to allocate tasks to (>0)
 *   this_id       : IN  : the id (base 0) of the task calling us for work
 *   first_task_id : OUT : the id (base 0) of the first task for this worker
 *   num           : OUT : the number of tasks assigned to this worker
 */
inline void taskAllocate(int total_tasks, int total_workers, int this_id,
                         int * first_task_id, int * num)
{
    int l_num;
    int leftovers = total_tasks % total_workers; /* num of "leftover" tasks */
    if (this_id < leftovers)
    {
        l_num = total_tasks / total_workers + 1; /* do one of the leftovers */
        *first_task_id = l_num * this_id;
    }
    else
    {
        l_num = total_tasks / total_workers;
        *first_task_id = l_num * this_id + leftovers;
    }
    *num = l_num;
}

int main(int argc, char * argv[])
{
    vector<char> files[2];
    num_threads = 1;
    int file_to_read = 0;

    for (int i = 1; i < argc; i++)
    {
        if ( ! strcmp("-n", argv[i]) )
        {
            if (i == argc - 1)
                usage(argv[0]);
            i++;
            num_threads = atoi(argv[i]);
        }
        else
        {
            if (file_to_read < 2)
                readFile(argv[i], files[file_to_read]);
            else
                usage(argv[0]);
            file_to_read++;
        }
    }

    if (file_to_read != 2)
        usage(argv[0]);

    s = &files[0];
    t = &files[1];
    pthread_t * threads = new pthread_t[num_threads];
    matrix = new int[(files[0].size() + 1) * (files[1].size() + 1)];
    pthread_barrier_init(&barrier, NULL, num_threads);

    for (int i = 0; i < num_threads; i++)
    {
        int * arg = new int;
        *arg = i;
        int ret = pthread_create(&threads[i], NULL, &calcSimMatrixThread, arg);
        if (ret)
        {
            cerr << "Error " << ret << " when creating thread!" << endl;
            return -4;
        }
    }

    for (int i = 0; i < num_threads; i++)
    {
        pthread_join(threads[i], NULL);
    }

    pthread_barrier_destroy(&barrier);

    delete[] matrix;
    delete[] threads;
    return 0;
}

/* Read a file into a vector of non-space characters */
bool readFile(char * fileName, vector<char> & v)
{
    ifstream in(fileName);
    if (!in.is_open())
        return false;
    for(;;)
    {
        char chr;
        in >> chr;
        if (in.eof())
            break;
        v.push_back(chr);
    }
    return true;
}

/* Compute portions of the similarity matrix between two character arrays */
void * calcSimMatrixThread(void * arg)
{
    int * realarg = (int *) arg;
    int id = *realarg;
    int s_size = s->size();
    int t_size = t->size();
    int (*F)[s_size][t_size] = (int (*) [s_size][t_size]) matrix;
    int max_x = 0, max_y = 0, max_val = 0;
    int first_task_id, num_tasks;
    taskAllocate(t_size+1, num_threads, id, &first_task_id, &num_tasks);
    for (int i = 0, idx = first_task_id; i < num_tasks; i++, idx++)
        (*F)[0][idx] = 0;               /* set first row to 0's */
    taskAllocate(s_size+1, num_threads, id, &first_task_id, &num_tasks);
    for (int i = 0, idx = first_task_id; i < num_tasks; i++, idx++)
        (*F)[idx][0] = 0;               /* set first column to 0's */
    pthread_barrier_wait(&barrier);
    for (int i = 1; i <= s_size; i++)
    {
        for (int j = 1; j <= t_size; j++)
        {
            /* Compute the value for the matrix */
            (*F)[i][j] =
                max(
                    max(
                        (*F)[i][j-1] - 2,
                        (*F)[i-1][j-1] +
                            (eq(s->at(i-1), t->at(j-1)) ? 1 : -1)
                    ),
                    max(
                        (*F)[i-1][j] - 2,
                        0
                    )
                );
            /* See if we found a new maximum value */
            if ((*F)[i][j] > max_val)
            {
                max_val = (*F)[i][j];
                max_x = i;
                max_y = j;
            }
            else if ((*F)[i][j] == max_val)
            {
                /* If we found a value the same as our current maximum
                 * value, see if it has a greater i+j value */
                if ( (i + j) > (max_x + max_y) )
                {
                    max_x = i;
                    max_y = j;
                }
            }
        }
    }

    cout << "Maximum value is " << max_val << " at position ("
        << max_x << ", " << max_y << ")" << endl;

    delete realarg;
    return NULL;
}