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


#define STRTOL_OVERFLOW(n) (((n == LONG_MIN) || (n == LONG_MAX)) && \
			    (errno == ERANGE))

long seq_len;
long per_core_seq_len;
long nb_seqs;
int nb_cores = 8; /* Bossa */
//int nb_cores = 12; /* Quad Hexa */

/* General note on variable names: _ denotes a minus sign */

void fibo_node(long fibi_2, long fibi_1, long *fibk_1, long *fibk)
{
	long i, fibj_2, fibj_1, fibj = 0;

	fibj_2 = fibi_2;
	fibj_1 = fibi_1;
	for (i = 0; i < per_core_seq_len; i++)
	{
		fibj = fibj_2 + fibj_1;
		fibj_2 = fibj_1;
		fibj_1 = fibj;
	}
	*fibk_1 = fibj_1;
	*fibk = fibj;
}

int main(int argc, char *argv[])
{
	long i, fib1_1, fib1, fib2_1, fib2, fib3_1, fib3, fib4_1, fib4, fib5_1;
	long fib5, fib6_1, fib6, fib7_1, fib7, fib8_1, fib8;
	char **arg_p;

	argc--;
	arg_p = argv;
	while(arg_p++, argc--)
	{
		if ((*arg_p)[0] == '-')
		{
			if ((!(*arg_p)[1]) || (*arg_p)[2])
			{
				printf("Unsupported option: %s\n", *arg_p);
				return -1;
			}
			switch((*arg_p)[1])
			{
				char *endptr;

				case 'l':
					argc--, arg_p++;
					seq_len = strtol(*arg_p, &endptr, 10);
					if ((endptr == *arg_p) ||
					    (STRTOL_OVERFLOW(seq_len)))
					{
						fprintf(stderr, "Invalid length: "
								"%s\n", *arg_p);
						return -1;
					}
					if (seq_len % nb_cores)
					{
						fprintf(stderr, "The length of "
							"the sequence must be "
							"divisible by %d\n",
							nb_cores);
						return -1;
					}
					per_core_seq_len = seq_len / nb_cores;
					break;

				case 'n':
					argc--, arg_p++;
					nb_seqs = strtol(*arg_p, &endptr, 10);
					if ((endptr == *arg_p) ||
					    (STRTOL_OVERFLOW(nb_seqs)))
					{
						fprintf(stderr, "Invalid number"
								"of sequences: "
								"%s\n", *arg_p);
						return -1;
					}
					break;

				default:
					printf("Unsupported option: %s\n", *arg_p);
					return -1;
			}
		}
	}

	printf("Number of sequence: %ld, sequence lengths: %ld\n", nb_seqs, seq_len);

#pragma omp parallel num_threads (1) default (none) \
	shared (nb_seqs, seq_len, fib8_1, fib8) \
	private (i, fib1_1, fib1, fib2_1, fib2, fib3_1, fib3, fib4_1, fib4, fib5_1, fib5, fib6_1, fib6, fib7_1, fib7)
	//private (i, fib1_1, fib1, fib2_1, fib2, fib3_1, fib3, fib4_1, fib4, fib5_1, fib5, fib6_1, fib6, fib7_1, fib7, fib8_1, fib8, fib9_1, fib9, fib10_1, fib10, fib11_1, fib11)
	{
#pragma omp single
		{
			for (i = 0; i < nb_seqs; i ++)
			{

#pragma omp task output (fib1_1, fib1) shared (i)
				fibo_node(i, i + 1, &fib1_1, &fib1);
#pragma omp task input (fib1_1, fib1) output (fib2_1, fib2)
				fibo_node(fib1_1, fib1, &fib2_1, &fib2);


#pragma omp task input (fib2_1, fib2) output (fib3_1, fib3)
				fibo_node(fib2_1, fib2, &fib3_1, &fib3);
#pragma omp task input (fib3_1, fib3) output (fib4_1, fib4)
				fibo_node(fib3_1, fib3, &fib4_1, &fib4);


#pragma omp task input (fib4_1, fib4) output (fib5_1, fib5)
				fibo_node(fib4_1, fib4, &fib5_1, &fib5);
#pragma omp task input (fib5_1, fib5) output (fib6_1, fib6)
				fibo_node(fib5_1, fib5, &fib6_1, &fib6);


#pragma omp task input (fib6_1, fib6) output (fib7_1, fib7)
				fibo_node(fib6_1, fib6, &fib7_1, &fib7);
#pragma omp task input (fib7_1, fib7) shared (fib8_1, fib8)
				fibo_node(fib7_1, fib7, &fib8_1, &fib8);
#if 0
#pragma omp task input (fib8_1, fib8) output (fib9_1, fib9)
				fibo_node(fib8_1, fib8, &fib9_1, &fib9);
#pragma omp task input (fib9_1, fib9) output (fib10_1, fib10)
				fibo_node(fib9_1, fib9, &fib10_1, &fib10);


#pragma omp task input (fib10_1, fib10) output (fib11_1, fib11)
				fibo_node(fib10_1, fib10, &fib11_1, &fib11);
#pragma omp task input (fib11_1, fib11) output (fib12_1, fib12)
				fibo_node(fib11_1, fib11, &fib12_1, &fib12);
#endif
				if (!(i % (nb_seqs / 100)))
					printf("fib_0=%ld, fib_1=%ld => "
						"fib_%ld=%ld\n", i, i + 1,
						seq_len, fib8);
			}
		}
	}
	return 0;
}