/*
	Example program that shows how to create a multithreaded program in which each thread will treat a part of an array.
*/

#include <thread>
#include <vector>
#include <time.h>
#include <chrono>
#include <iostream> // cout, ...

using namespace std;
using namespace std::chrono;

const int MAX_NBR_ITERATIONS = 100;
#define N (1024*1024) 


_int64 elementWiseMultiplication(int n, float* arr, float* arr2, float* result);
_int64 elementWiseMultiplicationIterative(int n, int nbrIterations, float* arr, float* arr2, float* result);


int main()
{
	cout << "Considering arrays of size " << N << endl;

	// allocation and initialization
	srand(time(NULL)); // random initial seed
	float* arr = new float[N], *arr2 = new float[N], *result_seq = new float[N], *result_par = new float[N], *arr4 = new float[N], *arr5 = new float[N], *arr6 = new float[N];
	for (int i = 0; i < N; i++) {
		arr[i] = rand();
		arr2[i] = rand();
		result_seq[i] = 0; // actually, initialization is not necessary
		result_par[i] = 0;
		arr4[i] = rand();
		arr5[i] = rand();
		arr6[i] = 0;
	}

	// flush caches
	_int64 run_time_seq = elementWiseMultiplication(N, arr4, arr5, arr6);
	cout << "Runtime for basic version: " << run_time_seq << "us " << (run_time_seq * 1000 / N) << "ns/op" << endl;

	// warm up

	for (int nbr_iterations = 10; nbr_iterations <= MAX_NBR_ITERATIONS; nbr_iterations+=10) {

		_int64 run_time = elementWiseMultiplicationIterative(N, nbr_iterations, arr, arr2, result_par);
		cout << "Runtime for " << nbr_iterations << " iterations: " << run_time << "us " << (run_time * 1000 / (float) N / nbr_iterations) << "ns/op - Slowdown = "<< ((double)run_time/run_time_seq) <<endl;
	}
	

	// clean up memory
	delete arr, arr2, result_seq, result_par, arr4, arr5, arr6;

	cout << endl << "Press ENTER to close window...";
	char c = cin.get();
	return 0;
}



//  ==== ALL VERSIONS ====
_int64 elementWiseMultiplication(int n, float* arr, float* arr2, float* result) {
	time_point<system_clock> now = system_clock::now();

	for (int i = 0; i < n; ++i) {
		result[i] = arr[i] * arr2[i];
	}

	time_point<system_clock> epoch = system_clock::now();
	microseconds us = duration_cast<microseconds>(epoch - now);

	return us.count();
}
_int64 elementWiseMultiplicationIterative(int n, int nbrIterations, float* arr, float* arr2, float* result) {
	time_point<system_clock> now = system_clock::now();

	for (int i = 0; i < n; ++i) {
		float x = arr[i];
		for (int it = 0;it < nbrIterations;it++) {
			x = x * arr2[i];
		}
		result[i] = x;
	}

	time_point<system_clock> epoch = system_clock::now();
	microseconds us = duration_cast<microseconds>(epoch - now);

	return us.count();
}


void checkIfResultsAreTheSame(string name, float* arr3, float* arr4, int n, bool PRINT_DATA) {
	int nbr_diff = 0;
	for (int i = 0; i < n; i++) {
		float rel_diff = arr3[i] == arr4[i] ? 0 : (arr3[i] - arr4[i]) / (arr3[i] == 0 ? arr4[i] : arr3[i]);
		if (rel_diff > 0.0001) {
			nbr_diff++;
			if (PRINT_DATA  && i < 10) cout << arr3[i] << "<>" << arr4[i] << " (" << rel_diff << "), ";
		}
		else {
			if (PRINT_DATA  && i < 10) cout << arr3[i] << "==" << arr4[i] << " (" << rel_diff << "), ";
		}
		arr4[i] = 0; // we reset arr4 for further usage
	}
	if (PRINT_DATA) cout << endl;
	if (nbr_diff > 0)
		cout << "Problem with algorithm : " << nbr_diff << " differences with Original Algorithm." << endl;
} // " << name << "