#include <iostream>
#include "string.h"
#include "asterics_hpc.h"
#include "sgemm_intrinsics_pitch.h"

using namespace std;

///Get the number of cycles per elements of the reduction
/**	@param nbElement : number of elements of the tables
 * 	@param nbRepetition : number of repetition to evaluate the function reduction
*/
void evaluateSgemm(long unsigned int nbElement, long unsigned int nbRepetition){
	float * matX = asterics_malloc2f(nbElement, nbElement);
	float * matY = asterics_malloc2f(nbElement, nbElement);
	float * matOut = asterics_malloc2f(nbElement, nbElement);
	
	long unsigned int pitch(getPitch(nbElement));
	long unsigned int sizeCol(nbElement + pitch);
	
	for(long unsigned int i(0lu); i < nbElement; ++i){
		for(long unsigned int j(0lu); j < nbElement; ++j){
			long unsigned int index(i*nbElement + j);
			matX[i*sizeCol + j] = (float)(index*32lu%17lu);
			matY[i*sizeCol + j] = (float)(index*77lu%7lu);
		}
	}
	long unsigned int beginTime(rdtsc());
	for(long unsigned int i(0lu); i < nbRepetition; ++i){
		sgemm(matOut, matX, matY, nbElement, pitch);
	}
	long unsigned int elapsedTime((double)(rdtsc() - beginTime)/((double)nbRepetition));
	
	double cyclePerElement(((double)elapsedTime)/((double)(nbElement*nbElement)));
	cout << "evaluateSgemm : nbElement = "<<nbElement<<", cyclePerElement = " << cyclePerElement << " cy/el, elapsedTime = " << elapsedTime << " cy" << endl;
	cerr << nbElement << "\t" << cyclePerElement << "\t" << elapsedTime << endl;
	
	asterics_free(matOut);
	asterics_free(matY);
	asterics_free(matX);
}

int main(int argc, char** argv){
	cout << "SGEMM Intrinsics Pitch" << endl;
	evaluateSgemm(10lu, 100000lu);
	evaluateSgemm(16lu, 100000lu);
	evaluateSgemm(24lu, 100000lu);
	evaluateSgemm(32lu, 100000lu);
	evaluateSgemm(40lu, 100000lu);
	evaluateSgemm(56lu, 10000lu);
	evaluateSgemm(80lu, 10000lu);
	evaluateSgemm(90lu, 10000lu);
	evaluateSgemm(104lu, 10000lu);
	return 0;
}