workloads/realworld/async/lavaMD/main.c

//========================================================================================================================================================================================================200
//======================================================================================================================================================150
//====================================================================================================100
//==================================================50

//========================================================================================================================================================================================================200
//	UPDATE
//========================================================================================================================================================================================================200

//	14 APR 2011 Lukasz G. Szafaryn

//========================================================================================================================================================================================================200
//	DEFINE/INCLUDE
//========================================================================================================================================================================================================200

//======================================================================================================================================================150
//	LIBRARIES
//======================================================================================================================================================150

#include <stdio.h>					// (in path known to compiler)			needed by printf
#include <stdlib.h>					// (in path known to compiler)			needed by malloc
#include <stdbool.h>				// (in path known to compiler)			needed by true/false

//======================================================================================================================================================150
//	UTILITIES
//======================================================================================================================================================150

#include "./util/timer/timer.h"			// (in path specified here)
#include "./util/num/num.h"				// (in path specified here)

//======================================================================================================================================================150
//	MAIN FUNCTION HEADER
//======================================================================================================================================================150

#include "./main.h"						// (in the current directory)

//======================================================================================================================================================150
//	KERNEL
//======================================================================================================================================================150

#include "./kernel/kernel_gpu_cuda_wrapper.h"	// (in library path specified here)

//========================================================================================================================================================================================================200
//	MAIN FUNCTION
//========================================================================================================================================================================================================200
#define _POSIX_C_SOURCE 200809L
#include <stdio.h>
#include <time.h>
#include <stdint.h>
extern inline __attribute__((always_inline)) unsigned long rdtsc()
{
	   unsigned long a, d;

	      __asm__ volatile("rdtsc" : "=a" (a), "=d" (d));

	         return (a | (d << 32));
}

extern inline __attribute__((always_inline)) unsigned long rdtsp() {
		struct timespec tms;
		    if (clock_gettime(CLOCK_REALTIME, &tms)) {
			            return -1;
				        }
		        unsigned long ns = tms.tv_sec * 1000000000;
			    ns += tms.tv_nsec;
			        return ns;
}


int 
main(	int argc, 
		char *argv [])
{
	uint64_t start_tsc = rdtsc();
	uint64_t start_tsp = rdtsp();
	printf("start_tsc %llu start_tsp %llu\n", start_tsc, start_tsp);
	
	printf("thread block size of kernel = %d \n", NUMBER_THREADS);
	//======================================================================================================================================================150
	//	CPU/MCPU VARIABLES
	//======================================================================================================================================================150

	// timer
	long long time0;

	time0 = get_time();

	// timer
	long long time1;
	long long time2;
	long long time3;
	long long time4;
	long long time5;
	long long time6;
	long long time7;

	// counters
	int i, j, k, l, m, n;

	// system memory
	par_str par_cpu;
	dim_str dim_cpu;
	box_str* box_cpu;
	FOUR_VECTOR* rv_cpu;
	fp* qv_cpu;
	FOUR_VECTOR* fv_cpu;
	int nh;

	time1 = get_time();

	//======================================================================================================================================================150
	//	CHECK INPUT ARGUMENTS
	//======================================================================================================================================================150

	// assing default values
	dim_cpu.boxes1d_arg = 1;

	// go through arguments
	dim_cpu.boxes1d_arg = atoi(argv[1]);
	int nblocks = atoi(argv[2]);

	// Print configuration
	printf("Configuration used: boxes1d = %d\n", dim_cpu.boxes1d_arg);

	time2 = get_time();

	//======================================================================================================================================================150
	//	INPUTS
	//======================================================================================================================================================150

	par_cpu.alpha = 0.5;

	time3 = get_time();

	//======================================================================================================================================================150
	//	DIMENSIONS
	//======================================================================================================================================================150

	// total number of boxes
	dim_cpu.number_boxes = dim_cpu.boxes1d_arg * dim_cpu.boxes1d_arg * dim_cpu.boxes1d_arg;

	// how many particles space has in each direction
	dim_cpu.space_elem = dim_cpu.number_boxes * NUMBER_PAR_PER_BOX;
	dim_cpu.space_mem = dim_cpu.space_elem * sizeof(FOUR_VECTOR);
	dim_cpu.space_mem2 = dim_cpu.space_elem * sizeof(fp);

	// box array
	dim_cpu.box_mem = dim_cpu.number_boxes * sizeof(box_str);

	time4 = get_time();

	//======================================================================================================================================================150
	//	SYSTEM MEMORY
	//======================================================================================================================================================150

	//====================================================================================================100
	//	BOX
	//====================================================================================================100

	// allocate boxes
	box_cpu = (box_str*)malloc(dim_cpu.box_mem);

	// initialize number of home boxes
	nh = 0;

	// home boxes in z direction
	for(i=0; i<dim_cpu.boxes1d_arg; i++){
		// home boxes in y direction
		for(j=0; j<dim_cpu.boxes1d_arg; j++){
			// home boxes in x direction
			for(k=0; k<dim_cpu.boxes1d_arg; k++){

				// current home box
				box_cpu[nh].x = k;
				box_cpu[nh].y = j;
				box_cpu[nh].z = i;
				box_cpu[nh].number = nh;
				box_cpu[nh].offset = nh * NUMBER_PAR_PER_BOX;

				// initialize number of neighbor boxes
				box_cpu[nh].nn = 0;

				// neighbor boxes in z direction
				for(l=-1; l<2; l++){
					// neighbor boxes in y direction
					for(m=-1; m<2; m++){
						// neighbor boxes in x direction
						for(n=-1; n<2; n++){

							// check if (this neighbor exists) and (it is not the same as home box)
							if(		(((i+l)>=0 && (j+m)>=0 && (k+n)>=0)==true && ((i+l)<dim_cpu.boxes1d_arg && (j+m)<dim_cpu.boxes1d_arg && (k+n)<dim_cpu.boxes1d_arg)==true)	&&
									(l==0 && m==0 && n==0)==false	){

								// current neighbor box
								box_cpu[nh].nei[box_cpu[nh].nn].x = (k+n);
								box_cpu[nh].nei[box_cpu[nh].nn].y = (j+m);
								box_cpu[nh].nei[box_cpu[nh].nn].z = (i+l);
								box_cpu[nh].nei[box_cpu[nh].nn].number =	(box_cpu[nh].nei[box_cpu[nh].nn].z * dim_cpu.boxes1d_arg * dim_cpu.boxes1d_arg) + 
																			(box_cpu[nh].nei[box_cpu[nh].nn].y * dim_cpu.boxes1d_arg) + 
																			 box_cpu[nh].nei[box_cpu[nh].nn].x;
								box_cpu[nh].nei[box_cpu[nh].nn].offset = box_cpu[nh].nei[box_cpu[nh].nn].number * NUMBER_PAR_PER_BOX;

								// increment neighbor box
								box_cpu[nh].nn = box_cpu[nh].nn + 1;

							}

						} // neighbor boxes in x direction
					} // neighbor boxes in y direction
				} // neighbor boxes in z direction

				// increment home box
				nh = nh + 1;

			} // home boxes in x direction
		} // home boxes in y direction
	} // home boxes in z direction

	//====================================================================================================100
	//	PARAMETERS, DISTANCE, CHARGE AND FORCE
	//====================================================================================================100

	// random generator seed set to random value - time in this case
	srand(time(NULL));

	// input (distances)
	rv_cpu = (FOUR_VECTOR*)malloc(dim_cpu.space_mem);
	for(i=0; i<dim_cpu.space_elem; i=i+1){
		rv_cpu[i].v = (rand()%10 + 1) / 10.0;			// get a number in the range 0.1 - 1.0
		rv_cpu[i].x = (rand()%10 + 1) / 10.0;			// get a number in the range 0.1 - 1.0
		rv_cpu[i].y = (rand()%10 + 1) / 10.0;			// get a number in the range 0.1 - 1.0
		rv_cpu[i].z = (rand()%10 + 1) / 10.0;			// get a number in the range 0.1 - 1.0
	}

	// input (charge)
	qv_cpu = (fp*)malloc(dim_cpu.space_mem2);
	for(i=0; i<dim_cpu.space_elem; i=i+1){
		qv_cpu[i] = (rand()%10 + 1) / 10.0;			// get a number in the range 0.1 - 1.0
	}

	// output (forces)
	fv_cpu = (FOUR_VECTOR*)malloc(dim_cpu.space_mem);
	for(i=0; i<dim_cpu.space_elem; i=i+1){
		fv_cpu[i].v = 0;								// set to 0, because kernels keeps adding to initial value
		fv_cpu[i].x = 0;								// set to 0, because kernels keeps adding to initial value
		fv_cpu[i].y = 0;								// set to 0, because kernels keeps adding to initial value
		fv_cpu[i].z = 0;								// set to 0, because kernels keeps adding to initial value
	}

	time5 = get_time();

	//======================================================================================================================================================150
	//	KERNEL
	//======================================================================================================================================================150

	//====================================================================================================100
	//	GPU_CUDA
	//====================================================================================================100

	kernel_gpu_cuda_wrapper(par_cpu,
							dim_cpu,
							box_cpu,
							rv_cpu,
							qv_cpu,
							fv_cpu,
							nblocks);

	time6 = get_time();

	//======================================================================================================================================================150
	//	SYSTEM MEMORY DEALLOCATION
	//======================================================================================================================================================150

	// dump results
#ifdef OUTPUT
        FILE *fptr;
	fptr = fopen("result.txt", "w");	
	for(i=0; i<dim_cpu.space_elem; i=i+1){
        	fprintf(fptr, "%f, %f, %f, %f\n", fv_cpu[i].v, fv_cpu[i].x, fv_cpu[i].y, fv_cpu[i].z);
	}
	fclose(fptr);
#endif       	



	free(rv_cpu);
	free(qv_cpu);
	free(fv_cpu);
	free(box_cpu);

	time7 = get_time();

	//======================================================================================================================================================150
	//	DISPLAY TIMING
	//======================================================================================================================================================150

	// printf("Time spent in different stages of the application:\n");

	// printf("%15.12f s, %15.12f % : VARIABLES\n",						(float) (time1-time0) / 1000000, (float) (time1-time0) / (float) (time7-time0) * 100);
	// printf("%15.12f s, %15.12f % : INPUT ARGUMENTS\n", 					(float) (time2-time1) / 1000000, (float) (time2-time1) / (float) (time7-time0) * 100);
	// printf("%15.12f s, %15.12f % : INPUTS\n",							(float) (time3-time2) / 1000000, (float) (time3-time2) / (float) (time7-time0) * 100);
	// printf("%15.12f s, %15.12f % : dim_cpu\n", 							(float) (time4-time3) / 1000000, (float) (time4-time3) / (float) (time7-time0) * 100);
	// printf("%15.12f s, %15.12f % : SYS MEM: ALO\n",						(float) (time5-time4) / 1000000, (float) (time5-time4) / (float) (time7-time0) * 100);

	// printf("%15.12f s, %15.12f % : KERNEL: COMPUTE\n",					(float) (time6-time5) / 1000000, (float) (time6-time5) / (float) (time7-time0) * 100);

	// printf("%15.12f s, %15.12f % : SYS MEM: FRE\n", 					(float) (time7-time6) / 1000000, (float) (time7-time6) / (float) (time7-time0) * 100);

	// printf("Total time:\n");
	// printf("%.12f s\n", 												(float) (time7-time0) / 1000000);

	//======================================================================================================================================================150
	//	RETURN
	//======================================================================================================================================================150

	return 0.0;																					// always returns 0.0

}