int
main( int argc,
char *argv [])
{
//======================================================================================================================================================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.arch_arg = 0;
dim_cpu.cores_arg = 1;
dim_cpu.boxes1d_arg = 1;
// go through arguments
if(argc==3){
for(dim_cpu.cur_arg=1; dim_cpu.cur_arg<argc; dim_cpu.cur_arg++){
// check if -boxes1d
if(strcmp(argv[dim_cpu.cur_arg], "-boxes1d")==0){
// check if value provided
if(argc>=dim_cpu.cur_arg+1){
// check if value is a number
if(isInteger(argv[dim_cpu.cur_arg+1])==1){
dim_cpu.boxes1d_arg = atoi(argv[dim_cpu.cur_arg+1]);
if(dim_cpu.boxes1d_arg<0){
printf("ERROR: Wrong value to -boxes1d argument, cannot be <=0\n");
return 0;
}
dim_cpu.cur_arg = dim_cpu.cur_arg+1;
}
// value is not a number
else{
printf("ERROR: Value to -boxes1d argument in not a number\n");
return 0;
}
}
// value not provided
else{
printf("ERROR: Missing value to -boxes1d argument\n");
return 0;
}
}
// unknown
else{
printf("ERROR: Unknown argument\n");
return 0;
}
}
// Print configuration
printf("Configuration used: arch = %d, cores = %d, boxes1d = %d\n", dim_cpu.arch_arg, dim_cpu.cores_arg, dim_cpu.boxes1d_arg);
}
else{
printf("Provide boxes1d argument, example: -boxes1d 16");
return 0;
}
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; // 8*8*8=512
// how many particles space has in each direction
dim_cpu.space_elem = dim_cpu.number_boxes * NUMBER_PAR_PER_BOX; //512*100=51,200
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].v = 0.1; // 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].x = 0.2; // 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].y = 0.3; // 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
// rv_cpu[i].z = 0.4; // 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
// qv_cpu[i] = 0.5; // 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_OPENCL
//====================================================================================================100
kernel_gpu_opencl_wrapper( par_cpu,
dim_cpu,
box_cpu,
rv_cpu,
qv_cpu,
fv_cpu);
time6 = get_time();
//======================================================================================================================================================150
// SYSTEM MEMORY DEALLOCATION
//======================================================================================================================================================150
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
}
int
main( int argc,
char* argv []){
printf("WG size of kernel = %d \n", NUMBER_THREADS);
//======================================================================================================================================================150
// VARIABLES
//======================================================================================================================================================150
// time
long long time0;
long long time1;
long long time2;
long long time3;
long long time4;
long long time5;
long long time6;
// inputs image, input paramenters
fp* image_ori; // originalinput image
int image_ori_rows;
int image_ori_cols;
long image_ori_elem;
// inputs image, input paramenters
fp* image; // input image
int Nr,Nc; // IMAGE nbr of rows/cols/elements
long Ne;
// algorithm parameters
int niter; // nbr of iterations
fp lambda; // update step size
// size of IMAGE
int r1,r2,c1,c2; // row/col coordinates of uniform ROI
long NeROI; // ROI nbr of elements
// surrounding pixel indicies
int* iN;
int* iS;
int* jE;
int* jW;
// counters
int iter; // primary loop
long i; // image row
long j; // image col
// memory sizes
int mem_size_i;
int mem_size_j;
time0 = get_time();
//======================================================================================================================================================150
// INPUT ARGUMENTS
//======================================================================================================================================================150
if(argc != 5){
printf("ERROR: wrong number of arguments\n");
return 0;
}
else{
niter = atoi(argv[1]);
lambda = atof(argv[2]);
Nr = atoi(argv[3]); // it is 502 in the original image
Nc = atoi(argv[4]); // it is 458 in the original image
}
time1 = get_time();
//======================================================================================================================================================150
// READ INPUT FROM FILE
//======================================================================================================================================================150
//====================================================================================================100
// READ IMAGE (SIZE OF IMAGE HAS TO BE KNOWN)
//====================================================================================================100
image_ori_rows = 502;
image_ori_cols = 458;
image_ori_elem = image_ori_rows * image_ori_cols;
image_ori = (fp*)malloc(sizeof(fp) * image_ori_elem);
read_graphics( "../../data/srad/image.pgm",
image_ori,
image_ori_rows,
image_ori_cols,
1);
//====================================================================================================100
// RESIZE IMAGE (ASSUMING COLUMN MAJOR STORAGE OF image_orig)
//====================================================================================================100
Ne = Nr*Nc;
image = (fp*)malloc(sizeof(fp) * Ne);
resize( image_ori,
image_ori_rows,
image_ori_cols,
image,
Nr,
Nc,
1);
//====================================================================================================100
// End
//====================================================================================================100
time2 = get_time();
//======================================================================================================================================================150
// SETUP
//======================================================================================================================================================150
// variables
r1 = 0; // top row index of ROI
r2 = Nr - 1; // bottom row index of ROI
c1 = 0; // left column index of ROI
c2 = Nc - 1; // right column index of ROI
// ROI image size
NeROI = (r2-r1+1)*(c2-c1+1); // number of elements in ROI, ROI size
// allocate variables for surrounding pixels
mem_size_i = sizeof(int) * Nr; //
iN = (int *)malloc(mem_size_i) ; // north surrounding element
iS = (int *)malloc(mem_size_i) ; // south surrounding element
mem_size_j = sizeof(int) * Nc; //
jW = (int *)malloc(mem_size_j) ; // west surrounding element
jE = (int *)malloc(mem_size_j) ; // east surrounding element
// N/S/W/E indices of surrounding pixels (every element of IMAGE)
for (i=0; i<Nr; i++) {
iN[i] = i-1; // holds index of IMAGE row above
iS[i] = i+1; // holds index of IMAGE row below
}
for (j=0; j<Nc; j++) {
jW[j] = j-1; // holds index of IMAGE column on the left
jE[j] = j+1; // holds index of IMAGE column on the right
}
// N/S/W/E boundary conditions, fix surrounding indices outside boundary of image
iN[0] = 0; // changes IMAGE top row index from -1 to 0
iS[Nr-1] = Nr-1; // changes IMAGE bottom row index from Nr to Nr-1
jW[0] = 0; // changes IMAGE leftmost column index from -1 to 0
jE[Nc-1] = Nc-1; // changes IMAGE rightmost column index from Nc to Nc-1
time3= get_time();
//======================================================================================================================================================150
// KERNEL
//======================================================================================================================================================150
kernel_gpu_opencl_wrapper( image, // input image
Nr, // IMAGE nbr of rows
Nc, // IMAGE nbr of cols
Ne, // IMAGE nbr of elem
niter, // nbr of iterations
lambda, // update step size
NeROI, // ROI nbr of elements
iN,
iS,
jE,
jW,
iter, // primary loop
mem_size_i,
mem_size_j);
time4 = get_time();
//======================================================================================================================================================150
// WRITE OUTPUT IMAGE TO FILE
//======================================================================================================================================================150
write_graphics( "./output/image_out.pgm",
image,
Nr,
Nc,
1,
255);
time5 = get_time();
//======================================================================================================================================================150
// FREE MEMORY
//======================================================================================================================================================150
free(image_ori);
free(image);
free(iN);
free(iS);
free(jW);
free(jE);
time6 = get_time();
//======================================================================================================================================================150
// DISPLAY TIMING
//======================================================================================================================================================150
printf("Time spent in different stages of the application:\n");
printf("%.12f s, %.12f % : READ COMMAND LINE PARAMETERS\n", (fp) (time1-time0) / 1000000, (fp) (time1-time0) / (fp) (time5-time0) * 100);
printf("%.12f s, %.12f % : READ AND RESIZE INPUT IMAGE FROM FILE\n", (fp) (time2-time1) / 1000000, (fp) (time2-time1) / (fp) (time5-time0) * 100);
printf("%.12f s, %.12f % : SETUP\n", (fp) (time3-time2) / 1000000, (fp) (time3-time2) / (fp) (time5-time0) * 100);
printf("%.12f s, %.12f % : KERNEL\n", (fp) (time4-time3) / 1000000, (fp) (time4-time3) / (fp) (time5-time0) * 100);
printf("%.12f s, %.12f % : WRITE OUTPUT IMAGE TO FILE\n", (fp) (time5-time4) / 1000000, (fp) (time5-time4) / (fp) (time5-time0) * 100);
printf("%.12f s, %.12f % : FREE MEMORY\n", (fp) (time6-time5) / 1000000, (fp) (time6-time5) / (fp) (time5-time0) * 100);
printf("Total time:\n");
printf("%.12f s\n", (fp) (time5-time0) / 1000000);
}