int main(int argc,char *argv[]) {
double *A0_naive, *A0_test;
double *Anext_naive, *Anext_test;
double *Afinal_naive, *Afinal_test;
int nx,ny,nz,tx,ty,tz,timesteps;
int i,j;
ticks t1, t2;
double spt;
/* parse command line options */
if (argc < 8) {
printf("\nUSAGE:\n%s <grid x> <grid y> <grid z> <block x> <block y> <block z> <timesteps>\n", argv[0]);
printf("\nTIME SKEWING CONSTRAINTS:\nIn each dimension, <grid size - 2> should be a multiple of <block size>.\n");
printf("\nCIRCULAR QUEUE CONSTRAINTS:\n<grid y - 2> should be a multiple of <block y>. The block sizes in the other dimensions are ignored.\n\n");
return EXIT_FAILURE;
}
nx = atoi(argv[1]);
ny = atoi(argv[2]);
nz = atoi(argv[3]);
tx = atoi(argv[4]);
ty = atoi(argv[5]);
tz = atoi(argv[6]);
timesteps = atoi(argv[7]);
printf("%dx%dx%d, blocking: %dx%dx%d, timesteps: %d\n",
nx,ny,nz,tx,ty,tz,timesteps);
#ifdef HAVE_PAPI
PAPI_library_init(PAPI_VER_CURRENT);
#endif
/* find conversion factor from ticks to seconds */
spt = seconds_per_tick();
// allocate arrays
A0_naive=(double*)malloc(sizeof(double)*nx*ny*nz);
A0_test=(double*)malloc(sizeof(double)*nx*ny*nz);
Anext_naive=(double*)malloc(sizeof(double)*nx*ny*nz);
Anext_test=(double*)malloc(sizeof(double)*nx*ny*nz);
// Run Naive Code
StencilInit(nx,ny,nz,A0_naive);
StencilInit(nx,ny,nz,Anext_naive);
StencilProbe_naive(A0_naive, Anext_naive, nx, ny, nz, tx, ty, tz, timesteps);
if (timesteps%2 == 0) {
Afinal_naive = A0_naive;
}
else {
Afinal_naive = Anext_naive;
}
printf("USING TIMER: %s \t SECONDS PER TICK:%g \n", TIMER_DESC, spt);
// Test Rivera Blocking
StencilInit(nx,ny,nz,A0_test);
StencilInit(nx,ny,nz,Anext_test);
printf("Checking Rivera (single-timestep) blocking...\n");
StencilProbe_rivera(A0_test, Anext_test, nx, ny, nz, tx, ty, tz, timesteps);
if (timesteps%2 == 0) {
Afinal_test = A0_test;
}
else {
Afinal_test = Anext_test;
}
check_vals(Afinal_naive, Afinal_test, nx, ny, nz);
// Test Cache-Oblivious Blocking
StencilInit(nx,ny,nz,A0_test);
StencilInit(nx,ny,nz,Anext_test);
printf("Checking Cache-Oblivious blocking...\n");
StencilProbe_oblivious(A0_test, Anext_test, nx, ny, nz, tx, ty, tz, timesteps);
if (timesteps%2 == 0) {
Afinal_test = A0_test;
}
else {
Afinal_test = Anext_test;
}
check_vals(Afinal_naive, Afinal_test, nx, ny, nz);
// Test Time-Skewed Blocking
StencilInit(nx,ny,nz,A0_test);
StencilInit(nx,ny,nz,Anext_test);
printf("Checking Time-Skewed blocking...\n");
StencilProbe_timeskew(A0_test, Anext_test, nx, ny, nz, tx, ty, tz, timesteps);
if (timesteps%2 == 0) {
Afinal_test = A0_test;
}
else {
Afinal_test = Anext_test;
}
check_vals(Afinal_naive, Afinal_test, nx, ny, nz);
// Test Circular-Queue Blocking
StencilInit(nx,ny,nz,A0_test);
StencilInit(nx,ny,nz,Anext_test);
if (timesteps > 1) {
CircularQueueInit(nx, ty, timesteps);
}
printf("Checking Circular-Queue blocking...\n");
StencilProbe_circqueue(A0_test, Anext_test, nx, ny, nz, tx, ty, tz, timesteps);
Afinal_test = Anext_test;
check_vals(Afinal_naive, Afinal_test, nx, ny, nz);
/* free arrays */
free(Anext_naive);
free(A0_naive);
free(Anext_test);
free(A0_test);
return EXIT_SUCCESS;
}
int main(int argc,char *argv[])
{
float* vx0;
float* vy0;
float* vz0;
float* txx0;
float* tyy0;
float* tzz0;
float* txy0;
float* txz0;
float* tyz0;
float* fx;
float* fy;
float* fz;
int nx,ny,nz,tx,ty,tz,timesteps;
int i,j,k;
float spt;
int block;
uint64_t start, exec_time;
/* parse command line options */
nx = atoi(argv[1]);
ny = atoi(argv[2]);
nz = atoi(argv[3]);
tx = 1;
ty = 1;
tz = 1;
timesteps = atoi(argv[4]);
// printf("%dx%dx%d, blocking: %dx%dx%d, timesteps: %d\n",
// nx,ny,nz,tx,ty,tz,timesteps);
/* allocate arrays */
vx0=(float*)malloc(sizeof(float)*nx*ny*nz);
vy0=(float*)malloc(sizeof(float)*nx*ny*nz);
vz0=(float*)malloc(sizeof(float)*nx*ny*nz);
txx0=(float*)malloc(sizeof(float)*nx*ny*nz);
tyy0=(float*)malloc(sizeof(float)*nx*ny*nz);
tzz0=(float*)malloc(sizeof(float)*nx*ny*nz);
txy0=(float*)malloc(sizeof(float)*nx*ny*nz);
txz0=(float*)malloc(sizeof(float)*nx*ny*nz);
tyz0=(float*)malloc(sizeof(float)*nx*ny*nz);
fx=(float*)malloc(sizeof(float)*nx*ny*nz);
fy=(float*)malloc(sizeof(float)*nx*ny*nz);
fz=(float*)malloc(sizeof(float)*nx*ny*nz);
/* initialize arrays to all ones */
StencilInit(nx,ny,nz,vx0);
StencilInit(nx,ny,nz,vy0);
StencilInit(nx,ny,nz,vz0);
StencilInit(nx,ny,nz,txx0);
StencilInit(nx,ny,nz,tyy0);
StencilInit(nx,ny,nz,tzz0);
StencilInit(nx,ny,nz,txy0);
StencilInit(nx,ny,nz,txz0);
StencilInit(nx,ny,nz,tyz0);
StencilInit(nx,ny,nz,fx);
StencilInit(nx,ny,nz,fy);
StencilInit(nx,ny,nz,fz);
start = get_time();
/* stencil function */
StencilProbe(
vx0,
vy0,
vz0,
txx0,
tyy0,
tzz0,
txy0,
txz0,
tyz0,
fx,
fy,
fz,
nx, ny, nz, tx, ty, tz, timesteps);
exec_time = diff_time(start, get_time());
//printf("#call of pair of stencil(prev/next) :%d \n",timesteps/2);
printf ("%d;%d;%d;%d;%d;" "%" PRIu64, nx, ny, nz, timesteps, omp_thread_count(), exec_time);
//printf ("%d " "%" PRIu64, omp_thread_count(), exec_time/1000/1000);
/*
for (k = 2; k < nz - 2; k++) {
for (j = 2; j < ny - 2; j++) {
for (i = 2; i < nx - 2; i++) {
*/
#ifdef PROUT
i=13;
j=15;
k=37;
printf("\n");
printf("i=%d - j=%d - k=%d - %f - %f \n",i,j,k,vx0[Index3D (nx, ny, i, j, k)],txx0[Index3D (nx, ny, i, j, k)]);
#endif
/*
}
}
}
*/
/* free arrays */
free(vx0);
free(vy0);
free(vz0);
free(fx);
free(fy);
free(fz);
free(txx0);
free(tyy0);
free(tzz0);
free(txy0);
free(txz0);
free(tyz0);
}
int main(int argc,char *argv[])
{
double *Anext;
double *A0;
int nx,ny,nz,tx,ty,tz,timesteps;
int i;
ticks t1, t2;
double spt;
/* parse command line options */
if (argc < 8) {
printf("\nUSAGE:\n%s <grid x> <grid y> <grid z> <block x> <block y> <block z> <timesteps>\n", argv[0]);
printf("\nTIME SKEWING CONSTRAINTS:\nIn each dimension, <grid size - 2> should be a multiple of <block size>.\n");
printf("\nCIRCULAR QUEUE CONSTRAINTS:\n<grid y - 2> should be a multiple of <block y>. The block sizes in the other dimensions are ignored.\n\n");
return EXIT_FAILURE;
}
nx = atoi(argv[1]);
ny = atoi(argv[2]);
nz = atoi(argv[3]);
tx = atoi(argv[4]);
ty = atoi(argv[5]);
tz = atoi(argv[6]);
timesteps = atoi(argv[7]);
printf("%dx%dx%d, blocking: %dx%dx%d, timesteps: %d\n",
nx,ny,nz,tx,ty,tz,timesteps);
#ifdef HAVE_PAPI
PAPI_library_init(PAPI_VER_CURRENT);
#endif
/* find conversion factor from ticks to seconds */
spt = seconds_per_tick();
/* allocate arrays */
Anext=(double*)malloc(sizeof(double)*nx*ny*nz);
A0=(double*)malloc(sizeof(double)*nx*ny*nz);
printf("USING TIMER: %s \t SECONDS PER TICK:%g \n", TIMER_DESC, spt);
for (i=0;i<NUM_TRIALS;i++) {
/* initialize arrays to all ones */
StencilInit(nx,ny,nz,Anext);
StencilInit(nx,ny,nz,A0);
#ifdef CIRCULARQUEUEPROBE
if (timesteps > 1) {
CircularQueueInit(nx, ty, timesteps);
}
#endif
// clear_cache();
t1 = getticks();
/* stencil function */
StencilProbe(A0, Anext, nx, ny, nz, tx, ty, tz, timesteps);
t2 = getticks();
printf("elapsed ticks: %g time:%g \n", elapsed(t2, t1), spt * elapsed(t2,t1));
#ifdef COMPUTECHECKSUM
extern double CheckSum(int, int, int, double*);
double r1=CheckSum(nx, ny, nz, A0);
double r2=CheckSum(nx, ny, nz, Anext);
printf("Checksums: A0=%g, Anext=%g\n", r1, r2);
#endif
}
/* free arrays */
free(Anext);
free(A0);
return EXIT_SUCCESS;
}
