int main(int argc, char* argv[]) {
char* input = "Gdkkn\x1FGR@\x1FVnqkc";
size_t strlength = strlen(input);
char *output = (char*) malloc_global(strlength + 1);
char *secode = (char*) malloc_global(strlength + 1);
char *output2 = (char*) malloc_global(strlength + 1);
SNK_INIT_LPARM(lparm,strlength);
decode(input,output,lparm);
output[strlength] = '\0';
printf("Decoded :%s\n",output);
/* Show we can call multiple functions in the .cl file */
super_encode(output,secode,lparm);
printf("Super encoded :%s\n",secode);
super_decode(secode,output2,lparm);
printf("Super decoded :%s\n",output2);
/* Show we can call same function multiple times */
decode(secode,output,lparm);
decode(output,output2,lparm);
printf("Decoded twice :%s\n",output2);
free(output);
free(secode);
free(output2);
return 0;
}
int main(int argc, char **argv) {
if(argc < 2) {
usage();
return -1;
}
//init the filter array
float filter[49] =
{1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1};
//operate the params of cmd
const char* inputFileName;
const char* outputFileName;
inputFileName = (argv[1]);
outputFileName = (argv[2]);
//the image height and width
int imageHeight, imageWidth;
//read the bmp image to the memory
float* inputImage = readBmpImage(inputFileName, &imageWidth, &imageHeight);
//to check the read is succ
printf("the width of the image is %d, the height of the image is %d\n", imageWidth, imageHeight);
//calculate the datasize
int dataSize = imageHeight * imageWidth * sizeof(float);
//output image
float *outputImage = NULL;
outputImage = (float*)malloc(dataSize);
//init lparm
SNK_INIT_LPARM(lparm, (imageWidth * imageHeight));
lparm->ndim=2;
lparm->gdims[0]=imageWidth;
lparm->gdims[1]=imageHeight;
lparm->ldims[0]=1;
lparm->ldims[1]=1;
//call the blur function
int filterWidth = 7;
blur(outputImage, inputImage, imageWidth, imageHeight, filter, filterWidth, lparm);
//store the image
storeBmpImage(outputImage, outputFileName, imageHeight, imageWidth, inputFileName);
free(outputImage);
return 0;
}
int main(int argc, char **argv) {
if(argc < 2) {
usage();
return -1;
}
//operate the params of cmd
const char* inputFileName;
const char* outputFileName;
inputFileName = (argv[1]);
outputFileName = (argv[2]);
printf("1233556\n");
//calculate the sin(theta) & cos(theta)
float sinTheta = sinf(THETA);
float cosTheta = cosf(THETA);
//the image height and width
int imageHeight, imageWidth;
//read the bmp image to the memory
float* inputImage = readBmpImage(inputFileName, &imageWidth, &imageHeight);
//to check the read is succ
printf("the width of the image is %d, the height of the image is %d\n", imageWidth, imageHeight);
//calculate the datasize
int dataSize = imageHeight * imageWidth * sizeof(float);
//output image
float *outputImage = NULL;
outputImage = (float*)malloc(dataSize);
//init lparm
SNK_INIT_LPARM(lparm, (imageWidth * imageHeight));
lparm->ndim=2;
lparm->gdims[0]=imageWidth;
lparm->gdims[1]=imageHeight;
lparm->ldims[0]=1;
lparm->ldims[1]=1;
//call the rotate function
image_rotate(outputImage,inputImage,imageWidth,imageHeight,sinTheta,cosTheta,lparm);
//store the image
storeBmpImage(outputImage, outputFileName, imageHeight, imageWidth, inputFileName);
free(outputImage);
return 0;
}
// ------------------------------------
// 更新table, 傳回sse, 存入點之變動數
float update_table_cl(int* ch_pt)
{
int i, j, k, min_k;
float dis, t_sse=0.0;
float data_ker[DCNT*DIM];
float cent_ker[K*DIM];
int cent_c_ker[DCNT];
//float min_dis;
int chpt[DCNT];
float min_dis[DCNT];
unsigned int k_ker=K,dim_ker=DIM,dcnt_ker=DCNT;
*ch_pt=0; // 變動點數設0
memset(cent_c, 0, sizeof(cent_c)); // 各叢聚資料數清0
memset(dis_k, 0, sizeof(dis_k)); // 各叢聚距離和清0
/*OpenCL*/
size_t work_size;
for (i=0; i<DCNT;++i) {
for(k=0; k<DIM;++k) {
data_ker[i*DIM+k]=data[i][k];
}
}
for (i=0; i<K;++i) {
for(k=0; k<DIM;++k) {
cent_ker[i*DIM+k]=cent[i][k];
}
}
//err = 0;
/*snack
err |= clEnqueueWriteBuffer(queue, cl_data, CL_TRUE, 0,
sizeof(float) * DCNT * DIM, data_ker, 0, 0, 0);
err |= clEnqueueWriteBuffer(queue, cl_cent, CL_TRUE, 0,
sizeof(float) * K * DIM, cent_ker, 0, 0, 0);
err |= clEnqueueWriteBuffer(queue, cl_table, CL_TRUE, 0,
sizeof(int) * DCNT, table, 0, 0, 0);
err |= clSetKernelArg(kernel, 3, sizeof(unsigned int), &k_ker);
err |= clSetKernelArg(kernel, 4, sizeof(unsigned int), &dim_ker);
err |= clSetKernelArg(kernel, 5, sizeof(unsigned int), &dcnt_ker);
err |= clEnqueueWriteBuffer(queue, cl_chpt, CL_TRUE, 0,
sizeof(int) * DCNT, chpt, 0, 0, 0);
err |= clEnqueueWriteBuffer(queue, cl_cent_c, CL_TRUE, 0,
sizeof(int) * DCNT, cent_c_ker, 0, 0, 0);
err |= clEnqueueWriteBuffer(queue, cl_min_dis, CL_TRUE, 0,
sizeof(float) * DCNT, min_dis, 0, 0, 0);
if (err) {
printf("ERROR! WRITE BUFFER\n");
exit(1);
}
work_size = DCNT;
err = clEnqueueNDRangeKernel(queue, kernel, 1, 0, &work_size, 0, 0, 0, 0);
if (err == CL_SUCCESS) {
err |= clEnqueueReadBuffer(queue, cl_chpt, CL_TRUE, 0, sizeof(cl_int)*DCNT, &chpt[0], 0, 0, 0);
err |= clEnqueueReadBuffer(queue, cl_table, CL_TRUE, 0, sizeof(cl_int)*DCNT, &table[0], 0, 0, 0);
err |= clEnqueueReadBuffer(queue, cl_cent_c, CL_TRUE, 0, sizeof(cl_int)*DCNT, ¢_c_ker[0], 0, 0, 0);
err |= clEnqueueReadBuffer(queue, cl_min_dis, CL_TRUE, 0, sizeof(cl_float)*DCNT, &min_dis[0], 0, 0, 0);
}
*/
/*
float * data_ker=(float*)malloc(DCNT*DIM);
float * cent_ker=(float*)malloc(K*DIM);
int * table=(int*)malloc(DCNT);
unsigned int * k_ker=(unsigned int*)malloc(K);
unsigned int * dim_ker=(unsigned int*)malloc(DIM);
unsigned int * dcnt_ker=(unsigned int*)malloc(DCNT);
int * chpt=(int*)malloc(DCNT);
int * cent_c_ker=(int*)malloc(DCNT);
float * min_dis=(float*)malloc(DCNT);
*/
//our snack code
SNK_INIT_LPARM(lparm,DCNT);
cal_diskernel(data_ker,cent_ker,table,K,DIM,DCNT,chpt,cent_c_ker,min_dis,lparm);
for(i=0;i<DCNT;i++) {
*ch_pt+=chpt[i];
++cent_c[cent_c_ker[i]];
t_sse+=min_dis[i];
for(j=0;j<DIM;j++) {
dis_k[table[i]][j]+=data[i][j];
}
}
return t_sse;
}
int main(int argc, char **argv)
{
struct timeval tstart, tstop;
unsigned long int meas[NB_MEASURE];
unsigned long int meas2[NB_MEASURE];
unsigned long int duration = 0;
unsigned long int var = 0;
unsigned int size = 1;
int verbose = 0;
int i = 0;
int c;
int prefillingcache = 0;
int coef = 1;
int nbopt = 0;
int flops = 0;
while ((c = getopt(argc, argv, "cf")) != -1)
switch (c) {
case 'c':
prefillingcache = 1;
nbopt++;
break;
case 'f':
flops = 1;
nbopt++;
break;
default:
abort();
}
if (2 <= argc) {
int pos = nbopt + 1;
if (NULL != strcasestr(argv[pos], "kB"))
coef = 1024;
else if (NULL != strcasestr(argv[pos], "MB"))
coef = 1024 * 1024;
else if (NULL != strcasestr(argv[pos], "GB"))
coef = 1024 * 1024 * 1024;
size *= coef * atoi(argv[pos]);
}
if (1 == coef) {
size *= sizeof(float);
}
//Setup kernel arguments
float *in = (float *) malloc(size);
float *out = (float *) malloc(size);
memset(out, 0, size);
memset(in, 0, size);
for (i = 0; i < (size / sizeof(float)); i++)
in[i] = i;
// Mesure NBR_COPY copy vector
//gettimeofday(&tstart, NULL);
//for (int i = 0; i < LOOPS; i++) {
SNK_INIT_LPARM(lparm, size / sizeof(float));
//Fill Caches
if (prefillingcache) {
printf("Pre-filling cache option SET\n");
for (i = 0; i < 64; i++) {
if(flops) flops_3(in, out, lparm)
else vcopy(in, out, lparm);
}
}
for (i = 0; i < NB_MEASURE; i++) {
gettimeofday(&tstart, NULL);
vcopy(in, out, lparm);
gettimeofday(&tstop, NULL);
meas[i] =
((tstop.tv_sec - tstart.tv_sec) * 1000000L + tstop.tv_usec) -
tstart.tv_usec;
//meas2[i] = meas[i] * meas[i];
}
for (i = 0; i < NB_MEASURE; i++) {
duration += meas[i];
//var += meas2[i];
}
duration /= NB_MEASURE;
for (i = 0; i < NB_MEASURE; i++) {
var += ((meas[i] - duration) * (meas[i] - duration));
}
var /= NB_MEASURE;
//var -= duration * duration;
if(flops){
printf
("HSA: Vector of %lu integer of %d-bytes = %lu Bytes takes %lu usec [+/-var %lu] => Speed = %.3f\n",
(size / sizeof(float)), (int) sizeof(float), size, duration, var, (float)(3.0 * size/sizeof(float) / duration * 1000000));
}
else{
printf
("HSA: Vector of %lu integer of %d-bytes = %lu Bytes takes %lu usec [+/-var %lu]\n",
(size / sizeof(float)), (int) sizeof(float), size, duration, var);
}
//Validate
bool valid = true;
int failIndex = 0;
for (i = 0; i < (size / sizeof(float)); i++) {
if (verbose && i < 10)
printf("in[%d]=%d, out[%d]=%d, ", i, in[i], i, out[i]);
if (out[i] != in[i]) {
failIndex = i;
valid = false;
break;
}
}
if (valid) {
if (verbose)
printf("passed validation\n");
} else
printf("VALIDATION FAILED!\nBad index: %d\n", failIndex);
free(in);
free(out);
return 0;
}
