void PrpFromPrfFixed::computeBlock(const vector<byte> & inBytes, int inOff, int inLen, vector<byte>& outBytes, int outOff, int outLen) {
if (!isKeySet())
throw IllegalStateException("secret key isn't set");
if ((inOff > inBytes.size()) || (inOff + inLen > inBytes.size()))
throw out_of_range("wrong offset for the given input buffer");
if ((outOff > outBytes.size()) || (outOff + outLen > outBytes.size()))
throw out_of_range("wrong offset for the given output buffer");
// if the input and output length are equal to the blockSize, call the computeBlock that doesn't take length arguments.
if (inLen == outLen && inLen == getBlockSize())
computeBlock(inBytes, inOff, outBytes, outOff);
else
throw out_of_range("input and output lengths should be equal to Block size");
}
void PrpFromPrfVarying::computeBlock(const vector<byte> & inBytes, int inOff, int inLen, vector<byte>& outBytes, int outOff, int outLen) {
if (!isKeySet())
throw new IllegalStateException("secret key isn't set");
// checks that the offsets and lengths are correct
if ((inOff > inBytes.size()) || (inOff + inLen > inBytes.size()))
throw out_of_range("wrong offset for the given input buffer");
if ((outOff > outBytes.size()) || (outOff + outLen > outBytes.size()))
throw out_of_range("wrong offset for the given output buffer");
//if the input and output lengths are equal, call the computeBlock which takes just one length argument
if (inLen == outLen)
computeBlock(inBytes, inOff, inLen, outBytes, outOff);
else throw out_of_range("input and output lengths should be equal");
}
void *th_electric_field(void *argthinfo) {
Thinfo *thinfo = (Thinfo *) argthinfo;
float * charge = thinfo->charge;
float * coord1 = thinfo->coord1;
float * coord2 = thinfo->coord2;
float * coord3 = thinfo->coord3;
float grid_span = thinfo->grid_span;
int grid_size = thinfo->grid_size;
int totalElements = thinfo->totalElements;
int block_size = thinfo->block_size;
int x_start = thinfo->x_start;
int x_end = thinfo->x_end;
int num_threads = thinfo->num_threads;
fftw_real *grid = thinfo->grid;
float phi;
int block_ini, block_fin;
float x_centre, y_centre, z_centre;
int x, y, z, i, k;
__m128 _phi_tmp;
__m128 _phiSet;
float *phiSet = (float *) &_phiSet;
float * distance;
float * epsilon;
if ((posix_memalign((void**)&distance, 16, 4*sizeof(float))!=0)) {
printf("No memory.\n");
exit(-1);
}
if ((posix_memalign((void**)&epsilon, 16, 4*sizeof(float))!=0)) {
printf("No memory.\n");
exit(-1);
}
for (block_ini = 0; block_ini < totalElements-(block_size-1); block_ini += block_size) {
block_fin = block_ini + block_size;
for (x = x_start; x < x_end; x++) {
x_centre = gcentre(x ,grid_span, grid_size ) ;
for (y = 0; y < grid_size; y++) {
y_centre = gcentre(y ,grid_span ,grid_size ) ;
for (z = 0; z < grid_size; z++) {
z_centre = gcentre( z , grid_span , grid_size ) ;
phi = 0 ;
computeBlock(block_ini, block_fin);
grid[gaddress(x,y,z,grid_size)] += (fftw_real)phi;
}
}
}
//Changing block. Must wait for others threads to finish.
pthread_mutex_lock(&mut);
ready_threads++;
if (ready_threads==num_threads) {
ready_threads=0;
pthread_cond_broadcast(&cond);
} else {
pthread_cond_wait(&cond, &mut);
}
pthread_mutex_unlock(&mut);
}
//Last block is not a multiple of block_size
if (block_ini != totalElements) {
for (x = x_start; x < x_end; x++ ) {
x_centre = gcentre(x, grid_span, grid_size) ;
for (y = 0; y < grid_size; y++) {
y_centre = gcentre(y, grid_span, grid_size) ;
for (z = 0; z < grid_size; z++) {
z_centre = gcentre(z, grid_span, grid_size) ;
phi = 0 ;
computeEndBlock(block_ini, totalElements);
grid[gaddress(x,y,z,grid_size)] += (fftw_real)phi ;
}
}
}
}
pthread_exit(NULL);
return ;
}
