// Black-Scholes formula for both call and put
HEMI_KERNEL(BlackScholes)
(float *callResult, float *putResult, const float *stockPrice,
const float *optionStrike, const float *optionYears, float Riskfree,
float Volatility, int optN)
{
int offset = hemiGetElementOffset();
int stride = hemiGetElementStride();
for(int opt = offset; opt < optN; opt += stride)
{
float S = stockPrice[opt];
float X = optionStrike[opt];
float T = optionYears[opt];
float R = Riskfree;
float V = Volatility;
float sqrtT = sqrtf(T);
float d1 = (logf(S / X) + (R + 0.5f * V * V) * T) / (V * sqrtT);
float d2 = d1 - V * sqrtT;
float CNDD1 = CND(d1);
float CNDD2 = CND(d2);
//Calculate Call and Put simultaneously
float expRT = expf(- R * T);
callResult[opt] = S * CNDD1 - X * expRT * CNDD2;
putResult[opt] = X * expRT * (1.0f - CNDD2) - S * (1.0f - CNDD1);
}
}
__global__ void init_device_rngs(int nthreads, unsigned long long seed,
curandStateXORWOW* state) {
int idx = hemiGetElementOffset();
if (idx >= nthreads) {
return;
}
curand_init(seed, idx, 0, &state[idx]);
}
HEMI_DEV_CALLABLE_INLINE
void pick_new_vector_device(int nthreads, RNGState* rng,
const float* sigma,
const double* current_vector,
double* proposed_vector) {
int offset = hemiGetElementOffset();
int stride = hemiGetElementStride();
for (int i=offset; i<(int)nthreads; i+=stride) {
#ifdef HEMI_DEV_CODE
double u = curand_normal(&rng[i]);
proposed_vector[i] = current_vector[i] + sigma[i] * u;
#else
double u = gRandom->Gaus(current_vector[i], sigma[i]);
proposed_vector[i] = u;
#endif
}
}
// append all steps to jump buffer
int count = counter[0];
for (unsigned i=0; i<nparameters; i++) {
jump_buffer[count * (nparameters + 1) + i] = v_current[i];
}
jump_buffer[count * (nparameters + 1) + nparameters] = nll_current[0];
counter[0] = count + 1;
}
HEMI_KERNEL(nll_event_chunks)(const float* __restrict__ lut,
const int* __restrict__ dataweights,
const double* __restrict__ pars,
const size_t ne, const size_t ns,
double* sums) {
int offset = hemiGetElementOffset();
int stride = hemiGetElementStride();
double sum = 0;
for (int i=offset; i<(int)ne; i+=stride) {
double s = 0;
for (size_t j=0; j<ns; j++) {
float v = lut[j * ne + i];
s += pars[j] * (!isnan(v) ? v : 0); // handle nans from empty hists
}
sum += log(s)*dataweights[i];
}
if (!isnan(sum)) {
sums[offset] = sum;
}
}