REAL
value(const unsigned i,
const REAL s0,
const REAL t,
const REAL alpha,
const REAL nu,
const REAL beta,
const unsigned int numX,
const unsigned int numY,
const unsigned int numT)
{
REAL strike;
PrivGlobs globs(numX, numY, numT);
strike = 0.001*i;
initGrid(s0, alpha, nu, t, numX, numY, numT, globs);
initOperator(globs.myX, globs.myDxx);
initOperator(globs.myY, globs.myDyy);
setPayoff(strike, globs);
for(int i = globs.myTimeline.size()-2; i >= 0; i--) {
updateParams(i,alpha,beta,nu,globs);
rollback(i, globs);
}
return globs.myResult[globs.myXindex][globs.myYindex];
}
void run_GPU(
const unsigned int& outer,
const unsigned int& numX,
const unsigned int& numY,
const unsigned int& numT,
const REAL& s0,
const REAL& t,
const REAL& alpha,
const REAL& nu,
const REAL& beta,
REAL* res // [outer] RESULT
) {
REAL strike = 0.001*i;
PrivGlobs globs(numX, numY, numT);
// Outerloop - Technically parallelizable, but restricts further
// parallization further in.
// If strike and globs is privatized, the loop can be parallelized.
// Value is the limiting factor since most of the actual work is deeper in
// the function.
for( unsigned i = 0; i < outer; ++ i ) {
res[i] = value( globs, s0, strike, t,
alpha, nu, beta,
numX, numY, numT );
}
}
void run_OrigCPU(
const unsigned int& outer,
const unsigned int& numX,
const unsigned int& numY,
const unsigned int& numT,
const REAL& s0,
const REAL& t,
const REAL& alpha,
const REAL& nu,
const REAL& beta,
REAL* res // [outer] RESULT
) {
REAL strike;
PrivGlobs globs(numX, numY, numT);
for( unsigned i = 0; i < outer; ++ i ) {
strike = 0.001*i;
res[i] = value( globs, s0, strike, t,
alpha, nu, beta,
numX, numY, numT );
}
}