• This module is part of a multi-threaded program(using openmp).

• I am using intel c++ compiler, icpc, with -O3

• My ultimate aim is to run this code on Xeon-Phi.

Edit: Function

• This module is part of a multi-threaded program  (using OpenMO).

• I am using Intel C++ compiler, icpc, with -O3.

• My ultimate aim is to run this code on Xeon-Phi.

Function

Edit: Function

__declspec(vector) void roeflux(double rhol, double rhor, double ul, double ur, double pl, double pr,
                double *f_0, double *f_1, double *f_2){
  double el = pl/(GAMMA_M) + 0.5f*rhol*ul*ul;
  double er = pr/(GAMMA_M) + 0.5f*rhor*ur*ur;
  double hl = (el + pl)/rhol;
  double hr = (er + pr)/rhor;

  double sqrtrhol = sqrt(rhol);
  double sqrtrhor = sqrt(rhor);
  double den_inverse = 1/(sqrtrhol + sqrtrhor);
  double uavg = (sqrtrhol*ul + sqrtrhor*ur)*den_inverse;
  double havg = (sqrtrhol*hl + sqrtrhor*hr)*den_inverse;
  double cavg = sqrt(GAMMA_M*(havg - 0.5f*uavg*uavg));
  double cavg_inverse = 1.0f/cavg;

  double d1 = rhor - rhol;
  double d2 = rhor*ur - rhol*ul;
  double d3 = er - el;
  
  double alpha_2 = GAMMA_M*((havg - uavg*uavg)*d1 + uavg*d2 - d3)*cavg_inverse*cavg_inverse;
  double alpha_3 = 0.5f*(d2 + (cavg - uavg)*d1 - cavg*alpha_2)*cavg_inverse;
  double alpha_1 = d1 - alpha_2 - alpha_3;
  
  double lambda_1 =  fabs(uavg - cavg);
  double lambda_2 =  fabs(uavg);
  double lambda_3 =  fabs(uavg + cavg);

  double f1 = lambda_1*alpha_1 + lambda_2*alpha_2 + lambda_3*alpha_3;
  double f2 = lambda_1*alpha_1*(uavg-cavg) + lambda_2*alpha_2*uavg + lambda_3*alpha_3*(uavg+cavg);
  double f3 = lambda_1*alpha_1*(havg-cavg*uavg) + 0.5f*lambda_2*alpha_2*uavg*uavg + lambda_3*alpha_3*(havg+cavg*uavg);

  *f_0 = 0.5f*((rhol*ul + rhor*ur) - f1);
  *f_1 = 0.5f*((rhol*ul*ul + pl + rhor*ur*ur + pr) - f2);
  *f_2 = 0.5f*(ul*hl*rhol + ur*hr*rhor - f3);
}