Looking at the the FLOP calculations in the harmonic bond calculation, it looks like some of the numbers may be off (unless I am missing something, like is it assumed that the compiler is smart enough to not do two identical calculations). So the numbers being fed to the profiler is slightly inaccurate.
//I count 15, not 14
// FLOPS: 14 / MEM TRANSFER: 2 Scalars
Scalar rsq = dx*dx+dy*dy+dz*dz; // 5
Scalar r = sqrt(rsq); //1
Scalar forcemag_divr = m_K[bond.type] * (m_r_0[bond.type] / r - Scalar(1.0)); //3
Scalar bond_eng = Scalar(0.5) * Scalar(0.5) * m_K[bond.type] * (m_r_0[bond.type] - r) * (m_r_0[bond.type] - r); //6
Carolyn - does the 1.0/6.0 not count? I count 3 FLOPS.
// calculate the virial (FLOPS: 2)
Scalar bond_virial = -Scalar(1.0/6.0) * rsq * forcemag_divr;
