Closed ye-luo closed 5 years ago
What is the motivation for splitting the throughput into diffusion and PP parts? The determinant update in the diffusion part has the N^3 scaling, so putting the N^3 factor into the diffusion throughput makes sense. Putting an N^3 factor in the PP throughput makes it consistent mathematically, but I'm not sure it is meaningful on it's own?
Mathematically NLPP scales as Nelec^2. Fixed printout.
I got asked why miniQMC runs slower with OpenMP threads than serial. So I think we need to print out throughput.
Now the following is printed: