NodalProjector: fix comments (#1396)

Src/LinearSolvers/Projections/AMReX_NodalProjector.H

@@ -7,22 +7,41 @@
 #include <AMReX_MLNodeLaplacian.H>
 #include <AMReX_MLMG.H>
 
+//
+//
+// ***************************  DEFAULT MODE  ***************************
 //
 // Solves
 //
 //     div(sigma*grad(phi)) = div(vel) + S_nd + S_cc
 //
 // and  performs the projection
 //
-//     vel = vel - ( sigma / alpha ) * grad(phi)
+//     vel = vel - sigma * grad(phi)
 //
-// vel, sigma, alpha, and S_cc are cell-centered variables, while
+// vel, sigma, and S_cc are cell-centered variables, while
 // phi and S_nd are nodal-centered variables.
 //
+// ***************************  CUSTOM MODE   ***************************
+//
+// Solves
+//
+//     div(sigma*grad(phi)) = rhs
+//
+// and performs the projection
+//
+//     vel = vel - (sigma/alpha) * grad(phi)
+//
+// alpha is a cell-centered variable, while rhs is nodal-centered.
+//
+// In this mode, the user provides rhs and alpha
+//
 // By default alpha is assumed to be 1. Use setAlpha to change the default.
 //
-// The user can provide a custom RHS instead of letting NodalProjector to compute
-// one. Use setCustomRHS to provide a custom RHS.
+// Use setCustomRHS to provide a custom RHS, else "div(vel) + S_nd + S_cc"
+// is used.
+//
+// Example: rhs = div(alpha*vel)
 //
 namespace amrex {
 

Src/LinearSolvers/Projections/AMReX_NodalProjector.cpp

@@ -241,7 +241,7 @@ NodalProjector::project ( Real a_rtol, Real a_atol )
     // phi comes out already averaged-down and ready to be used by caller if needed
     m_mlmg -> solve( GetVecOfPtrs(m_phi), GetVecOfConstPtrs(m_rhs), a_rtol, a_atol );
 
-    // Get fluxes -- fluxes = -  (alpha/beta) * grad(phi)
+    // Get fluxes -- fluxes = - sigma * grad(phi)
     m_mlmg -> getFluxes( GetVecOfPtrs(m_fluxes) );
 
     // At this time, the fluxes are "correct" only on regions not covered by finer grids.
@@ -258,15 +258,15 @@ NodalProjector::project ( Real a_rtol, Real a_atol )
     {
         if (m_has_alpha)
         {
-            // fluxes -> fluxes/alpha = -grad(phi)/beta
+            // fluxes -> fluxes/alpha = - ( sigma / alpha ) * grad(phi)
             for (int n = 0; n < AMREX_SPACEDIM; ++n)
             {
                 MultiFab::Divide( m_fluxes[lev], *m_alpha[lev], 0, n, 1, 0 );
             }
         }
 
         //
-        // vel = vel + fluxes = vel - grad(phi) / beta
+        // vel = vel + fluxes = vel - ( sigma / alpha ) * grad(phi)
         //
         // Since we already averaged-down the velocity field and -grad(phi),
         // we perform the projection by simply adding the two of them.