Gravity - hyschive/gamer-fork GitHub Wiki

Compilation Options

Related options: --gravity,   --pot_scheme,   --store_pot_ghost,   --unsplit_gravity,   --comoving  

Runtime Parameters

Parameters described on this page: OPT__BC_POT,   GFUNC_COEFF0,   NEWTON_G,   SOR_OMEGA,   SOR_MAX_ITER,   SOR_MIN_ITER,   MG_MAX_ITER,   MG_NPRE_SMOOTH,   MG_NPOST_SMOOTH,   MG_TOLERATED_ERROR,   OPT__GRA_P5_GRADIENT,   OPT__SELF_GRAVITY,   OPT__EXT_ACC,   OPT__EXT_POT,   EXT_POT_TABLE_NAME,   EXT_POT_TABLE_NPOINT_X,   EXT_POT_TABLE_NPOINT_Y,   EXT_POT_TABLE_NPOINT_Z,   EXT_POT_TABLE_DH,   EXT_POT_TABLE_EDGEL_X,   EXT_POT_TABLE_EDGEL_Y,   EXT_POT_TABLE_EDGEL_Z

Other related parameters: DT__GRAVITY,   POT_GPU_NPGROUP  

Parameters below are shown in the format:   Name   (Valid Values)   [Default Value]

  • GFUNC_COEFF0   (≥0.0; <0.0 → set to default)   [depend]

    • Description: Green's function coefficient for calculating the "self" gravitational potential of each cell (i.e., the gravitational potential contribution from the mass within the same cell). The default value depends on the particle interpolation scheme (PAR_INTERP). Set GFUNC_COEFF0=0.0 to exclude the self-potential.
    • Restriction: Only applicable to the isolated boundary condition.

  • NEWTON_G   (>0.0)   [conform to the unit system set by OPT__UNIT]

    • Description: Gravitational constant.
    • Restriction: It will be overwritten by the default value when OPT__UNIT is on; no default when OPT__UNIT is off.

  • SOR_OMEGA   (≥0.0; <0.0 → set to default)   [1.69]

    • Description: Parameter of the SOR Poisson solver for optimizing the convergence rate.
    • Restriction: Only applicable when adopting the compilation option --pot_scheme=SOR.

  • SOR_MAX_ITER   (≥0; <0 → set to default)   [single precision=60, double precision=110]

    • Description: Maximum number of iterations in the SOR Poisson solver. The default value depends on the adopted floating-point accuracy (FLOAT8).
    • Restriction: Only applicable when adopting the compilation option --pot_scheme=SOR.

  • SOR_MIN_ITER   (≥3; <0 → set to default)   [10]

    • Description: Minimum number of iterations in the SOR Poisson solver.
    • Restriction: Only applicable when adopting the compilation option --pot_scheme=SOR.

  • MG_MAX_ITER   (≥0; <0 → set to default)   [single precision=10, double precision=20]

    • Description: Maximum number of iterations in the multigrid Poisson solver. The default value depends on the adopted floating-point accuracy (--double).
    • Restriction: Only applicable when adopting the compilation option --pot_scheme=MG.

  • MG_NPRE_SMOOTH   (≥0; <0 → set to default)   [3]

    • Description: Number of pre-smoothing steps in the multigrid Poisson solver.
    • Restriction: Only applicable when adopting the compilation option --pot_scheme=MG.

  • MG_NPOST_SMOOTH   (≥0; <0 → set to default)   [3]

    • Description: Number of post-smoothing steps in the multigrid Poisson solver.
    • Restriction: Only applicable when adopting the compilation option --pot_scheme=MG.

  • MG_TOLERATED_ERROR   (≥0.0; <0.0 → set to default)   [single precision=1e-6, double precision=1e-15]

    • Description: Maximum tolerable error in the multigrid Poisson solver.
    • Restriction: Only applicable when adopting the compilation option --pot_scheme=MG.

  • OPT__GRA_P5_GRADIENT   (0=off, 1=on)   [0]

    • Description: Use 5-point instead of 3-point gradient for calculating the gravitational acceleration from potential. This functionality is only experimental.
    • Restriction: Must manually set #define GRA_GHOST_SIZE 2 (and #define USG_GHOST_SIZE 2 as well when adopting the compilation option --unsplit_gravity) in the header file Macro.h. Unsupported for particle update.

  • OPT__SELF_GRAVITY   (0=off, 1=on)   [1]

    • Description: Enable self-gravity.
    • Restriction:

  • OPT__EXT_ACC   (0=off, 1=on)   [0]

    • Description: Add external acceleration. See External Acceleration/Potential for how to specify external acceleration.
    • Restriction: Not applicable to the wave dark matter simulations (--model=ELBDM).

  • OPT__EXT_POT   (0=off, 1=function, 2=table)   [0]

  • EXT_POT_TABLE_NAME   (string)   [none]

  • EXT_POT_TABLE_NPOINT_X   (>1)   [none]

    • Description: For OPT__EXT_POT=2: table size (i.e., number of data points) along the x direction.
    • Restriction:

  • EXT_POT_TABLE_DH   (>0.0)   [none]

    • Description: For OPT__EXT_POT=2: spatial interval between adjacent tabular data.
    • Restriction:

  • EXT_POT_TABLE_EDGEL_X   (cover the simulation domain plus ghost zones)   [none]

    • Description: For OPT__EXT_POT=2: starting x coordinate of the tabular data.
    • Restriction: Must cover the simulation domain plus gravity ghost zones. Specifically, EXT_POT_TABLE_EDGEL_X must be smaller or equal to DomainLeftEdge - 1*RootLevelCell and EXT_POT_TABLE_EDGEL_X + (EXT_POT_TABLE_NPOINT_X-1)*EXT_POT_TABLE_DH must be larger or equal to DomainRightEdge + 1*RootLevelCell.

  • EXT_POT_TABLE_EDGEL_Y   (cover the simulation domain)   [none]

  • EXT_POT_TABLE_EDGEL_Z   (cover the simulation domain)   [none]

Remarks

External Acceleration/Potential

Using Analytical Function:

Follow the steps below to add external acceleration with an analytical function. Get familiar with the general procedure of Adding Problem-specific Functionalities first and be aware that adding external acceleration/potential takes a few extra steps in order to utilize GPUs.

  1. Enable OPT__EXT_ACC.
  2. Go to your new test problem folder and copy the built-in point-mass acceleration routine as a template: 
    cp ../../../SelfGravity/CPU_Gravity/CPU_ExtAcc_PointMass.cpp ExtAcc_NewProblem.cpp
  3. Edit ExtAcc_NewProblem.cpp to
    • Replace the keyword PointMass by NewProblem (or whatever appropriate).
    • Edit the routine SetExtAccAuxArray_NewProblem() to set the auxiliary array AuxArray[], which will be automatically passed to ExtAcc_NewProblem() as the input array UserArray[]. The array size is set by EXT_ACC_NAUX_MAX in include/Macro.h (default 20).
    • Edit the routine ExtAcc_NewProblem(), which should return the external acceleration Acc[] at (x, y, z, Time).
  4. Link ExtAcc_NewProblem.cpp to a corresponding CUDA source file: 
    ln -s ExtAcc_NewProblem.cpp ExtAcc_NewProblem.cu
    This avoids writing redundant codes for CPU and GPU runs separately.
  5. Edit your problem source file (e.g., Init_TestProb_Hydro_NewProblem.cpp) to
    • Declare a function prototype on top of the file: 
      void Init_ExtAcc_NewProblem();
    • Set up a function pointer in your problem initialization function (e.g., Init_TestProb_Hydro_NewProblem()): 
      # ifdef GRAVITY
  Init_ExtAcc_Ptr = Init_ExtAcc_NewProblem
# endif

Adding external potential can be done in a very similar way. Enable OPT__EXT_POT=1 and then follow the procedure above and replace Acc by Pot. A point-mass example can be found at src/SelfGravity/CPU_Poisson/CPU_ExtPot_PointMass.cpp.

Caution: wave dark matter simulations (i.e., --model=ELBDM) does not support external acceleration. Just use external potential.

Using Table:

External potential can also be set via interpolating a table loaded from disk. Set OPT__EXT_POT=2 and all EXT_POT_TABLE_* parameters. The following C++ example creates such a table with a point-mass potential.

#include <cstdio>
#include <cmath>

typedef double real; // this must be consistent with the compilation option FLOAT8

int main()
{
// parameters
   const char  *Filename = "ExtPotTable";       // EXT_POT_TABLE_NAME
   const int    NX       = 131;                 // EXT_POT_TABLE_NPOINT_X
   const int    NY       = 131;                 // EXT_POT_TABLE_NPOINT_Y
   const int    NZ       = 131;                 // EXT_POT_TABLE_NPOINT_Z
   const double dh       = 0.0078125;           // EXT_POT_TABLE_NPOINT_DH
   const double EdgeL[3] = { -dh, -dh, -dh };   // EXT_POT_TABLE_EDGEL_X/Y/Z

// set the external potential array
   const double G  = 1.0;                       // gravitational constant
   const double M  = 1.0;                       // particle mass
   const double cx = EdgeL[0] + 0.5*(NX-1)*dh;  // centre
   const double cy = EdgeL[1] + 0.5*(NY-1)*dh;
   const double cz = EdgeL[2] + 0.5*(NZ-1)*dh;

   double dx, dy, dz, r;

   real (*ExtPot)[NY][NX] = new real [NZ][NY][NX];

   for (int k=0; k<NZ; k++)  {  dz = EdgeL[2] + k*dh - cz;
   for (int j=0; j<NY; j++)  {  dy = EdgeL[1] + j*dh - cy;
   for (int i=0; i<NX; i++)  {  dx = EdgeL[0] + i*dh - cx;

      r               = sqrt( dx*dx + dy*dy + dz*dz );
      ExtPot[k][j][i] = -G*M/r;

   }}}

// output
   FILE *File = fopen( Filename, "wb" );
   fwrite( ExtPot, sizeof(real), (long)NX*NY*NZ, File );
   fclose( File );

   delete [] ExtPot;
}

[Optional] To implement a user-specified interpolation routine, use src/SelfGravity/CPU_Poisson/CPU_ExtPot_Tabular.cpp as a template and follow the steps [2-5] in Using Analytical Function.

Caution: external acceleration does not support this functionality yet.


Links

Add a custom footer
⚠️ **GitHub.com Fallback** ⚠️