Gerris Flow Solver Programming Course for Dummies

From Gerris

(Difference between revisions)
Jump to: navigation, search
Revision as of 23:20, 30 October 2007
Zaleski (Talk | contribs)
(Answer to the exercise)
← Previous diff
Revision as of 23:25, 30 October 2007
Zaleski (Talk | contribs)
(Various Gerris techniques - removed a duplicated section)
Next diff →
Line 731: Line 731:
To be written. To be written.
-=== Various Gerris techniques === 
-==== Cast to the parent class ==== 
- 
-The example shows how one can extract information from an object if one knows in which 
-parent it is stored. For example, if we want to use the information of the event variable 
-which came from the GfsEvent class, we can have access to it by means 
-of the following variable "parentdata": 
- 
-<source c> 
-GfsEvent * parentdata = GFS_EVENT (event); 
-</source c> 
- 
-In this way, we load the information of the "event" object coming from 
-its parent GfsEvent into this variable and we can either use it or 
-manipulate if we want. For example 
-<code> parentdata->step </code> will give us the temporal step which have been 
-read in the input. 
- 
-==== <code>GFS_VARIABLE</code> macro ==== 
- 
-This macro is used everywhere in the code to access the value of a GFS_VARIABLE in a cell. An example of its use is found in the ''Daniel'' object above:  
-<source c> 
-GFS_VARIABLE (cell, v->i) = sin (2.*(*m)*M_PI*pos.x)*cos (2.*(*m)*M_PI*pos.y); 
-</source c> 
-This expression replaces 
-<source c> 
-GFS_STATE (cell)->v = sin (2.*init->m*M_PI*pos.x)*cos (2.*init->m*M_PI*pos.y); 
-</source c> 
-which was used in the tutorial.  
- 
-To use this macro, there are two points of view.  
- 
-* In the "magical" approach we just now that <code>GFS_VARIABLE(somecell,somevariable->i)</code>allows us to write to the variable <code>somevariable</code>in the cell <code>somecell</code>. For us ''dummies'' a question may arise: what value should the index ''i'' have ? Is it the cell number ? Well ''i'' is not an index: it is a member of ''each'' GfsVariable. We do not have to set it: it is already there in the variable.  
- 
-* In the rational approach, we would like to read the code and understand the definition of <code>GFS_VARIABLE</code>. However the definition of the <code>GFS_VARIABLE</code> macro is a bit baffling. It can be found in <code>fluid.h</code>. (It is a good idea to try the following: find any place in the code where the <code>GFS_VARIABLE</code> macro is used ---that should be easy--- and then click on it. Type M-. in emacs ---or the equivalent in vim--- and get to the definition.) 
-<source c> 
-#define GFS_VARIABLE(cell, index) ((&GFS_STATE (cell)->place_holder)[index]) 
-</source c> 
-Notice that we have a special member of the cell state that tells us where the variable is stored. The macro <code>GFS_STATE</code> is defined by 
-<source c> 
-#define GFS_STATE(cell) ((GfsStateVector *) (cell)->data) 
-</source c> 
- 
-==== The <code> FttCell</code> class ==== 
- 
-This class is defined as follows:  
-<source c> 
-struct _FttCell { 
- /*< public >*/ 
- guint flags; 
- gpointer data; 
- 
- /*< private >*/ 
- struct _FttOct * parent, * children; 
-}; 
-</source c> 
- 
-The data member of this class is a gpointer. Remember from last session that it is 
-an untyped pointer, so there is no difficulty to cast to a <code>GfsStateVector</code> 
-by using 
-<source c> 
-struct _GfsStateVector { 
- /* temporary face variables */ 
- GfsFaceStateVector f[FTT_NEIGHBORS]; 
- 
- /* solid boundaries */ 
- GfsSolidVector * solid; 
- 
- gdouble place_holder; 
-}; 
-</source c> 
- 
-To be written. 
=== Various Gerris techniques === === Various Gerris techniques ===

Revision as of 23:25, 30 October 2007

Contents

Preamble


This course material is about Gerris, a general-purpose fluid mechanics code developped by Stephane Popinet at NIWA, Wellington, New Zealand. Gerris is a free, GPL-licensed, open source code available at [http://gfs.sf.net] .

The intended audience is typical first-year science or engineering graduate students with either very little experience of C or with some Fortran knowledge, but willing to work hard and learn. The student should know simple C data types, pointers and functions but not structures.

The course is being taught as of writing (october 2007) in Paris. In the actual course a lot of talking is done in addition to the material here. Each session is 30 minutes + 15 minutes of questions.

Session 1


Introduction


This course is about how Gerris ([[1]] ) is programmed. It is intended to help in understanding the Gerris source code and learning how to modify it usefully.

This course is not about the numerical methods in Gerris (however it would be good for students of this course to learn about them, for instance in the J. Comput. Phys. article) .

Gerris is closely linked to Gts, the GNU triangulated surface library, also written by Stephane Popinet

Gerris and Gts are programmed in a style analogous to that of Glib, Gnome and GTK . It is a style of C programming that offers several advantages:

  • Most aspects of Object-Oriented Programming (OOP) , such as the existence of classes with their own methods and inheritance.
  • The ability to interface to other programming languages. (As far as I know, this feature is not used in Gerris/Gts)

To implement Object-Oriented Programming, Gerris/Gts uses its own “Object system”. This system is analogous to the Glib object system (Gobject), but not identical to it. Learning more about Gobject can be very useful.

References and reading material

  • Introduction and in-depth discussion of the Gobject system (in order of decreasing ease of reading):
  1. [http://en.wikipedia.org/wiki/GObject]
  2. [http://docs.programmers.ch/index.php/HOWTO_gobject]
  3. [http://library.gnome.org/devel/gobject/stable/index.html]
  • Introduction to the Gerris/Gts object system :

You can find one in the [Gerris Tutorial, Section 5.1 on Objects].

  • Introduction to C :

No preferences here. Use a recent edition of a good text and/or manual. Read the part about structures and pointers to structures.

Useful tips


Navigate using the TAGS file in emacs or vim

- Create the TAGS file :
% cd src
% make tags
  1. In emacs:
    1. open any *.c or *.h file with emacs.
    2. Position the cursor on a function or variable.
    3. do ESC . to find its definition. (or M-. using the emacs Meta key convention )
    4. do ESC * to return to the previous location (s) .
  2. In vim:
    1. to be written ...


No order in which to read the code

There is no good order in which to read the code (I have not found it yet).


Keep a C precedence and associativity table nearby

A lot of macros and functions such g_assert come from the Glib. Keep a bookmark to the Glib documentation: [[2]]


C basics

Introduction to Structures


struct Point {
char name;
double x, y;
};

An example of usage

main()
{
struct Point my_point; /* declaration */
 
my_point.x = 0. ;
my_point.y = 1.;
my_point.name = ‘A’;
}

name, x and y are members of the structure of type “Point” called my_point. We also give it a name that can be exported (printed, passed to other functions) as a character. This example shows why it is useful to use structures to store several relevant informations or data together.

An example: the structure GfsNorm in Gerris

struct _GfsNorm {
gdouble bias, first, second, infty, w;
};
typedef struct _GfsNorm GfsNorm

From domain.h.

GfsNorm gfs_domain_norm_residual (GfsDomain * domain,
FttTraverseFlags flags,
gint max_depth,
gdouble dt,
GfsVariable * res)
{
GfsNorm n;
gpointer data[2];
 
g_return_val_if_fail (domain != NULL, n);
g_return_val_if_fail (res != NULL, n);
 
gfs_norm_init (&n);
data[0] = res;
data[1] = &n;
gfs_domain_cell_traverse (domain, FTT_PRE_ORDER, flags, max_depth,
(FttCellTraverseFunc) add_norm_residual, data);
#ifdef HAVE_MPI
domain_norm_reduce (domain, &n);
#endif /* HAVE_MPI */
gfs_norm_update (&n);
 
dt *= dt;
n.bias *= dt;
n.first *= dt;
n.second *= dt;
n.infty *= dt;
return n;
}

From domain.c

Notice the use of

  • glib basic types gdouble, gpointer
  • glib functions g_return_val_if_fail


Note

This is what the [glib documentation] says about gpointer

typedef void* gpointer;

An untyped pointer. gpointer looks better and is easier to use than void*.

Structures that are related to other structures

More complex relationships between structures create the need for inheritance.

Below one defines two structures which have some similarities. The Square is a kind of Quadrangle.

struct _Quadrangle {
Point A, B, C, D;
};
 
typedef struct _Quadrangle Quadrangle

Notice that we use typedef: this creates an "alias" for the structure name and will make it able to refer to itself in its list of member names.

/* Related structure */
 
struct _Square {
Point A, B, C, D;
};
 
typedef struct _Square Square

In the following pseudo-code we use the structures, for instance we define a function that computes the area.

area_Quadrangle (Quadrangle X )
{
Triangle ABC;
ABC = create_triangle(X.A,X.B,X.C);

return area_Triangle (ABC) + area_Triangle (CDA)  ;
}

If we use squares, we may need a very similar function to compute the area of squares.

area_Square ( Square Y )
{

return area_Triangle (ABC) + area_Triangle (CDA)  ;
}


Elementary classes

  • We want to avoid such repetitions.
  • Notion of parent structure: Quadrangle is the parent of Square.

One shall from now on use the “class” terminology. A class is a structure with some functions attached to it, the functions are called methods. This should become clearer in what follows.

Simple example : structures with inheritance

struct _Square { 
Quadrangle parent;
double longueur_du_cote;
};
 
typedef struct _Square Square
 
main()
{
Square MySquare ; /* Déclaration of Square */
double area;
 
Initialize_Square(&MySquare) ; /*Initialisation */
area = area_Square ( MySquare ) ;
}

Problem: for each class we need a new function giving the area area_Square, area_Quadrangle, etc… So we will define the function "area_square" quite simply from "area_quadrangle". Functions that apply to a class are called the methods of the class.

In C++ the methods are listed when the class is defined. In the Gerris/Glib Object system the situation is somewhat different. A structure like the above "Square" structure is created. Then a separated C structure is created, the class structure that contains the methods and other information about the class.


This is the skeleton of such a class in C:

struct _Square { 
Quadrangle parent;
double side_length;
};
 
typedef struct _Square Square
typedef struct _SquareClass SquareClass
 
struct _SquareClass {
/*< private >*/
QuadrangleClass parent_class;
 
/*< public >*/
/* add extra methods here */
};

The gtstemplate tool
  • Automatic class creation: Usage of the Gts library utility function gtstemplate to create a Gts Object :
% gtstemplate
Usage: gtstemplate [OPTIONS] Class ParentClass
Options:
        [--no-extra-data]
        [--no-extra-method]
        [--overload=METHOD]
% gtstemplate Square Quadrangle

Here is the result:

/* Square: Header */
 
typedef struct _Square Square;
 
struct _Square {
/*< private >*/
Quadrangle parent;
 
/*< public >*/
/* add extra data here (if public) */
};
 
typedef struct _SquareClass SquareClass;
 
struct _SquareClass {
/*< private >*/
QuadrangleClass parent_class;
 
/*< public >*/
/* add extra methods here */
};
 
#define SQUARE(obj) GTS_OBJECT_CAST (obj,\
Square,\
Square_class ())
#define SQUARE_CLASS(klass) GTS_OBJECT_CLASS_CAST (klass,\
SquareClass,\
Square_class())
#define IS_SQUARE(obj) (gts_object_is_from_class (obj,\
Square_class ()))
 
SquareClass * Square_class (void);
Square * Square_new (SquareClass * klass);
 
/* Square: Object */
 
static void Square_class_init (SquareClass * klass)
{
/* define new methods and overload inherited methods here */
 
}
static void Square_init (Square * object)
{
/* initialize object here */
}
 
SquareClass * Square_class (void)
{
static SquareClass * klass = NULL;
 
if (klass == NULL) {
GtsObjectClassInfo Square_info = {
"Square",
sizeof (Square),
sizeof (SquareClass),
(GtsObjectClassInitFunc) Square_class_init,
(GtsObjectInitFunc) Square_init,
(GtsArgSetFunc) NULL,
(GtsArgGetFunc) NULL
};
klass = gts_object_class_new (GTS_OBJECT_CLASS (Quadrangle_class ()),
&Square_info);
}
 
return klass;
}
 
Square * Square_new (SquareClass * klass)
{
Square * object; object = Square (gts_object_new (GTS_OBJECT_CLASS (klass)));
 
 
 
return object;
}
A simpler, actual example

Typical class usage in Gerris is less complex. An actual example from Gerris: the SourceScalar class from source.h :

/* GfsSourceScalar: Header */
 
typedef struct _GfsSourceScalar GfsSourceScalar;
 
struct _GfsSourceScalar {
/*< private >*/
GfsSourceGeneric parent;
 
/*< public >*/
GfsVariable * v;
};
 
#define GFS_SOURCE_SCALAR(obj) GTS_OBJECT_CAST (obj,\
GfsSourceScalar,\
gfs_source_scalar_class ())
#define GFS_IS_SOURCE_SCALAR(obj) (gts_object_is_from_class (obj,\
gfs_source_scalar_class ()))
 
GfsSourceGenericClass * gfs_source_scalar_class (void);

Notice that we do not use all the possibilities in the template. The class structure is a SourceGenericClass, there is no need to create a specific SourceScalarClass .

Object manipulation: the read method

Many Gfs objects have a read method. From source.c

static void source_scalar_read (GtsObject ** o, GtsFile * fp)
{
GfsSourceScalar * source;
GfsDomain * domain;
 
if (GTS_OBJECT_CLASS (gfs_source_scalar_class ())->parent_class->read)
(* GTS_OBJECT_CLASS (gfs_source_scalar_class ())->parent_class->read)
(o, fp);
if (fp->type == GTS_ERROR)
return;

This is standard Gerris “read” code that uses the parent class read method to read first the parameters that are defined in the parent class. It will initialize the variables of the parent object.

source = GFS_SOURCE_SCALAR (*o);

This “casts” the “generic” object **o into an object of type “source_scalar”.

domain =  GFS_DOMAIN (gfs_object_simulation (source));
if (fp->type != GTS_STRING) {
gts_file_error (fp, "expecting a string (GfsVariable)");
return;
}
source->v = gfs_variable_from_name (domain->variables,
fp->token->str);
if (source->v == NULL) {
gts_file_error (fp, "unknown variable `%s'", fp->token->str);
return;
}

We expect a string: the name of the variable. If the name of the variable is not found in the list of domain gfs_variables, an error results.

If it is found, the member v of the structure source becomes this gfs_variable.

if (source->v->sources == NULL)
source->v->sources =
gts_container_new (GTS_CONTAINER_CLASS (gts_slist_container_class ()));
gts_container_add (source->v->sources, GTS_CONTAINEE (source));

The variable then aquires a new source in its list of sources v->sources . gts_containers are defined in the gts library.

gts_file_next_token (fp);
}

Reading assignements and exercises

Session 2

Answer to the exercise

Here is an example class written by Daniel Fuster. The beginning looks like the tutorial and the class generated by gtstemplate, but there is a difference. Can you see it ?

/* Daniel: Header */
#include <math.h>
#include <stdlib.h>
#include <gfs.h>
 
typedef struct _Daniel Daniel;
 
struct _Daniel {
/*< private >*/
GfsGenericInit parent;
 
/*< public >*/
gdouble m;
};
 
typedef struct _DanielClass DanielClass;
 
struct _DanielClass {
 
GfsGenericInitClass parent_class;
/*< public >*/
/* add extra methods here */
};
 
#define DANIEL(obj) GTS_OBJECT_CAST (obj,\
Daniel,\
daniel_class ())
#define DANIEL_CLASS(klass) GTS_OBJECT_CLASS_CAST (klass,\
DanielClass,\
daniel_class())
#define IS_DANIEL(obj) (gts_object_is_from_class (obj,\
daniel_class ()))
 
DanielClass * daniel_class (void);

Did you notice the change ? We are using GfsGenericInit as the parent class of Daniel, not GfsInit. Can you explain why ? Now we define the read, write etc.. functions

/* Daniel: Object */
 
static void daniel_read (GtsObject ** o, GtsFile * fp)
{
/* call read method of parent */
if (GTS_OBJECT_CLASS (daniel_class ())->parent_class->read)
(* GTS_OBJECT_CLASS (daniel_class ())->parent_class->read)
(o, fp);
if (fp->type == GTS_ERROR)
return;
 
if (fp->type != GTS_INT && fp->type != GTS_FLOAT) {
gts_file_error (fp, "expecting a number (m)");
return;
}
DANIEL (*o)->m = atof (fp->token->str);
 
/* do not forget to prepare for next read */
gts_file_next_token (fp);
 
}
 
static void daniel_write (GtsObject * o, FILE * fp)
{
/* call write method of parent */
if (GTS_OBJECT_CLASS (daniel_class ())->parent_class->write)
(* GTS_OBJECT_CLASS (daniel_class ())->parent_class->write)
(o, fp);
 
fprintf (fp, " %g", DANIEL (o)->m);
 
/* do object specific write here */
}

Now we have a problem following the tutorial. The tutorial says:

static void init_velocity (FttCell * cell,
InitPeriodic * init)
{
FttVector pos;
 
ftt_cell_pos (cell, &pos);
GFS_STATE (cell)->u =
- cos (2.*init->m*M_PI*pos.x)*sin (2.*init->m*M_PI*pos.y);
GFS_STATE (cell)->v =
sin (2.*init->m*M_PI*pos.x)*cos (2.*init->m*M_PI*pos.y);
}
 
static gboolean init_periodic_event (GfsEvent * event, GfsSimulation * sim)
{
if ((* GFS_EVENT_CLASS (GTS_OBJECT_CLASS (init_periodic_class ())\
->parent_class)->event) (event, sim)) {
gfs_domain_cell_traverse (GFS_DOMAIN (sim),
FTT_PRE_ORDER, FTT_TRAVERSE_LEAFS, -1,
(FttCellTraverseFunc) init_velocity,
event);
return TRUE;
}
return FALSE;
}

but when we try to do something similar it does not work. The problem is with passing the event object to the init_velocity function. Daniel fixed it as follows:

static void init_velocity (FttCell * cell,
gpointer * data )
{
FttVector pos;
GfsVariable * v;
gdouble * m = data[0];
 
ftt_cell_pos (cell, &pos);
v = data[1];
GFS_VARIABLE (cell, v->i) = sin (2.*(*m)*M_PI*pos.x)*cos (2.*(*m)*M_PI*pos.y);
}
 
static gboolean daniel_event (GfsEvent * event, GfsSimulation * sim)
{
gpointer data[2];
Daniel * inputdata = DANIEL (event);
 
if ((* GFS_EVENT_CLASS (GTS_OBJECT_CLASS (daniel_class ())->parent_class)->event) (event, sim)) {
/* do object-specific event here */
 
GfsDomain * domain = GFS_DOMAIN (sim);
 
data[0] = &(inputdata->m);
 
data[1] = gfs_variable_from_name (domain->variables, "U");
 
gfs_domain_cell_traverse (GFS_DOMAIN (sim),
FTT_PRE_ORDER, FTT_TRAVERSE_LEAFS, -1,
(FttCellTraverseFunc) init_velocity,
data);
return TRUE;
}
return FALSE;
}

This works, and is the way things are usually written in Gerris.

static void daniel_class_init (DanielClass * klass)
{
/* define new methods and overload inherited methods here */
 
GFS_EVENT_CLASS (klass)->event = daniel_event;
GTS_OBJECT_CLASS (klass)->read = daniel_read;
GTS_OBJECT_CLASS (klass)->write = daniel_write;
}
 
static void daniel_init (Daniel * object)
{
/* initialize object here */
object->m = 1.;
}
 
DanielClass * daniel_class (void)
{
static DanielClass * klass = NULL;
 
if (klass == NULL) {
GtsObjectClassInfo daniel_info = {
"Daniel",
sizeof (Daniel),
sizeof (DanielClass),
(GtsObjectClassInitFunc) daniel_class_init,
(GtsObjectInitFunc) daniel_init,
(GtsArgSetFunc) NULL,
(GtsArgGetFunc) NULL
};
klass = gts_object_class_new (GTS_OBJECT_CLASS (gfs_init_class ()),
&daniel_info);
}
 
return klass;
}
/* Initialize module */
 
const gchar * g_module_check_init (void)
{
daniel_class ();
return NULL;
}

Various Gerris techniques

Cast to the parent class

The example shows how one can extract information from an object if one knows in which parent it is stored. For example, if we want to use the information of the event variable which came from the GfsEvent class, we can have access to it by means of the following variable "parentdata":

GfsEvent * parentdata = GFS_EVENT (event);

In this way, we load the information of the "event" object coming from its parent GfsEvent into this variable and we can either use it or manipulate if we want. For example parentdata->step will give us the temporal step which have been read in the input.

GFS_VARIABLE macro

This macro is used everywhere in the code to access the value of a GFS_VARIABLE in a cell. An example of its use is found in the Daniel object above:

GFS_VARIABLE (cell, v->i) = sin (2.*(*m)*M_PI*pos.x)*cos (2.*(*m)*M_PI*pos.y);

This expression replaces

GFS_STATE (cell)->v = sin (2.*init->m*M_PI*pos.x)*cos (2.*init->m*M_PI*pos.y);

which was used in the tutorial.

To use this macro, there are two points of view.

  • In the "magical" approach we just now that GFS_VARIABLE(somecell,somevariable->i)allows us to write to the variable somevariablein the cell somecell. For us dummies a question may arise: what value should the index i have ? Is it the cell number ? Well i is not an index: it is a member of each GfsVariable. We do not have to set it: it is already there in the variable.
  • In the rational approach, we would like to read the code and understand the definition of GFS_VARIABLE. However the definition of the GFS_VARIABLE macro is a bit baffling. It can be found in fluid.h. (It is a good idea to try the following: find any place in the code where the GFS_VARIABLE macro is used ---that should be easy--- and then click on it. Type M-. in emacs ---or the equivalent in vim--- and get to the definition.)
#define GFS_VARIABLE(cell, index)     ((&GFS_STATE (cell)->place_holder)[index])

Notice that we have a special member of the cell state that tells us where the variable is stored. The macro GFS_STATE is defined by

#define GFS_STATE(cell)               ((GfsStateVector *) (cell)->data)

The FttCell class

This class is defined as follows:

struct _FttCell {
/*< public >*/
guint flags;
gpointer data;
 
/*< private >*/
struct _FttOct * parent, * children;
};

The data member of this class is a gpointer. Remember from last session that it is an untyped pointer, so there is no difficulty to cast to a GfsStateVector by using

struct _GfsStateVector {
/* temporary face variables */
GfsFaceStateVector f[FTT_NEIGHBORS];
 
/* solid boundaries */
GfsSolidVector * solid;
 
gdouble place_holder;
};

To be written.


Various Gerris techniques

Cast to the parent class

The example shows how one can extract information from an object if one knows in which parent it is stored. For example, if we want to use the information of the event variable which came from the GfsEvent class, we can have access to it by means of the following variable "parentdata":

GfsEvent * parentdata = GFS_EVENT (event);

In this way, we load the information of the "event" object coming from its parent GfsEvent into this variable and we can either use it or manipulate if we want. For example parentdata->step will give us the temporal step which have been read in the input.

GFS_VARIABLE macro

This macro is used everywhere in the code to access the value of a GFS_VARIABLE in a cell. An example of its use is found in the Daniel object above:

GFS_VARIABLE (cell, v->i) = sin (2.*(*m)*M_PI*pos.x)*cos (2.*(*m)*M_PI*pos.y);

This expression replaces

GFS_STATE (cell)->v = sin (2.*init->m*M_PI*pos.x)*cos (2.*init->m*M_PI*pos.y);

which was used in the tutorial.

To use this macro, there are two points of view.

  • In the "magical" approach we just now that GFS_VARIABLE(somecell,somevariable->i)allows us to write to the variable somevariablein the cell somecell. For us dummies a question may arise: what value should the index i have ? Is it the cell number ? Well i is not an index: it is a member of each GfsVariable. We do not have to set it: it is already there in the variable.
  • In the rational approach, we would like to read the code and understand the definition of GFS_VARIABLE. However the definition of the GFS_VARIABLE macro is a bit baffling. It can be found in fluid.h. (It is a good idea to try the following: find any place in the code where the GFS_VARIABLE macro is used ---that should be easy--- and then click on it. Type M-. in emacs ---or the equivalent in vim--- and get to the definition.)
#define GFS_VARIABLE(cell, index)     ((&GFS_STATE (cell)->place_holder)[index])

Notice that we have a special member of the cell state that tells us where the variable is stored. The macro GFS_STATE is defined by

#define GFS_STATE(cell)               ((GfsStateVector *) (cell)->data)

The FttCell class

This class is defined as follows:

struct _FttCell {
/*< public >*/
guint flags;
gpointer data;
 
/*< private >*/
struct _FttOct * parent, * children;
};

The data member of this class is a gpointer. Remember from last session that it is an untyped pointer, so there is no difficulty to cast to a GfsStateVector by using

struct _GfsStateVector {
/* temporary face variables */
GfsFaceStateVector f[FTT_NEIGHBORS];
 
/* solid boundaries */
GfsSolidVector * solid;
 
gdouble place_holder;
};

To be written.

Personal tools
communication