[flo90000]

Hello

I have an error "Process 60025:received signal SIGSEV" . I find it's due to a program try to access a memory location which is not allowed. So I allow a User defined memory location and a user defined node memory location but without success. It's my UDF:

 #include "udf.h"
 #include "unsteady.h"
 #include "math.h"
 
DEFINE_GRID_MOTION(position_piston,domain,dt,time,dtime)
 {
	#if PARALLEL
        static Thread *tf;
	static face_t f;
	static Node *v;
	static int n;
	static real t;
	static real L,A,WS;
	L = 0.06;  /* m Connecting rod lenght*/
        A = 0.054; /* m piston stroke */
	WS = 60; /* rpm Crank shaft speed */
	t = CURRENT_TIME;
	begin_f_loop(f,tf)
	{
		f_node_loop(f,tf,n)
		 {
			 v = F_NODE(f,tf,n);
			 
			 if (NODE_POS_NEED_UPDATE (v)) /*the current node has not been previously visited when looping through previous faces */
			 {				 
			   NODE_Y(v)=L+((A/2)*(1-cos(t*WS*M_PI/30)))-sqrt((L*L)-((A*A)*(sin(t*WS*M_PI/30)*sin(t*WS*M_PI/30))/4));
			 }
		 }
	 }
	 end_f_loop(f,tf);
	 #endif
 }

thanks for your help

[`e`]

You haven't used user-defined memory and therefore don't require allocating memory for UDM. Why are you restricting the DEFINE_GRID_MOTION macro to parallel (with the #if PARALLEL preprocessor directive)? Are you running Fluent in serial or parallel? The host process would still execute the code; the only thing this conditional statement is achieving is avoiding mesh motion in serial mode. Face looping macros are skipped over by the host node anyway; you don't require any special parallel considerations for this macro.

If you're checking if the node has already been updated, then you should tell Fluent that it has been updated (shouldn't practically matter as the node movement is absolute and not dependent on its current position):

if (NODE_POS_NEED_UPDATE (v)) /*the current node has not been previously visited when looping through previous faces */
{				 
	NODE_Y(v)=L+((A/2)*(1-cos(t*WS*M_PI/30)))-sqrt((L*L)-((A*A)*(sin(t*WS*M_PI/30)*sin(t*WS*M_PI/30))/4));
	NODE_POS_UPDATED(v);
}

You've declared a face thread pointer (*tf) but haven't initialised this value: that's the cause of your error. The face thread is given by DEFINE_GRID_MOTION via dt and can be accessed with:

Thread *tf = DT_THREAD((Dynamic_Thread *)dt);

Lastly, add trailing dots to your integers to avoid erroneous type casting operations.

[flo90000]

Originally Posted by `e`

You haven't used user-defined memory and therefore don't require allocating memory for UDM. Why are you restricting the DEFINE_GRID_MOTION macro to parallel (with the #if PARALLEL preprocessor directive)? Are you running Fluent in serial or parallel? The host process would still execute the code; the only thing this conditional statement is achieving is avoiding mesh motion in serial mode. Face looping macros are skipped over by the host node anyway; you don't require any special parallel considerations for this macro.

If you're checking if the node has already been updated, then you should tell Fluent that it has been updated (shouldn't practically matter as the node movement is absolute and not dependent on its current position):

if (NODE_POS_NEED_UPDATE (v)) /*the current node has not been previously visited when looping through previous faces */
{				 
	NODE_Y(v)=L+((A/2)*(1-cos(t*WS*M_PI/30)))-sqrt((L*L)-((A*A)*(sin(t*WS*M_PI/30)*sin(t*WS*M_PI/30))/4));
	NODE_POS_UPDATED(v);
}

You've declared a face thread pointer (*tf) but haven't initialised this value: that's the cause of your error. The face thread is given by DEFINE_GRID_MOTION via dt and can be accessed with:

Thread *tf = DT_THREAD((Dynamic_Thread *)dt);

Lastly, add trailing dots to your integers to avoid erroneous type casting operations.

I run the simulation in parallel with 16 processes and I found some compiler directives to write in the UDF to parallelize it. That's why I put "#if PARALLEL". I change how you suggested me and tried with the three compiler directives. And I have another error when I want to load the UDF library. Fluent tells me that the UDF library I am trying to load is not compiled for parallel use.

[`e`]

Which three compiler directives are you using now? You don't need to use any for this UDF code. Read the UDF manual on parallelising if you're interested. Ensure you're recompiling correctly by checking my compiling process for Fluent UDFs with an existing library loaded.