NewtonBodyGetAABB

[dehseth]

Can anybody explain NewtonBodyGetAABB function? I am confused with second and third paramerters.

Definition says:
# dFloat *p0 - pointer to an array of at least three floats to hold minimum value for the AABB.
# dFloat *p1 - pointer to an array of at least three floats to hold maximum value for the AABB.

At least And I tried in code it returns 6 floats for each parameter? Well I expect 3 floats (x,y,z), a min point and a max point of box.

How can I define a box using 6 parameters, and is tehere any chance in a different version can return 9 floats instead of 6

[JernejL]

Interesting, it is possible in theory that if overwrites 6 floats (24 bytes) instead of 3 (12 bytes) if the documentation is wrong (let's wait for Julio to respond).

But are you sure you use the correct build of newton? if you use the double precision build, it should write 24 bytes as doubles are 8 bytes long..

[dehseth]

I did not build it. I am using current lib and dll in sdk. There's a dFloat type def which is float and 4 bytes-length. And as I see double prec. is commented out:

Code: Select all

//#define __USE_DOUBLE_PRECISION__
#ifdef __USE_DOUBLE_PRECISION__
   #define dFloat double
#else
   #define dFloat float
#endif

[Julio Jerez]

the function copy the minimu into the three adjecent floats at poiter p0
and copy the thre float for thr max point of teh aabb into thr three adjecent float at p1
it does no override memory, beyond those limits.

[dehseth]

but it does...

here's my code:

Code: Select all

float floorMin[6], floorMax[6];
   NewtonBodyGetAABB(bodyFloor, &floorMin[0], &floorMax[0]);

And here's how memory changed after I called function:

for floorMin: 2.0000000 -2.7519531 0.00000000 -1.5000000 0.00000000 -1.5000000
for flooarMax: 2.0000000 2.7519531 0.00000000 1.5000000 0.00000000 1.5000000

And if I change my arrays size to 3 and run them:

Code: Select all

float floorMin[3], floorMax[3];
   NewtonBodyGetAABB(bodyFloor, &floorMin[0], &floorMax[0]);

On exit I got this message:

Run-Time Check Failure #2 - Stack around the variable 'floorMax' was corrupted.

It is corrupted cause each writes 6 floats...

[Julio Jerez]

that is not possible, this is the function

Code: Select all

void NewtonBodyGetAABB(const NewtonBody* bodyPtr, dFloat* p0, dFloat* p1)   
{
   dgBody *body;

   dgVector vector0;
   dgVector vector1;

   TRACE_FUNTION(__FUNCTION__);
   body = (dgBody *)bodyPtr;
   body->GetAABB (vector0, vector1);

   p0[0] = vector0.m_x;
   p0[1] = vector0.m_y;
   p0[2] = vector0.m_z;

   p1[0] = vector1.m_x;
   p1[1] = vector1.m_y;
   p1[2] = vector1.m_z;

}

the only way that can happens is if you are linking to the single presition library but you are using the double presition dll.

[dehseth]

the only way that can happens is if you are linking to the single presition library but you are using the double presition dll.

yeah you are right, I was using this double dll presition

thank you

[Julio Jerez]

double presition is for people how use very large worlds.
when using the exact solver the double presition is also more acurate, In the iterative mode it does not make difference at the solver level but the collision is much much better.

for you I believe you single presision is the right one.

[dehseth]

I was just editing my text and I saw you replied

sorry it's kinda late in here and I didn't understand single means 4-byte float

thanks again...