ComputeSphereVisibility: a patch

[paulo lesgaz]

Hi,

here is a patch for a first try to implement ComputeSphereVisibility. Any feedback is welcome.

David

[Frank Richter]

On 14.11.2008 20:27, paulo lesgaz wrote:
> Hi,
>
> here is a patch for a first try to implement ComputeSphereVisibility.
> Any feedback is welcome.

I think you can simplify the sphere-plane intersection. Just compute the
signed distance D of the sphere center from the plane. If D > r, the
sphere is visible. If -r <= D <= r, the sphere is partially visible. if
D < -r, it's invisible.

Also, wine people like tests. You should probably add some.

-f.r.

[paulo lesgaz]

Hello,

can anyone test the the following tests in a windows box, please?
The option to be given at the executable is  d3d.

Thanks in advance

David

--- En date de : Sam 15.11.08, Frank Richter  a écrit :
De: Frank Richter 
Objet: Re: ComputeSphereVisibility: a patch
À: "paulo lesgaz" 
Cc: [e-mail]
Date: Samedi 15 Novembre 2008, 19h14

On 14.11.2008 20:27, paulo lesgaz wrote:
> Hi,
>
> here is a patch for a first try to implement ComputeSphereVisibility.
> Any feedback is welcome.

I think you can simplify the sphere-plane intersection. Just compute the
signed distance D of the sphere center from the plane. If D > r, the
sphere is visible. If -r <= D <= r, the sphere is partially visible. if
D < -r, it's invisible.

Also, wine people like tests. You should probably add some.

-f.r.

[Henri Verbeet]

2008/11/15 Frank Richter :
> On 14.11.2008 20:27, paulo lesgaz wrote:
> > Hi,
> >
> > here is a patch for a first try to implement ComputeSphereVisibility.
> > Any feedback is welcome.
>
> I think you can simplify the sphere-plane intersection. Just compute the
> signed distance D of the sphere center from the plane. If D > r, the
> sphere is visible. If -r <= D <= r, the sphere is partially visible. if
> D < -r, it's invisible.

Yeah, that's basically what I described in my earlier mail.

[paulo lesgaz]

Hi,

here is a new patch for ComputeSphereVisibility implementing Henri and Franck's idea.

The problem is that the test takes the value 0x3f once  (if the test is correct), and that my function can not never obtain such a value.

Any idea to fix this problem.
Here is the only one link talking about ComputeSphereVisibility

http://209.85.135.104/search?q=cache:ZkQTOaa7h7UJ:salaam.cs.buap.mx/EBOOKS/I...

Thanks for the help.

David
--- En date de : Lun 17.11.08, Henri Verbeet  a écrit :
De: Henri Verbeet 
Objet: Re: ComputeSphereVisibility: a patch
À: "Frank Richter" 
Cc: "paulo lesgaz" , [e-mail]
Date: Lundi 17 Novembre 2008, 0h21

2008/11/15 Frank Richter :
> On 14.11.2008 20:27, paulo lesgaz wrote:
> > Hi,
> >
> > here is a patch for a first try to implement ComputeSphereVisibility.
> > Any feedback is welcome.
>
> I think you can simplify the sphere-plane intersection. Just compute the
> signed distance D of the sphere center from the plane. If D > r, the
> sphere is visible. If -r <= D <= r, the sphere is partially visible.

if
> D < -r, it's invisible.

Yeah, that's basically what I described in my earlier mail.

[Henri Verbeet]

2008/11/17 paulo lesgaz :
> The problem is that the test takes the value 0x3f once  (if the test is
> correct), and that my function can not never obtain such a value.

Looks like you need to set one of the
D3DVIS_INSIDE_FRUSTUM/D3DVIS_INTERSECT_FRUSTUM/D3DVIS_OUTSIDE_FRUSTUM
flags, depending on which applies. The values of the flags also
suggest INSIDE/OUTSIDE applies to the sphere centers rather than the
sphere as a whole. Ie, you can have both D3DVIS_INSIDE_LEFT and
D3DVIS_INTERSECT_LEFT set. You would need to write tests for this, of
course.
> -
>    vec.u1.x = m._14 + m._11;
> -
>    vec.u2.y = m._24 + m._21;
> -
>    vec.u3.z = m._34 + m._31;
> -
>    origin_plane = m._44 + m._41;

Note that the planes you get from this aren't necessarily normalized.