Problems with transform matrix.

[u3dreal]

Hi all i'm implementing Instances at the moment
and i'm a bit confused here...

it seems that i need the inverse matrix for the rotations to work correct
and for the scaling i need the normal matrix...

Any ideas Nick ??? Scaling seems to be the wrong way round with invers matrix.. any way to inverse the rotation in Indigo ??
Or could you give me more detail on how the indigo matrix works ??



So far the transform and rotation works good with inverse matrix...

cheers
u3dreal

[u3dreal]

No help c'mon guys..

this would help everyone..

cheers

u3dreal

[u3dreal]

I use inverse matrix now and it works perfect with rotation and translation... but scale seems to be
the other way round .. when i scale the object down..indigo renders it bigger...

Any idea Nick ??

cheers

u3dreal

[OnoSendai]

Hmm..
The matrix should be defined in row major order
http://en.wikipedia.org/wiki/Row-major_order

Any scale should be set with the <scale> element.

[u3dreal]

Thanks Nick...

i found the rotation is not presented the right way in the normal max matrix you have to extract that first...

Now it works ...



But i came up with another question...

Is it necessary to go recalculate the tree for each
instance ?? Couldn't it be reloaded from disk ??

Or does the tree have to be build for every position of the instances ??

cheers

u3dreal

[OnoSendai]

Yes, there is a different tree for each instance, because the verts are stored in world space.
Someday i'll implement proper internal instancing, where the verts are stored in object space. Then the tree and mesh will only be stored once in memory and in the tree cache.

[psor]

This is looking promissing u3dreal! And I hope Nick will give us 'real'
instance support in a future version ... hehe. ;D


take care
psor