Is there any advantage to using port groups (other than accounting for
arbitrary dynamic connections)?
I think it's time the Softimage wiki got integrated into the standard
documentation that ships with the product. Since there is no more development,
there's no point in having information scattered in several places around the
internet.
Another issue:
I am writing a custom operator (deformer) in C++ which is experiencing an
update issue. It takes several inputs, and has a single output. The deformed
mesh has both an input and output port with the output port defined first. I
can read and write the point positions of the deformed mesh producing desired
results - except for one detail.
One feature of my deformer is to allow the user to 'push' the points along the
geometry's normal using a weight map just like the native push deformer.
However, the nature of my operator requires the point positions be written to
the output port to update the deformed mesh before I can perform the push
calculations as I need updated normals to do that. The problem I am
experiencing is the 2nd attempt to write the point positions during the
_Update() callback do not update the point positions. The weird part is I had
the double update working for the push in previous versions of my plugin, but
the geometry updates were one cycle behind. Fixing the one cycle lag (caused
by storing primitives in plugin userdata) broke the push map feature. 2nd
attempt to write to the output port results in nothing happening.
So....how can I deform a mesh, then perform a push on it after it's been
deformed? Do I have to compute the normals myself?
Pseudo code:
//-----------------------------------------
Primitive oPrimitive = oCustomOperator.GetInputValue(
portInputDeformedMesh );
MATH::CVector3Array aPointPositions =
oPrimitive.GetGeometry().GetPoints().GetPositionArray();
for ( LONG I = 0; I < aPointPositions.GetCount(); i++ )
{
MATH::CVector3 oPointPosition =
aPointPositions[i];
// deform points math here
...
// record the modified point position
aPointPositions[i] = oPointPosition;
}
// Write results to mesh
Primitive oPrimitive = oCustomOperator.GetOutputTarget();
oPrimitive.GetGeometry().GetPoints().PutPositionArray( aPointPositions );
// good until here
// Get weight map
ClusterProperty oPushNormalMap = oCustomOperator.GetInputValue(
portInputWeightMap );
CFloatArray aWeightMapValues;
oPushNormalMap.GetValues( aWeightMapValues );
// re-get points and normals in current deformed state
// ( I have also tried using the input object's primitive - same result)
aPointPositions = oPrimitive.GetGeometry().GetPoints().GetPositionArray();
aPointNormals = oPrimitive.GetGeometry().GetPoints().GetNormalArray();
// do push math
for ( LONG I = 0; I < aPointPositions.GetCount(); i++ )
{
MATH::CVector3 oPointPosition =
aPointPositions[i];
MATH::Cvector3 oPointNormal = aPointNormals[i];
// push points math here
...
// record the modified point position
aPointPositions[i] = oPointPosition;
aPointNormals[i] = oPointNormal;
}
// update the deformed mesh (again)
oPrimitive.GetGeometry().GetPoints().PutPositionArray( aPointPositions );
//----------------------------
Thanks,
Matt
From: [email protected]
[mailto:[email protected]] On Behalf Of Hsiao Ming Chia
Sent: Thursday, April 24, 2014 5:35 AM
To: [email protected]
Subject: Re: SDK: port groups in custom operators
Hi Matt,
There is no performance advantage with port groups.
In fact, there is extra logic invoked when connecting to a port group but once
connected, they are just as fast.
http://softimage.wiki.softimage.com/index.php?title=Custom_Operators_(XSISDK)#1._PortGroups_make_it_easier_to_connect_an_operator_when_it_is_created
Thanks,
Hsiao Ming
-------- Original message --------
From: Matt Lind <[email protected]<mailto:[email protected]>>
Date:
To: [email protected]<mailto:[email protected]>
Subject: SDK: port groups in custom operators
Other than the situation of needing an arbitrary number of input or output
connections, is there any technical reason to put ports into different port
groups for a custom operator? Is there a performance advantage? If so, what's
the best strategy?
Thanks,
Matt
