...
The RIX_PATTERNCREATE() macro defines the CreateRixPattern() method, which is called by the renderer to create an instance of the pattern plugin. Generally, the implementation of this method should simply return a new allocated copy of your pattern class. Similarly, the RIX_PATTERNDESTROY() macro defines the DestroyRixPattern() method called by the renderer to delete an instance of the pattern plugin; a typical implementation of this method is to delete the passed in pattern pointer:
...
| Code Block | ||
|---|---|---|
| ||
RIX_PATTERNCREATE |
...
{ |
...
return new MyPattern(); |
...
} |
...
RIX_PATTERNDESTROY |
...
{ |
...
delete ((MyPattern*)pattern); |
...
} |
| Anchor | ||||
|---|---|---|---|---|
|
ComputeOutputParams() is the heart of a pattern plugin: it evaluates the input parameters, and computes the pattern output. It is called once per graph execution, and all outputs must be computed during this single invocation. The number and type of outputs should match the number and type of outputs declared in the parameter table. The domain of evaluation of this function is a shading context, which is of type RixShadingContext, defined in RixShading.h.
...
After reading input values, output values need to be set up. First, memory buffers for the requested outputs should be allocated using the RixShadingContext memory allocation services. These buffers should then be bound to the requested OutputSpec outputs parameter passed to ComputeOutputParams(), and the type and detail information about those outputs filled in as well. This information should match the declarations from the parameter table. The following code is boilerplate that can be used: it reads the plugin's parameter table, loops through and allocates the appropriate buffers, and sets the detail and type assuming that the output is always a varying color or float (typical of most patterns).
| Code Block | ||
|---|---|---|
| ||
// Find the number of outputs
RixSCParamInfo const* paramTable = GetParamTable();
int numOutputs = -1;
while (paramTable[++numOutputs].access == k_RixSCOutput) {}
// Allocate and bind our outputs
RixShadingContext::Allocator pool(sctx);
OutputSpec* out = pool.AllocForPattern<OutputSpec>(numOutputs);
*outputs = out;
*noutputs = numOutputs;
// looping through the different output ids
for (int i = 0; i < numOutputs; ++i)
{
out[i].paramId = i;
out[i].detail = k_RixSCInvalidDetail;
out[i].value = NULL;
type = paramTable[i].type; // we know this
sctx->GetParamInfo(i, &type, &cinfo);
if(cinfo == k_RixSCNetworkValue)
{
if( type == k_RixSCColor )
{
out[i].detail = k_RixSCVarying;
out[i].value = pool.AllocForPattern<RtColorRGB>(sctx->numPts);
}
else if( type == k_RixSCFloat )
{
out[i].detail = k_RixSCVarying;
out[i].value = pool.AllocForPattern<RtFloat>(sctx->numPts);
}
}
} |
...
Finally, the pattern can now actually compute the values that go into the output buffers. This is typically done by using the inputs and looping through the number of shaded points RixShadingContext::numPts to compute some values that are stored in the allocated output buffers.
| Code Block | ||
|---|---|---|
| ||
RtColorRGB* outColor = (RtColorRGB*) out[k_resultRGB].value; |
...
for (int i=0; i<sctx->numPts; i++) { // Compute some output values based on your input. Here we assume // outColor is the memory buffer allocated for an output parameter, // and inputColor and inputFloat are two inputs that were returned from // EvalParam. if (style == 1) { outColor[i] = inputColor[i] * inputFloat[i]; } } |
...
In the simple example above, outColor is assigned the buffer that was allocated corresponding to the private enumeration value k_resultRGB, which matches the position of that output in the parameter table. We (So long as the output parameters are at the beginning of the parameter table, reuse of this enumeration is valid for this purpose.) We assume the style variable was a uniform RtInt input value, so there is only one value for all the points in the shading context. Meanwhile, the inputColor and inputFloat variable were varying instead of uniform, so they are pointers to an array of RtColorRGB values and array of RtFloat values respectively, one for each shaded point in the shading context.
...
Display "patternTest" "framebuffer" "rgba"
Quantize "rgba" 255 0 255 0
Format 128 128 1
Projection "perspective" "fov" [45]
Hider "raytrace" "string integrationmode" ["path"]
Integrator "PxrPathTracer" "integrator"
WorldBegin
AttributeBegin
Attribute "identifier" "name" ["sphere1"]
Translate 0 0 2.75
Pattern "PxrCustomPattern" "customPattern"
Bxdf "PxrDiffuse" "smooth"
"reference color diffuseColor" "customPattern:outColor"
Sphere 1.0 -1.0 1.0 360.0
AttributeEnd
WorldEndCreating a Pattern args File
...
Texture Baking
RenderMan can optionally bake pattern outputs to 2D or 3D textures by evaluating those patterns over an output manifold. Pattern plug-ins that wish to bake outputs should provide custom implementations of the RixPattern::Bake2dOutput or RixPattern::Bake3dOutput methods that return true. When in bake mode, RenderMan queries these methods to describe the output manifold and to initialize display drivers. For 2d atlas/UDIM outputs that set RixPattern::Bake2dSpec::atlas to true, RenderMan will query RixPattern::Bake2dOutput once for each UV tile.
It is possible to write a generalized baking node that bakes the output of arbitrary upstream pattern graphs. For example, see PxrBakeTexture and PxrBakePointCloud pattern plug-ins:
| Code Block | ||
|---|---|---|
| ||
Hider "bake"
Format 512 512 1
Display "render.exr" "openexr" "rgba"
Projection "perspective" "fov" [30]
Translate 0 0 5
WorldBegin
AttributeBegin
Pattern "PxrFractal" "pattern"
Pattern "PxrBakeTexture" "baked" "reference color inputRGB" ["pattern:resultRGB"]
"string filename" ["bake.tif"] "string display" ["tiff"]
"string primVar" ["st"] "int resolutionX" [512] "int resolutionY" [512]
Bxdf "PxrDiffuse" "default" "reference color diffuseColor" ["baked:resultRGB"]
Sphere 1 -1 1 360 "varying float[2] st" [0 0 1 0 0 1 1 1]
AttributeEnd
WorldEnd |