Tutorial Node: Attributes With CUDA Data

This is a tutorial node. It performs different functions on the GPU to illustrate different types of data access. The first adds inputs ‘a’ and ‘b’ to yield output ‘sum’, all of which are on the GPU. The second is a sample expansion deformation that multiplies every point on a set of input points, stored on the GPU, by a constant value, stored on the CPU, to yield a set of output points, also on the GPU. The third is an assortment of different data types illustrating how different data is passed to the GPU. This particular node uses CUDA for its GPU computations, as indicated in the memory type value. Normal use case for GPU compute is large amounts of data. For testing purposes this node only handles a very small amount but the principle is the same.

Installation

To use this node enable omni.graph.tutorials in the Extension Manager.

Inputs

Name	Type	Descripton	Default
A (inputs:a)	float	First value to be added in algorithm 1	0.0
B (inputs:b)	float	Second value to be added in algorithm 1	0.0
Color (inputs:color)	colord[3]	Input with three doubles as a color for algorithm 3	[1.0, 0.5, 1.0]
Half (inputs:half)	half	Input of type half for algorithm 3	1.0
Matrix (inputs:matrix)	matrixd[4]	Input with 16 doubles interpreted as a double-precision 4d matrix	[[1.0, 0.0, 0.0, 0.0], [0.0, 1.0, 0.0, 0.0], [0.0, 0.0, 1.0, 0.0], [0.0, 0.0, 0.0, 1.0]]
Multiplier (inputs:multiplier)	float[3]	Amplitude of the expansion for the input points in algorithm 2	[1.0, 1.0, 1.0]
Points (inputs:points)	float[3][]	Points to be moved by algorithm 2	[]

Outputs

Name	Type	Descripton	Default
Color (outputs:color)	colord[3]	Output with three doubles as a color for algorithm 3	None
Half (outputs:half)	half	Output of type half for algorithm 3	None
Matrix (outputs:matrix)	matrixd[4]	Output with 16 doubles interpreted as a double-precision 4d matrix	None
Points (outputs:points)	float[3][]	Final positions of points from algorithm 2	None
Sum (outputs:sum)	float	Sum of the two inputs from algorithm 1	None

Metadata

Name	Value
Unique ID	omni.graph.tutorials.CudaData
Version	1
Extension	omni.graph.tutorials
Icon	ogn/icons/omni.graph.tutorials.CudaData.svg
Has State?	False
Implementation Language	C++
Default Memory Type	cuda
Generated Code Exclusions	None
uiName	Tutorial Node: Attributes With CUDA Data
__memoryType	cuda
Categories	tutorials
Generated Class Name	OgnTutorialCudaDataDatabase
Python Module	omni.graph.tutorials

See the accompanying tutorial and annotated code at Tutorial 8 - GPU Data Node