See the books “Users Guide” CNRS 2005 Visualization Summer School.Open source VTK is a C++ toolkit, modular components assembled in “pipelines” ParaView is an end-user application integrating many VTK features ParaView hides the complexity of pipeline editing.
parallel data access: piece-wise data + parallel processing Streaming data: VTK offers piece-wise data access for intermediate or accumulated representations. client-server model: share the memory cost between two machines (The network bandwith might become the bottleneck) The renderer drives the request for data updates Why is this important? Scaling to large data, and planning a distributed visualization should be driven bottom-to-topĭata Sources Data Filters Data Mappers Graphics Renderingĭata access on-demand data loading: (was first popularized with EnSight). Any change in the data triggers a downstream chain of updates VTK is an event-flow environment. (see picture on the right) => EnSight, ParaView hide it from the user while maintaining its functionality CNRS 2005 Visualization Summer Schoolĭata Flow or Event Flow? AVS/Express is a data-flow environment. IRIS Explorer, AVS, MayaVi make this graph visible. The Visualization Pipeline The Visualization pipeline can most often be represented by a graph (DAG) of modules connected to each other with data “flowing” between them. The Visualization Pipeline Data Sources: reading from data files, or generating data online Data Filters: construct objects used to understand the data examples: warp scalar, contour, shrink, streamline, probe, gradient Data Mappers: convert filters’s output to geometric primitives examples: polygonal models (triangle strips) Rendering: OpenGL or off-screen + interaction modes The Visualization pipeline is an active collection of modules allowing users to create/delete derived quantities and objects to represent their data. Strategies to deal with large data Data access It is partitioned over multiple domains.
“it is big if it does not fit on the computer you have access to” Can be large because: it is time-dependent. I’ll accept Jim Ahrens (LANL) own definition. Large Data Volumes Definition: What is considered “large”? History of visualization softwares at CSCS In the last 10 years, IRIS Explorer ›VS5, AVS/Express ‾nSight ›mira ‼FXPost (ANSYS) In-house packages ‼OVISE Most examples in this class VTK Most practice in our Center ParaView CNRS 2005 Visualization Summer School
Outline Historical perspective Some strategies to deal with large data How do VTK and ParaView fit in our portfolio? VTK and ParaView’s best features for Large Data Data Streaming, data parallelism ParaView’s Compositor Distributed Visualization Parallel Visualization Large data volumes Dr.
0 kommentar(er)
