Database Reference
In-Depth Information
Figure 9.10
(See color insert following page 224.) H5Part readers are in-
cluded with visualization and data analysis tools in use by the accelerator
modeling community. This image shows H5Part data loaded into VisIt for
comparative analysis of multiple variables at different simulation timesteps.
to include support for block-structured fields and unstructured meshes.
It is an open source project, and source code may be downloaded from
https://codeforge.lbl.gov/projects/h5part. Several accelerator modeling codes
use H5Part for writing data in parallel, and H5Part readers have been created
and integrated into visual data analysis applications in widespread use by the
accelerator modeling community (see Figures 9.10 and 9.11).
Recent efforts have focused on merging the capabilities of HDF5 FastQuery,
which provides a high-level API interface for advanced index/query capabil-
ities, into H5Part. The objective here is to encapsulate the complexity of
index/query APIs and to integrate that capability into what appears to the
application as an I/O layer. This work has proven very useful in enabling
rapid visual data exploration in several projects, including advanced accelera-
tor design using laser wakefield acceleration (see Figure 9.12). Here, the tightly
integrated index/query and I/O provides the fundamental scientific data man-
agement infrastructure needed to implement query-driven visualization: A
visual interface enables rapid exploration of high-level data characteristics,
and enables visual specification of “interesting” data subsets, which are then
quickly extracted from a very large dataset and used by downstream visual
