China
Africa
Explore chapters and articles related to this topic
Getting Started with FLAME
Published in Mariam Kiran, X-Machines for Agent-Based Modeling, 2017
FLAME uses C, XML and interacts with generated files. The models can be enhanced by using the same principle and working with additional libraries. For example,More C standard libraries and custom libraries: C Math functions, C memory accessible variables (include the basic int, char, float) or all C functions. Custom libraries are user-defined libraries.MPICH-2 libraries: Implementation of MPI, MPICH2 provides MPI implementations for important platforms and massively parallel processors. It is open-source and freely available for use in parallel programming environments. OpenMP can be used with MPI to allow hybrid parallelization for loop-level parallelism. More information on integrating MPICH with Windows and Linux platforms can be found at [78].OpenGL libraries: Open Graphics Library provides access to functions for high quality graphical image in 2D or 3D. OpenGL is concerned with manipulation of frame buffer for drawing and rendering of images. It can be integrated with C language for its functionality.Libxml2: With XML input and storage format, Libxml2 is an XML C parser toolkit that can be used across various platforms. It provides a variety of language bindings and wrappers making it useful with various languages. It provides support for Document Object Model as well.Interfacing with SBML: Libsbml allows manipulation of various SBML (systems biology markup language) files and data streams. Written in C and C++, it is used as a library for various programming languages (like C/C++, Java, Lisp, Perl, Matlab) and makes the code portable to different platforms of Windows or Linux.HDF5: Hierarchical Data Format 5 is a library used to store various data. It can allow data to be stored as dataset or in groups. A dataset is a multidimensional array of data elements whereas a group is a structure for organizing objects. Using these two storage mechanisms, one can generate any kind of required data structure like images, arrays of vectors or grid structures.GraphViz graph library: FLAME is already using GraphViz for generation of dotty diagrams or graphs showing function dependencies in parallel activity. It can be used for more outputs on networking structures, depicting hierarchy, clusters and more.Sqlite3: A small C library supports the SQL database engine to store data into a single disk file. These files can be shared as a database between various machines.
Optimising the Termofluids CFD code for petascale simulations
Published in International Journal of Computational Fluid Dynamics, 2016
R. Borrell, J. Chiva, O. Lehmkuhl, G. Oyarzun, I. Rodríguez, A. Oliva
Another relevant aspect that influences the performance of the time-integration process are the checkpointing input and output (IO) operations. Since simulations are generally completed by multiple executions, checkpoints are used to restart simulations form the last point, from a specific point of interest, or from the last point preceding a failure. In TF, the IO operations are managed by means of the HDF5 library (The HDF Group, 1997–2015). Achieving performance on the parallel IO operations with HDF5 library relies on taking advantage of collective operations. However, there are many intrinsic hardware constraints such as the bandwidth of the parallel file system that cannot be overcome. In particular, our layout of data on the hierarchical data format of the HDF5 library consists on one collective data-set for each scalar field and a contiguous region within it reserved to each parallel process engaged on the simulation. Our goal regarding the IO operations of the checkpointing process is that those are fast enough and generate an acceptable overhead.
Deep learning for identifying environmental risk factors of acute respiratory diseases in Beijing, China: implications for population with different age and gender
Published in International Journal of Environmental Health Research, 2020
DNN was adopted to train the deep learning model in this study. We based on the open source deep learning framework Convolutional Architecture for Fast Feature Embedding (Caffe) (Caffe 0000) to construct deep learning models. The data format was needed to convert into the format of Hierarchical Data File (HDF5) (HDF group 0000), which could be recognized by Caffe. HDF5 was a self-describing, multi-object file format, which could maintain high input/output efficiency and speed up data loading. And HDF5 was developed by the National Center for Supercomputer Applications of America to meet research needs in various fields (HDF5 0000).