 |
kspaceFirstOrder3D-OMP
1.2
The C++ implementation of the k-wave toolbox for the time-domain simulation of acoustic wave fields in 3D
|
|
kspaceFirstOrder3D-OMP
1.2
The C++ implementation of the k-wave toolbox for the time-domain simulation of acoustic wave fields in 3D
|
The header file containing the HDF5 related classes. More...
#include <hdf5.h>#include <hdf5_hl.h>#include <string>#include <map>#include <Utils/MatrixNames.h>#include <Utils/DimensionSizes.h>
Go to the source code of this file.
Classes | |
| class | Hdf5File |
| Class wrapping the HDF5 routines. More... | |
The CUDA/C++ code has been designed as a standalone application which is not dependent on MATLAB libraries or a MEX interface. This is of particular importance when using servers and supercomputers without MATLAB support. For this reason, simulation data must be transferred between the CUDA/C++ code and MATLAB using external input and output files. These files are stored using the Hierarchical Data Format HDF5. This is a data model, library, and file format for storing and managing data. It supports a variety of datatypes, and is designed for flexible and efficient I/O and for high volume and complex data. The HDF5 technology suite includes tools and applications for managing, manipulating, viewing, and analysing data in the HDF5 format.
Each HDF5 file is a container for storing a variety of scientific data and is composed of two primary types of objects: groups and datasets. An HDF5 group is a structure containing zero or more HDF5 objects, together with supporting metadata. A HDF5 group can be seen as a disk folder. An HDF5 dataset is a multidimensional array of data elements, together with supporting metadata. A HDF5 dataset can be seen as a disk file. Any HDF5 group or dataset may also have an associated attribute list. An HDF5 attribute is a user-defined HDF5 structure that provides extra information about an HDF5 object. More information can be obtained from the HDF5 documentation.
kspaceFirstOrder3D-CUDA v1.2 uses the file format introduced in version 1.1. The code is happy to work with both versions (1.0 and 1.1), however when working with an input file of version 1.0, some features are not supported, namely the cuboid sensor mask, and u_non_staggered_raw. When running from within the actual MATLAB K-Wave Toolbox, the files will always be generated in version 1.1.
The HDF5 input file for the CUDA/C++ simulation code contains a file header with brief description of the simulation stored in string attributes, and the root group '/' which stores all the simulation properties in the form of 3D datasets (a complete list of input datasets is given bellow). The HDF5 checkpoint file contains the same file header as the input file and the root group '/' with a few datasets keeping the actual simulation state. The HDF5 output file contains a file header with the simulation description as well as performance statistics, such as the simulation time and memory consumption, stored in string attributes.
The results of the simulation are stored in the root group '/' in the form of 3D datasets. If the linear sensor mask is used, all output quantities are stored as datasets in the root group. If the cuboid corners sensor mask is used, the sampled quantities form private groups containing datasets on per cuboid basis.
+----------------------------------------------------------------------------------------------------------------------+ | Input File / Checkpoint File Header | +----------------------------------------------------------------------------------------------------------------------+ | created_by Short description of the tool that created this file | | creation_date Date when the file was created | | file_description Short description of the content of the file (e.g. simulation name) | | file_type Type of the file (input) | | major_version Major version of the file definition (1) | | minor_version Minor version of the file definition (1) | +----------------------------------------------------------------------------------------------------------------------+
+----------------------------------------------------------------------------------------------------------------------+ | Output File Header | +----------------------------------------------------------------------------------------------------------------------+ | created_by Short description of the tool that created this file | | creation_date Date when the file was created | | file_description Short description of the content of the file (e.g. simulation name) | | file_type Type of the file (output) | | major_version Major version of the file definition (1) | | minor_version Minor version of the file definition (1) | +----------------------------------------------------------------------------------------------------------------------+ | host_names List of hosts (computer names) the simulation was executed on | | number_of_cpu_cores Number of CPU cores used for the simulation | | data_loading_phase_execution_time Time taken to load data from the file | | pre-processing_phase_execution_time Time taken to pre-process data | | simulation_phase_execution_time Time taken to run the simulation | | post-processing_phase_execution_time Time taken to complete the post-processing phase | | total_execution_time Total execution time | | peak_core_memory_in_use Peak memory required per core during the simulation | | total_memory_in_use Total Peak memory in use | +----------------------------------------------------------------------------------------------------------------------+
The input and checkpoint file stores all quantities as three dimensional datasets with dimension sizes designed by (Nx, Ny, Nz). In order to support scalars and 1D and 2D arrays, the unused dimensions are set to 1. For example, scalar variables are stored with a dimension size of (1,1,1), 1D vectors oriented in y-direction are stored with a dimension size of (1, Ny, 1), and so on. If the dataset stores a complex variable, the real and imaginary parts are stored in an interleaved layout and the lowest used dimension size is doubled (i.e., Nx for a 3D matrix, Ny for a 1D vector oriented in the y-direction). The datasets are physically stored in row-major order (in contrast to column-major order used by MATLAB) using either the 'H5T_IEEE_F32LE' data type for floating point datasets or 'H5T_STD_U64LE' for integer based datasets. All the datasets are store under the root group.
The output file of version 1.0 could only store recorded quantities as 3D datasets under the root group. However, with version 1.1 and the new cuboid corner sensor mask, the sampled quantities may be laid out as 4D quantities stored under specific groups. The dimensions are always (Nx, Ny, Nz, Nt), every sampled cuboid is stored as a distinct dataset and the datasets are grouped under a group named by the quantity stored. This makes the file clearly readable and easy to parse.
In order to enable compression and more efficient data processing, big datasets are not stored as monolithic blocks but broken into chunks that may be compressed by the ZIP library and stored separately. The chunk size is defined as follows:
(1M elements, 1, 1) in the case of 1D variables - index sensor mask (8MB blocks). (Nx, Ny, 1) in the case of 3D variables (one 2D slab). (Nx, Ny, Nz, 1) in the case of 4D variables (one time step). (N_sensor_points, 1, 1) in the case of the output time series (one time step of the simulation).All datasets have two attributes that specify the content of the dataset. The 'data_type' attribute specifies the data type of the dataset. The admissible values are either 'float' or 'long'. The 'domain_type' attribute specifies the domain of the dataset. The admissible values are either 'real' for the real domain or 'complex' for the complex domain. The C++ code reads these attributes and checks their values.
+----------------------------------------------------------------------------------------------------------------------+ | Input File Datasets | +----------------------------------------------------------------------------------------------------------------------+ | Name Size Data type Domain Type Condition of Presence | +----------------------------------------------------------------------------------------------------------------------+ | 1. Simulation Flags | +----------------------------------------------------------------------------------------------------------------------+ | ux_source_flag (1, 1, 1) long real | | uy_source_flag (1, 1, 1) long real | | uz_source_flag (1, 1, 1) long real | | p_source_flag (1, 1, 1) long real | | p0_source_flag (1, 1, 1) long real | | transducer_source_flag (1, 1, 1) long real | | nonuniform_grid_flag (1, 1, 1) long real must be set to 0 | | nonlinear_flag (1, 1, 1) long real | | absorbing_flag (1, 1, 1) long real | +----------------------------------------------------------------------------------------------------------------------+ 2. Grid Properties | +----------------------------------------------------------------------------------------------------------------------+ | Nx (1, 1, 1) long real | | Ny (1, 1, 1) long real | | Nz (1, 1, 1) long real | | Nt (1, 1, 1) long real | | dt (1, 1, 1) float real | | dx (1, 1, 1) float real | | dy (1, 1, 1) float real | | dz (1, 1, 1) float real | +----------------------------------------------------------------------------------------------------------------------+ | 3. Medium Properties | +----------------------------------------------------------------------------------------------------------------------+ | 3.1 Regular Medium Properties | | rho0 (Nx, Ny, Nz) float real heterogenous | | (1, 1, 1) float real homogenous | | rho0_sgx (Nx, Ny, Nz) float real heterogenous | | (1, 1, 1) float real homogenous | | rho0_sgy (Nx, Ny, Nz) float real heterogenous | | (1, 1, 1) float real homogenous | | rho0_sgz (Nx, Ny, Nz) float real heterogenous | | (1, 1, 1) float real homogenous | | c0 (Nx, Ny, Nz) float real heterogenous | | (1, 1, 1) float real homogenous | | c_ref (1, 1, 1) float real | | | | 3.2 Nonlinear Medium Properties (defined if (nonlinear_flag == 1)) | | BonA (Nx, Ny, Nz) float real heterogenous | | (1, 1, 1) float real homogenous | | | | 3.3 Absorbing Medium Properties (defined if (absorbing_flag == 1)) | | alpha_coef (Nx, Ny, Nz) float real heterogenous | | (1, 1, 1) float real homogenous | | alpha_power (1, 1, 1) float real | +----------------------------------------------------------------------------------------------------------------------+ | 4. Sensor Variables | +----------------------------------------------------------------------------------------------------------------------+ | sensor_mask_type (1, 1, 1) long real file version 1.1 | | (0 = index, 1 = corners) | | sensor_mask_index (Nsens, 1, 1) long real file version 1.0 always, | | file version 1.1 and sensor_mask_type == 0 | | sensor_mask_corners (Ncubes, 6, 1) long real file version 1.1 and sensor_mask_type == 1 | +----------------------------------------------------------------------------------------------------------------------+ | 5 Source Properties | +----------------------------------------------------------------------------------------------------------------------+ | 5.1 Velocity Source Terms (defined if (ux_source_flag == 1 || uy_source_flag == 1 || uz_source_flag == 1)) | | u_source_mode (1, 1, 1) long real | | u_source_many (1, 1, 1) long real | | u_source_index (Nsrc, 1, 1) long real | | ux_source_input (1, Nt_src, 1) float real u_source_many == 0 | | (Nsrc, Nt_src, 1) float real u_source_many == 1 | | uy_source_input (1, Nt_src, 1) float real u_source_many == 0 | | (Nsrc, Nt_src, 1) float real u_source_many == 1 | | uz_source_input (1, Nt_src, 1) float real u_source_many == 0 | | (Nt_src, Nsrc, 1) float real u_source_many == 1 | | | | 5.2 Pressure Source Terms (defined if (p_source_flag == 1)) | | p_source_mode (1, 1, 1) long real | | p_source_many (1, 1, 1) long real | | p_source_index (Nsrc, 1, 1) long real | | p_source_input (Nsrc, Nt_src, 1) float real p_source_many == 1 | | (1, Nt_src, 1) float real p_source_many == 0 | | | | 5.3 Transducer Source Terms (defined if (transducer_source_flag == 1)) | | u_source_index (Nsrc, 1, 1) long real | | transducer_source_input (Nt_src, 1, 1) float real | | delay_mask (Nsrc, 1, 1) float real | | | | 5.4 IVP Source Terms (defined if ( p0_source_flag == 1)) | | p0_source_input (Nx, Ny, Nz) float real | +----------------------------------------------------------------------------------------------------------------------+ | 6. K-space and Shift Variables | +----------------------------------------------------------------------------------------------------------------------+ | ddx_k_shift_pos_r (Nx/2 + 1, 1, 1) float complex | | ddx_k_shift_neg_r (Nx/2 + 1, 1, 1) float complex | | ddy_k_shift_pos (1, Ny, 1) float complex | | ddy_k_shift_neg (1, Ny, 1) float complex | | ddz_k_shift_pos (1, 1, Nz) float complex | | ddz_k_shift_neg (1, 1, Nz) float complex | | x_shift_neg_r (Nx/2 + 1, 1, 1) float complex file version 1.1 | | y_shift_neg_r (1, Ny/2 + 1, 1) float complex file version 1.1 | | z_shift_neg_r (1, 1, Nz/2) float complex file version 1.1 | +----------------------------------------------------------------------------------------------------------------------+ | 7. PML Variables | +----------------------------------------------------------------------------------------------------------------------+ | pml_x_size (1, 1, 1) long real | | pml_y_size (1, 1, 1) long real | | pml_z_size (1, 1, 1) long real | | pml_x_alpha (1, 1, 1) float real | | pml_y_alpha (1, 1, 1) float real | | pml_z_alpha (1, 1, 1) float real | | | | pml_x (Nx, 1, 1) float real | | pml_x_sgx (Nx, 1, 1) float real | | pml_y (1, Ny, 1) float real | | pml_y_sgy (1, Ny, 1) float real | | pml_z (1, 1, Nz) float real | | pml_z_sgz (1, 1, Nz) float real | +----------------------------------------------------------------------------------------------------------------------+
+----------------------------------------------------------------------------------------------------------------------+ | Checkpoint File Datasets | +----------------------------------------------------------------------------------------------------------------------+ | Name Size Data type Domain Type Condition of Presence | +----------------------------------------------------------------------------------------------------------------------+ | 1. Grid Properties | +----------------------------------------------------------------------------------------------------------------------+ | Nx (1, 1, 1) long real | | Ny (1, 1, 1) long real | | Nz (1, 1, 1) long real | | Nt (1, 1, 1) long real | | t_index (1, 1, 1) long real | +----------------------------------------------------------------------------------------------------------------------+ | 2. Simulation State | +----------------------------------------------------------------------------------------------------------------------+ | p (Nx, Ny, Nz) float real | | ux_sgx (Nx, Ny, Nz) float real | | uy_sgy (Nx, Ny, Nz) float real | | uz_sgz (Nx, Ny, Nz) float real | | rhox (Nx, Ny, Nz) float real | | rhoy (Nx, Ny, Nz) float real | | rhoz (Nx, Ny, Nz) float real | +----------------------------------------------------------------------------------------------------------------------+
+----------------------------------------------------------------------------------------------------------------------+ | Output File Datasets | +----------------------------------------------------------------------------------------------------------------------+ | Name Size Data type Domain Type Condition of Presence | +----------------------------------------------------------------------------------------------------------------------+ | 1. Simulation Flags | +----------------------------------------------------------------------------------------------------------------------+ | ux_source_flag (1, 1, 1) long real | | uy_source_flag (1, 1, 1) long real | | uz_source_flag (1, 1, 1) long real | | p_source_flag (1, 1, 1) long real | | p0_source_flag (1, 1, 1) long real | | transducer_source_flag (1, 1, 1) long real | | nonuniform_grid_flag (1, 1, 1) long real | | nonlinear_flag (1, 1, 1) long real | | absorbing_flag (1, 1, 1) long real | | u_source_mode (1, 1, 1) long real if u_source | | u_source_many (1, 1, 1) long real if u_source | | p_source_mode (1, 1, 1) long real if p_source | | p_source_many (1, 1, 1) long real if p_source | +----------------------------------------------------------------------------------------------------------------------+ | 2. Grid Properties | +----------------------------------------------------------------------------------------------------------------------+ | Nx (1, 1, 1) long real | | Ny (1, 1, 1) long real | | Nz (1, 1, 1) long real | | Nt (1, 1, 1) long real | | dt (1, 1, 1) float real | | dx (1, 1, 1) float real | | dy (1, 1, 1) float real | | dz (1, 1, 1) float real | +----------------------------------------------------------------------------------------------------------------------+ | 3. PML Variables | +----------------------------------------------------------------------------------------------------------------------+ | pml_x_size (1, 1, 1) long real | | pml_y_size (1, 1, 1) long real | | pml_z_size (1, 1, 1) long real | | pml_x_alpha (1, 1, 1) float real | | pml_y_alpha (1, 1, 1) float real | | pml_z_alpha (1, 1, 1) float real | | | | pml_x (Nx, 1, 1) float real | | pml_x_sgx (Nx, 1, 1) float real | | pml_y (1, Ny, 1) float real | | pml_y_sgy (1, Ny, 1) float real | | pml_z (1, 1, Nz) float real | | pml_z_sgz (1, 1, Nz) float real | +----------------------------------------------------------------------------------------------------------------------+ | 4. Sensor Variables (present if --copy_sensor_mask) | +----------------------------------------------------------------------------------------------------------------------+ | sensor_mask_type (1, 1, 1) long real file version 1.1 and --copy_sensor_mask | | sensor_mask_index (Nsens, 1, 1) long real file version 1.1 and sensor_mask_type == 0 | | sensor_mask_corners (Ncubes, 6, 1) long real file version 1.1 and sensor_mask_type == 1 | +----------------------------------------------------------------------------------------------------------------------+ | 5a. Simulation Results: if sensor_mask_type == 0 (index), or File version == 1.0 | +----------------------------------------------------------------------------------------------------------------------+ | p (Nsens, Nt - s, 1) float real -p or --p_raw | | p_rms (Nsens, 1, 1) float real --p_rms | | p_max (Nsens, 1, 1) float real --p_max | | p_min (Nsens, 1, 1) float real --p_min | | p_max_all (Nx, Ny, Nz) float real --p_max_all | | p_min_all (Nx, Ny, Nz) float real --p_min_all | | p_final (Nx, Ny, Nz) float real --p_final | | | | ux (Nsens, Nt - s, 1) float real -u or --u_raw | | uy (Nsens, Nt - s, 1) float real -u or --u_raw | | uz (Nsens, Nt - s, 1) float real -u or --u_raw | | | | ux_non_staggered (Nsens, Nt - s, 1) float real --u_non_staggered_raw (File version ==1.1) | | uy_non_staggered (Nsens, Nt - s, 1) float real --u_non_staggered_raw (File version ==1.1) | | uz_non_staggered (Nsens, Nt - s, 1) float real --u_non_staggered_raw (File version ==1.1) | | | | ux_rms (Nsens, 1, 1) float real --u_rms | | uy_rms (Nsens, 1, 1) float real --u_rms | | uz_rms (Nsens, 1, 1) float real --u_rms | | | | ux_max (Nsens, 1, 1) float real --u_max | | uy_max (Nsens, 1, 1) float real --u_max | | uz_max (Nsens, 1, 1) float real --u_max | | | | ux_min (Nsens, 1, 1) float real --u_min | | uy_min (Nsens, 1, 1) float real --u_min | | uz_min (Nsens, 1, 1) float real --u_min | | | | ux_max_all (Nx, Ny, Nz) float real --u_max_all | | uy_max_all (Nx, Ny, Nz) float real --u_max_all | | uz_max_all (Nx, Ny, Nz) float real --u_max_all | | | | ux_min_all (Nx, Ny, Nz) float real --u_min_all | | uy_min_all (Nx, Ny, Nz) float real --u_min_all | | uz_min_all (Nx, Ny, Nz) float real --u_min_all | | | | ux_final (Nx, Ny, Nz) float real --u_final | | uy_final (Nx, Ny, Nz) float real --u_final | | uz_final (Nx, Ny, Nz) float real --u_final | +----------------------------------------------------------------------------------------------------------------------+ | 5b. Simulation Results: if sensor_mask_type == 1 (corners) and file version == 1.1 | +----------------------------------------------------------------------------------------------------------------------+ | /p group of datasets, one per cuboid -p or --p_raw | | /p/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | /p/2 (Cx, Cy, Cz, Nt-s) float real 2nd sampled cuboid, etc. | | | | /p_rms group of datasets, one per cuboid --p_rms | | /p_rms/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | | | /p_max group of datasets, one per cuboid --p_max | | /p_max/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | | | /p_min group of datasets, one per cuboid --p_min | | /p_min/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | | | p_max_all (Nx, Ny, Nz) float real --p_max_all | | p_min_all (Nx, Ny, Nz) float real --p_min_all | | p_final (Nx, Ny, Nz) float real --p_final | | | | | | /ux group of datasets, one per cuboid -u or --u_raw | | /ux/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | /uy group of datasets, one per cuboid -u or --u_raw | | /uy/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | /uz group of datasets, one per cuboid -u or --u_raw | | /uz/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | | | /ux_non_staggered group of datasets, one per cuboid --u_non_staggered_raw | | /ux_non_staggered/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | /uy_non_staggered group of datasets, one per cuboid --u_non_staggered_raw | | /uy_non_staggered/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | /uz_non_staggered group of datasets, one per cuboid --u_non_staggered_raw | | /uz_non_staggered/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | | | /ux_rms group of datasets, one per cuboid --u_rms | | /ux_rms/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | /uy_rms group of datasets, one per cuboid --u_rms | | /uy_rms/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | /uz_rms group of datasets, one per cuboid --u_rms | | /uy_rms/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | | | /ux_max group of datasets, one per cuboid --u_max | | /ux_max/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | /uy_max group of datasets, one per cuboid --u_max | | /ux_max/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | /uz_max group of datasets, one per cuboid --u_max | | /ux_max/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | | | /ux_min group of datasets, one per cuboid --u_min | | /ux_min/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | /uy_min group of datasets, one per cuboid --u_min | | /ux_min/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | /uz_min group of datasets, one per cuboid --u_min | | /ux_min/1 (Cx, Cy, Cz, Nt-s) float real 1st sampled cuboid | | | | ux_max_all (Nx, Ny, Nz) float real --u_max_all | | uy_max_all (Nx, Ny, Nz) float real --u_max_all | | uz_max_all (Nx, Ny, Nz) float real --u_max_all | | | | ux_min_all (Nx, Ny, Nz) float real --u_min_all | | uy_min_all (Nx, Ny, Nz) float real --u_min_all | | uz_min_all (Nx, Ny, Nz) float real --u_min_all | | | | ux_final (Nx, Ny, Nz) float real --u_final | | uy_final (Nx, Ny, Nz) float real --u_final | | uz_final (Nx, Ny, Nz) float real --u_final | +----------------------------------------------------------------------------------------------------------------------+
This file is part of the C++ extension of the k-Wave Toolbox.
This file is part of the k-Wave. k-Wave is free software: you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
k-Wave is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public License along with k-Wave. If not, see http://www.gnu.org/licenses/.
Definition in file Hdf5File.h.
1.8.13