Tests HDF5 serial performance
Syntax:
h5perf_serial [-h | --help]
h5perf_serial [options]
Description:
h5perf_serial provides tools for testing the performance of the HDF5 Library in serial mode.See h5perf_serial, a Serial File System Benchmarking Tool for a complete description of this tool.
The following environment variable can be set to control the specfied aspect of h5perf_serial behavior:
HDF5_NOCLEANUP | If set, h5perf_serial does not remove data files.
(Default: Data files are removed.) |
HDF5_PREFIX | Sets the prefix for output data files. |
Options and Parameters:
The term size specifier is used as follows in this section:
A size specifier is an integer greater than or equal to 0 (zero) followed by a size indicator:
K
| for kilobytes (1024 bytes) |
M | for kilobytes (1024 bytes) |
G | for gigabytes (1073741824 bytes) |
Example:
37M specifies 37 megabytes or 38797312 bytes.
-A api_list | Specifies which APIs to test. api_list is a comma-separated list with the following valid values: hdf5 | HDF5 library APIs | posix | POSIX APIs |
(Default: All APIs are monitored.) Example: -A hdf5,posix specifies that the HDF5 and POSIX APIs are to be monitored. |
-c chunk_size_list | Specifies chunked storage and defines chunks dimensions and sizes.
(Default: Chunking is off.)chunk_size_list is a comma-separated list of size specifiers. For example, a chunk_size_list value of
2K,4K,6M
specifies that chunking is turned on and that chunk size is 2 kilobytes by 4 kilobytes by 6 megabytes. |
-e dataset_size_list | Specifies dataset dimensionality and dataset dimension sizes.
(Default dataset size is 100x200, or 100,200.)dataset_size_list is a comma-separated list of size specifiers, which are defined above. For example, a dataset_size_list value of
2K,4K,6M
specifies a 2 kilobytes by 4 kilobytes by 6 megabytes dataset. |
-i iterations | Specifies the number of iterations to perform.
(Default: A single iteration, 1, is performed.)iterations is an integer specifying the number of iterations. |
-r access_order | Specifies dimension access order.
(Default: 1,2)access_order is a comma-separated list of integers specifying the order of access. For example,
-r 1,3,2
specifies the traversal of dimension 1 first, then dimension 3, and finally dimension 2. |
-t | Selects extendable HDF5 dataset dimensions.
(Default: Datasets are fixed size.) |
-v file_driver | Selects HDF5 driver to be used for HDF5 file access.
(Default: sec2)Valid values are as follows:
sec2
stdio
core
split
multi
family
direct |
-w | Specifies the performance of write tests only, read performance will not be tested.
(Default: Both write and read tests are performed.) |
-x buffer_size_list | Specifies transfer buffer dimensions and sizes.
(Default: 10,20) |
Exit Status:
| 0 | Succeeded. |
| > 0 | An error occurred. |
History:
| Release | Change
|
|---|
| 1.8.1 | Tool introduced in this release. |
--- Last Modified: August 28, 2019 | 10:43 AM
