data_base ❭ isf_data_base ❭ IO ❭ LoaderDumper ❭ shared_numpy_store

shared_numpy_store¶

Read and write numpy arrays to and from shared memory.

Shared memory is a memory space that is shared between multiple processes.

Note

This module is speicalized for use with SLURM HPC systems. To use with other systems, set the JOB_SHMTMPDIR environment variable to a directory in shared memory.

Classes¶

`Uninterruptible`	Context manager to create an uninterruptible section of code.
`SharedNumpyStore`	Store numpy arrays on disk and share them between processes.
`Loader`	Loader for `SharedNumpyStore` objects.

Functions¶

`shared_array_from_numpy`(arr, name)	Takes an array in memory and puts it into shared memory
`shared_array_from_disk`(path, shape, dtype, name, start_row, end_row)	Loads a numpy array from disk and puts it into shared memory
`memmap_from_disk`(path, shape, dtype, name, start_row, end_row)	Memory map a numpy array on disk.
`shared_array_from_shared_mem_name`(fname, shape, dtype)	Loads an existing shared array by its name
`check`(obj)	Check whether the object can be saved with this dumper.
`dump`(obj, savedir)	Dump the shared numpy store in the specified directory.