crc32c 2.6

A python package implementing the crc32c algorithm in hardware and software

This package implements the crc32c checksum algorithm. It automatically chooses between a hardware-based implementation (using the CRC32C SSE 4.2 instruction of Intel CPUs, and the crc32* instructions on ARMv8 CPUs), or a software-based one when no hardware support can be found.

Because crc32c is in PyPI, you can install it with:

pip install crc32c

Supported platforms are Linux and OSX using the gcc and clang compilers, and Windows using the Visual Studio compiler. Other compilers in Windows (MinGW for instance) might work. Binary wheels are also provided in PyPI for major platforms/architectures.

The project is using certain gcc/clang compiler extensions to support building hardware-specific functions that might not be supported by older compiler versions.

Usage

The core function exposed by this module is crc32c(data, value=0, gil_release_mode=-1). It computes the CRC32C checksum of data starting with an initial value checksum, similarly to how the built-in binascii.crc32 works. It can thus be used like this:

print(crc32c.crc32c(b'hello world'))
# 3381945770
crc = crc32c.crc32c(b'hello')
print(crc32c.crc32c(b' world', value=crc))
# 3381945770

In older versions, the function exposed by this module was called crc32. That name is still present but deprecated, and will be removed in new versions of the library.

The gil_release_mode keyword argument specifies whether a call of this library shall release or keep the Global Interpreter Lock. It can be set to the following values:

• Negative: Only release the GIL when data >= 32KiB

• 0: Never release the GIL

• Positive: Always release the GIL

On top of the crc32c function, a CRC32CHash class is also offered. It is modelled after the “hash objects” of the hashlib module of the standard library:

crc32c_hash = crc32c.CRC32CHash()
crc32c_hash.update(b'hello')
crc32c_hash.update(b' world')
print(crc32c_hash.digest())
# b'\xc9\x94e\xaa'

For more details see the documentation on hash objects.

Additionally one can consult the following module-level values:

• hardware_based indicates if the algorithm in use is software- or hardware-based.

• big_endian indicates whether the platform is big endian or not.

Implementation details

By default, if your CPU doesn’t have CRC32C hardware support, the package will fallback to use a software implementation of the crc32c checksum algorithm. This behavior can be changed by setting the CRC32C_SW_MODE environment variable to one of the following values:

• auto: same as if unset, will eventually be discontinued.

• force: use software implementation regardless of hardware support.

• none: fail to import the module with an ImportError if no hardware support is found (old 1.x default behavior).

The software algorithm is based on Intel’s slice-by-8 package, with some adaptations done by Evan Jones and packaging provided by Ferry Toth. Further adaptations were required to make the code more portable and fit for inclusion within this python package.

The Intel SSE 4.2 algorithm is based on Mark Adler’s code, with some modifications required to make the code more portable and fit for inclusion within this python package.

The ARMv8 hardware implementation is based on Google’s crc32c C++ library.

Copyright

This package is copyrighted:

ICRAR - International Centre for Radio Astronomy Research
(c) UWA - The University of Western Australia, 2017
Copyright by UWA (in the framework of the ICRAR)

The original slice-by-8 software algorithm is copyrighted by:

Copyright (c) 2004-2006 Intel Corporation - All Rights Reserved

Further adaptations to the slice-by-8 algorithm previous to the inclusion in this package are copyrighted by:

Copyright 2008,2009,2010 Massachusetts Institute of Technology.

The original Intel SSE 4.2 crc32c algorithm is copyrighted by:

Copyright (C) 2013 Mark Adler

The crc32c ARMv8 hardware code is copyrighted by:

Copyright 2017 The CRC32C Authors

A copy of the AUTHORS file from Google’s crc32c project as it was at the time of copying the code is included in this repository.

License

This package is licensed under the LGPL-2.1 license.

The original slice-by-8 software algorithm is licensed under the 2-clause BSD licence.

Further modifications to the slice-by-8 software algorithm are licensed under a 3-clause BSD licence

The original Intel SSE 4.2 crc32c algorithm’s code is licensed under a custom license embedded in the crc32c_adler.c file.

The original crc32c ARMv8 hardware code is licensed under a 3-clause BSD license.