csnappy

csnappy copied to clipboard

Google Snappy, available at https://google.github.io/snappy/ is a compression library designed for speed rather than compression ratios.

It is not a new concept by far. The Linux kernel currently uses LZO as the default fast compressor. Snappy works faster, though achieves less compression.

Google's code is written in C with a lot of C++. Some of the more interesting features that rely on C++ appear to have been elided from the open source version of Snappy.

Goals: To get the codebase into a shape that can be accepted into the mainline Linux kernel and used with zram (http://code.google.com/p/compcache/). Being able to compress vmlinux and initrd with Snappy is a secondary goal. Same for support in Squashfs and other parts of the kernel that currently support LZO.

Results: I cut out or ported to plain ANSI C the necessary code and headers. To cause less confusion, I call this project (and files) csnappy. The API looks right, but I welcome comments. The code has been tested in kernel-space using a patched zram and it works.

The code has been tested in a qemu emulating a PowerPC Mac and ARMv5TE running Debian Wheezy.

I also use an ARMv6 Android phone for testing.

Testing on other hardware or platforms is welcome.

Note: The userspace tester is a hack, as is the shared library.

Someone wrote a perl wrapper for csnappy: http://search.cpan.org/dist/Compress-Snappy/ https://github.com/gray/compress-snappy

Patch for upstream snappy tester is available: snappy_tester.patch Patch for linux kernel is available: kernel_3_2_10.patch

Benchmark in userspace: userspace_benchmark.txt Benchmark in kernel space with zram: zram_benchmark.txt