libjpeg-turbo is a high-speed version of libjpeg for x86 and x86-64 processors. It uses SIMD instructions (MMX, SSE, SSE2) to accelerate baseline JPEG compression and decompression. libjpeg-turbo is generally 2-4 times as fast as the unmodified version of libjpeg. It also includes a wrapper library for the TurboJPEG API used by VirtualGL and TurboVNC. It was originally based on libjpeg/SIMD but has improved support for Mac OS X, 64-bit platforms, 32-bit and big endian pixel formats (RGBA/BGRA/ABGR/ARGB), accelerated Huffman encoding/decoding, and various other fixes.
The ROHC library provides an easy and robust way for applications to reduce their bandwidth usage on network links with limited capacity or expensive costs. Headers of network packets are compressed with the ROHC protocol and algorithms. The ROHC protocol is very efficient for compressing VoIP streams which use RTP as a transport protocol. ROHC is also suitable for compressing IP-only (IPv4 or IPv6) flows, UDP flows, and many others.
crunch/crnlib is a lossy texture compression tool and library for developers that distributes and uses content in the DXT1/5/N or 3DC/BC5 compressed mipmapped texture formats. crnlib can compress mipmapped 2D textures and cubemaps to .8-1.25 bits/texel, and normal maps to 1.75-2 bits/texel. crnlib's quality and performance is competitive to transform-based solutions, or other offline/real-time DXTn compressors such as squish or ATI_Compress. crnlib implements a new form of "clustered" DXTn compression, with a compressed texture data format that was carefully designed to be very quickly transcodable directly to raw DXTn texture bits with no intermediate recompression step or individual pixel-level operations. The typical single threaded transcode to DXTn rate is equivalent to 100-250 megatexels/sec. Fast random access to individual mipmap levels is supported.