I've come across quite a lot of firmware on embedded Linux devices that uses LZMA (the xz compression algorithm) to compress the kernel, u-boot, and/or filesyst

I've come across quite a lot of firmware on embedded Linux devices that uses LZMA (the xz compression algorithm) to compress the kernel, u-boot, and/or filesystems. One memory optimisation for these, as they are typically being decompressed straight into RAM, is for the decompressor to refer to its output as the dictionary rather than building a separate one, as would be the case in decompressing to the network or disk.