# mblzp
An implementation of the McIlroy-Tamayo Lempel-Ziv algorithm written in Malbolge.
Calgary Corpus (3,251,493 bytes) ratings:
- `mblzp`: 1,909,033 bytes ("vanilla" McIlroy-Tamayo LZP).
- `FSE`: 1'829'383 bytes ("vanilla" FSE entropy coding).
- `mblzp` + `FSE`: 1'447'671 bytes (McIlroy-Tamayo LZP with FSE entropy coding).
- `lz4 -1`: 1'690'181 bytes ("vanilla" LZSS w. minimum match of 4).
- `lz4 -9`: 1'257'610 bytes.
- `gzip -1`: 1'238'908 bytes (LZSS with Huffman entropy coding).
- `gzip -9`: 1'059'335 bytes.
_Remark:_ The LZP algorithm employed by this program is generally considered to be a preprocessor to a more sophisticated entropy coder. Hence some compression is left on the table. Ddespite being significantly less complicated, when paired with an entropy coder (FSE), the McIlroy-Tamayo LZP algorithm is able to perform within approx. 5% of the compression ratio of `gzip -1`.
_Remark 2:_ This is not a serious data compressor. Don't do this at home.
Nominally, it requires approximately 20MB of physical memory to operate. The program can be run using `fast20` as follows:
```
$ fast20 lz-c.mb < in > out
$ fast20 lz-d.mb < out > org
```
`fast20` is best compiled as:
```
$ clang-15 fast20.c -march=native -mtune=native -flto -Ofast -static
```
However, your mileage may vary as the performance of fast20 is quite unpredictable.
List of all files in the repository:
- fast20.c - the source code to the fast20 malbolge interpreter, originally bundled with malbolge-lisp
- fast20_x86-64 - a precompiled binary of the fast20 interpreter for x86-64
- README - this file
- lz-c.mb (4.71MB) - The compressor
- lz-d.mb (4.58MB) - The decompressor
- lz-c.sq.bz3 (691 bytes) - The compressor, reimplemented in SUBLEQ and compressed with bzip3. Original size 133KB. Compression ratio 0.52%, 0.04bpb.
- lz-d.sq.bz3 (835 bytes) - The decompressor, reimplemented in SUBLEQ and compressed with bzip3. Original size 134KB. Compression ratio 0.62%, 0.05 bpb.
- calgary.tar - The Calgary Corpus, a collection of files used for benchmarking compression algorithms
- calgary-lzp.tar - The Calgary Corpus, compressed with mblzp
Benchmarking on a AMD Ryzen 7 5700U CPU demonstrates the compression speed of approx. 50 bytes per second. Decompression is slower.
_Remark 3:_ This compression algorithm, alongside others, is explained in great detail in my upcoming book "Data Compression in Depth". Stay tuned and follow https://palaiologos.rocks/ for updates.