What's the command to compress the first 256 bytes (0-255), and the commando to the next 256 bytes (256-511) and so on? The docs aren't clear about this.
There is no way to specify how many bytes should be compressed, only how many bytes to use as a prefix.
albs_br wrote:
Also, on my tests, only standard decompressor is working, the other three aren't.
why do we want to compress each chunk separately? Can't the decompression routine not just be run until it decompresses X bytes, and then stop (saving the state to some memory location), to then resume it later to decompress another chunk?
There is no way to specify how many bytes should be compressed, only how many bytes to use as a prefix.
That's sad, as my need is to read and spit 256 bytes to VRAM. Background image for just one level of msx-wings is taking 256kb, compressed it will be under 80kb, but I need to decompress in steps of 256 bytes, one at each 3 frames. I have some spare cpu left. No way to decompress all before level starts as it would need lots of RAM. Target is standard 64kb of RAM.
That's not a very unusual need. Many people should have had this need before. Or am I the first?
I suppose you need to write your own decompressor then. One that uses a circular lookback buffer of certain size (maybe 1, 2 or 4 kb whatever you can spare or give decent results) and use the compressor with a maximum/fixed lookback offset setting.
Does that mean that from the dozens of packars available (for example, one list: https://github.com/uniabis/z80depacker), none of them are able to do partial depackings as I need?
Step and Jäästä used a custom static Huffman decoder to unpack each string shown on the screen separately (so even single words). The encoder is implemented in Python. However:
- this uses a fixed decompression tree, which takes a lot of space in case of several compressed different bytes. This structure should be included in the source code (maybe 2KB of data at most?).
- it allows compressing at most 254 different values (no escape characters)
- not all cases can be compressed with this, because I use only one byte to signify jumps in the code tree.
FWIW, the source codes are at github; the compressor in /python and decompressor is /src/main.asm:Decompress
