TC 5.0dev2 released!

[encode]

What's new:
+ Improved literal/match length coding
+ Optimized and minimized source code
+ Simplified decoder

Download TC 5.0dev2 (23 KB)

[ggf31416]

<u>sharks12.txt</u> 5.582.655 bytes
TC 5.0dev2 e Process Time = 2.015 - 1.851.546 bytes
TC 5.0dev2 d Process Time = 2.218

<u>gcc-html.tar</u> 2.754.560 bytes
TC 5.0dev2 e Process Time = 0.453 - 404.957 bytes
TC 5.0dev2 d Process Time = 0.500

<u>TC 5.0dev1 and TC 4.3 Results</u>

[encode]

Performance on Large Text Compression Benchmark:

Layout:
<compressed size> (<comp. time>/<decomp. time>)

ENWIK8:
32,417,139 bytes (25 sec/25 sec)

ENWIK9:
283,039,249 bytes (270 sec/244 sec)

System:
P4 3.0 GHz, 1 GB RAM, Windows XP Professional SP2

[encode]

New tc used lzw+ari?

[encode]

No... Look at the map of versions of TC and algorithms used:

TC 1.0-1.1 - QLZ (My own simple algorithm)
TC >1.2 - Byte-oriented huffman (My own simple algorithm)
TC 2.x - PPMCB variant
TC 3.x - LZC variant
TC 4.x - LZC variant with large dictionary
TC 5.x - LZP variant

Release date of the first version - <u>summer of 2004</u>

[encode]

Also note, probably I add this new fast engine to PIMPLE:
<u>
Estimated preset map:</u>

Fastest - Variant of current TC engine
Normal - Current Fastest preset
High - Current Normal preset
Maximum - Current High preset but with Model Order of 5
Extreme - Maximum preset but with 256 MB Model Size

Or I can add just support for TC files - and you will able to choose between PIM and TC compression...

Or I can add just support for TC files - and you will able to choose between PIM and TC compression...
it's better...
Than differs TC from LZPX?

[encode]

Current TC (dev2) is just some kind of simplified and optimized LZPX... Currently I'm experimenting with engine to find what can I do to improve compression, but keeping speed and simplicity.

Also note, probably I add this new fast engine to PIMPLE:

Estimated preset map:

Fastest - Variant of current TC engine
Normal - Current Fastest preset
High - Current Normal preset
Maximum - Current High preset but with Model Order of 5
Extreme - Maximum preset but with 256 MB Model Size

I like this idea best.

Current TC (dev2) is just some kind of simplified and optimized LZPX... Currently I'm experimenting with engine to find what can I do to improve compression, but keeping speed and simplicity.

Hi,
I think you can implement changes similar to LZPXj 1.0a-e. It would improve compression on files containing precompressed data.
This changes were simple and speed was not affected. If you want I can do this if you send me source.

Hi,
I think you can implement changes similar to LZPXj 1.0a-e. It would improve compression on files containing precompressed data.
This changes were simple and speed was not affected. If you want I can do this if you send me source

This seems like a good idea.

[encode]

Currently, I won't add this feature, at least temporarily. Look at the features and phylosophy of this compressor:
+ Developed and maintained only by me
+ As simple as it possible
+ Well structured, optimized and small source code
+ Only the most significant part of the engine will be included - i.e. if some trick improves compression just a little - I won't include this feature
Currently, I improve the base of the engine - matching portion, index table and literal/match length coding. - To make this parts simple, efficient and well structured.

[encode]

Also special thanks to Jan for "simplelzpxj".

Also special thanks to Jan for "simplelzpxj".

What is this?

[encode]

Also special thanks to Jan for "simplelzpxj".

What is this?

It is just improved LZPX 1.5b. (Improved compression on precompressed data)