BALZ with Optimal Parsing

[encode]

Hi folks!

Please welcome the brand new release of the BALZ file compressor. It's kind of experimental & closed source release (since it contains a code from my new upcoming LZ-based file compressor).
As a note, the BALZ v1.20 and earlier versions use Flexible Parsing with some improvements.
And of course this new version is fully compatible with older open source versions!

The results:

english.dic
balz120 cx -> 766,554 bytes
balz150 cx -> 630,224 bytes

fp.log
balz120 cx -> 577,291 bytes
balz150 cx -> 542,684 bytes

world95.txt
balz120 cx -> 588,480 bytes
balz150 cx -> 579,160 bytes

calgary.tar
balz120 cx -> 871,048 bytes
balz150 cx -> 859,493 bytes

ENWIK8
balz120 cx -> 28,232,824 bytes
balz150 cx -> 27,831,025 bytes

ENWIK9
balz120 cx -> 245,218,274 bytes
balz150 cx -> 241,575,273 bytes

[comp1]

Good work Ilya!

Thanks for sharing.

[encode]

UPDATE:
Uploaded a faster compile!

[kaitz]

newsgroups archive - news051f1.tar
cx option

36472832 -> 6739213 in 10.958s v1.20
36472832 -> 6646471 in 72.3 sec v1.50

[Sportman]

enwik10:
2,559,864,772 bytes, 456.650 sec. - 201.812 sec., balz c (v1.50)
2,340,167,057 bytes, 6,080.800 sec. - 194.725 sec., balz cx (v1.50)

[lz77]

By the way: I'm just comparing (un)compression rates and speed some LZ based compressors with timer.exe 3.01 by I. Pavlov. I select "User Time" from output of timer.exe. For example, if I repeat the test 50 times and only 1 time I see a minimum result of 2.3 seconds should I choose it for comparison?

Here are results on I3 5005U (can AVX2) 2 HGz, RAM DDR3L 1.6GHz, 1 core, 1 thread:

    enwik8  compress/uncompress time         | ratio

my program   2.250/0.562                     | 0.40452
zstd         1.234/0.265                     | 0.40749
brotli       0.921/0.687                     | 0.41952
lzturbo      1.515/0.515                     | 0.35583
lz4x32_1.4   0.718/0.203                     | 0.53511
lz4x64_1.9.2 0.437/0.031                     | 0.57274
ulz          1.671/0.109                     | 0.47674

If I port my program on asm it may work 2 times faster. My program is pure LZ-type compressor by my idea.
​Pardon for my English.