Quote:
Originally Posted by kj911
Second LOLZ-clone/based compressor, after panker1992 publicly released TARZ? Nice! 
|
Benchmark: same precompressed input for every compressor
Ryzen 7 5800X3D (8C/16T), 64 GB RAM, 16 threads, every output verified byte for byte
Code:
| Call of Duty 2 (5.29 GB) | Mass Effect (8.61 GB)
compressor | ratio | compress | decompress | ratio | compress | decompress
---------------+---------+----------+--------------+---------+----------+-----------
LOLZ ULTRA | 51.28% | 1462 s | 13.0 s | 33.49% | 2390 s | 19.3 s
ROFZ -l1 | 51.20% | 409 s | 46.8 s | 33.83% | 790 s | 85.1 s
ROFZ -l2 | 50.67% | 510 s | 46.8 s | 33.57% | 952 s | 86.9 s
ROFZ -l3 | 50.46% | 669 s | 47.6 s | 33.41% | 1278 s | 86.1 s
ROFZ -l4 | 50.30% | 797 s | 47.8 s | 33.32% | 1700 s | 86.2 s
ROFZ -l5 | 49.21% | 1424 s | 65.4 s | 32.74% | 3095 s | 112.7 s
- Call of Duty 2: xtool 0.9.9 -mzlib -mpreflate -dd3 -> srep 4.90g -m3f -a0/0
- Mass Effect: uelr ulv -> xtool 0.9.9 -mzlib -mzstd -mpreflate -dd3 -> srep 4.90g -m3f -a0/0
- LOLZ ULTRA: -dtb1 -d263 -mtt1 -mt16 -mc1023, ROFZ: rofz c -lN -m85%
Where ROFZ beats LOLZ ULTRA (0.12.2, single-type samples of 32–64 MB, each file compressed on its own, every output verified)
ROFZ -l5 (smallest) – smaller than LOLZ:- PCM audio (WAV 16-bit): −21.1%
- x86-64 executables: −9.5%
- Structured binary data (records, tables): −7.2%
- Text: −5.1%
- BC3 / DXT5 textures: −5.1%
- Raw images (uncompressed pixels): −3.9%
- Maps (BSP): −3.3%
- BC7 textures: −2.9%
- BC1 / DXT1 textures: −2.4%
- x86 executables (32-bit): −2.2%
- UE3 packages (Mass Effect, after uelr): −1.0%
- Already compressed data: −0.6%
- Animations: −0.4%
I am still working on optimization