DOS ain't dead - Ilia M.'s Gzip 1.2.4 "hack" (improved -9)

Rugxulo

Usono,
20.05.2009, 05:30

Ilia M.'s Gzip 1.2.4 "hack" (improved -9) (Announce)

I finally figured out how to compile Ilia M.'s Gzip 1.2.4 hack with OpenWatcom 1.8 (deflate_max improves -9 without being nearly as slow as 7-Zip). By default his only comes with a Win32 .EXE, which is annoying in pure DOS even with HXRT (needs LFNs). So I've built both DOS and Win32 binaries.

gzhakwat.zip (311k)

Archive: gzhakwat.zip Zip file size: 318072 bytes, number of entries: 7 -rw-a-- 3.0 fat 33957 t- defX 07-Aug-14 00:06 deflate.c -rw-a-- 3.0 fat 1018 t- defX 09-May-19 22:03 enwik8_d.txt -rw-a-- 3.0 fat 1013 t- defX 09-May-19 22:03 enwik8_e.txt -rwxa-- 3.0 fat 1299 t- defX 09-May-19 21:58 wmake-gz.bat -rwxa-- 3.0 fat 54584 b- defX 09-May-19 22:04 gziphakd.exe -rwxa-- 3.0 fat 46080 b- defX 09-May-19 22:03 gziphakw.exe -rw-a-- 3.0 fat 211474 b- stor 09-May-19 22:13 gzip124s.tgz 7 files, 349425 bytes uncompressed, 317356 bytes compressed: 9.2%

> http://www.cs.fit.edu/~mmahoney/compression/text.html#3211
>
> .3211 gzip124hack
> gzip124hack is a modified version of gzip 1.2.4 by Ilia Muraviev,
> Aug. 13, 2007. It uses LZ77. It is a file compressor like gzip,
> except that it does not delete the input file. It improves
> compression by using LZ77 lazy matching with 2 byte lookahead.
> The compressed format is compatible with gzip. -9 selects
> maximum compression.

Basically, it compresses better than typical Gzip:

enwik8 enwik9 gzip124hack 1.2.4 -9 36,273,716 321,050,648 gzip 1.3.5 -9 36,445,248 322,591,995

If you really want to read about it, see the news://comp.compression thread from two years ago.

Rugxulo

Usono,
20.05.2009, 08:16

@ Rugxulo

Ilia M.'s Gzip 1.2.4 "hack" (improved -9)

Post reply

> I finally figured out how to compile Ilia M.'s Gzip 1.2.4 hack with
> OpenWatcom 1.8 (deflate_max improves -9 without being nearly as slow as
> 7-Zip). By default his only comes with a Win32 .EXE, which is annoying in
> pure DOS even with HXRT (needs LFNs). So I've built both DOS and Win32
> binaries.

Note that 1.2.4 is fairly old, and obviously 7-Zip compresses Deflate better, and you can use AdvanceDef (advdef -z4) to recompress your .gz files in place instead of manually unpacking / repacking. BUT ... if you only want a slight bit better compression with reasonable speed, the 1.2.4 hack will work okay.

P.S. I figured out that you can get the same results with DJGPP (GCC) only if you do -DNO_ASM and remove match.S from the makefile (since DJGPP optimizes better). Of course, it doesn't by default have the LFN-aware patch that the latest DJGPP port of Gzip (1.3.12) has. But whatever, just FYI. ;-)