Best Compression Methods for 'Specific' Games. Q&A

[KaktoR]

For my earlier test, about unpacking PCK files, encoding wem's and such, here is a comparisson for ACOrigins audio files. Excluding processing banks.


SFX + English audio

original
input 3.795.077.936 bytes

Compressed 41 files, 3,795,077,936 => 3,041,411,552 bytes. Ratio 80.14%
Compression time: cpu 2.05 sec/real 223.01 sec = 1%. Speed 17.02 mB/s
All OK

Time 4 minutes

---

diffz+ww files + nulled wem blovks inside PCK files
input 7.422.151.129 bytes)

Compressed 148,175 files, 7,422,151,129 => 2,979,433,554 bytes. Ratio 40.14%
Compression time: cpu 23.94 sec/real 763.29 sec = 3%. Speed 9.72 mB/s
All OK

Time 1 hour 20 minutes

~25%-33% +time per language

[L33THAK0R]

Originally Posted by KaktoR

For my earlier test, about unpacking PCK files, encoding wem's and such, here is a comparisson for ACOrigins audio files. Excluding processing banks.


SFX + English audio

original
input 3.795.077.936 bytes

Compressed 41 files, 3,795,077,936 => 3,041,411,552 bytes. Ratio 80.14%
Compression time: cpu 2.05 sec/real 223.01 sec = 1%. Speed 17.02 mB/s
All OK

Time 4 minutes

---

diffz+ww files + nulled wem blovks inside PCK files
input 7.422.151.129 bytes)

Compressed 148,175 files, 7,422,151,129 => 2,979,433,554 bytes. Ratio 40.14%
Compression time: cpu 23.94 sec/real 763.29 sec = 3%. Speed 9.72 mB/s
All OK

Time 1 hour 20 minutes

~25%-33% +time per language

I wonder if it'd be faster and result in a better ratio to instead extract all the relevant files and pack each PCK as one singular block, then patch the block once its been decoded and reverted back to its original form with the original PCK file. I've found especially with smaller blanked out sections within an archive the final compressor sometimes doesn't achieve as optimal a result as it would when working on a singular patch file. I had this exact issue with Just Cause 3, when I initially patched each FSB (of which there were thousands) and then wrote the now patched FSBs into the archives, which resulted in a fairly lackluster result, compared to combining each valid file extracted from the FSBs into one singular block and creating a hdiffz patch between that block and the original "*.arc" archive.