LZ4 Depacker

[42bs]

Similar to the depacker collection in new_bjl, I started also one for the Lynx.

Started with lz4.

 

https://github.com/42Bastian/new_bll/tree/master/demos/depacker

[sage]

you may want to check:

 

https://github.com/bspruck/exolynx

https://github.com/bspruck/pucrunch

 

for more advanced unpacking codes for lynx.

[42bs]

Did you compare one of those with for example ZX0 (which uses also Elias coding)?

[42bs]

Compared pucrunch with zx0:
 

Original:    7381 bytes
LZ4:         3959 bytes
TurboPacker: 3749 bytes
ZX0:         3074 bytes
Pucrunch:    3460 bytes
Shrinkler:   2866 bytes
upkr:        2778 bytes

 

Edited October 18, 2022 by 42bs
add lz4 and tp

[sage]

I always used the one which was (a) working on the lynx (b) fast enough.

And depending on use case, stream depacking vs load-and in place depack.

 

pucrunch was easiest to use for a while and easily integrated into file-loading. now exo is my favorite.

Edited October 18, 2022 by sage

[42bs]

1 hour ago, sage said:

I always used the one which was (a) working on the lynx (b) fast enough.

And depending on use case, stream depacking vs load-and in place depack.

That's why want to build a library of depackers, as none fits all.

[42bs]

On 10/18/2022 at 9:55 AM, sage said:

you may want to check:

 

https://github.com/bspruck/exolynx

https://github.com/bspruck/pucrunch

 

for more advanced unpacking codes for lynx.

I tried exomizer, but the unpacked data misses the first two bytes?! As if they were not written.

I "faked" it, by setting `zp_dest_lo` two bytes higher and stored the first two bytes _after_ decrunch by hand and the sprite is ok.

 

I used the file directly from github

 

[sage]

strange. i has been a while since i used it for bad apple video. maybe i added a workaround or used some special command line parameters? Maybe I just forgot.

 

you use "level" as type?

Edited October 23, 2022 by sage

[42bs]

I will start over from original C code and see how this behaves.

Not the fastest though: 341ms.

 

[42bs]

28 minutes ago, sage said:

strange. i has been a while since i used it for bad apple video. maybe i added a workaround or used some special command line parameters? Maybe I just forgot.

 

you use "level" as type?

Tried both, level and raw.

[sage]

Just checked my script, you are right. the first two bytes are not packed 😛

I added two zero bytes in front of each binary i wanted to pack.

Long live the documentation :facepalm:

 

 

 

[42bs]

22 minutes ago, sage said:

Just checked my script, you are right. the first two bytes are not packed 😛

I added two zero bytes in front of each binary i wanted to pack.

Long live the documentation :facepalm:

Strange. There is no hint in the original sources. But good to know. Thanks for checking.

[42bs]

On 10/18/2022 at 9:55 AM, sage said:

you may want to check:

 

https://github.com/bspruck/exolynx

https://github.com/bspruck/pucrunch

 

for more advanced unpacking codes for lynx.

Exo speed improved by 15%, but size also increased by 8%

[42bs]

On 10/24/2022 at 8:55 PM, sage said:

Just checked my script, you are right. the first two bytes are not packed 😛

I added two zero bytes in front of each binary i wanted to pack.

Long live the documentation :facepalm:

IIRC A8 files have the load address in the first two bytes, that's likely the reason those are skipped.

[Ecernosoft]

On 10/23/2022 at 7:47 AM, 42bs said:

I will start over from original C code and see how this behaves.

Not the fastest though: 341ms.

 

C? Why not assembly?

 

(Just asking)

Edited November 27, 2022 by Ecernosoft

[42bs]

5 hours ago, Ecernosoft said:

C? Why not assembly?

 

(Just asking)

I used the C code to understand the depacker. Not assembly code someone else wrote which is already optimized and thus harder to understand. I did not compile the C code.

 

Edited November 27, 2022 by 42bs

[Ecernosoft]

14 hours ago, 42bs said:

I used the C code to understand the depacker. Not assembly code someone else wrote which is already optimized and thus harder to understand. I did not compile the C code.

 

That's about what I thought. 

What's funny is I have a hard time understanding C and it's Void and struct stuff and all. I took Computer science (The second semester we learn C++) for that reason. Yet, I can understand assembly fine, as long as it's source code and NOT a dissasembly. (That's for 6502 and it's variants, as well as some Z80)

[sage]

Just want to mention another packer/unpacker for the lynx. This code is used in/for Lemmings:

 

https://github.com/tobiasvl/rnc_propack

[Heaven/TQA]

On 10/26/2022 at 8:54 AM, 42bs said:

IIRC A8 files have the load address in the first two bytes, that's likely the reason those are skipped.

CBM machines  a8 have 6 or 4…