K
kuwanger
I've been working on some python code to have an open source apack
compression (http://www.ibsensoftware.com/products_aPACK.html is a link
to the original proprietary version). Just to clarify the legality of
all of this, from decompressor source written by Dwedit in arm/thumb
assembly I ended up writing unapack.c and then unapack.py
(http://kuwanger.malfunct.net/unapack.py). The compressor is based on
studying the C/Python versions I wrote. So, I've got a working
compressor (one in C and one in Python). But, my Python version is
incredibly slow. And my attempts to have better compression in my C
version has produced an incredibly messy/slow code, which is why I'd
rather focus on making my current Python version faster.
So, here is what I have so far,
"http://kuwanger.malfunct.net/apack.hash.py". Currently it's setup
under profiling and does a lot of printing out. But even without that,
the search part of the algorithm is incredibly slow (~60s to compress
80K on my 1.6GHz Athlon) while adding on attempts to optimize the
compress slows it down further (~90s to compress 80K on my 1.6GHz
Athlon).
So, I'd be curious if others could look at the source and perhaps make
some suggestions, such as if I'm just going about it horribly wrong.
I'd also greatly welcome someone else joining me on this project, as my
efforts to optimize haven't produced results nearly as good as the
official compressor. Comments are greatly welcome.
Kuwanger
compression (http://www.ibsensoftware.com/products_aPACK.html is a link
to the original proprietary version). Just to clarify the legality of
all of this, from decompressor source written by Dwedit in arm/thumb
assembly I ended up writing unapack.c and then unapack.py
(http://kuwanger.malfunct.net/unapack.py). The compressor is based on
studying the C/Python versions I wrote. So, I've got a working
compressor (one in C and one in Python). But, my Python version is
incredibly slow. And my attempts to have better compression in my C
version has produced an incredibly messy/slow code, which is why I'd
rather focus on making my current Python version faster.
So, here is what I have so far,
"http://kuwanger.malfunct.net/apack.hash.py". Currently it's setup
under profiling and does a lot of printing out. But even without that,
the search part of the algorithm is incredibly slow (~60s to compress
80K on my 1.6GHz Athlon) while adding on attempts to optimize the
compress slows it down further (~90s to compress 80K on my 1.6GHz
Athlon).
So, I'd be curious if others could look at the source and perhaps make
some suggestions, such as if I'm just going about it horribly wrong.
I'd also greatly welcome someone else joining me on this project, as my
efforts to optimize haven't produced results nearly as good as the
official compressor. Comments are greatly welcome.
Kuwanger