[A83] Re: Is compression possible?


[Prev][Next][Index][Thread]

[A83] Re: Is compression possible?






On Wed, 24 Apr 2002, Joe Pemberton wrote:

>
>
> >
> >
> > The exprience on the TI-86 would seem to suggest that compression is
> > feasible.  For the TI-86, there are already (at least)  two compression
> > programs:  Lite86 II (which only decompresses on-calc) and Bloat (which
> > has an on-calc compressor as well).  The decompression code is only a few
> > dozen bytes.  Even programs that are only a few kilobytes in size can be
> > made substantially smaller if they made into auto-extracting compressed
> > programs.  Since the program runs as normal, this is no problem for users.
> >
> > Since the TI-83 Plus has more memory than a TI-86, programs for it should
> > be expected to (one way or another) end up larger than the TI-86
> > equivalents, and thus even more suitable for compression.  Of course this
> > might not be the case for the regular TI-83, where not only do programs
> > have to be smaller, but you can't really find a lot of temporary storage
> > to use.
> >
> >
> Those programs used very large buffers (16k+) to compress and decompress
> things (which is a nice thing about the 86).  I was talking to Kirk Meyer
> about this very subject and he was pretty sure that implementing the
> compression algorithm from lite86 at least would not be feasible on the 83+.
> The 83se, however, has those extra ram pages that could be used.

Decompression, though, is no problem since there is already Z80 code to do
that using only source and destination.  So as long as you consider all of
the RAM to be temporary space because everything is stored in archive, it
would work fine if compressed size + uncompressed size is no larger than
the amount of free RAM.  And if you could modify it to decompress straight
from the archive, not much extra memory would be needed at all.

It seems as if you could use the whole RAM for compression too.  It is
even possible to compress using only one buffer.  But unfortunately speed
is probably a real problem if you have to search backwards a long way for
repeated data and don't have enough storage for lookup tables.




References: