Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.compression > #3236

Re: LZSS for 8-bit computers: I'm doing well but want to do better

Show all headers | View raw

On 10/16/2015 10:29 AM, Robert Wessel wrote:
> On Tue, 13 Oct 2015 16:06:23 -0700 (PDT), Harry Potter
> <rose.joseph12@yahoo.com> wrote:
>
>> On Tuesday, October 13, 2015 at 7:04:26 AM UTC-4, Harry Potter wrote:
>>> I tweaked the program some more, and it seems to work *better* than *7-Zip*!!!   :D
>>
>> I was wrong.  Two bug-fixes *killed* the compression ratio!  :/
>
>
> You've done this repeatedly.  Why isn't checking the decompression
> part of your basic checkout of anything before you publish?  Given
> what you're doing I'd expect you'd have put together a good sized and
> fully automated test to see if things actually work.
>

yep. write compressor and decompressor in a pair, check stats for one 
while also verifying the other produces the expected results.

for data compression, for the files tested, it makes sense to verify 
that round-trip things remain intact. ex: compress, then decompress, and 
run a memory comparison.

also maybe try to run enough assorted data to detect any bugs.


lossy compression is harder. simple is using PSNR as the main heuristic, 
and manually evaluate/investigate any cases where there is a sudden 
change in PSNR (either for bugs or for artifacts).


benchmarks may also be worthwhile to verify whether or not its speed is 
acceptably fast for the use case, ... likewise, one may also track 
whether features have an adverse effect on speed (disproportionate to 
its effect on compression, preferably avoiding features which don't help 
much with compression but which tend to waste a lot of CPU cycles).


granted, different tradeoffs can be made for use-case, for example, what 
is good for a PC may not be good for an embedded ARM controller. 
likewise goes for offline vs real-time processing (offline can generally 
afford to be a bit slower, and something which is run only rarely can be 
slower than something which needs to be run continuously, ...).

or such...

Back to comp.compression | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-10-12 14:08 -0700
  Re: LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-10-13 04:04 -0700
    Re: LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-10-13 16:06 -0700
      Re: LZSS for 8-bit computers: I'm doing well but want to do better Robert Wessel <robertwessel2@yahoo.com> - 2015-10-16 10:29 -0500
        Re: LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-10-16 13:32 -0700
        Re: LZSS for 8-bit computers: I'm doing well but want to do better BGB <cr88192@hotmail.com> - 2015-10-16 17:20 -0500
          Re: LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-10-16 15:56 -0700
            Re: LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-10-17 12:12 -0700
              Re: LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-10-17 12:14 -0700
                Re: LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-10-17 14:01 -0700
                Re: LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-10-18 15:29 -0700
                Re: LZSS for 8-bit computers: I'm doing well but want to do better BGB <cr88192@hotmail.com> - 2015-10-19 10:01 -0500
                Re: LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-10-19 08:15 -0700
                Re: LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-10-20 06:00 -0700
                Re: LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-10-21 07:18 -0700
                Re: LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-10-24 06:22 -0700
                Re: LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-10-24 12:38 -0700
                Re: LZSS for 8-bit computers: I'm doing well but want to do better glen herrmannsfeldt <gah@ugcs.caltech.edu> - 2015-10-25 05:45 +0000
                Re: LZSS for 8-bit computers: I'm doing well but want to do better Harry Potter <rose.joseph12@yahoo.com> - 2015-11-09 03:40 -0800

csiph-web