Results 1 to 14 of 14

Thread: TC 5.1dev4 is here! [!]

  1. #1
    The Founder encode's Avatar
    Join Date
    May 2006
    Location
    Moscow, Russia
    Posts
    3,982
    Thanks
    377
    Thanked 351 Times in 139 Posts
    Okay, here are the new version of TC. It uses a fast PAQ inspired encoder instead of small PPM. Enjoy!

    Link:
    Download TC 5.1dev4 (25 KB)


  2. #2
    The Founder encode's Avatar
    Join Date
    May 2006
    Location
    Moscow, Russia
    Posts
    3,982
    Thanks
    377
    Thanked 351 Times in 139 Posts
    Note, this is a DRAFT version - i.e. the most of code must be improved/optimized. Also, if you compare this version against other compressors (LZPXJ, for example) consider, this TC have no filters/preprocessors.


  3. #3
    Moderator

    Join Date
    May 2008
    Location
    Tristan da Cunha
    Posts
    2,034
    Thanks
    0
    Thanked 4 Times in 4 Posts
    Thanks Ilia!

    It uses a fast PAQ inspired encoder instead of small PPM. Enjoy!

    This looks Interesting!

  4. #4
    Moderator

    Join Date
    May 2008
    Location
    Tristan da Cunha
    Posts
    2,034
    Thanks
    0
    Thanked 4 Times in 4 Posts
    Test file: ENWIK8

    TC 5.1dev3

    28.6 MB (30,037,190 bytes)

    Kernel Time = 1.281 = 00:00:01.281 = 0%
    User Time = 181.968 = 00:03:01.968 = 98%
    Process Time = 183.250 = 00:03:03.250 = 98%
    Global Time = 185.672 = 00:03:05.672 = 100%


    TC 5.1dev4

    27.5 MB (28,846,242 bytes)

    Kernel Time = 1.250 = 00:00:01.250 = 0%
    User Time = 257.468 = 00:04:17.468 = 98%
    Process Time = 258.718 = 00:04:18.718 = 98%
    Global Time = 262.078 = 00:04:22.078 = 100%


    PX 1.0

    23.8 MB (24,971,871 bytes)

    Kernel Time = 1.500 = 00:00:01.500 = 0%
    User Time = 389.703 = 00:06:29.703 = 98%
    Process Time = 391.203 = 00:06:31.203 = 98%
    Global Time = 396.141 = 00:06:36.141 = 100%


    PIMPLE 1.0b10c

    20.9 MB (21,947,483 bytes)

    Kernel Time = 7.468 = 00:00:07.468 = 0%
    User Time = 1557.953 = 00:25:57.953 = 98%
    Process Time = 1565.421 = 00:26:05.421 = 98%
    Global Time = 1585.359 = 00:26:25.359 = 100%


    LZPX 1.5b

    31.0 MB (32,553,957 bytes)

    Kernel Time = 1.140 = 00:00:01.140 = 2%
    User Time = 37.812 = 00:00:37.812 = 95%
    Process Time = 38.953 = 00:00:38.953 = 98%
    Global Time = 39.406 = 00:00:39.406 = 100%


    LZPXJ 1.2g

    24.6 MB (25,833,249 bytes)

    Kernel Time = 1.156 = 00:00:01.156 = 0%
    User Time = 196.281 = 00:03:16.281 = 98%
    Process Time = 197.437 = 00:03:17.437 = 98%
    Global Time = 199.891 = 00:03:19.891 = 100%

  5. #5
    The Founder encode's Avatar
    Join Date
    May 2006
    Location
    Moscow, Russia
    Posts
    3,982
    Thanks
    377
    Thanked 351 Times in 139 Posts
    Testing compression of dev4 on text data... Hmm... TC for binary data!

  6. #6
    Moderator

    Join Date
    May 2008
    Location
    Tristan da Cunha
    Posts
    2,034
    Thanks
    0
    Thanked 4 Times in 4 Posts
    Testing compression of dev4 on text data... Hmm... TC for binary data!

    I thought the results from dev4 on this file were really good!

  7. #7
    The Founder encode's Avatar
    Join Date
    May 2006
    Location
    Moscow, Russia
    Posts
    3,982
    Thanks
    377
    Thanked 351 Times in 139 Posts
    The difference in performance on ENWIK8 of dev3 and dev4 is really small. Try to compress any EXE file. The difference is killer. Furthermore, TC without any filters simply beats LZPXJ with EXE filter on the most EXE files. So results on ENWIK8 not shows the full power of the new TC. The reason TC not performs on text data as well as on binary data is low order CM (PAQ inspired encoder), I simply can add a higher order model to beat many programs on the most files, but the decompression speed will be really affected. Anyway, probably I really add this, along with the new EXE filter to make TC one of the top compressors and to keep myself as a well known author.


  8. #8
    Moderator

    Join Date
    May 2008
    Location
    Tristan da Cunha
    Posts
    2,034
    Thanks
    0
    Thanked 4 Times in 4 Posts
    Anyway, probably I really add this, along with the new EXE filter to make TC one of the top compressors and to keep myself as a well known author.

    Excellent news!

  9. #9
    The Founder encode's Avatar
    Join Date
    May 2006
    Location
    Moscow, Russia
    Posts
    3,982
    Thanks
    377
    Thanked 351 Times in 139 Posts
    News about TC. I completely rewrite the entire program, including arithmetic encoder and modeller. Also I add the unique combination of semi-stationary and stationary models. Plus I add the new EXE filter. As a result, now TC have a higher compression compared to dev4 and really higher compresison on executables (or TAR files with executable files). Estimated release date: <u>Friday 13</u>


  10. #10
    Moderator

    Join Date
    May 2008
    Location
    Tristan da Cunha
    Posts
    2,034
    Thanks
    0
    Thanked 4 Times in 4 Posts
    News about TC. I completely rewrite the entire program, including arithmetic encoder and modeller. Also I add the unique combination of semi-stationary and stationary models. Plus I add the new EXE filter. As a result, now TC have a higher compression compared to dev4 and a really higher compresison on executables (or TAR files with executable files).

    The power of that new engine should be amazing!


    Estimated release date: Friday 13

    That's Scary!!!

  11. #11
    The Founder encode's Avatar
    Join Date
    May 2006
    Location
    Moscow, Russia
    Posts
    3,982
    Thanks
    377
    Thanked 351 Times in 139 Posts
    The power of that new engine should be amazing!

    Do not expect too much. But note, this engine is the best engine ever made by myself. The performance on binary and executables is capable to PIMPLE. Furthermore on some files this new TC easily outperfotms PIMPLE and even 7-Zip (and WinRAR). One disadvantage of this engine is relatively low compression on text files, as a result new TC not performs on ENWIK9 as well as other powerful compressors. (DARK as example). But if you try to compress one TAR with lots of different files, you'll see the true power of TC!


  12. #12
    Moderator

    Join Date
    May 2008
    Location
    Tristan da Cunha
    Posts
    2,034
    Thanks
    0
    Thanked 4 Times in 4 Posts
    Surely this new engine must give TC a place in the top ten over at Werner's site!

  13. #13
    Guest
    I want TC compressing significantly better especially this types of data:

    random(precompressed):
    TC 5.1dev4 464101 -> 496624 (+7%)
    ZIP 464101 -> 464356 (+0%)
    RAR 464101 -> 464171 (+0%)

    wav(stereo):
    TC 5.1dev4 6994092 -> 6829129 (-2.5%)
    ZIP 6994092 -> 6708743 (-4%)
    RAR 6994092 -> 5347958 (-23.5%)

  14. #14
    The Founder encode's Avatar
    Join Date
    May 2006
    Location
    Moscow, Russia
    Posts
    3,982
    Thanks
    377
    Thanked 351 Times in 139 Posts
    Surely this new engine must give TC a place in the top ten over at Werner's site!

    AFAIK, at least new dev5 outperforms WinZip 10.0 and DARK at SFC. And it is one of the top compressors tested without extra tuning for each filetype (For example, programs witch uses extra tuning are 7-Zip, WinRAR and many others.) Also this is not the last version of TC - i.e. next versions will have again higher compression.

    I want TC compressing significantly better especially this types of data:
    random(precompressed):


    In this test RAR and ZIP just stores the compressed data. dev5 will have the same problems (expanded size) with random data. Anyway, in next dev6 version I will add the SSE feature. Note with SSE, TC can compress a10.jpg and many more precompressed data. Also such kind of data will not be expanded too much in worst case.

    wav(stereo):

    Looks like this wave file have a lots of HI FREQs, as a result this file is a hard compressible. Note RAR uses special audio compression and I'm not planning to include such audio compression to TC. Anyway, again with SSE, TC will have really higher compression on such files.


Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •