Page 23 of 32 FirstFirst ... 132122232425 ... LastLast
Results 661 to 690 of 956

Thread: Paq8pxd dict

  1. #661
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,155
    Thanks
    706
    Thanked 455 Times in 352 Posts
    4 corpuses scores for paq8pxd v69 fixed - there are proper scores for all files now.
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	paq8pxd_v69_4_corpuses_fixed.jpg 
Views:	67 
Size:	1.81 MB 
ID:	7086  

  2. Thanks:

    kaitz (20th November 2019)

  3. #662
    Member
    Join Date
    Apr 2018
    Location
    Indonesia
    Posts
    74
    Thanks
    15
    Thanked 5 Times in 5 Posts
    Darek may you test bigm on enwik8 or yours benchmark ?
    Thanx

  4. #663
    Member
    Join Date
    May 2008
    Location
    Estonia
    Posts
    509
    Thanks
    208
    Thanked 347 Times in 185 Posts
    paq8pxd_v69
    paq8pxd was forked from paq8px_v69
    Code:
    -Online wrt is present and fixed, now actually decodes word if code  is larger then 1 byte.
    -WordModel form paq8px_v182
      this one is not active in any models, you have to manually add it
      in source. Replaced in model array with blankmodel for now.
      Online wrt is needed so one can detect presence of BT and ET
      for pdf mode in new WordModel.
    -Combined models into arrays to ease mixer inputs/sets/context clutter,
      let models to return usage.
    -Colored console (Windows only!), based on type stream.
    -JPEG compress improvment (APM), small gain.
    -Detect LZW in PDF and ad to CMP stream.
    AVX2 and SSE41 compiles.
    Last source release has other extension for archive.

    Darek tests should still stand, did not change much.
    Attached Files Attached Files
    KZo


  5. Thanks (3):

    brispuss (21st November 2019),Darek (20th November 2019),moisesmcardona (28th November 2019)

  6. #664
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,155
    Thanks
    706
    Thanked 455 Times in 352 Posts
    AVX2 version is about 10% faster on my laptop.

  7. #665
    Member
    Join Date
    Aug 2008
    Location
    NZ
    Posts
    59
    Thanks
    30
    Thanked 11 Times in 7 Posts
    Thanks for the SSE41 compile, as not all of us have AVX2 capable CPU's.

  8. #666
    Member
    Join Date
    May 2008
    Location
    Estonia
    Posts
    509
    Thanks
    208
    Thanked 347 Times in 185 Posts
    Code:
    Change StationaryMap
    Change im24 model, compresses faster.
    detect tiff change
    On cathedral.ppm its about 20kb better and about 100 sec faster.
    Did not test other files on that set.
    Attached Files Attached Files
    KZo


  9. Thanks (2):

    brispuss (3rd February 2020),Darek (3rd February 2020)

  10. #667
    Member
    Join Date
    May 2008
    Location
    Estonia
    Posts
    509
    Thanks
    208
    Thanked 347 Times in 185 Posts
    paq8pxd_v71
    Code:
    small change in im8 compression
    ​reduce compression loss in v70
    Attached Files Attached Files
    KZo


  11. Thanks (2):

    Darek (4th February 2020),Mike (4th February 2020)

  12. #668
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,155
    Thanks
    706
    Thanked 455 Times in 352 Posts
    My testset and 4 corpuses scores for paq8pxd v70. AVX2 version is about 25% faster than previous! Great!
    There are some loses for almost every file except A10.jpg, rafale.bmp. For A.TIF and B.TGA from my testset it set best scores overall and due to these scores total testset score is the best for paq8pxd versions!
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	paq8pxd_v70.jpg 
Views:	42 
Size:	770.7 KB 
ID:	7332   Click image for larger version. 

Name:	4_Corpuses_paq8pxd_v70.jpg 
Views:	37 
Size:	1.86 MB 
ID:	7333  

  13. #669
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,155
    Thanks
    706
    Thanked 455 Times in 352 Posts
    Scores for my testset on paq8pxd_v71. Indeed for some files there is a gain, however for others there are more loses.
    I've made also comparison of v71 vs. v69 version. If not the A.TIF and B.TGA files there would be loses to v69 version.

    @kaitz -> could you implement audio model from paq8px_v183 into paq8pxd?
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	paq8pxd_v71.jpg 
Views:	48 
Size:	770.4 KB 
ID:	7334   Click image for larger version. 

Name:	paq8pxd_v71_compared_to_v69.jpg 
Views:	37 
Size:	771.0 KB 
ID:	7335  

  14. Thanks:

    kaitz (4th February 2020)

  15. #670
    Member
    Join Date
    May 2008
    Location
    Estonia
    Posts
    509
    Thanks
    208
    Thanked 347 Times in 185 Posts
    Quote Originally Posted by Darek View Post
    @kaitz -> could you implement audio model from paq8px_v183 into paq8pxd?
    I know, it has been in my todo list for a long time. Sorry. Will do my best.

    --
    I am sure that on maximumcompression last version will be best. I only test on -8, so...
    KZo


  16. Thanks:

    Darek (4th February 2020)

  17. #671
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,155
    Thanks
    706
    Thanked 455 Times in 352 Posts
    Here are 4 corpuses scores for paq8pxd_v71. Better numbers than v70, however still for these benchmarks (Calgary, Canterbury and Silesia) version v62 keep the first place from paq8pxd serie.
    MaximumCompression have the best score overall from all compressors!
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	4_Corpuses_paq8pxd_v71.jpg 
Views:	34 
Size:	1.97 MB 
ID:	7338  

  18. Thanks:

    kaitz (5th February 2020)

  19. #672
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,155
    Thanks
    706
    Thanked 455 Times in 352 Posts
    Some enwik results for paq8pxd v70 & v71 with comparisons to older versions:

    16'364'165 - enwik8 -s8 by Paq8pxd_v68_AVX2
    16'033'591 - enwik8 -s15 by Paq8pxd_v68_AVX2
    16'755'942 - enwik8.drt -s15 by Paq8pxd_v68_AVX2
    126'958'003 - enwik9_1423 -s15 by Paq8pxd_v68_AVX2


    16'362'891 - enwik8 -s8 by Paq8pxd_v69
    16'032'657 - enwik8 -s15 by Paq8pxd_v69
    16'697'316 - enwik8.drt -s15 by Paq8pxd_v69
    126'950'607 estimated - enwik9_1423 -s15 by Paq8pxd_v69
    130'759'126 estimated
    - enwik9_1423.drt -s15 by Paq8pxd_v69


    16'366'538 - enwik8 -s8 by Paq8pxd_v70_AVX2
    16'033'811 - enwik8 -s15 by Paq8pxd_v70_AVX2
    16'699'314 - enwik8.drt -s15 by Paq8pxd_v70_AVX2
    127'125'708 estimated - enwik9_1423 -s15 by Paq8pxd_v70_AVX2 - for pure enwik9 (not DRT) after block segmentation and 0.16% passed - program quit
    132'767'857 estimated
    - enwik9_1423.drt -s15 by Paq8pxd_v70_AVX2


    16'363'164 - enwik8 -s8 by Paq8pxd_v71_AVX2
    16'031'278 - enwik8 -s15 by Paq8pxd_v71_AVX2
    16'696'326 - enwik8.drt -s15 by Paq8pxd_v71_AVX2
    127'456'638 estimated - enwik9_1423 -s15 by Paq8pxd_v71_AVX2 - for pure enwik9 (not DRT) after block segmentation and 0.16% passed - program quit - attached screen
    132'744'101 - enwik9_1423.drt -s15 by Paq8pxd_v71_AVX2
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	paq8pxd_v70_v71_issue.jpg 
Views:	32 
Size:	72.3 KB 
ID:	7365  

  20. #673
    Member
    Join Date
    May 2008
    Location
    Estonia
    Posts
    509
    Thanks
    208
    Thanked 347 Times in 185 Posts
    Before your post i was messing with wrton and its actually not working. Its a mess. Works if debug build, fails after wordlist is read. % seems to indicate that.
    In todolist
    KZo


  21. #674
    Member
    Join Date
    May 2008
    Location
    Estonia
    Posts
    509
    Thanks
    208
    Thanked 347 Times in 185 Posts
    1. Statemap update
    If we are say at pos 0xffff in processing input then we switch statemap to mode where we only update prediction.
    Assuming update limit has been reached then update is something:
    inline void update1(int y, int limit) {
    assert(cxt>=0 && cxt<N);
    assert(y==0 || y==1);
    U32 *p=&t[cxt], p0=p[0];
    int pr=p0>>13; // prediction
    p0+=(((y<<19)-pr));
    p[0]=p0;
    ​}

    This gives us full 32 bits for prediction. Selecting switch pos needs more testing and is arbitrary on this example.

    2. Mixer final layer.
    Current mixer has final mixer prediction layer where there is no context. If we change it to accept final layer context then we can get small improvement on compression.
    Selecting this context is not easy, but...
    Or if we had separate component of inputs into mixer then we can use one layer mixers and can do same thing. Like in pxv.




    First test:
    on mill.jpg about 20 sec faster 200 bytes worse.
    on rafale.bmp about 7 sec faster 200 bytes worse.
    KZo


  22. #675
    Member DZgas's Avatar
    Join Date
    Feb 2020
    Location
    Russia
    Posts
    30
    Thanks
    14
    Thanked 8 Times in 6 Posts
    My text compress micro-test of all versions paq8pxd that I could download here and run non-AVX.
    Code:
    file     |  html |
    original | 32 793 bytes
    zip      |  6 621|
    7z       |  6 369|
    paq8pxd  
              deflaut
    v1       |  4 607|
    2        |- 4 606|
    3        |=      |
    4        |- 4 578|
    5        |=      |
    7        |- 4 572|best up to v55
    10       |+ 4 591|
    11       |+ 4 595|
                text
    19       |  4 700|
    20       |- 4 684|
    21       |=      |
    23       |+ 4 692|
    24       |=      |
    25       |- 4 691|
    26       |+ 4 725|
    27       |=      |
    28       |=      |
    31       |=      |
    32       |=      |
    34       |- 4 719|
    35       |=      |
    36       |- 4 645|
    37       |=      |
    38       |=      |
    39       |=      |
    40       |+ 4 647|
    41       |- 4 643|
    42       |+ 4 644|
    43       |- 4 633|
    44       |+ 4 648|
    45       |+ 4 654|
    49       |- 4 604|
    50       |=      |
    51       |- 4 594|
    52       |- 4 592|
    53       |=      |
    54       |- 4 589|
    55       |- 4 571|
    56       |- 4 570|
    57       |=      |
    58       |- 4 550|
    67       |- 4 514|
    69       |- 4 491|
    69f      |+ 4 492|
    70       |+ 4 493|
    71       |+ 4 495|

  23. Thanks:

    Darek (11th February 2020)

  24. #676
    Member
    Join Date
    May 2008
    Location
    Estonia
    Posts
    509
    Thanks
    208
    Thanked 347 Times in 185 Posts
    I was looking ways to improve compression time on jpeg. In pxd dmc model is before jpeg. Gives better compression if file is larger. On mill.jpeg its 4kb worse if not used, but 30sec faster.
    So trying other options i found a way.
    DMC removed, no additional main jpeg model contexts, some other changes. a10.jpg -> 627388
    On 16MB jpeg its 82 faster and improvement stands (6kb better).
    In attached image is current status.

    I remembered that i added new audio model to working version of pxd_v69 but only mono compression was better. So i removed it in final version.
    Also im8 compression is allot worse then in px version. Cant figure out why.


    Click image for larger version. 

Name:	mill.v72.PNG 
Views:	37 
Size:	65.4 KB 
ID:	7392
    KZo


  25. Thanks:

    Darek (11th February 2020)

  26. #677
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,155
    Thanks
    706
    Thanked 455 Times in 352 Posts
    Quote Originally Posted by DZgas View Post
    My text compress micro-test of all versions paq8pxd that I could download here and run non-AVX.
    Code:
    file     |  html |
    original | 32 793 bytes
    zip      |  6 621|
    7z       |  6 369|
    paq8pxd  
              deflaut
    v1       |  4 607|
    2        |- 4 606|
    3        |=      |
    4        |- 4 578|
    5        |=      |
    7        |- 4 572|best up to v55
    10       |+ 4 591|
    11       |+ 4 595|
                text
    19       |  4 700|
    20       |- 4 684|
    21       |=      |
    23       |+ 4 692|
    24       |=      |
    25       |- 4 691|
    26       |+ 4 725|
    27       |=      |
    28       |=      |
    31       |=      |
    32       |=      |
    34       |- 4 719|
    35       |=      |
    36       |- 4 645|
    37       |=      |
    38       |=      |
    39       |=      |
    40       |+ 4 647|
    41       |- 4 643|
    42       |+ 4 644|
    43       |- 4 633|
    44       |+ 4 648|
    45       |+ 4 654|
    49       |- 4 604|
    50       |=      |
    51       |- 4 594|
    52       |- 4 592|
    53       |=      |
    54       |- 4 589|
    55       |- 4 571|
    56       |- 4 570|
    57       |=      |
    58       |- 4 550|
    67       |- 4 514|
    69       |- 4 491|
    69f      |+ 4 492|
    70       |+ 4 493|
    71       |+ 4 495|
    Hmmm, I know there is a lot's of time but such small file (after compression) could be to small to observe improvenments and gives =
    But from other side it's nice comparison to "standard" paq8px version.

  27. #678
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,155
    Thanks
    706
    Thanked 455 Times in 352 Posts
    > Also im8 compression is allot worse then in px version. Cant figure out why.
    But new changes on im24 compression gives you the best overall scores for paq8pxd_v70 on my testset files.

  28. #679
    Member
    Join Date
    May 2008
    Location
    Estonia
    Posts
    509
    Thanks
    208
    Thanked 347 Times in 185 Posts
    paq8pxd_v72
    Code:
        - Change StationaryMap, StateMap
        - Change jpegModel
        - Change wordModel1
        - remove wavModel
        - added audio8, audio16
    There my be some tiny loss on 24 bit image compression.
    Audio models now work, was tired probably when i could not get it working.

    --
    Its always possible to add ppm,lstm models back.
    Attached Files Attached Files
    KZo


  29. Thanks (2):

    Darek (12th February 2020),moisesmcardona (12th February 2020)

  30. #680
    Member
    Join Date
    Jun 2009
    Location
    Puerto Rico
    Posts
    233
    Thanks
    127
    Thanked 45 Times in 35 Posts
    Edit: @kaitz beat me XD

  31. #681
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,155
    Thanks
    706
    Thanked 455 Times in 352 Posts
    Scores for my testset on paq8pxd_v72. Very nice gain on audio files - about 111KB on total set.
    Also fine gains for textual files, however there are some loses on other kind of files -> especially K.WAD loses 10.5KB.
    In general, very good audio model implementation - thanks for this!
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	paq8pxd_v72.jpg 
Views:	32 
Size:	771.3 KB 
ID:	7405  

  32. Thanks:

    kaitz (13th February 2020)

  33. #682
    Member
    Join Date
    Aug 2008
    Location
    Planet Earth
    Posts
    982
    Thanks
    96
    Thanked 396 Times in 276 Posts
    Quote Originally Posted by kaitz View Post
    paq8pxd_v72
    enwik8:
    16,013,638 bytes, 6029.396 sec., paq8pxd_v72_AVX2 -s15

  34. Thanks (2):

    Darek (13th February 2020),kaitz (13th February 2020)

  35. #683
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,155
    Thanks
    706
    Thanked 455 Times in 352 Posts
    And 4 corpuses scores for paq8pxd_v72.
    Best scores from paq8pxd family for Canterbury and MaximumCompression
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	4_Corpuses_paq8pxd_v72.jpg 
Views:	33 
Size:	2.01 MB 
ID:	7407  

  36. Thanks:

    kaitz (14th February 2020)

  37. #684
    Member
    Join Date
    Aug 2008
    Location
    NZ
    Posts
    59
    Thanks
    30
    Thanked 11 Times in 7 Posts
    Did some brief tests which may be of use.

    Used jpg fileset from here, consisting of 171 8 bit jpg images of slightly varying sizes.

    Test run under Windows 7 64 bit, using i5-3570K CPU, and 8 GB RAM. Used SSE41 compiles of paq8pxd as I can't run AVX2 compiles on the i5-3570k CPU.

    Each jpg file compressed individually via a batch file.

    Total original jpg files size = 64,469,752 bytes.

    Code:
    Compressor     Resulting total files size (bytes)    Compression time (seconds)    Compression options
    paq8pxd v69            51,365,725                              7,753                      -s9
    paq8pxd v72            51,338,132                              7,533                      -s9

    So, there was a slight improvement in compression for v72 with respect to v69. Also, the compression time was slightly reduced.

  38. #685
    Member
    Join Date
    May 2008
    Location
    Estonia
    Posts
    509
    Thanks
    208
    Thanked 347 Times in 185 Posts
    And if you compress as tar/zip (uncompressed), what is compressed size and speed then?
    KZo


  39. #686
    Member
    Join Date
    May 2008
    Location
    Estonia
    Posts
    509
    Thanks
    208
    Thanked 347 Times in 185 Posts
    Another testing.
    Wanted to see how well context actually work (wordmodel). ContextMap collects info and if threshold is reached context is permanently disabled, stats collection also.




    enwik6
    i(0)=431965, i(3)=324985, i(24)=488462, i(27)=207541, i(33)=493349, i(34)=157440, i(35)=168725, i(36)=179219, i(37)=555562, i(38)=558076, i(45)=425289, i(58)=399287, i(60)=230295, i(61)=210033,
    book1
    i(0)=520011, i(3)=491253, i(24)=394110, i(25)=220564, i(26)=4994, i(32)=132856, i(33)=490038, i(34)=76312, i(35)=77780, i(36)=80227, i(37)=463017, i(38)=461574, i(45)=421269, i(58)=405311, i(60)=256524, i(61)=131212,
    below bad contexts:
    + is enwik6
    - is book1

    in book i(26) is still ok sortof.
    -+ cm.set(hash(++i,x.spafdo, x.spaces,ccword));
    -+ cm.set(hash(++i,x.spaces, (x.words&255), (numbers&255)));
    -+ cm.set(hash(++i,h, word1,word2,lastUpper<x.wordlen));
    - cm.set(hash(++i,text0&0xffffff));
    - cm.set(text0&0xfffff);/// i(26)=4994, book1
    + cm.set(hash(++i,word0,number0, data0,xword0));
    - cm.set(hash(++i,word0, cword0,isfword));
    -+ cm.set(hash(++i,word0,buf(1), word2,isfword));
    -+ cm.set(hash(++i,word0,buf(1), word3));
    -+ cm.set(hash(++i,word0,buf(1), word4));
    -+ cm.set(hash(++i,word0,buf(1), word5));
    -+ cm.set(hash(++i,word0,buf(1), word1,word3));
    -+ cm.set(hash(++i,word0,buf(1), word2,word3));
    -+ cm.set(hash(++i,nl1-nl2,x.col,buf(1),above));
    - cm.set(hash(++i,h, llog(wordGap), mask&0x1FF, ));
    + cm.set(hash(x.col,x.wordlen1,above,above1,x.c4&0xfF)); else cm.set(); //wordlist
    -+ cm.set(hash(++i,x.col,above^above1,above2 , ((islink)<<8)|)); //wordlist((istemplate)<<9)|
    -+ cm.set(hash((*pWord).Hash[2], h));



    book1 compressed
    Code:
    ​ 183314 (pxd v72)                   100 sec
     183752 (pxd vXX skip if i(x)>2024)  88 sec
     183288 (pxd vXX no skip)            99 sec
     184490 (px  v183)                  139 sec
    KZo


  40. #687
    Member
    Join Date
    Aug 2008
    Location
    NZ
    Posts
    59
    Thanks
    30
    Thanked 11 Times in 7 Posts
    Quote Originally Posted by kaitz View Post
    And if you compress as tar/zip (uncompressed), what is compressed size and speed then?
    I've run further tests, this time "tarring" all 171 jpg images first.

    Again, there was an improvement in compression, and compression time was reduced a bit of v72 with respect to to v69.

    The "tarred" files produced slightly better compression with respect to compressing each file individually.

    Code:
    Compressor               Total file(s) size (bytes)    Compression time (seconds)    Compression options
    Original 171 jpg files         64,469,752
    paq8pxd v69                    51,365,725                          7,753                      -s9
    paq8pxd v72                    51,338,132                          7,533                      -s9
    
    Tarred jpg files               64,605,696
    paq8pxd v69                    50,571,934                          7,897                      -s9
    paq8pxd v72                    50,552,930                          7,756                      -s9

  41. Thanks (2):

    Darek (15th February 2020),kaitz (16th February 2020)

  42. #688
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,155
    Thanks
    706
    Thanked 455 Times in 352 Posts
    Some enwik scores for paq8pxd_v72:

    16'309'641 - enwik8 -s8 by Paq8pxd_v61
    15'968'477 - enwik8 -s15 by Paq8pxd_v61
    16'570'543 - enwik8.drt -s15 by Paq8pxd_v61
    126'587'796 - enwik9_1423 -s15 by Paq8pxd_v61 - best score for paq8pxd serie

    16'309'012 - enwik8 -s8 by Paq8pxd_v63
    15'967'201 - enwik8 -s15 by Paq8pxd_v63
    16'637'302 - enwik8.drt -s15 by Paq8pxd_v63
    126'597'584 - enwik9_1423 -s15 by Paq8pxd_v63


    16'374'223 - enwik8 -s8 by Paq8pxd_v67_AVX2
    16'048'070 - enwik8 -s15 by Paq8pxd_v67_AVX2
    16'774'998 - enwik8.drt -s15 by Paq8pxd_v67_AVX2
    127'063'602 - enwik9_1423 -s15 by Paq8pxd_v67_AVX2


    16'364'165 - enwik8 -s8 by Paq8pxd_v68_AVX2
    16'033'591 - enwik8 -s15 by Paq8pxd_v68_AVX2
    16'755'942 - enwik8.drt -s15 by Paq8pxd_v68_AVX2
    126'958'003 - enwik9_1423 -s15 by Paq8pxd_v68_AVX2


    16'358'450 - enwik8 -s8 by Paq8pxd_v72_AVX2 - My compress time 6'780s.
    16'013'638 - enwik8 -s15 by Paq8pxd_v72_AVX2 - tested by Sportman, My compress time 6'811s -@Sportman - you have very fast machine!
    16'672'036 - enwik8.drt -s15 by Paq8pxd_v72_AVX2 - My compress time 9'280s.
    126'779'432 - enwik9_1423 -s15 by Paq8pxd_v72_AVX2 - very nice gain, however still about 200KB behind paq8pxd_v61 version. My compress time 67'740s.
    132'464'891 - enwik9_1423.drt -s15 by Paq8pxd_v72_AVX2 - also very nice gain, however still starting from paq8pxd_v57 DRT precompressed files got worse scores than pure enwik8/9 and the compression time is 50% higher. My compress time 93'207s.
    Last edited by Darek; 16th February 2020 at 13:22.

  43. Thanks:

    kaitz (16th February 2020)

  44. #689
    Member
    Join Date
    Sep 2015
    Location
    Italy
    Posts
    273
    Thanks
    114
    Thanked 159 Times in 116 Posts
    hbcount is always 0 in line 9905 (jpegModelx.p()):
    9869 if (hbcount==0) {
    [...]
    9905 cxt[34]=(!hbcount)?hash(mcupos, column, row, hc>>2):0; // MJPEG

  45. Thanks:

    kaitz (18th February 2020)

  46. #690
    Member
    Join Date
    May 2008
    Location
    Estonia
    Posts
    509
    Thanks
    208
    Thanked 347 Times in 185 Posts
    paq8pxd_v73
    Code:
        - Change text detection
        - Change wordModel1
        - Add mod_ppmd/chart model back (-x option)
    Attached Files Attached Files
    KZo


  47. Thanks (5):

    brispuss (18th February 2020),Darek (18th February 2020),Mike (18th February 2020),moisesmcardona (18th February 2020),Shelwien (18th February 2020)

Page 23 of 32 FirstFirst ... 132122232425 ... LastLast

Similar Threads

  1. FreeArc compression suite (4x4, Tornado, REP, Delta, Dict...)
    By Bulat Ziganshin in forum Data Compression
    Replies: 554
    Last Post: 26th September 2018, 02:41
  2. Dict preprocessor
    By pat357 in forum Data Compression
    Replies: 5
    Last Post: 2nd May 2014, 21:51

Tags for this Thread

Posting Permissions

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