Page 61 of 62 FirstFirst ... 115159606162 LastLast
Results 1,801 to 1,830 of 1842

Thread: paq8px

  1. #1801
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,028
    Thanks
    623
    Thanked 418 Times in 316 Posts
    @DZgas - this is similar to paq8pxe_v1_cl90 compile - I don't know if it's the same situation on all computers but there are much worse scores than other compiles .
    Is it works properly on your system ? https://encode.su/attachment.php?att...1&d=1581266141

    I've checked it. Scores for my testset are the same as paq8pxe_v1_cl90.
    Last edited by Darek; 11th February 2020 at 00:18.

  2. #1802
    Administrator Shelwien's Avatar
    Join Date
    May 2008
    Location
    Kharkov, Ukraine
    Posts
    3,774
    Thanks
    276
    Thanked 1,206 Times in 671 Posts
    As I said, you have to add "-simd sse2" option - or compile it for avx2 target.

  3. #1803
    Member
    Join Date
    Jun 2009
    Location
    Puerto Rico
    Posts
    192
    Thanks
    91
    Thanked 19 Times in 15 Posts
    I compiled both v184 and v185 using MinGW.

    Code:
    cmake -G "MSYS Makefiles" -DUSE_TEXTMODEL=ON -DUSE_AUDIOMODEL=ON -DUSE_ZLIB=ON ..
    make -j32
    The compiled exe is in the build folder.
    Attached Files Attached Files

  4. Thanks:

    Darek (11th February 2020)

  5. #1804
    Member DZgas's Avatar
    Join Date
    Feb 2020
    Location
    Russia
    Posts
    18
    Thanks
    4
    Thanked 7 Times in 5 Posts
    @Darek
    Oh I used MinGW...I think I forgot something...
    Everything works fine for me, compresses than 183fix1 (text only):

    paq8px_v183fix1
    Total input size : 123465
    Total archive size : 3094
    Time 137.65 sec, used 845 MB

    my paq8px_v185
    Total input size : 123465
    Total archive size : 3094
    Time138.58 sec, used 845 MB

    I was able to extract the 185 text file using 183fix1, identical.

    @moisesmcardona
    Is it avx? I can't run this.

  6. Thanks (2):

    Darek (11th February 2020),Eppie (11th February 2020)

  7. #1805
    Member
    Join Date
    Jun 2009
    Location
    Puerto Rico
    Posts
    192
    Thanks
    91
    Thanked 19 Times in 15 Posts
    Quote Originally Posted by DZgas View Post
    @moisesmcardona
    Is it avx? I can't run this.
    Yes. The CMake file compiles using the native architecture. It was compiled on a Ryzen 9 3950X CPU.

  8. #1806
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,028
    Thanks
    623
    Thanked 418 Times in 316 Posts
    Scores for paq8px v184 and v185 on my testset.

    paq8px v184 version got some loses to v183fix1, however it got the best overall result in WAV file. From other side I.EXE crashed in the middle of the process - then value for this file in the table is the same as v183fix1.
    paq8px v185 have similar compression to paq8pxe_v1 ic19.
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	paq8px_v184_v185.jpg 
Views:	40 
Size:	173.9 KB 
ID:	7391  

  9. #1807
    Member DZgas's Avatar
    Join Date
    Feb 2020
    Location
    Russia
    Posts
    18
    Thanks
    4
    Thanked 7 Times in 5 Posts
    @Darek
    Maybe I took a small file (124 kb) but the EXE compression v183fix1 is identical to my v185, and also when I decompress v185 files using the v183fix1, I get a file identical to the original file...
    I always use LEVEL -6
    Two your EXE files are identical too, give your I.EXE test file. I want to compressing it too

  10. #1808
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,028
    Thanks
    623
    Thanked 418 Times in 316 Posts
    Maybe there are matter of used options. I've found recently best options for my testfiles and 4 corpuses.
    Some of it uses exe files as training file and due to change of the build there would be different scores. I'm testing 4 corpuses now.
    My best options are as in attached table. I.EXE also in attached zip file.
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	paq8px_best_options.jpg 
Views:	38 
Size:	214.9 KB 
ID:	7393  
    Attached Files Attached Files
    • File Type: zip I.zip (356.9 KB, 19 views)

  11. #1809
    Member DZgas's Avatar
    Join Date
    Feb 2020
    Location
    Russia
    Posts
    18
    Thanks
    4
    Thanked 7 Times in 5 Posts
    @Darek
    Ah As I thought, compression result - The Identical files
    Code:
    Creating archive I.EXE.paq8px185 in single file mode...
    Block segmentation:
     0           | default          |     46725 bytes [0 - 46724]
     1           | exe              |      6258 bytes [46725 - 52982]
     2           | default          |    218622 bytes [52983 - 271604]
     3           | exe              |     16182 bytes [271605 - 287786]
     4           | default          |     17778 bytes [287787 - 305564]
     5           | exe              |    248307 bytes [305565 - 553871]
     6           | default          |    161621 bytes [553872 - 715492]
    -----------------------
    Total input size     : 715493
    Total archive size   : 194628
    Time 1171.02 sec, used 846 MB (887390382 bytes) of memory
    
    Creating archive I.EXE.paq8px183fix1 in single file mode...
    Block segmentation:
     0           | default          |     46725 bytes [0 - 46724]
     1           | exe              |      6258 bytes [46725 - 52982]
     2           | default          |    218622 bytes [52983 - 271604]
     3           | exe              |     16182 bytes [271605 - 287786]
     4           | default          |     17778 bytes [287787 - 305564]
     5           | exe              |    248307 bytes [305565 - 553871]
     6           | default          |    161621 bytes [553872 - 715492]
    -----------------------
    Total input size     : 715493
    Total archive size   : 194628
    Time 1123.28 sec, used 846 MB (887399271 bytes) of memory
    I used the paq8px_v185 that I compiled. Did you use it or moisesmcardona's paq8px_v185-AVX?
    Try to do the test with the level -6

  12. #1810
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,028
    Thanks
    623
    Thanked 418 Times in 316 Posts
    Ok, for I.exe I've used -9et option -> it's probably the reason of the difference. And yes, I've used moisesmcardona's builds.
    Did you try v184 version? This version is crashing for I.EXE (paq8px v185 is ok) and also crashes on AcroRd32.exe and MSO97.DLL from MaximumCompression.
    I've tried to use -6et and pure -6 options - nope - the same crash... as on attached screen.
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	paq8px_v184_crash.jpg 
Views:	27 
Size:	135.2 KB 
ID:	7396  

  13. #1811
    Member DZgas's Avatar
    Join Date
    Feb 2020
    Location
    Russia
    Posts
    18
    Thanks
    4
    Thanked 7 Times in 5 Posts
    @Darek
    I can not run v184, im am non-AVX user.
    I do not think that v184 can be tested due to problems.
    Quote Originally Posted by Eppie View Post
    The latest version there is v184, which is a large refactoring of paq8px to make it more modularized. In the process of refactoring, I broke some things (windows compilation, compression ratio for text files, non-AVX2 systems). Therefore, I would consider v184 to be unworthy of testing.
    I never used -e -t because they affect of tests and results(is seen in your example). Why are you using this?

  14. Thanks:

    Darek (12th February 2020)

  15. #1812
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,028
    Thanks
    623
    Thanked 418 Times in 316 Posts
    @DZgas - you have right that for pure comparison between versions of the same compressor it is better to not to use switches.

    My tests try to find maximum compression from each compressor/version which means also use the most optimal switches - especially for textual files it gives much better results on paq8px serie.
    Use of maximum options gives also proper comparison for other programs like cmix, cmv, emma, nncp or even paq8pxd - best score to best score.

  16. #1813
    Member DZgas's Avatar
    Join Date
    Feb 2020
    Location
    Russia
    Posts
    18
    Thanks
    4
    Thanked 7 Times in 5 Posts
    Ok.

    Code:
                     exe
    original     | 393 728|
    v183fix1 -6  | 371 238|
    v185     -6  | 371 238|
    v183fix1 -6e | 371 156|
    v185     -6e | 371 173|
    Last edited by DZgas; 12th February 2020 at 17:08.

  17. Thanks:

    Darek (12th February 2020)

  18. #1814
    Member
    Join Date
    Dec 2008
    Location
    Poland, Warsaw
    Posts
    1,028
    Thanks
    623
    Thanked 418 Times in 316 Posts
    for textual files difference of use -t option are crucial
    S.DOC:
    original 141'417
    v185 -9 23'877
    v185 -9t 22'741
    v185 -9ta 22'700

    R.DOC:
    original 120'201
    v185 -9 27'068
    v185 -9t 24'823
    v185 -9ta 24'759

  19. Thanks:

    DZgas (13th February 2020)

  20. #1815
    Member
    Join Date
    Sep 2015
    Location
    Italy
    Posts
    268
    Thanks
    111
    Thanked 153 Times in 112 Posts
    In JpegModel the line
    m1=MixerFactory::CreateMixer(sh, N+1, 2050, 3);
    should be
    m1=MixerFactory::CreateMixer(sh, N+1+2, 2050, 3);
    because MJPEGMap adds 2 additional inputs to the mixer
    MJPEGMap.mix(*m1);

    To reproduce the error I ran the following test with paq8px_v183fix1, but I suppose this problem is still present in the current version.
    Commenting out the line
    //#define NDEBUG // Remove (comment out) this line for debugging (turns on Array bound checks and asserts)
    and compressing a JPEG file with -simd none, e.g.
    paq8px_v183fix1 -9 -simd none A10.jpg
    the assert(nx<N) in Mixer.add() is triggered:
    Assertion failed!
    Program: C:\paq8px_v183fix1\paq8px_v183fix1.exe
    File: paq8px_v183fix1.cpp, Line 1735
    Expression: nx<N

  21. #1816
    Member
    Join Date
    Feb 2020
    Location
    Chelyabinsk, Russia
    Posts
    15
    Thanks
    3
    Thanked 2 Times in 2 Posts
    Test file PIF.mht

    PAQ8pxd (all versions) - ok

    PAQ8px to v71 - ok
    FP8 to v4 - ok

    PAQ8px over v72 (to v132) - error (file not created)
    FP8 over v5 (to v6) - error (file not created)

    i3-4130, 16 GB, Win 8.1
    Attached Files Attached Files

  22. #1817
    Member
    Join Date
    Sep 2015
    Location
    Italy
    Posts
    268
    Thanks
    111
    Thanked 153 Times in 112 Posts
    Quote Originally Posted by User View Post
    Test file PIF.mht
    [...]
    PAQ8px over v72 (to v132) - error (file not created)
    Also paq8px_v183fix1 has this problem: it executes an infinite loop of decoding of a base64 segment.
    I compared paq8px and paq8pxd: paq8pxd fixed this problem by adding the variable "g" in decode_base64():
    int tlf=0,g=0;
    [...]
    inn[i]=0,g=1;
    [...]
    if (g) break; //if past eof, break

    197.959 paq8px_v183fix1+g -9
    106.607 paq8pxd_74_SSE4 -s9 (much faster than paq8px_v183fix1+g, ~~3x)

  23. Thanks (2):

    schnaader (25th February 2020),User (25th February 2020)

  24. #1818
    Member
    Join Date
    May 2008
    Location
    Estonia
    Posts
    471
    Thanks
    198
    Thanked 296 Times in 161 Posts
    KZo


  25. #1819
    Member
    Join Date
    May 2008
    Location
    Estonia
    Posts
    471
    Thanks
    198
    Thanked 296 Times in 161 Posts
    Quote Originally Posted by Mauro Vezzosi View Post
    197.959 paq8px_v183fix1+g -9
    106.607 paq8pxd_74_SSE4 -s9 (much faster than paq8px_v183fix1+g, ~~3x)
    PIF.mht -> px version detects large part as text, its base64. This is main speed difference.
    KZo


  26. Thanks:

    Mauro Vezzosi (2nd March 2020)

  27. #1820
    Member
    Join Date
    Feb 2020
    Location
    Chelyabinsk, Russia
    Posts
    15
    Thanks
    3
    Thanked 2 Times in 2 Posts
    All versions PAQ8px, PAQ8pxd, FP8 do not detect jpg in this file
    Attached Files Attached Files

  28. #1821
    Member
    Join Date
    May 2008
    Location
    Estonia
    Posts
    471
    Thanks
    198
    Thanked 296 Times in 161 Posts
    Quote Originally Posted by User View Post
    All versions PAQ8px, PAQ8pxd, FP8 do not detect jpg in this file
    Progressive?
    KZo


  29. #1822
    Programmer schnaader's Avatar
    Join Date
    May 2008
    Location
    Hessen, Germany
    Posts
    593
    Thanks
    233
    Thanked 226 Times in 107 Posts
    Quote Originally Posted by kaitz View Post
    Quote Originally Posted by User View Post
    All versions PAQ8px, PAQ8pxd, FP8 do not detect jpg in this file
    Progressive?
    Yes, it's a progressive JPG image. @User: PAQ derivates only support baseline JPG, see the column "Prog" in this table from http://mattmahoney.net/dc/#jpeg:

    Code:
    Program              a10.jpg  dscn3974.jpg  Comp  Base  Prog  Embed Non-JPEG
    -------              -------  ------------  ----  ----  ----  ----- --------
    Original size        842,468   1,114,198
    Stuffit 11 (best)    638,540     834,079     1     yes   yes   no    yes
    paq8o8 -6            638,206     827,071    30.8   yes   no    yes   yes
    paq8fthis_fast -6    673,112     862,834     9.4   yes   no    yes   yes
    PackJPG 2.3          697,822     873,261     1.5   yes   yes   no    no
    lprepaq 5            699,692   1,083,737     9.5   yes   no    yes   yes
    zpaq ocjpg_test2.cfg 716,043     916,353    26.9   yes   no    no    no
    jpg2dct | ppmd -o2   770,302     964,803     2.0   yes   yes   no    no  (lossy)
    rings 0.1            819,169   1,075,239     0.3   yes               yes
    ppmd -o2             833,336   1,094,853     1.2   no    no    no    yes
    The list has not been updated in a while, so I don't know if there has been some PAQ fork in the meantime that supports it, but I don't think so. Other programs that support progressive JPEGs not mentioned in this table are Brunsli, Lepton (using -allowprogressive) and Precomp. There are even more mentioned in this thread, but I'm not sure about their progressive capabilities.

    Also, looking at this ImageMagick forum entry, it could be possible to convert a progressive JPG to a baseline JPG without introducing additional image artifacts using jpegtran from Libjpeg.

    EDIT: Could verify the (pixel-perfect) lossless conversion to baseline using jpegtran. Just run "jpegtran img.jpg > img_baseline.jpg". Converted both the progressive and the baseline version to BMP, those were identical. Baseline version of the JPG attached.

    EDIT2: Just checked - the metadata is lost in the attached JPG. If this is not wanted, "-copy all" keeps all metadata and results in a bigger baseline JPG (1,003,584 bytes).
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	CIMG1186_baseline.JPG 
Views:	32 
Size:	933.9 KB 
ID:	7454  
    http://schnaader.info
    Damn kids. They're all alike.

  30. Thanks:

    User (29th February 2020)

  31. #1823
    Member
    Join Date
    Feb 2020
    Location
    Chelyabinsk, Russia
    Posts
    15
    Thanks
    3
    Thanked 2 Times in 2 Posts
    Yes, it turns out CIMG1186 - Progressive.
    Interestingly, CIMG1185, CIMG1187 and others are Baseline DCT.
    Not expected.

  32. #1824
    Programmer schnaader's Avatar
    Join Date
    May 2008
    Location
    Hessen, Germany
    Posts
    593
    Thanks
    233
    Thanked 226 Times in 107 Posts
    Quote Originally Posted by User View Post
    Yes, it turns out CIMG1186 - Progressive.
    Interestingly, CIMG1185, CIMG1187 and others are Baseline DCT.
    Not expected.
    Depends on the processing software (or camera, in this case?). Progressive is used to get a quick preview of the image with low connections, but it's not the only use case. It also leads to smaller JPGs sometimes (but not always), so it looks like the software's choice here could have been: Try to encode both baseline and progressive and keep the smaller one.
    http://schnaader.info
    Damn kids. They're all alike.

  33. #1825
    Member
    Join Date
    Feb 2020
    Location
    Chelyabinsk, Russia
    Posts
    15
    Thanks
    3
    Thanked 2 Times in 2 Posts
    Photos without processing, from the camera.

  34. #1826
    Member
    Join Date
    Feb 2020
    Location
    Chelyabinsk, Russia
    Posts
    15
    Thanks
    3
    Thanked 2 Times in 2 Posts
    Drag'n'drop
    In versions prior to paq8px_v132_fix1 - it worked.
    Starting with version paq8px_v137 - does not work.
    ---
    In paq8pxd and fp8 - it works.
    ---
    What is the reason?

  35. #1827
    Member DZgas's Avatar
    Join Date
    Feb 2020
    Location
    Russia
    Posts
    18
    Thanks
    4
    Thanked 7 Times in 5 Posts
    Drag and Drop and folder compression were removed after v137, so hardly anyone will be "worried" about them at paq8px...
    I used paq8px before for my "work", but after removing these functions, mostly folder compression, I shift to paq8pxd.

  36. #1828
    Member DZgas's Avatar
    Join Date
    Feb 2020
    Location
    Russia
    Posts
    18
    Thanks
    4
    Thanked 7 Times in 5 Posts
    Although this does not apply to paq8px, it seemed interesting to me: to use compression paq8px in practice, so that the file size would be small so that decoding would not be long... I decided to take the text, compress and transcode to base64, and additionally in the QR-code...
    p.s the QR code can be written in binary format, although what is it needed if it can be read only by specialized software...base32(base10 or base45 is best for QR-code), base64 and base91 can read any QR scanner on smartphones.

    Small 1 bit picture compressed using paq8px v185 (I used notepad++ for base64 encoding/decoding)
    Code:
    cGFxOHB4BgD99bOyeRPmqGP97MX/xs+OM9RyIfSyDBaJFk+IN8s3w6RGV2mQA///5hMj3HN6Xkc5Gco1wIJARlmUSAh5ueRsqxLfnFujO25GLIclGVi7uUjZ9AZvMcdLvg
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	ENCODE.png 
Views:	24 
Size:	126.0 KB 
ID:	7459  
    Attached Files Attached Files
    Last edited by DZgas; 9th March 2020 at 11:52.

  37. #1829
    Member
    Join Date
    Feb 2020
    Location
    Chelyabinsk, Russia
    Posts
    15
    Thanks
    3
    Thanked 2 Times in 2 Posts
    I was comfortable using Drag and Drop.
    Using the command line is less convenient.
    But PAQCompress cannot be downloaded due to Avast antivirus - it believes that all PAQCompress contain a trojan.

  38. #1830
    Member DZgas's Avatar
    Join Date
    Feb 2020
    Location
    Russia
    Posts
    18
    Thanks
    4
    Thanked 7 Times in 5 Posts
    @User use paq8pxd elsie you need Drag and Drop! And delete Avast antivirus сhёrtov.

Page 61 of 62 FirstFirst ... 115159606162 LastLast

Similar Threads

  1. FrontPAQ - GUI frontend for PAQ8PF and PAQ8PX
    By LovePimple in forum Download Area
    Replies: 26
    Last Post: 17th January 2019, 13:36
  2. Alternative paq8px builds
    By M4ST3R in forum Download Area
    Replies: 20
    Last Post: 25th June 2010, 16:19
  3. Optimized paq7asm.asm code not compatible with paq8px?
    By M4ST3R in forum Data Compression
    Replies: 7
    Last Post: 3rd June 2009, 15:34

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
  •