Login
Username:

Password:

Remember me



Lost Password?

Register now!
Sections
Who's Online
61 user(s) are online (49 user(s) are browsing Forums)

Members: 0
Guests: 61

more...
Support us!
Recent OS4 Files
OS4Depot.net





New xpk sub-lib for testing (XZ-based)
Amigans Defender
Joined:
2006/11/17 22:40
From England
Posts: 3228
I have written a new XPK compression library based on XZ/LZMA2. It needs some testing and tweaking.

***update 13th Sept***
Due to an administrative error I'd given this library a reserved name. I've renamed it xpkL2XZ.library - please download the new one and delete any copies of xpkXZL2.library.


The beta version is here: xpkL2XZ.lha
Do not use it on any files you care about!

You also need lzma.library if you don't already have it (if you use my XAD XZ client, then you already have it).

Some stats: (from initial version)
File 'NetSurf' with a size of 14759179 bytes.
Type  Num Version P    CSize     CTime      CSpd     UTime      USpd  Rate
XZL2
:   4 50.1       4218832      4.02   3671437      2.05   7199599  71.5
XZL2
:   9 50.1       4124232      3.00   4919726      2.04   7234891  72.1
XZL2
:  14 50.1       4133708      0.04 368979475      2.00   7379589  72.0
XZL2
:  19 50.1       4038692      0.03 491972633      2.06   7164650  72.7
XZL2
:  24 50.1       4092364      2.00   7379589      3.06   4823261  72.3
XZL2
:  29 50.1       3996984      2.00   7379589      3.05   4839075  73.0
XZL2
:  34 50.1       4076088      0.01 1475917900      3.03   4871016  72.4
XZL2
:  39 50.1       3981024      6.00   2459863      4.01   3680593  73.1
XZL2
:  44 50.1       3691932      5.06   2916833      4.00   3689794  75.0
XZL2
:  49 50.1       3592376      4.02   3671437      4.05   3644241  75.7
XZL2
:  54 50.1       3653624      2.06   7164650      5.06   2916833  75.3
XZL2
:  59 50.1       3557036      2.06   7164650      5.04   2928408  75.9
XZL2
:  64 50.1       3642824      1.05  14056360      5.05   2922609  75.4
XZL2
:  69 50.1       3544684      1.04  14191518      5.03   2934230  76.0
XZL2
:  74 50.1       3642824      1.06  13923753      1.04  14191518  75.4
XZL2
:  79 50.1       3544684      1.04  14191518      1.01  14613048  76.0
XZL2
100 50.1       3544684      1.05  14056360      1.01  14613048  76.0


For comparison, here's Joerg's xpkLZMA on the same file:

File 'NetSurf' with a size of 14759179 bytes.
Type  Num Version P    CSize     CTime      CSpd     UTime      USpd  Rate
LZMA
:  19 52.2       4440712      0.01 1475917900      2.06   7164650  99.1
LZMA
:  39 52.2       4120080      6.03   2447625      2.00   7379589  72.1
LZMA
:  59 52.2       3907816      0.04 368979475      1.00  14759179  73.6
LZMA
:  69 52.2       3667064      5.06   2916833      1.04  14191518  75.2
LZMA
:  79 52.2       3405416      6.01   2455770      1.04  14191518  77.0
LZMA
:  89 52.2       3240052      0.01 1475917900      1.06  13923753  78.1
LZMA
100 52.2       3204660      3.01   4903381      1.05  14056360  78.3


As you can see, I have some tweaking to do with the settings. Mine uses LZMA2 which should, theoretically, be better than LZMA in all situations. xBench's timings are unfortunately all over the place so I'm not even sure which one is quicker.

Note that I've disabled modes above 79 as default settings up that high were crashing my SAM. This is probably indicative of a bug, or it may have just been my impatience as it ran into virtual memory and appeared to freeze.

Modes ending in 5, 6, 7, 8, and 9 have a PowerPC filter enabled on them. For data files this actually makes the compressed files slightly bigger. For executables it makes them quite a bit smaller. The file in the benchmarks above was an executable. The compressed sizes are a bit non-linear when going up the modes because of this filter.

Feedback required please!


Edited by Chris on 2014/9/12 13:40:47
Edited by Chris on 2014/9/12 23:37:22
   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Amigans Defender
Joined:
2006/11/17 22:40
From England
Posts: 3228
I think I've solved the compression problem:
XZL2:  79 50.1       3074516      3.02   4887145      2.03   7270531  79.2


The max chunk sized I'd defined (read: copy'n'pasted) was too small. Mode 79 (7+PPC) is now performing better than mode 100 of xpkLZMA.

New version in the same place. (although everybody seems to be getting "old" results - I might not have replaced it, will bumprev and re-upload tonight)


Edited by Chris on 2014/9/12 13:41:20
Edited by Chris on 2014/9/12 13:42:46
   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Not too shy to talk
Joined:
2007/2/6 13:57
From Donostia (SPAIN)
Posts: 412
Tested here on my SAM460ex:
#xbench NetSurf xzl2
File 'NetSurf' with a size of 14759179 bytes.
Type  Num Version P    CSize     CTime      CSpd     UTime      USpd  Rate
XZL2
:   4 50.1       4218832      2.02   7306524      1.03  14329300  71.5
XZL2
:   9 50.1       4124232      2.03   7270531      1.01  14613048  72.1
XZL2
:  14 50.1       4133708      2.00   7379589      1.01  14613048  72.0
XZL2
:  19 50.1       4038692      2.03   7270531      1.03  14329300  72.7
XZL2
:  24 50.1       4092364      1.02  14469783      1.01  14613048  72.3
XZL2
:  29 50.1       3996984      1.01  14613048      1.03  14329300  73.0
XZL2
:  34 50.1       4076088      0.03 491972633      2.00   7379589  72.4
XZL2
:  39 50.1       3981024      0.02 737958950      2.02   7306524  73.1
XZL2
:  44 50.1       3691932      0.02 737958950      2.01   7342875  75.0
XZL2
:  49 50.1       3592376      3.01   4903381      2.03   7270531  75.7
XZL2
:  54 50.1       3653624      2.01   7342875      3.01   4903381  75.3
XZL2
:  59 50.1       3557036      2.00   7379589      3.02   4887145  75.9
XZL2
:  64 50.1       3642824      0.01 1475917900      3.00   4919726  75.4
XZL2
:  69 50.1       3544684      0.02 737958950      3.02   4887145  76.0
XZL2
:  74 50.1       3642824      2.00   7379589      0.01 1475917900  75.4
XZL2
:  79 50.1       3544684      2.00   7379589      0.02 737958950  76.0
XZL2
100 50.1       3544684      2.00   7379589      0.03 491972633  76.0
#


BTW xBench locks the whole system during such test? Anything can be done in such respect? (will try to see xbench source code)


Edited by jabirulo on 2014/9/12 14:29:06
   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Amigans Defender
Joined:
2006/11/17 22:40
From England
Posts: 3228
@jabirulo

I think xBench raises the task priority to get more accurate results. If you compress with XPKatana it doesn't do this.

It looks like you've tested with my initial version, please can you download the current one. I should have bumped the version number...

Also I need people to decrunch and check the data isn't corrupt.

   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Quite a regular
Joined:
2007/2/27 10:47
From Gravity well
Posts: 654
Checked a little!

7.RAM Disk:> sha1sum smallfil
3ea96ebe54427724633db240e41bda04b8798196 smallfil
7.RAM Disk:> xpack method xzl2 file smallfil minsize 100
xz-packed smallfil ( 285 bytes, 100% done, 51% CF)
7.RAM Disk:> xpack smallfil
xz-unpacked smallfil ( 285 bytes, 100% done, 68% CF)
7.RAM Disk:> sha1sum smallfil
3ea96ebe54427724633db240e41bda04b8798196 smallfil

also

7.RAM Disk:> xpack method xzl2.79 file smallfil minsize 100
xz-packed smallfil ( 285 bytes, 100% done, 51% CF)
7.RAM Disk:> xpack smallfil
xz-unpacked smallfil ( 285 bytes, 100% done, 68% CF)
7.RAM Disk:> sha1sum smallfil
3ea96ebe54427724633db240e41bda04b8798196 smallfil
7.RAM Disk:> xpack method xzl2.0 file smallfil minsize 100
xz-packed smallfil ( 285 bytes, 100% done, 50% CF)
7.RAM Disk:> xpack smallfil
xz-unpacked smallfil ( 285 bytes, 100% done, 67% CF)
7.RAM Disk:> sha1sum smallfil
3ea96ebe54427724633db240e41bda04b8798196 smallfil

   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Just can't stay away
Joined:
2006/11/24 18:52
From Gloucestershire, UK.
Posts: 1169
@Chris

Testing on my X1000:

xbench NetSurf xzl2
File 
'NetSurf' with a size of 14759179 bytes.
Type  Num Version P    CSize     CTime      CSpd     UTime      USpd  Rate
XZL2
:   4 50.1       4218832      3.47   4253365      0.26  56766073  71.5
XZL2
:   9 50.1       4124232      3.42   4315549      0.26  56766073  72.1
XZL2
:  14 50.1       4133708      3.27   4513510      0.23  64170343  72.0
XZL2
:  19 50.1       4038692      3.22   4583595      0.22  67087177  72.7
XZL2
:  24 50.1       4092364      4.47   3301829      0.21  70281804  72.3
XZL2
:  29 50.1       3996984      4.41   3346752      0.21  70281804  73.0
XZL2
:  34 50.1       4076088      5.14   2871435      0.21  70281804  72.4
XZL2
:  39 50.1       3981024      5.07   2911080      0.20  73795895  73.1
XZL2
:  44 50.1       3691932     12.43   1187383      0.21  70281804  75.0
XZL2
:  49 50.1       3592376     12.33   1197013      0.20  73795895  75.7
XZL2
:  54 50.1       3653624     14.15   1043051      0.20  73795895  75.3
XZL2
:  59 50.1       3557036     14.09   1047493      0.20  73795895  75.9
XZL2
:  64 50.1       3642824     16.21    910498      0.20  73795895  75.4
XZL2
:  69 50.1       3544684     16.21    910498      0.20  73795895  76.0
XZL2
:  74 50.1       3642824     16.07    918430      0.20  73795895  75.4
XZL2
:  79 50.1       3544684     16.06    919002      0.20  73795895  76.0
XZL2
100 50.1       3544684     16.06    919002      0.20  73795895  76.0


hint: enclose your results with code /code in square brackets.


Edited by Severin on 2014/9/12 13:48:07
_________________
Amiga user since 1985
AOS4, A-EON, IBrowse & Alinea Betatester

Ps. I hate the new amigans website. <shudder>
   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Amigans Defender
Joined:
2006/11/17 22:40
From England
Posts: 3228
OK, new version.

Changes:
* Bumped version to 50.2
* Added new delta encoding modes. If the last digit of the mode is 1-4 the packer will apply a delta filter over that many bytes. 9 is still for PPC code.
* Mode 100 is now equivalent to mode 70 but with "extreme" compression set. This should make it quite a lot slower than mode 70 for a slight improvement in compression (according to the documentation, that's what extreme mode does!). Note that modes 80-99 should be the same as 100, but might not be, and whether you get delta or PPC filters applied is undocumented, so please assume a label "reserved for future use".


The below is a test using a PPM file. As the data is a photo in uncompressed RGBRGBRGB, I was expecting delta3 mode to be best, however delta1 seems to be better. I'm interesting in seeing compression results for a range of different filetypes (MODs and samples I imagine will benefit from the delta filters).

File 'test.ppm' with a size of 4494544 bytes.
Type  Num Version P    CSize     CTime      CSpd     UTime      USpd  Rate
XZL2
:   0 50.2       1884264      0.01 449454400      0.06  74909066  58.1
XZL2
:   1 50.2       1291256      5.03    893547      0.04 112363600  71.3
XZL2
:   2 50.2       1623216      6.05    742899      0.06  74909066  63.9
XZL2
:   3 50.2       1410700      4.00   1123636      0.01 449454400  68.7
XZL2
:   4 50.2       1951260      1.02   4406415      1.00   4494544  56.6
XZL2
:   9 50.2       1896784      0.01 449454400      1.02   4406415  57.8
XZL2
:  10 50.2       1799280      3.05   1473620      0.03 149818133  60.0
XZL2
:  11 50.2       1305988      1.05   4280518      0.02 224727200  71.0
XZL2
:  12 50.2       1581984      2.03   2214061      0.05  89890880  64.9
XZL2
:  13 50.2       1320536      1.02   4406415      0.01 449454400  70.7
XZL2
:  14 50.2       1889792      4.04   1112510      1.06   4240135  58.0
XZL2
:  19 50.2       1812096      3.07   1464020      0.06  74909066  59.7
XZL2
:  20 50.2       1791840      4.04   1112510      0.03 149818133  60.2
XZL2
:  21 50.2       1305780      5.05    890008      0.02 224727200  71.0
XZL2
:  22 50.2       1564300      6.03    745363      0.05  89890880  65.2
XZL2
:  23 50.2       1293244      0.06  74909066      0.06  74909066  71.3
XZL2
:  24 50.2       1873436      7.04    638429      1.06   4240135  58.4
XZL2
:  29 50.2       1804680      4.05   1109763      1.00   4494544  59.9
XZL2
:  30 50.2       1790000      4.02   1118045      0.05  89890880  60.2
XZL2
:  31 50.2       1304684      2.00   2247272      0.04 112363600  71.0
XZL2
:  32 50.2       1555852      2.04   2203207      0.06  74909066  65.4
XZL2
:  33 50.2       1276204      1.02   4406415      0.01 449454400  71.7
XZL2
:  34 50.2       1867704      2.02   2225021      1.01   4450043  58.5
XZL2
:  39 50.2       1802812      4.02   1118045      1.02   4406415  59.9
XZL2
:  40 50.2       1628808      4.01   1120833      0.01 449454400  63.8
XZL2
:  41 50.2       1046876      2.03   2214061      0.03 149818133  76.8
XZL2
:  42 50.2       1377328      0.03 149818133      0.03 149818133  69.4
XZL2
:  43 50.2       1129312      6.02    746601      0.05  89890880  74.9
XZL2
:  44 50.2       1712688      0.07  64207771      1.04   4321676  61.9
XZL2
:  49 50.2       1642168      4.00   1123636      0.04 112363600  63.5
XZL2
:  50 50.2       1616796      6.03    745363      0.03 149818133  64.1
XZL2
:  51 50.2        959900      2.06   2181817      1.02   4406415  78.7
XZL2
:  52 50.2       1338832      5.06    888249      0.05  89890880  70.3
XZL2
:  53 50.2       1067612      0.01 449454400      0.06  74909066  76.3
XZL2
:  54 50.2       1691436      3.06   1468805      1.06   4240135  62.4
XZL2
:  59 50.2       1630232      6.00    749090      1.00   4494544  63.8
XZL2
:  60 50.2       1614948      7.00    642077      0.03 149818133  64.1
XZL2
:  61 50.2        953568      4.05   1109763      1.02   4406415  78.8
XZL2
:  62 50.2       1335824      7.00    642077      0.04 112363600  70.3
XZL2
:  63 50.2       1058212      5.06    888249      0.06  74909066  76.5
XZL2
:  64 50.2       1689224      4.06   1107030      1.06   4240135  62.5
XZL2
:  69 50.2       1628684      7.03    639337      0.06  74909066  63.8
XZL2
:  70 50.2       1614948      7.01    641161      1.01   4450043  64.1
XZL2
:  71 50.2        953568      5.05    890008      1.00   4494544  78.8
XZL2
:  72 50.2       1335824      0.04 112363600      1.03   4363634  70.3
XZL2
:  73 50.2       1058212      6.00    749090      0.04 112363600  76.5
XZL2
:  74 50.2       1689224      4.05   1109763      1.04   4321676  62.5
XZL2
:  79 50.2       1628684      0.02 224727200      1.05   4280518  63.8
XZL2
100 50.2       1614744      3.01   1493203      1.01   4450043  64.1


edit I've just switched YAM to using XZL2 70% and moving stuff into folders is smooth again, and should be better compressed to boot. (right, now I have to ensure I don't break it!)


Edited by Chris on 2014/9/12 18:45:49
Edited by Chris on 2014/9/12 19:15:25
   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Amigans Defender
Joined:
2006/11/17 22:40
From England
Posts: 3228
New version, new name!

Apparently XPK IDs that start with X are reserved. As such, I've renamed this packer L2XZ. Please delete the old version and download the new one. Files packed with XZL2 should be unpacked before deleting the old library. Technically, you could fix them with FileX, but given that I warned not to use this for important data...

Other changes:
* Modified the pack/unpack code so it is split up how XPK likes it. This should stop the LZMA (un)packer being initialised multiple times, which might save a few CPU cycles.
* Put some reasonable (but still meaningless) values in for pack speed/compression for all modes.
* Took the opportunity to reset to v1, rather than v50.

   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Just can't stay away
Joined:
2006/11/24 18:52
From Gloucestershire, UK.
Posts: 1169
Latest Tst results on X1000, it's seems awfully slow compared to the 460. Does it use altivec?

File 'netsurf' with a size of 14759179 bytes.
Type  Num Version P    CSize     CTime      CSpd     UTime      USpd  Rate
L2XZ
:   0  1.1       4047912      3.56   4145836      0.24  61496579  72.6
L2XZ
:   1  1.1       5045124      4.04   3653262      1.16  12723430  95.0
L2XZ
:   2  1.1       5856088      4.41   3346752      1.26  11713634  89.5
L2XZ
:   3  1.1       6091668      4.10   3599799      1.34  11014312  87.9
L2XZ
:   4  1.1       5905040      4.23   3489167      1.35  10932725  89.1
L2XZ
:   9  1.1       3950304      3.46   4265658      0.24  61496579  73.3
L2XZ
:  10  1.1       3826088      4.49   3287122      0.17  86818700  74.1
L2XZ
:  11  1.1       4774392      5.25   2811272      1.09  13540531  96.8
L2XZ
:  12  1.1       5561964      5.10   2893956      1.19  12402671  91.5
L2XZ
:  13  1.1       5774240      6.27   2353936      1.27  11621400  90.0
L2XZ
:  14  1.1       5699804      6.50   2270642      1.28  11530608  90.5
L2XZ
:  19  1.1       3725896      4.36   3385132      0.16  92244868  74.8
L2XZ
:  20  1.1       3760664      6.00   2459863      0.15  98394526  74.6
L2XZ
:  21  1.1       4706140      6.00   2459863      1.08  13665906  97.3
L2XZ
:  22  1.1       5494960      7.00   2108454      1.18  12507778  91.9
L2XZ
:  23  1.1       5710804      7.44   1983760      1.26  11713634  90.5
L2XZ
:  24  1.1       5647056      7.17   2058462      1.27  11621400  90.9
L2XZ
:  29  1.1       3660352      5.17   2854773      0.15  98394526  75.2
L2XZ
:  30  1.1       3727020      8.11   1819874      0.15  98394526  74.8
L2XZ
:  31  1.1       4671544      8.38   1761238      1.07  13793625  97.5
L2XZ
:  32  1.1       5457988      8.33   1771810      1.17  12614682  92.2
L2XZ
:  33  1.1       5675496      8.51   1734333      1.25  11807343  90.7
L2XZ
:  34  1.1       5618044      9.09   1623672      1.27  11621400  91.1
L2XZ
:  39  1.1       3624972      7.21   2047042      0.14 105422707  75.5
L2XZ
:  40  1.1       3298792     14.28   1033555      0.15  98394526  77.7
L2XZ
:  41  1.1       4337516     13.50   1093272      1.07  13793625  99.8
L2XZ
:  42  1.1       4931808     13.05   1130971      1.16  12723430  95.7
L2XZ
:  43  1.1       5271684     13.59   1086032      1.26  11713634  93.4
L2XZ
:  44  1.1       5135084     14.38   1026368      1.27  11621400  94.4
L2XZ
:  49  1.1       3195440     14.63   1008829      0.14 105422707  78.4
L2XZ
:  50  1.1       3183696     18.17    812282      0.14 105422707  78.5
L2XZ
:  51  1.1       4266436     16.24    908816      1.07  13793625  71.1
L2XZ
:  52  1.1       4854752     15.55    949143      1.16  12723430  96.3
L2XZ
:  53  1.1       5204352     15.33    962764      1.26  11713634  93.9
L2XZ
:  54  1.1       5069992     16.60    889107      1.27  11621400  94.8
L2XZ
:  59  1.1       3083384     18.01    819499      0.13 113532146  79.2
L2XZ
:  60  1.1       3179568     20.23    729568      0.14 105422707  78.5
L2XZ
:  61  1.1       4248648     17.24    856100      1.06  13923753  71.3
L2XZ
:  62  1.1       4840632     17.07    864626      1.16  12723430  96.4
L2XZ
:  63  1.1       5192156     17.50    843381      1.26  11713634  94.0
L2XZ
:  64  1.1       5058832     18.06    817230      1.26  11713634  94.9
L2XZ
:  69  1.1       3064520     20.09    734653      0.13 113532146  79.3
L2XZ
:  70  1.1       3174336     20.32    726337      0.14 105422707  78.5
L2XZ
:  71  1.1       4245332     17.34    851163      1.07  13793625  71.3
L2XZ
:  72  1.1       4843180     17.08    864120      1.16  12723430  96.3
L2XZ
:  73  1.1       5190168     17.42    847254      1.26  11713634  94.0
L2XZ
:  74  1.1       5055176     18.02    819044      1.27  11621400  94.9
L2XZ
:  79  1.1       3074516     20.21    730290      0.13 113532146  79.2
L2XZ
100  1.1       3170928     24.42    604388      0.15  98394526  78.6

_________________
Amiga user since 1985
AOS4, A-EON, IBrowse & Alinea Betatester

Ps. I hate the new amigans website. <shudder>
   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Amigans Defender
Joined:
2006/11/17 22:40
From England
Posts: 3228
@Severin

That does seem a bit slow (and the compression ratios are wrong!). Can you try with this version of lzma.library: http://homepage.ntlworld.com/cdyoung/tmp/lzma_lib_x1000.lha

   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Just can't stay away
Joined:
2006/11/24 18:52
From Gloucestershire, UK.
Posts: 1169
@Chris

That library just gives:

L2XZ:Out of memory

for every test with xbench.

_________________
Amiga user since 1985
AOS4, A-EON, IBrowse & Alinea Betatester

Ps. I hate the new amigans website. <shudder>
   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Amigans Defender
Joined:
2006/11/17 22:40
From England
Posts: 3228
@Severin

Doesn't surprise me, I'm unable to test it here as I don't have an X1000.

I've built it again with different options. I'm a little suspicious as the filesize is exactly the same, but the md5sum is different (c5d7f5db302bca702a4047e38b3b0489 if you want to check you have the new one). It's in the same place. Let me know how you get on!

   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Just can't stay away
Joined:
2006/11/24 18:52
From Gloucestershire, UK.
Posts: 1169
@Chris

Tried the new version, same problem.

_________________
Amiga user since 1985
AOS4, A-EON, IBrowse & Alinea Betatester

Ps. I hate the new amigans website. <shudder>
   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Amigans Defender
Joined:
2006/11/17 22:40
From England
Posts: 3228
@Severin

Hmm, unless anybody has any ideas it looks like it's going to be slow on the X1000 then. Basically the broken version I built for the ppc970, as per some information I saw online (there isn't a PA6T target). Something in that target is causing the code in the library to return an error. Without an X1000 in front of me there's little chance of debugging it.

I might try building a ppc440 specific version in case there's something wrong with the source code, as this is the first time I've compiled liblzma since whatever the date is on that library.

edit a version built specifically for the PPC440 is working correctly (but no faster than the generic version), so probably the 970 isn't compatible enough with the PA6T. Will take suggestions of suitable gcc switches.


Edited by Chris on 2014/9/13 20:13:39
   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Just can't stay away
Joined:
2012/8/14 12:51
From France
Posts: 1494
...


Edited by zzd10h on 2014/9/14 11:13:59
   Report Go to top

Re: New xpk sub-lib for testing (XZ-based)
Not too shy to talk
Joined:
2007/2/6 13:57
From Donostia (SPAIN)
Posts: 412
Ok, tests with XPKatana on my SAM460:

--- XPKatana 1.8 Benchmark Results ---
Pack Mode40
Source
RAM Disk:NetSurf/NetSurf
Size
:   14.759.179 bytes

-----
 
BZP2    Speed:  694.805 bytes/sec   Size4.757.240 bytes   Efficiency68%
 
L2XZ    Speed:  225.680 bytes/sec   Size3.298.792 bytes   Efficiency78%
 
LZMA    Speed:  279.587 bytes/sec   Size3.907.816 bytes   Efficiency74%
-----
   
Best PackerL2XZ  -  Size3.298.792 bytes  -  Ratio78%
Fastest PackerBZP2  -  Speed694.805 bytes/sec


--- XPKatana 1.8 Benchmark Results ---
Pack Mode60
Source
RAM Disk:NetSurf/NetSurf
Size
:   14.759.179 bytes

-----
 
BZP2    Speed:  635.657 bytes/sec   Size4.724.188 bytes   Efficiency68%
 
L2XZ    Speed:  165.906 bytes/sec   Size3.179.568 bytes   Efficiency79%
 
LZMA    Speed:  230.781 bytes/sec   Size3.667.064 bytes   Efficiency76%
-----
   
Best PackerL2XZ  -  Size3.179.568 bytes  -  Ratio79%
Fastest PackerBZP2  -  Speed635.657 bytes/sec


--- XPKatana 1.8 Benchmark Results ---
Pack Mode79
Source
RAM Disk:NetSurf/NetSurf
Size
:   14.759.179 bytes

-----
 
BZP2    Speed:  643.671 bytes/sec   Size4.719.568 bytes   Efficiency69%
 
L2XZ    Speed:  164.605 bytes/sec   Size3.074.516 bytes   Efficiency80%
 
LZMA    Speed:  191.541 bytes/sec   Size3.405.416 bytes   Efficiency77%
-----
   
Best PackerL2XZ  -  Size3.074.516 bytes  -  Ratio80%
Fastest PackerBZP2  -  Speed643.671 bytes/sec


Edited by jabirulo on 2014/9/14 9:14:37
   Report Go to top





[Advanced Search]


Powered by XOOPS 2.0 © 2001-2016 The XOOPS Project