@DATABASE SDHC_1.1 @WORDWRAP @WIDTH 80 @NODE MAIN "xpkSDHC V1.1" @{b}xpkSDHC.library V1.1 Contents@{ub} -------- @{"Introduction" LINK Introduction} @{"Usage" LINK Usage} @{"Effiency" LINK Effiency} @{"Comparision" LINK Comparision} @{"History" LINK History} @{"ToDo" LINK ToDo} @{"Author" LINK Author} @{"Disclaimer" LINK Disclaimer} @ENDNODE @NODE "Introduction" xpk@{b}SDHC@{ub}.library (V1.1) ---------------------- This library is a standard sublibrary for the XPK-system. It was programmed to compress especially samples of nearly all kinds better than already available sublibraries. Besides the compression of 8-bit-samples like IFF-8SVX you can also compress 16-bit-Samples, e.g. in AIFF-Format. @{b}SDHC@{ub} is an abbreviation of "@{b}S@{ub}ample @{b}D@{ub}elta @{b}H@{ub}uffman @{b}C@{ub}ompression". The H for Huffman is somewhat obsolet, for compression GZIP is used instead of HFMN. The xpk@{b}SDHC@{ub}.library normally achieves equal or even better results than xpkSMPL.library, but with major decrease in speed. You also could replace xpkDLTA.library, because it's possible to save the coded, but uncompressed data. In the following text most times I used only the 4-letter-name (@{b}SDHC@{ub}) of a sublibrary instead of the whole name like xpk@{b}SDHC@{ub}.library. @{"Next" LINK Usage} @ENDNODE @NODE "Usage" @{b}Usage@{ub} ----- To use @{b}SDHC@{ub} you must copy the sublibrary to the Libs:Compressors directory. After doing this you can use it immediately. To run @{b}SDHC@{ub} Kickstart V2.0 or better is required. To use the compressing modes of @{b}SDHC@{ub} you also have to install the xpkGZIP.library. The newest version of xpkGZIP.library available in Aminet is V1.0. The normal working process of @{b}SDHC@{ub} is: Data input -> @{"coding" LINK Coding} -> @{"compressing" LINK Compression} -> Data output You can use different kinds of coding. You can also switch the compression off. @{"Next" LINK Effiency} @ENDNODE @NODE "Coding" @{u}Data coding@{uu} SDHC provides different modes for different kinds of samples. These are: @{u}type of data example mode name@{uu} 8 bit IFF-8SVX B byte 16 bit AIFF W word 16 bit, stereo AIFF SW stereo word Each of these three modes the data will be delta-coded. Here you can choose between two levels: simple or double delta-coding. In the mode-descriptor this is visualized by "1" or "2". Sometimes the double delta-coding improves the compression up to 10%. To replace the xpkDLTA.library you have to choose the mode "B1", which means 8-bit data with simple delta-coding. @{"Next" LINK Compression} @ENDNODE @NODE "Compression" @{u}Data Compression@{uu} @{b}SDHC@{ub} uses the xpkGZIP.library for compression. This sublibrary can be found in Aminet (util/pack/xpkGZIP). If the sublibrary isn't installed, @{b}SDHC@{ub} answers with the XPK-errormessage "Can't find required XPK library". To compress data you have to choose a mode, which has a "P" in it's descriptor. @{"Next" LINK Modes} @ENDNODE @NODE "Modes" @{u}Modes of Operation@{uu} The table gives you the number und description of all modes, like xQuery shows you. @{u}number description in words@{uu} 0 .. 7 B1 byte, simple delta 8 .. 15 B2 byte, double delta 16 .. 23 W1 word, simple delta 24 .. 31 W2 word, double delta 32 .. 39 SW1 stereo word, simple delta 40 .. 47 SW2 stereo word, double delta 48 .. 55 P B1 packed, byte, simple delta 56 .. 63 P B2 packed, byte, double delta 64 .. 71 P W1 packed, word, simple delta 72 .. 79 P W2 packed, word, double delta 80 .. 87 PSW1 packed, stereo word, simple delta 88 .. 100 PSW2 packed, stereo word, double delta @{"Next" LINK Effiency} @ENDNODE @NODE "Effiency" @{b}Effiency@{ub} -------- To define the effiency of a data compression is rather complicated. A file is compressed ideal, if it is represented by only 1 bit (!). That is the case, if the data are equal to the reference-data of the compression-scheme. Because @{b}SDHC@{ub} should compress any sample-data, the compressionrate can't be compared with the ideal case. The delta coding is a relatively simple calculation, which is done very fast. The xpkGZIP.library is responsible for compression. Because of that speed and compression-rate depends on this sublibrary. The values xQuery gives you aren't of much use. I don't own neither a A3000/25 MHz nor the program "AmigaVision". The present values are found out by using xBench with the program-file "PPaint V7.1". If there is anyone, who could make the standard test of the XPK-system and send me the results, I would be glad. @{"Next" LINK Comparision} @ENDNODE @NODE "Comparision" @{b}Facts@{ub} ----- At first the complete output of xQuery: Packer : SDHC Name : Sample Delta Huffman Compressor Descr. : Coder for audio data DefMode: 50 DefChunk: 128 Kb MaxChunk: 1023 Kb Pack Unpack Pack Unpack Name Mode Ratio Speed Speed Mem Mem Description SDHC 0..7 0.0% 1150K/s 1300K/s 0K 0K B1 SDHC 8..15 0.0% 700K/s 800K/s 0K 0K B2 SDHC 16..23 0.0% 1700K/s 1800K/s 0K 0K W1 SDHC 24..31 0.0% 1050K/s 1200K/s 0K 0K W2 SDHC 32..39 0.0% 1700K/s 1800K/s 0K 0K SW1 SDHC 40..47 0.0% 1050K/s 1200K/s 0K 0K SW2 SDHC 48..55 37.0% 15K/s 200K/s 256K 128K P B1 SDHC 56..63 28.9% 15K/s 200K/s 256K 128K P B2 SDHC 64..71 29.7% 15K/s 200K/s 256K 128K P W1 SDHC 72..79 20.4% 15K/s 200K/s 256K 128K P W2 SDHC 80..87 27.8% 15K/s 200K/s 256K 128K PSW1 SDHC 88..100 16.1% 15K/s 200K/s 256K 128K PSW2 The values for required memory are given for working on DefChunk-sized blocks. The exact memory requirement depends on the chunksize. When compressing, @{b}SDHC@{ub} needs 2*chunksize and on decompression 1*chunksize. @{"Next" LINK Comparision2} @ENDNODE @NODE "Comparision2" @{b}Comparision with other Sublibraries@{ub} ----------------------------------- For comparision different data/samples are used. All tests are done with xBench. The computer used is an Amiga 500, equipped with 68030+68882/50 MHz, 1+16 MB RAM, harddisk, CD-ROM and ZIP-drive. All data are copied to RAM-disk. As competitors I choosed: @{u}type purpose@{uu} RAKE.100 all kind of data SMPL 8-Bit-Samples SQSH 8-Bit-Samples GZIP.70 all kind of data (mode 70 is used in @{b}SDHC@{ub} too) @{u}Test 1@{uu} Simple testfile, content doesn't matter Type Num Version P CSize CTime CSpd USize UTime USpd Rate NONE: 100 1.2 568256 0.12 4733900 568068 0.12 4733900 0.0 SDHC: 7 1.1 568196 0.24 2366950 568068 0.21 2705085 0.0 SDHC: 15 1.1 568196 0.41 1385531 568068 0.36 1577966 0.0 SDHC: 23 1.1 568196 0.17 3341576 568068 0.15 3787120 0.0 SDHC: 31 1.1 568196 0.27 2103955 568068 0.23 2469860 0.0 SDHC: 39 1.1 568196 0.17 3341576 568068 0.15 3787120 0.0 SDHC: 47 1.1 568196 0.27 2103955 568068 0.23 2469860 0.0 This test should show the speed of the delta-coding routines. Speed varies from a third up to three quarter of speed of simple copying, like NONE does. @{u}Test 2@{uu} AIFF-file: 617454 bytes, 16 bit per sample, 2 tracks, 22050 Hz (part of Oasis: Wonderwall, "I don't believe...") Type Num Version P CSize CTime CSpd USize UTime USpd Rate RAKE: 100 1.7 617632 3.07 201125 617454 0.18 3430300 0.0 SMPL: 100 1.0 617640 2.00 308727 617454 0.14 4410385 0.0 SQSH: 100 1.10 617652 9.29 66464 617454 0.13 4749646 0.0 GZIP: 70 1.0 587720 23.06 26775 617454 2.51 245997 4.9 SDHC: 55 1.1 618112 20.82 29656 617454 0.66 935536 0.0 SDHC: 63 1.1 618112 20.96 29458 617454 0.81 762288 0.0 SDHC: 71 1.1 573956 24.42 25284 617454 2.70 228686 7.1 SDHC: 79 1.1 595048 24.58 25120 617454 2.78 222105 3.7 SDHC: 87 1.1 555084 25.11 24589 617454 2.71 227842 10.2 SDHC: 100 1.1 550268 25.37 24337 617454 2.80 220519 10.9 As you could see, the "PSW"-modes of @{b}SDHC@{ub} are the best in the test. The double delta-coding gives an advantage of nearly 1%. The only compressor other than @{b}SDHC@{ub}, which gives a compression at all, is GZIP. @{u}Test 3@{uu} IFF-8SVX-file: 220548 bytes, 8 bit per sample, 1 track, 22050 Hz (part of Rammstein: Engel, 10s from beginning) Type Num Version P CSize CTime CSpd USize UTime USpd Rate RAKE: 100 1.7 75700 2.09 105525 220548 0.14 1575342 65.7 SMPL: 100 1.0 106868 0.53 416128 220548 0.26 848261 51.6 SQSH: 100 1.10 76928 4.88 45194 220548 0.24 918950 65.2 GZIP: 70 1.0 64612 5.99 36819 220548 0.42 525114 70.8 SDHC: 55 1.1 67088 7.64 28867 220548 0.49 450097 69.6 SDHC: 63 1.1 71032 10.30 21412 220548 0.57 386926 67.8 SDHC: 71 1.1 85540 6.05 36454 220548 0.55 400996 61.3 SDHC: 79 1.1 99456 6.26 35231 220548 0.66 334163 55.0 SDHC: 87 1.1 104224 5.79 38091 220548 0.64 344606 52.8 SDHC: 100 1.1 141760 6.59 33467 220548 0.82 268960 35.8 Against the direct competitors, SMPL and SQSH, @{b}SDHC@{ub} shows a better compression, but is also slower. GZIP without Delta-coding is even better, perhaps because this sample is very quiet. @{u}Test 4@{uu} AIFF-file, 441054 bytes, 16 bit per sample, 2 tracks, 22050 Hz (5 seconds of speech) Type Num Version P CSize CTime CSpd USize UTime USpd Rate RAKE: 100 1.7 394016 2.27 194296 441054 0.23 1917626 10.7 SMPL: 100 1.0 415012 1.41 312804 441054 0.16 2756587 6.0 SQSH: 100 1.10 404960 10.74 41066 441054 0.40 1102635 8.2 GZIP: 70 1.0 354152 16.65 26489 441054 1.68 262532 19.8 SDHC: 55 1.1 405380 15.37 28695 441054 0.93 474251 8.1 SDHC: 63 1.1 406992 14.39 30650 441054 0.75 588072 7.8 SDHC: 71 1.1 346484 16.88 26128 441054 1.78 247783 21.5 SDHC: 79 1.1 360988 16.80 26253 441054 1.86 237125 18.2 SDHC: 87 1.1 341648 17.03 25898 441054 1.78 247783 22.6 SDHC: 100 1.1 345724 17.04 25883 441054 1.83 241013 21.7 This test shows the importance to choose the right mode for the data. According to the data you have to use the word-modes of @{b}SDHC@{ub}. The minimal difference between mono- and stereo-word-coding is probably due to a high correlation of both sound channels. @{u}Test 5@{uu} PGM-picture, 262159 bytes, 512*512 pixel, 8 bit grayscale Type Num Version P CSize CTime CSpd USize UTime USpd Rate RAKE: 100 1.7 193184 2.97 88269 262159 0.25 1048636 26.4 SMPL: 100 1.0 133288 0.68 385527 262159 0.33 794421 49.2 SQSH: 100 1.10 141976 4.25 61684 262159 0.30 873863 45.9 GZIP: 70 1.0 176984 10.36 25304 262159 0.92 284955 32.5 SDHC: 55 1.1 146056 13.03 20119 262159 0.92 284955 44.3 SDHC: 63 1.1 158176 10.93 23985 262159 1.05 249675 39.7 SDHC: 71 1.1 159940 11.09 23639 262159 0.96 273082 39.0 SDHC: 79 1.1 179012 9.84 26642 262159 1.11 236179 31.8 SDHC: 87 1.1 172732 10.48 25015 262159 1.02 257018 34.2 SDHC: 100 1.1 194736 9.85 26615 262159 1.18 222168 25.8 At the university I had to work with some PGM-pictures. When archiving them I used normally SMPL, sometimes also SQSH. Unfortunately @{b}SDHC@{ub} doesn't reach the results of SMPL or SQSH. But the Delta-coding improves the compression of GZIP by 12%. Another good compressor for this type of data is DMCB - but with only 2 KByte per second. @{u}Note@{uu} For V1.0 of @{b}SDHC@{ub} I got some error-messages, but I was unable to find the reason for this. After changing from HFMN to GZIP as compressor, I never got any error-messages. Unfortunately GZIP works much more slowly than HFMN, but on the other side most times it gives you a better compression. Old files packed with @{b}SDHC@{ub}/HFMN are also unpacked with this version, because decompression is done by xpkmaster.library. @{"Next" LINK History} @ENDNODE @NODE "History" @{b}History@{ub} ------- version date 1.0 08.03.98 first version 1.1 30.03.98 set a missing bit in PackerInfo-structure reworked Delta-coding-routines replaced HFMN by GZIP because of some strange error-messages @{"Next" LINK ToDo} @ENDNODE @NODE "ToDo" @{b}ToDo-List@{ub} --------- - (let) determine correct values for xQuery - other suggestions: ??? - only English AmigaGuide: May someone tell me the worst grammar mistakes? (of course with the right phrase, I wish to learn English! :-) It depends on my motivation and spare time, if and when I will program these points. @{"Next" LINK Author} @ENDNODE @NODE "Author" @{b}About me@{ub} -------- If you have any questions, remarks, suggestions or even bug reports please let me know. Gerd Wieczorek SMail: Berliner Str.1 14959 Trebbin BRD EMail: gwieifjc@sp.zrz.tu-berlin.de (checked only once per week, so don't await immediate response) @{"Next" LINK Disclaimer} @ENDNODE @NODE "Disclaimer" @{b}Disclaimer@{ub} ---------- This program is freeware, but gifts of (nearly) all kinds are welcome by the @{"author" LINK Author}. The entire risk as to the use of the program and the information presented is assumed by the user. In no event will I be liable for any damage resulting from any claim arising of the use of the program or the information presented herein, even if I have been advised of the possibilities of such damages. If anyone wants to include @{b}SDHC@{ub} into another archive, he should send me an e-mail before. @{"Back to Contents" LINK MAIN} @ENDNODE