The Datafile PD-CD 4

home *** CD-ROM | disk | FTP | other *** search

/ The Datafile PD-CD 4 / DATAFILE_PDCD4.iso / utilities / utilsc / computils / !CompUtils / Resources / !Changes next >

Wrap

Text File | 1996-04-13 | 12.4 KB | 251 lines

Note ==== This version has been finished in something a flurry. I'm really supposed to be doing work on my Third Year final project, but I'd got bored with it. :-) Anyway, now I have to get back to work but I thought I might as well release this now because it'll probably be months before I get this much free time to work on a project. Spelling mistakes, incomprehensible documentation, etc. are all to be expected. Don't hesitate to get in touch with me by email if you can't understand me. Anyway, there are now really, really, really simple decompressors for you in case you're one of these people who just wants to be able to compress all the samples they're using in a program and have it up and running again 5 minutes later. So now you've no excuse to waste everybody's download time with your 300K+ samples when they could be 160K or less. Oh yes, and please don't upload this to The Plasma Sphere - that's where I put the lastest versions. I'm sick of people uploading older versions than those already on the board, and then have people download them thinking they're new because they show up in the list of new files. Possible future improvements ============================ - Type 6 samples to support *storage* of 16bit samples. - StrongHelp files for documentation in addition to text files. - Type 5 to be added, based on Type 4. I've already got something in mind but a recent trial only brought compression down from 50% to about 46% so it needs a bit more work. - Sharing of tables in Expand modules to make the universal decoder much smaller. - All compressors to accept 16-bit linear and 8-bit VIDC as input as well as 8-bit linear. - Optional external 12-bit lin->VIDC and VIDC->16-bit lin tables for improved quality, plus a short utility of some sort to generate such tables. Otherwise the internal 8-bit tables will be used. - LoadCSamp can optionally produce a voice generator that repeats the sound automatically for background noises in games eg. engine noises. Version 1.14 ============ - Type 7 added. This is a slight variation of Type 6 with the sign of deltas coded in a new way so as to favour positive deltas and unbalance the tree. This yields a small increase in compression. By small I mean *small*! For example, a 460K sample compressed down to 300K with Type 6. With Type 7 it compressed to 296K, a saving of only 4K. I had expected more than this, but perhaps things will improve if I add the predictor I've been considering... Certainly this technique can be applied to Type 5 when I get round to doing it. - Type6 changed to Type6-7. I had to make the new variant into Type 7 to maintain compatibility with the old Type 6 decoder. The new feature is enabled by setting bit 2 of the flags, otherwise the standard Type 6 output is generated. Version 1.13 ============ - Slight changes to documentation here and there. - Addition of Type 4 decompressor and compressor. Type 5 is not yet ready. - Slight internal changes to all compressors and decompressors. - New reason code can be returned from Compress. Code 7 is used to request claiming of the phase 1 workspace. I'm having a bit of a rethink about this so it might change in the future, perhaps adding a similar mechanism for the global workspace. Version 1.12 ============ - Minor bug fix to Compress. If you try to compress a file of zero length then only the 7 byte header is actually written. Expand modified to accept these files. - Added optional table to Type6 for use with 8-bit samples to improve compression. Worst case it *could* increase the file size by 32 bytes, but almost always results in a slight improvement so is well worthwhile. Version 1.11 - beta-test release ============ - New sample type added (type 6). Slightly more efficient than types 0-3, requires less code for the decoder and is probably faster too. - From now on, all code that is modified will adopt the version number of the particular release of !CompUtils, rather than having individual histories. - Minor bug fix to Expand types 0-3. - C/APCS and ObjAsm versions of Compress/Expand/ExpCode. Note that these are patched versions of the raw code and therefore not as easy to use and feel a little hacky. Let me know if you have any problems with them. - New management of !CompUtils sources to provide automatic updating in future by means of Amu. Means it is now possible to develop new algorithms without an admin explosion. Took three days to do just this, which is why I've mentioned this, even though you aren't going to be remotely interested. :-( - Expand header offset removed from ExpCode header and replaced with the filetype for compressed samples. - Can now access files through a path variable, eg. CompUtils:Expand.Type6 - Loads of new documentation, including a reorganisation of same. - Compress and Expand now never return error codes 4 and 7. These are now reserved for future use. Therefore, 0-3 are valid return codes, 4-7 are soft errors, and 8 and up are hard errors. - Various minor bug fixes - Complete rewrite of the universal Expand. Somehow it hadn't been updated when it should and still expected a valid source buffer at the first call. It now behaves as a normal Expand utility again. - Type 6 samples now have a logarithm mode which is more efficient than previously. However, the linear mode is still recommended. - There is now a facility within Compress whereby multiple passes of the source file may be used in the scanning phase. Compress can now return reason code 4 (previously an unused error condition) which is a request to move back to the start of the source. Unlike reason code 3 it does *not* indicate the end of the phase, and the workspace should be left untouched. At present no compressor uses this facility but it is expected that Type6 might do in the near future. - Sample format now returned in Expand's header (but not in ExpCode's) - Loads of example code added - A module similar to LoadSample but it loads the sample into a block of RMA and installs a voice generator for it. Hey presto, compressed samples can be added to any program that wants them just by adding a few *commands to the !Run file. Version 1.10 - never released (complete rewrite) ============ This is a development version distributed to any users who may be considering using !CompUtils in their own programs. The aim is to let them try out the new versions of Expand and Compress before the proper release. The code may therefore still contain bugs. Please report any problems to me a.s.a.p. The next release will be a proper release and should contain code for type 4 samples (16bit ADPCM) and possibly 5 and 6 if I have the time. The algorithms for these have already been created, but not fully implemented. General ------- - New directory structure. - A program can reference CompUtils code by <CompUtils$Dir> during assembly. - Compression code added. - Separate code modules for compressing/expanding each sample type, plus a code module to decompress *any* known sample type. - Sprites directory added, containing both low res and high res sprites to be used for file icons. Any of these can be added to your own !Sprites file. - Histories started, simply because I keep getting letters from people who have ancient versions of this or !Compress and I can't remember what changes I've made. From now on all that's changed (and you lot get to see what a busy beaver I've been). Expand ------ - Expand code now handles returning to the user in a much improved manner. This means that you should *not* pass any source data to Expand until it asks for it. In addition, when a reason code of 0 is returned to indicate completion, the output buffer will *not* contain any data. Any outstanding data would have been passed using reason code 2 (destination buffer full) beforehand. All output buffers bar the last one will be completely full. The last one may be partially empty, but never completely empty. - When reason code 0 is returned from Expand, +28 gives the number of bytes that were in the last destination buffer returned, unless the buffer was completely full, in which case +28 is 0. - +44, the number of bytes left to be decompressed is now no longer divided by 16. After reason code 3 it is identical to +40. - +40 is the total number of output bytes not the total number of samples. This distinction is important now as Expand can output 16bit samples, where the number of bytes is twice the number of samples. - The output data no longer includes the sample period byte used for Armadeus files. This can be added manually using +48, which becomes valid after reason code 3. The output length calculated also omits this byte. - No data will be written to the output until after reason code 3 is returned. The destination buffer pointer and length values do not need to be set up until then. - The output buffer does not need to be a special length any more - any non-zero size will do. Similarly, the source buffer size is unrestricted. (Previous versions of Expand insisted the first source buffer passed was at least 320 or so bytes long, and that all destination buffers had to be a multiple of either 16 or 32 bytes.) - Bit 2 of the output flags is no longer relevant, so is now used to produce 16bit linear output (little endian - ie. lower byte first) instead of 8bit if linear output is selected. Sample types 0-3 only support 8bit samples internally, so Expand0-3 just outputs each byte twice, thus 8bit &FE becomes 16bit &FEFE. This is really provided for forward compatability. - Bit 3 of the output flags is no longer used. Previous this was used to downgrade the enhanced sample output flag when batches of samples were used (mainly when !Player is playing a sequence file). This is no longer supported. Instead, Expand returns information about the sample and how it intends to output it on reason code 3, as follows: bit 31 - if set, sample format stores sample as 16 bit. This specifies whether it would be worthwhile trying to output the sample as 16bit: if it is clear you gain nothing in quality but double the output sample size. Of course, it is too late to alter bit 3 now. You would have to restart the whole operation to change it. bit 30 - if set, sample file is storing the sample as linear signed rather than logarithmic. This is pretty much irrelevant. bit 29 - if set, Expand intends to use the enhance output format. In this case the output length is doubled and every other byte is formed by interpolating its two neighbours (to reduce distortion caused by the sampling waveform appearing in the output - a problem with low sample rates which results in a whining noise). bit 28 - if set, Expand is going to store the sample as 16bit rather than 8bit. It can be used in conjuction with bit 29, which would lead to an output sample four times larger than the original Armadeus file. - Output can be stored as enhanced regardless of whether or not the sample itself was compressed with the enhance flag set. (From now on, the enhance flag in the !Compress status window will described as a linear/log switch.) However, the output sample must be set to linear output to use either this feature or the 16bit output (or both). - The sample period in the header is unaffected by whether or not the sample is to be stored as enhanced, thus with enhanced output it will be twice as large as you might expect. Use bit 29 of the output flags if this is a problem to you. - Error codes 4-7 are considered soft errors. If you can determine and correct the problem encountered, then you may carry on as if nothing had happened. Error code 8 is a fatal error and you *must* abort the operation not continue or Expand will crash. ExpCode ------- Largely unchanged. A few small changes internally to let it use the new version of Expand. - The restrictions that applied to the flags previously now no longer hold. The flags are passed to Expand every time you call ExpCode, and the flags returned from Expand are passed back to you in ExpCode's header. You can therefore use the new status bits (28-31) for information. LoadSample ---------- Unchanged externally, but uses the new version of ExpCode. Compress -------- This did not exist previously, so all features are new! !Player/!Compress ----------------- These have not yet been updated and may crash horribly when used.