PMD 3 (P.acked-M.oD.ule) pattern packer by The Sirius Cybernetics Corporation [web]
PMD3-Infotext
*=-------------
As I want to release it finally (it seems to be 100% Bug-Free) I think it
is important to write some lines about the packer and the format. So that
maybe someone else than me will use it....
I. What is it?
*=-------------
PMD3 is my (Dynacore/TSCC) own little patterncompressed moduleformat, which
allows to depack the module in realtime while playing without (nearly) any
loss of CPU-time.
PMD3.PRG for 4-Channel mods and
PMD3_V8.PRG for 8-Channel mods
(Sorry that both are quite slow, but it's pure GfA-Basic code...)
II. What means PMD3?
*=-------------
This one is called PMD3 (P.acked-M.oD.ule, Version 3), because this is the
third change in the DEPACKER-Code.
The idea of a patterncompressor came from BLADE/Newcore, who did such a
thing for own purposes and released it together with the Octalyser.
I heard of it and began to code my own:
The first patternpacker was written for DYNABUSTERS+ to save the rare
memory. It was a simple run-lenght packer and so the compressionrates were
quite bad (around 80% of the orginal patternsize).
But then I had another idea, which allows a packrate of around 30%.
Well, this packer then was the used patternpacker for DYNABUSTERS+. As Scy
registered for Octalyser, I finally had the possibility to compare Blades
packer with mine.
Dissapointing for me: Blades one was better...
So I thought about another method, or how to improve my algorithms.
Well, I created a sequencelist, that saves any doubles -> And that really
helped a lot!!
So here is PMD3! And it beats Blades packer is all cases I compared them!!!
III. The Format
*=-------------
File-Description for
Module-Patterncompressor v3.0
Offsets |Lenght | Meaning
----------+---------+--------------------------------------------------
0 | 1080 |= Orginal ProTracker-Header
----------+---------+--------------------------------------------------
1080 | 4 | ID: 'PMD3' - 4 Channels
| | 'PMd3' - 8 Channels
| | 'pMD3' - 4 Channels / Delta-Samples
| | 'pMd3' - 8 Channels / Delta-Samples
----------+---------+--------------------------------------------------
1084 | 4 | Control-Len
1088 | 4 | Data-Len
----------+---------+--------------------------------------------------
1092 |p*trcks*8| For every pattern:
| | For all tracks: Control-Offs .l (C-Offs)
| | Data-Offs .l (D-Offs)
----------+---------+--------------------------------------------------
How you can find the other needed addresses:
1092+p*tracks*8 --> Control-Bytes (C-Adr)
1092+p*tracks*8+Control-Len --> Data-Longs (D-Adr)
1092+p*tracks*8+Control-Len+Data-Len --> Samples
----------+---------+--------------------------------------------------
(C-Adr)+(C-Offs) --> Address of the Control-Bytes-Start for Track
(D-Adr)+(D-Offs) --> Address of the Data-Longs-Start for Track
----------+---------+--------------------------------------------------
How the data is encoded:
The Controlbytes are coded that way:
76543210
||||||++--- Runlenght
++++++----- Data-Offset
Mask out the upper six Bits and you get how often you have to repeat the
Data-Long (like for DBRA, that means: 0=One time, 3=Four times!)
To get the Data-Long just mask out the lower two Bits and add that value to
the Data-Long-Startadr of the Track you want to depack.
Track is finished when 64 Longs are depacked.
A depacked pattern is absolutely identical to a normal pattern, so you
don't have to rewrite your player completely if you really want to use my
compressed format.
IV. The Sampleprepacker (PMD3_SPP.PRG)
*=-------------
Well, this prepacker just creates Deltasamples using a logharitmic table,
so that the sample will sound a bit different.
You can enter how many different delta-values will be used to create
deltasamples.
The deltas are stored as bytes. That's why it becomes just a kind of
prepacker, 'coz the size of a PMD3-Module doesn't get smaller by simply
using the PMD3_SPP.PRG
Just if you use a 'REAL' packer afterwards you can see how the samples can
be packed better than without it.
Here's a little algorithm to show how the deltasample can be converted to
eight bit samples again:
moveq #0,D0
delta_to_8bit: add.b (a0),d0 ;a0 points to samplestart
move.b d0,(a0)+
subq.l #1,d1 ;d1 is the samplelenght
bgt.s delta_to_8bit
V. Address
*=-------------
DYNACORE
ANDREAS JOHN
SCHULZENSTR.36
76771 HRDT
GERMANY
MAUSNET: Andreas John@KA
E-MAIL: andreas_john@ka.maus.de
CALL .tSCc. WHQ: NIGHTBOX: +49 (0) 8974995006 (mail to TSCC)
[ back to the prod ]