|
|
sources
mc680x0 fixedpoint math library for devpac, v1.4 |
|
about:
a project i coded a while ago trying to provide the mc680x0 cpu range with
fixed point versions of the most common floating point instructions offered
by motorola's mc6888x fpu series.
the implementation is mainly based on pade approximations.
requirements/specs.:
-a 680x0 cpu
-assembles with devpac/madmac
download (devpac version)
download (madmac version)
view relative error results
|
|
32 bytes sinus-table generator |
|
about:
this little snippet generates a sinus table using a parabolic
approximation. the result should be accurate enough for simple demo
screens and appear useful to somebody who needs a small sinus table
generator for byte or Ktros that are aimed at non fpu systems as
this is 'only' 32 bytes of plain 68000 code. another version that
needs 28 bytes but which produces a buggy table is also included.
requirements/specs.:
-a 680x0 cpu
download
|
|
truecolor tga-viewer |
|
about:
a command-line tool that let's you view 24bit targa images in an
interlaced 12bit mode on any 16/32bit atari machine except for the falcon.
of course, this involves alot of flicker as the effective refresh rate is only
20Hz. but remember it shows 4096 colors even on a plain ST.
the images' width may not exceed 320 pixels as the ST shifter only
supports vertical hardware scrolling (by multiples of 8 because of the
256byte alignment, however), which is why i only allowed a max. height
of 480 pixels.
requirements/specs.:
-ST(E)/MST(E)/TT030
-RGB/VGA monitor
-fastram accelerated
-smooth 1vbl scrolling
download
|
|
blitter example source |
|
about:
this one shows how gouraud and transparent gouraud triangles can be
mapped with hardware acceleration on the Falcon 030. there are two flags
inside the source, one called "method" to let you choose between cpu and
blitter mapping and a second one, "transparent" to choose if the triangles
should be mapped transparently. all 3 methods are provided as binaries, too.
please don't bother with the sometimes strange label names, i just cutted &
pasted the source directly from my 3d engine at some places. the performance
gets displayed as number of triangles/minute after you've pressed the space
key once. on a standard Falcon the speed difference is astonishing:
the blitter almost speeds it up by a factor of 2. for more information read
this article and take a look into the source.
requirements/specs.:
-Falcon 030
-RGB/VGA monitor
-fastram accelerated
download
|
|
lz77 packer/decruncher, v1.3 - final version |
|
about:
with a fairly effective packrate this utility reaches extremly high
decompression speed. the lz group of compression algorithms are based upon
the idea of encoding repeating strings similar to the way rle does with runs
of the same value. this explains why the bit operation bottleneck involved
in huffman coding/decoding falls away giving it its high speed.
as it does not make use of the patented LZW algorithm you may use this one
freely for storing your texture-, sound- and especially realtime animation data
in your demos, games etc without worrying about copyright laws :).
a format and useage description is included.
requirements/specs.:
-a 680x0 cpu
-incredibly fast decruncher
download
|
|
lz78 packer/decruncher, v1.0 |
|
about:
here comes an alternative to lz77. lz78 gives you a better compression ratio
while still providing a fairly good depacking speed. for several reasons the
the decoder has been optimized for size so it still could be faster, though.
there shouldn't be any legal issues neither, because the implementation differs
from the one of LZW.
lz78 is used to encode coefficients in my upcoming image codec called BIQ, but
have been some requests which is why i'm releasing this seperately. some credit
needs to be given to mark r. nelson for writing the original c source in 1989.
requirements/specs.:
-a 680x0 cpu
download
|
|
4ch protacker replay routine by lance (TT030 fix), optimized v1.1 |
|
about:
i've made up this little fix for lance's protracker replayer so that it works
on TTs as well. in my humble opinion this routine is quite stunning, as it uses
~30% cpu-time on the STE and even less on the TT while giving real 50Khz replay.
this unique speed can be achieved by using the STE's/TT's mircrowire hardware for
various volume tricks so that the cpu is left free for diffrent tasks which explains
why it doesn't work on the Falcon lacking that particular chip.
the replayer relies on a 50Hz interrupt which is realised using timer a in the example
source. i also optimized the code a bit to use 030 addressing and to use jumptables
to process pattern and efx commands. a nice module by tinker/lineout is included.
requirements/specs.:
-STE/TT030
-very fast 50Khz replayer
download
|
|
- 2002-2005 ray//.tscc. -
|
|
|