Chunks 2x2
Poisoned CyberJack / TBK
The new word in demomakinge:)
Greetings to all who consider themselves encoders - steep and
so-so! Today I will try to surprise you
:)
Modern coding (for
Spectrum, of coz) is mainly based on chankovyh engines. And this
correct - for otherwise it is very difficult to realize many
things. Explain the details will not - all of it was already
quite a long time. Especially, the choice is already made ...
Chunks are different. Contrary
popular belief, all Gopnik, versed in
code, as I have in Chinese cooking,
chunks - is not only the squares
4x4:) Roughly speaking, the chunks - this is
such a technology in which the screen is filled with blocks
having the same size, usually image is built not just on
screen, and in the buffer and we have random access to any part
of the image, regardless of others (in contrast to our planar
screen), and we usually draw the image in more colors than have
our ZX (15, not forgotten?) Since the campaign against
illiteracy complete dummies are not included in my
plans, at this muddled description I close this question - who
should be may find the information in other publications (Deja
Vu, Demo or Die, Buzz, etc)
So - chunks may be different. On
Speke mostly yuzayutsya size 4x4 (two types - pixel [Refresh,
Dogma, Inbetween] and Multicolor [Eye Ache 1 / 2,
Rage, Stellar Contour]), as
is the optimal ratio
between resolution and volume
screen. 8x8 species, as
purely attribute [any demos 95-97 years] and pixel-attribute
[Emergency, 63Bit3, Smile_4K], because of the low-resolution
strongly lose, so how to do something normal in
32h24 almost unreal. Oh, forgot
say - speaking about the code, I am
mind 3D - because it is the most difficult
(Hence interesting) application koderskih abilities.
All this has been tested very
long (97y year), and many have already nabilo mouth. Is there
no alternative?
And there is a solution - and also long been known. Remember
these things Higher State / 3SC, Over The Top
/ Enigma, Matricks / Baze? Well
Yes, I'm talking about a 2x2 chunk.
Many coders who have long
not new, now suspiciously
grins. Yes, yes, at first
view the benefits of such a size
chanok a little as the screen grows as much a factor of 4
compared to 4x4 (and, hence, the frame will be constructed to 4
times longer) - and in the memory of this screen to jostle it
is not so easy. And the use of tiny windows - is not out. And
even feature - a chunk of 2x2, we have 5 colors, it's awfully
small for the implementation of lighting in the 3D-parts (and
Phong, and Gourado, and even look Flat
just awful). To top
everything else and c2p brake, so
as there is more than reading and
less write (compared to
4x4, again).
But! I would not just waste your time (as well as their own,
which to me is much more expensive:)
if not sold chunks 2x2
Our Deme, which we unfortunately could not lead to
CAFe'02, and now can not in any way bring to end: (((That's
about This I tell you now ...
Most simply solve the problem
lack of 5-colors. Itself
dithering. Who does not know - in
some Deja Vu was a good
The article about it. Incidentally,
birdie. Relatively recently in
Psychoz # 11c was published
Article Dark / X-Trade of dithering
a 2x2 chunk. Despite all my
respect to the individual, the code
dithering there terribly weak. More
in 2000 to year we are on par with
Baze/3SC came up with the correct
routine c2p chanok for 2x2, and
its intro Matricks it is just
applied dithering in 16 colors
(Admittedly, c2p there it personal
more braking). C2p code like this:
sp - chunk screen
de - 1st line on spectrum screen (# 401F)
bc - 2nd line on spectrum screen (# 4120)
pop hl
ld a, (hl)
pop hl
ld l, (hl)
ld h, a
ldd
ld a, (hl)
ld (bc), a
Total: 10 +7 +10 +7 +4 +16 +7 +7 = 68tct
[Recall that two Dark'a familiarity drawn 90 94
tact, it is 92 cycles on average]
The trick you use LDD. A small feature - the picture is drawn
inverted horizontally, as SP is growing up, and BC and DE
reduced. Chunks of the form% --- xxxx0. For dither will have to
put a table of graphics in different banks and to work on one
screen. Explain I will not detail because it is all
well already passed stage:) Who
Still interesting and clear -
contact me explain.
As you can see, this speed is quite
normal - the whole screen is drawn in approximately 3 frames.
But problem is too big-screen remains so it is inadequate
solution.
Experimenting with different engines and filters, I came
head of a great idea:) And
that if we miscalculate the screen is not completely but only
partially and the remainder of the fill
program, say, a linear interpolation? Then, you can not use the
brake filter (a la Sair00s) for latent defects engine-that is,
we get benefits in both speed and
volume of the screen. There are some parallels with the Amiga
regime HAM8 (I understand), where the color of pixels depends
on the nodal and the latest demos are working to resolve not
640x480 or 320x240, and a lot less. So
I do not amizhnik (really), it can now be said to huge
stupidity - but it does not matter:) This
talking about the Spectrum.
What I mean by a linear
interpolation? I have not cheated
Each chunk screen, and in one
horizontally - and missed
get as an average between the two
already known. Type this:
so do the usual:
! C0! C1! C2! C3! C4! C5! C6! C7! C8!
and so do I:
! C0! -! C2! -! C4! -! C6! -! C8!
(C0 + c2) / 2 (c4 + c6) / 2
As a result, the picture gets
some blurring, which
personally I like:) Anyway
issue immediately chankovy screen
a little risky, because of the low resolution and simplified
shading image is often obtained, so to speak, scary:)
In the past demah Sair00s about the same interpolates two
the screen - I did a cost.
If interpolation conduct a separate operation, then any
benefits in speed, we did not find since it is not too rapid
(takes two from memory the numbers add up on the table
We get the result and puts it in
Memory - 40 cycles per chunk
go). I therefore distort and
combined with interpolation c2p:
sp - chunk screen
de - spectrum screen
h - previous chunk
pop bc
ld a, (bc)
ld l, a
ld a, (hl)
ld (de), a
inc h
inc d
ldi
dec d
ld h, b
Total: 10 +7 +4 +7 +7 +4 +4 +16 +4 +4 = 67tct
Mdaaaa:) I'll try to explain
available.
In order to draw the familiarity, (8x2 pixel), we must consider
three chunks:
! ! !
! - C0 - c1! - C2 - c3!
! ! !
^ ^ ^
We have 16 tables (as the chunk
previous familiarity may
be of any color), and the address of graphics is calculated as
follows (for the second familiarity):
c1 * # 0100 + c2 * # 0010 + c3
Chunks of the form% 110hhhh0 those # E0-# FE.
At the end of the sector is a table
junior addresses schedules, ie
from addresses # E0E0, # E2E0, etc. In the very same 32 sectors
is schedule.
Discrepancy came out - so the graphics and the multiplication
table located in the same address, their creeping occurs at each
other. I walked around it, throwing two
color:) That is, I do not have 16 real flowers, and 14. But
since more convenient to work in sixteen
colors, so I just combined 0 and 1, 14 and 15 colors (in the
table "multiplication" on display the same value). You can
actually do something and honest 1916 colors, but only by
slowing down c2p-so I chose this output.
Dithering I made with two cans with different chankovoy
graphics - for even rows and
odd. All this eats lots of memory, but still works on one
screen:)
What are we all this for you? We have
the following - in comparison with chunks 4x4 screen size has
doubled (rather than four), the screen is completely rebuilt in
about three frames, we do not need an additional filter, the
image quality much better than (Still would be a chunk fell by 4
times). That is, we lose in
speed, but we win as a. On this engine is quite
possible to do something that we
tried to prove in his Deme.
The annex to the magazine you will find the source generilki
chankovoy Graphics and c2p in text format, but unfortunately
for many, syntax Storm'a:) And also
small demonstration
engine.
respect to all spectrum's coders
and Adventurer's readers!!!
(C) Triebkraft
--------------------------------- I am, of course, with a
slightly stormozil his article (as well as with demoy
: (), And that though the thread compensate for this negative
aspect, bring to your attention an optimized version of
c2p-routine with interpolation. Wrote it JtN / / 4d, comments
will not:) The principle of operation is the same, and in
general - need to head to think more.
pop hl; 10
l, (hl); 7
ld a, h; 4
exa; 4
ld h, a; 4
ldd; 16
ld a, (hl); 7
ld (bc), a; 7
; = 59t !!!!!!!!!!!
