|
Adventurer #11
31 июля 2000 |
|
Exchange of experience - Game Making 2: The various methods for deriving sprites (based on the game Full Shit).
(C) DEMON / XPC / CPU
GAME MAKING # 2
Hail, wandered here ...
We are here again (see ADV # 10) on the shitty-din makinge>: ->
Well, let's start with the sprite vyvodilok all
plain and simple - they are:
- With per-pixel coordinates of O and znakomestovymi;
- With and without masochkoy masochki;
- With and without attributes.
Then there are combinations of these lazhi. Divided into
types by the method organization output (depends on the
structure storage infrastruc-tion in sprites) I will not,
still does not list all, not even
I remember how to store the sprites in the shitty
sprite generator of THD (goats), which was my first program of
its kind.
A few words about vyvodilkah sprites: I believe they must
meet certain conditions, for example, I did not need to
stack-fast 'Marketing, and to oherennnyh size razdekranchennye
because yuzalis both screens, then vyvodilka had to sit in the
main Memory Speck (# 5b00-# bfff), and apart from her
a lot of things were there.
I'll tell you about the three vyvodilkah; on Y
they pixel at X znakomestovye two of them and one pixel.
1. X - ZNAKOMESTOVAYA Unmasked
Everything is standard, and without any
fucking with a stack of bloat and other oddities, the simpler,
the less # proximity. But if you have the desire to fuck
interrupt and trace stack,
flag of your hands ...
So in 1932 LDI - it rulez, just 16
cycles per byte (and to my even-tact
Scorpion, too).
Set the coordinates in the DE, the size of
sprite sprite BC himself in HL, we do
CALL ssp, and he is on screen (if there are no
correctly, the screen is likely to
will be shit). Protsedurka this is nothing
no checks and returns nothing
as it is the lowest level of the nucleus bend. Do you want to -
or even redone better to do his will be proud of ...
WARNING! In the present listings
profanity - the Directive
XAS 'a.
; (C) DEMON / XPC
; Sprite x_Symbol Print without mask
; Average_Speed = 18.5 t / b (Scorp)
; Prog_Len = 120 b
;-------------; In: HL-sprite
; DE-coord (X-sim, Y-pix)
; BC-size (Y-pix, X-sym)
;------------- Ssp PUSH IX
LD IX, wo_r_ok +4; forward and
LD A, C; displace the loop
ADD A, A; by LDI
SUB 64; through JP (IX)
NEG
LD C, A
LD A, B
LD B, 0
ADD IX, BC
LD B, A
PUSH HL
CALL adres_y; calculation of addresses
EX DE, HL
POP HL
JR wo_r_ok; enter a loop ...
wo_rinc 28 +
LD A, E; 4
ADD A, 32, 8 (7)
LD E, A; 4
JR C, wo_r_ok; 8 (7) / 12
LD A, D; 4
SUB 8 8 (7)
LD D, A; 4
JP wo_r_ok; 10
wo_ylop INC D; 4
LD A, D; 4
AND 7, 8 (7)
JR Z, wo_rinc; 8 (7)
wo_r_ok PUSH DE; 10
LD C, D; Schaub was 4 C> 32
JP (IX); 8
! ASSM 32, 32-th instruction LDI
LDI; 16
! CONT
POP DE; 12 (11)
DJNZ wo_ylop; 14 (13)
POP IX
RET
;------------- Calculation of addresses in the screen
; In: DE-X (sym), Y (pix)
; Out: HL-adres in screen
;------------ Adres_y LD L, D
LD A, E
AND 7
LD H, A
LD A, E
AND # 38
RLCA
RLCA
OR L
LD L, A
LD A, E
AND # C0
RRA
RRA
RRA
OR H
LD H, A
LD A, (how_scr)
OR H
LD H, A
RET
;------------- VAR
how_scr DB # 40
Contacts listed in critical areas
the first figures are valid for the Scorpion, in
brackets - for everything else. By the way,
Scorpion on my tests show their 74,884 t in
Interrupt chetno-t-vym and from 71,141 to
71440 on odd-t-vym teams.
Partly how_scr is crucial in order to screen any currently
running program (not visible, and working!), # 40 for 5 and #
C0 for the 7 th, for example:
draw a 5-second screen (how_scr = # 40), and
see the 7 th, he suddenly interrupted: we define
the variable pag_or (see below) that
including a 7-th screen, change to how_scr # C0
and arrow to display a seventh, as All GFXprotsy, work through
that variable, then and addresses of all will be calculated for
the appropriate screen.
You can blather that it is stupid and should be
would simply be cut in the 5 th with # C000 (Sorry,
Elf), but when working with two screens, I
it came to this method.
2. X - ZNAKOMESTOVAYA WITH MASK
; (C) DEMON / XPC
; Sprite x_Symbol Printer with Mask
; Average_Speed = 50 t / b (Scorp)
; Len = 287 b
;------------; In: HL-sprite with mask
; DE-coord X (sym), Y (pix)
; BC-size Y (pix), X = (sym)
;------------ Ssp_m PUSH IX; truncated same
LD IX, wm_r_ok +3
LD A, C
ADD A, A
ADD A, A
ADD A, C
ADD A, C
ADD A, C
SUB 224
NEG
LD C, A
LD A, B
LD B, 0
ADD IX, BC
LD B, A
PUSH HL
CALL adres_y
LD D, B
LD B, H
LD C, L
POP HL
JR wm_r_ok; cycle again ...
wm_rinc; 28 +
LD A, C; 4
ADD A, 32, 8 (7)
LD C, A; 4
JR C, wm_r_ok; 4 (7) / 12
LD A, B; 4
SUB 8 8 (7)
LD B, A; 4
JP wm_r_ok; 10
wm_ylop INC B; 4
LD A, B; 4
AND 7, 8 (7)
JR Z, wm_rinc; 8 (7)
wm_r_ok LD E, C; 4
JP (IX); 8
! ASSM 1932
LD A, (BC); 8 (7) to a byte screen
AND (HL); 8 (7) impose a mask
INC HL; 6 and byte sprite
OR (HL); 8 (7) Assume the result
INC HL; 6
LD (BC), A; 8 (7) put in a screen
INC C; 4
! CONT
LD C, E 4
DEC D 4
JP NZ, wm_ylop; 14
POP IX
RET
Sprite dolzhon have the following
structure: the byte-mask, byte-date baytmaska, etc., as does
the Sprite Land of DR.
About two percent of these acts
the following rule:,
"BETTER wider and shorter than the already
Length ", ie the size of the sprite X = 32 and
Y = 10 draws faster than the size of
X = 10 Y = 32, although the size in bytes of the same.
3. Per-pixel WITH MASK
Made this procedure is mainly Elf,
XN0ByS something done too. When Kostya
(Elf) said that this is supposedly faster, then I have to
dispute it, darling, zafastil. Winning over the old in averaged
5000-7000 cycles, but if we take the sprite large sizes
(subject ate about 45000 cycles), then it (winning) is
proportional increases. And all through the use of undocumented
commands; think you can still speed, but broke ...
; FAST PIXEL'S SPRITE WITH MASKA ...
; By ELF / AURYN 'XN0ByS/SG
; ... FASTEST
; By DEMON / XPC '99
; IN: HL ADR_OF SPRITE
; D Coord X in pIXels
; E Coord Y in pIXels
; C-x_size symbol
; B-y_size pIXel
spr_pxm
PUSH HL
CALL adres_p
EX DE, HL
ADD A, A
PUSH BC
LD C, A
LD B, 0
LD HL, rol_tab
ADD HL, BC
LD A, (HL)
INC HL
LD H, (HL)
LD L, A
LD (pxspr +1), HL
POP IY
POP HL
pxspr JP bit_0
; In IY-Y 'X size
; HL-SPRITE WITH MASK
; DE-ADRES ON SCREEN
bit_0
LD C, LY
bit_0l1
LD B, C
LD LX, E
bit_0l2
LD A, (DE)
AND (HL)
INC HL
OR (HL)
LD (DE), A
INC HL
INC E
DJNZ bit_0l2
LD E, LX
CALL down_DE
DEC HY
JR NZ, bit_0l1
RET
bit_1
EX DE, HL
LD A, HY
LD HX, A
bit1_l1
LD HY, LY
LD A, L
EX AF, AF '
bit1_l2
LD BC, # ff
LD A, (DE)
SCF
RRA
RR C
AND (HL)
LD (HL), A
INC DE
LD A, (DE)
RRA
RR B
OR (HL)
LD (HL), A
INC L
INC DE
LD A, C
AND (HL)
OR B
LD (HL), A
DEC HY
JR NZ, bit1_l2
EX AF, AF '
LD L, A
CALL down_HL
DEC HX
JR NZ, bit1_l1
RET
bit_2
EX DE, HL
LD A, HY
LD HX, A
bit2_l1
LD HY, LY
LD A, L
EX AF, AF '
bit2_l2
LD BC, # ff
LD A, (DE)
SCF
! ASSM 2
RRA
RR c
! CONT
AND (HL)
LD (HL), A
INC DE
LD A, (DE)
! ASSM 2
RRA
RR B
! CONT
OR (HL)
LD (HL), A
INC L
INC DE
LD a, C
AND (HL)
OR B
LD (HL), A
DEC HY
JR NZ, bit2_l2
EX AF, AF '
LD L, A
CALL down_HL
DEC HX
JR NZ, bit2_l1
RET
bit_3
EX DE, HL
LD A, HY
LD HX, A
bit3_l1
LD HY, LY
LD A, L
EX AF, AF '
bit3_l2
LD BC, # ff
LD A, (DE)
SCF
! ASSM 3
RRA
RR C
! CONT
AND (HL)
LD (HL), A
INC DE
LD A, (DE)
! ASSM 3
RRA
RR B
! CONT
OR (HL)
LD (HL), A
INC L
INC DE
LD A, C
AND (HL)
OR B
LD (HL), A
DEC HY
JR NZ, bit3_l2
EX AF, AF '
LD L, A
CALL down_HL
DEC HX
JR NZ, bit3_l1
RET
bit_4
EX DE, HL
LD A, HY
LD HX, A
bit4_l1
LD HY, LY
LD A, L
EX AF, AF '
bit4_l2
LD BC, # ff
LD A, (DE)
SCF
! ASSM 4
RRA
RR C
! CONT
AND (HL)
LD (HL), A
INC DE
LD A, (DE)
! ASSM 4
RRA
RR B
! CONT
OR (HL)
LD (HL), A
INC L
INC DE
LD A, C
AND (HL)
OR B
LD (HL), A
DEC HY
JR NZ, bit4_l2
EX AF, AF '
LD L, A
CALL down_HL
DEC HX
JR NZ, bit4_l1
RET
bit_5
EX DE, HL
LD A, HY
LD HX, A
bit5_l1
LD HY, LY
LD A, L
EX AF, AF '
bit5_l2
LD A, (DE)
LD C, A
LD A, # ff
! ASSM 3
SLI C
RLA
! CONT
AND (HL)
LD (HL), A
INC DE
LD A, (DE)
LD B, A
XOR A
! ASSM 3
RL B
RLA
! CONT
OR (HL)
LD (HL), A
INC L
INC DE
LD A, C
AND (HL)
OR B
LD (HL), A
DEC HY
JR NZ, bit5_l2
EX AF, AF '
LD L, A
CALL down_HL
DEC HX
JR NZ, bit5_l1
RET
bit_6
EX DE, HL
LD A, HY
LD HX, A
bit6_l1
LD HY, LY
LD A, L
EX AF, AF '
bit6_l2
LD A, (DE)
LD C, A
LD A, # ff
! ASSM 2
SLI C
RLA
! CONT
AND (HL)
LD (HL), A
INC DE
LD A, (DE)
LD B, A
XOR A
! ASSM 2
RL B
RLA
! CONT
OR (HL)
LD (HL), A
INC L
INC DE
LD A, C
AND (HL)
OR B
LD (HL), A
DEC HY
JR NZ, bit6_l2
EX AF, AF '
LD L, A
CALL down_HL
DEC HX
JR NZ, bit6_l1
RET
bit_7
EX DE, HL
LD A, HY
LD HX, A
bit7_l1
LD HY, LY
LD A, L
EX AF, AF '
bit7_l2
LD A, (DE)
LD C, A
LD A, # ff
SLI C
RLA
AND (HL)
LD (HL), A
INC DE
LD A, (DE)
LD B, A
XOR A
RL B
RLA
OR (HL)
LD (HL), A
INC L
INC DE
LD A, C
AND (HL)
OR B
LD (HL), A
DEC HY
JR NZ, bit7_l2
EX AF, AF '
LD L, A
CALL down_HL
DEC HX
JR NZ, bit7_l1
RET
down_HL
INC H
LD A, H
AND 7
RET NZ
LD A, L
ADD A, 32
LD L, A
RET C
LD a, H
SUB 8
LD H, A
RET
down_DE
INC D
LD A, D
AND 7
RET NZ
LD A, E
ADD A, 32
LD E, A
RET C
LD A, D
SUB 8
LD D, A
RET
adres_p
LD A, E
AND A
RRA
SCF
RRA
AND A
RRA
XOR E
AND # f8
XOR E
LD H, A
LD A, D
RLCA
RLCA
RLCA
XOR E
AND # c7
XOR E
RLCA
RLCA
LD L, A
LD A, D
AND # 07
RET
rol_tab
DW bit_0, bit_1, bit_2, bit_3
DW bit_4, bit_5, bit_6, bit_7
The meaning of the following: computed
how many bytes have rollirovat
sprite with a mask, the table is taken
corresponding procedure, and the process cycle
begins. Sprite has the same structure as described above.
Perhaps in This procedure can be four-day clean up, if you
wish, although it works without problems and for quite some
time we yuzaetsya in this form.
All these vyvodilki work with monochrome sprites, because
output attributes up to a pixel in the Y and a
very problematic>: ->
Before proceeding to the next part of his narrative, I
want to tell about how I swap the memory and the screen.
;------------------- Page lister
; In: A-page logic nomber (0 ... 7)
; IF A = # FF THEN GOTO old_pag
pager EX AF, AF '
LD A, R
JP PE, pg_n_im
LD A, R
pg_n_im PUSH AF
DI
EX AF, AF '
AND A
JP M, old_pag; A = 0 ... 7
pg_f_op LD (sav_pag), A; for OLD_PAG
now_pag EQU $ +1
LD A, 7
LD (old_pag +1), A
sav_pag EQU $ +1
LD A, 0
LD (now_pag), a
pg_f_sw AND 7; for SCR_SWP
pag_or EQU $ +1
OR 0
OUT (# fd), A
LD (23388), A
POP AF
DI
RET PO
EI
RET
old_pag LD A, 0
JR pg_f_op
;---------------------- Screen Swaper
; In: A-0 is # 4000, other is # C000 viewing
scr_swp EX AF, AF '
LD A, R
JP PE, sw_n_im
LD A, R
sw_n_im PUSH AF
DI
EX AF, AF
OR A
LD A, (pag_or)
JR NZ, sw_is_7; IF A = 0 THEN scr # 5
RES 3, A
sw_f_7 LD (pag_or), A
LD A, (now_pag)
JR pg_f_sw
sw_is_7 SET 3, A
JR sw_f_7
;----------------- # FD mask detecter
; Out: A-Mask for pager
fd_mask LD A, # 10
LD BC, # 7ffd
OUT (C), A
LD HL, # ffff
LD E, (HL)
LD A, # 50
LD (HL), A
OUT (C), A
CP (HL)
LD A, # 10
OUT (C), A
LD (HL), E
RET NZ
LD A, # 50
RET
All of the above procedure shall constitute one package
for working with 128 KB.
- Proc fd_mask runs one
once at the beginning to determine the 512ti kilogram
component, if not, click memory enabled the 6 th bit
(For the Scorpion).
- Proc pager swap to 128 memory
remembers the current page now_pag,
the previous one in old_pag and all through the swap
pag_or. Marazm the register R and the definition of the state
of the interrupt is taken from Article Ivan Roshchina, what a
huge thanks to him. So makarom I never glitches were to
determine the current page. And why did I click on # FD, rather
than # 7FFD? Yes, that's like so.
- Scr_swp works through the pager,
so when I click the memory is not a headache about what to put
the mask to the screen.
Well now about the INT (interrupt). Ruleznaya very thing
that these same interrupt can be hung on them that you want
though the whole engine. But it's better to hang things such as
scanning buttons and mouse cursor movement (if present), some
chips, such as a mini-scroll with HELP, a small animation
somewhere in a corner of the screen and, of course, music and
effects at the end.
When working with the INT may
a lot of trouble, all is not even
predict, will talk about svoih.V ADV # 10
was given my interrupt handler,
similar to sitting in the "Full Shit", in which:
- A small percent of the switching
visible screen, yuzaet global variables NOW_PAG, PAGE_OR;
- Retained the current page number memory and defines the
visible screen, is preserved, and then changing
variable how_scr;
- Cut one page # 7, because it
all sitting percent. Keep in mind that
INT-table contains 257 bytes and is located in the 2 nd bank to
somehow not offend the original Spec. A possible
question: "Why is the 7 th bank?" The answer is: yuzayutsya
both screens, seats in # 5B00-# BFFF little, and in the seventh
minus screen and screen location is as much as 3 kilos.
- Restores the screen under the cursor, but if you
previously worked switching the visible screen, then the old
crap is not restored;
- Claudia scanned for pressing the buttons "6,7,8,9,0,
Break, Q, A, O, P, Space" and mouse movements (on the PC -
Emulsion some golyun with our driver: HRC
MUST DIE);
- There is recourse to procedure under the name
INT_GFX. This is different chips, consists mainly of piles of
vectors, which vary depending on the action Gamers, for
example: there menyuha, where the object is spinning (it should
cut off the bulb on the lamp, and then hang a rotating object,
and if you call a pocket the 4-th item; cartoons also played
by INT) or dialog box, all on INT
by x ..;
- Memorized piece at a new location
if there was a movement (aka the old)
the cursor is drawn, after all the bells and whistles,
so as not to fall under any sprite
or text;
- In the last turn is played for fun and Mouzon change the
attributes on traffic lights;
- Restoring the memory page
variable how_scr and back.
The first glitch came when I decided
All zaoptimizit: remove my own pct graphics output from the
interrupt, so everything was done through a single procedure.
And then began ... Began to appear on the screen
shit, without any regularity, sometimes even all of the Vistula
or cleared. And when everything works without Trace problems.
Such bugs hate, because cause once-and not clear. And all the
garbage was due to the fact that in the old version vyvodilki
saved local variables (The size of a sprite) and in contact
with INT they Killy, because INT also yuzal these
Proc. CONCLUSION: If the percent yuzayut all,
during its operation, all variables must be stored in registers
or pushing them into the stack, otherwise the interrupt will
have to keep all variables.
The second glitch came with switching
screens, but this is what I, in my opinion, has spoke ...
In general, if all goes well the prog
check, the bugs usually occurs
small and does not advise to hang on IM 1, although
as you like, even though the IM 0, a>: ->
Now the structure of "Full Shit",
which leaves much to be desired, but that
is, on the porridge and boiled ...
Can: on / off [T], all process; lock [C]
WWWWW INT_MODE 2
BEGIN <
> 1.SCR_SWP
pass 2.KEY_SCN
GO <[T] <3.CR_PROC
^
pass <4.INT_GFX
MENU <[T] <5.PLAY_MZ
pass n
TIME <[C] <o
CHK_CNT> [time?]
save all
and use
^
CLC_WAY CHK_MAP y
e
V is not no s
[Way?]> <[Thing?]
V
V
MAK_WAY RETURN WIN_PUT <
TO
"BEGIN"
use "GO"
GO_DRAW ACTIONS
^ ^
<
> [Doing]
V V
V V V
V
GFX_ENGINE DEATH HAPPY
CUT CUT
DRAW LOCATION:
1. BACKGROUND
2. HERO
3. UPGROUND
4. TURN ON "SCR_SWP"
use many primitive V
gfx_procs
LOW LEVEL PROCS <<
V V V
VIDEO RAM DISK
LOC_MAP
32 * 20 = 640 bytes
use it:
%%%%%% CAL_WAY - read
# # # O ##### O CHK_MAP - read
# # # About ###### ACTIONS - read 'write
# ########
POYASNILOVA
1) Legend:
[T] - threshold (# 0 or # C9);
[C] - threshold-counter (skips Th
denote n-th number of calls;
[Time?] - A condition (eg: time);
2) Podprogi:
GO: - per cent, is responsible for mixing
generalization of the hero;
CLC_WAY - miscalculation of the way, if not,
no movement;
MAK_WAY - otherwise makes a special pack
ravlyayuschy array;
GO_DRAW - Work with sprites mixing
absorption of Control. array.
MENU: - actions with objects;
CHK_MAP - checks the status of the cursor on
subject or not, "
if not, it ignores yuze
pa
WIN_PUT - otherwise it prompted action
and awaits action gamer;
ACTIONS - makes the selected action
conditions accompanying this business
displays a message and anima
tion, and according to them makes
go to acc. percent if
Bo returns to the main
cycle.
TIME: - constantly redraws u
rovoy screen;
CHK_CNT - periodically adds time
nuyu animation (the bus went out
Ball, a hero ...)
INT_MODE2 - controls all the processes
using the "input threshold" in
percent (NOP or RET), SWAPO
scans, paints and sings ...
GFX_ENGINE - always involved in the processing
Scoop the screen or output
anything on the screen.
LOW LEVEL - utilities, primitives, rabotayuRROCS ing it with
the resources of machine Ny.
LOC_MAP - an array of size 32 * 20
elements are the maps
location. Mainly used
characterizes the calculation of ways and
work with objects.
Now we will calculate the path of the procedure to the map
locations for the same asshole that you saw in the "Full
Shit ".
The algorithm is taken from Glory Mednonogova
(Which he, too, where something took, etc.
etc.) from his article in one of the ZX-FORMAT
'S. To start better read this
article, and then to poke your nose here, because I
question, then give the assembler option
this method ...
I must say that it is up to me
turned out to be somewhat problematic, because
My siskin (well, a character) has turned out not an element of
1x1, 1x4 and pedryukom. Therefore, I zae just ... Xia grind
prog for this asshole, that he somehow ok
walked on the location, the more so because of limitations on
memory, he could not walk on diagonal. In general, the process
of writing this piece of the engine stalled at all
the entire process.
Okay, here's Generally it-thing:
; (C) DEMON / XPC'99
;---------------------------; Wawe Algoritm for Search Way
; Main idea V. Mednonogov
; Realized by ME, of course ...
s_cod EQU # 50; Start COD
gc_mint EQU s_cod +45; MAX Iteration
;----------------
go_wawe RET; Turn
LD a, # c9
LD (go_wawe), a
;-------------------- Make Work DIM 34 * 22 = 748 b
; Okantovochka done, so once again do not check out
; Abroad array
LD b, 20
LD hl, # 6200; MAP_LOC
LD de, gc_wdim +35
gd_lop LD c, H
! ASSM 1932
LDI
! CONT
INC de
INC de
DJNZ gd_lop
;-------------------- Set Fin 'Start inside GO_DIM
; The copied code stake START and GOAL
gc_ctar LD bc, 0
LD a, C
SUB 4
LD c, A
CALL gc_ccrd
LD (hl), 128; Target Cod
gc_csrt LD bc, 0
LD a, C
SUB 4
LD c, A
CALL gc_ccrd
LD (hl), s_cod; Start Cod
INC hl
LD (hl), s_cod
INC hl
LD (hl), s_cod
INC hl
LD (hl), s_cod
LD lx, s_cod
;-------------------- MAIN ()
gc_main CALL gc_sway
JR nz, gc_found
DEC a
LD (go_way), a; # FF
RET
;----------------- Make way
; If found goal ...
gc_found EXX
LD hl, gcb_way-2
EXX
LD c, LX
INC lx
INC lx
LD de, 34
gc_b_lp DEC c
CALL gc_lb
JR z, gcb_fin
CALL gc_rb
JR z, gcb_fin
CALL gc_db
JR z, gcb_fin
CALL gc_ub
JR z, gcb_fin
INC c
LD a, LX
CP c
RET c
JR gc_b_lp +1
;----------------- Found iteration (-1)
gcb_fin LD a, B
EXX
LD (hl), A
DEC hl
EXX
LD (hl), # 81
JR gc_b_lp
;----------------- Make normal way_Map
gc_reway POP af
LD a, B
EXX
LD (hl), A
LD de, go_way
gcrw_lp LD b, (HL)
LD a, B
CP # ff
JR z, gcrw_end
CALL gcw_sub
LD (de), A
INC hl
INC de
JR gcrw_lp
gcrw_end LD (de), A
JP rewayer
;----------------- Invert way_step
gcw_sub LD a, "L"
CP b
LD a, "R"
RET z
CP b
LD a, "L"
RET z
LD a, "U"
CP b
LD a, "D"
RET z
LD a, "U"
RET
;----------------- Back_search
gc_lb LD b, "L"
DEC hl
LD a, (HL)
CP s_cod
JR z, gc_reway
CP c
RET z
INC hl
RET
;---------------- Gc_rb LD b, "R"
INC hl
LD a, (HL)
CP s_cod
JR z, gc_reway
CP c
RET z
DEC hl
RET
;---------------- Gc_ub LD b, "U"
AND a
SBC hl, DE
LD a, (HL)
CP s_cod
JR z, gc_reway
CP c
RET z
ADD hl, DE
RET
;---------------- Gc_db LD b, "D"
ADD hl, DE
LD a, (HL)
CP s_cod
JR z, gc_reway
CP c
RET z
AND a
SBC hl, DE
RET
;-------------------- Is Way being?
; If no way then A = 0
gc_sway LD a, LX
LD hl, gc_wdim +35
LD bc, 679
gc_cpir CPIR
JR z, gc_l1
INC lx
LD a, gc_mint
CP lx
JR nz, gc_sway
XOR a
RET
;----------------- I_Cod founded
gc_l1 PUSH hl, BC
DEC hl
LD (element), hl
CALL gc_left
CALL gc_right
CALL gc_down
CALL gc_up
POP bc, HL
LD a, LX
JR gc_cpir
;-------------------- STEP LEFT
gc_left
element EQU $ +1
LD hl, 0
DEC hl
LD a, (HL)
OR a
RET p
CP 128
JR z, glo_end
LD a, LX
INC a
LD (hl), A
RET
;-------------------- STEP RIGHT
gc_right LD hl, (element)
INC hl
gc_v_jr LD a, (HL)
OR a
RET p
CP 128
JR z, glo_end
LD a, LX
INC a
LD (hl), A
RET
;-------------------- STEP DOWN
gc_down LD hl, (element)
LD bc, 1934
ADD hl, BC
gcu_jr PUSH hl
CALL gc_c_4b
POP hl
JP m, gc_c_ok
RET
;-------------------- STEP UP
gc_up LD hl, (element)
LD bc, 1934
AND a
SBC hl, BC
JR gcu_jr
;-------------------- UP or DOWN way is OK!
gc_c_ok
LD a, 128
EX af, AF '
LD a, LX
INC a
! ASSM 4
EX af, AF
CP (hl)
JR z, glo_end
EX af, AF '
LD (hl), A
INC hl
! CONT
RET
;-------------------- Check 4 bytes
gc_c_4b LD a, (HL)
INC hl
AND (hl)
INC hl
AND (hl)
INC hl
AND (hl)
RET
;-------------------- It's START!!!
glo_end POP de, DE, DE
OR a
RET
;----------------- In BC-R (X, Y)> Out HL-Adr of ...
gc_ccrd INC c
INC b
LD a, B
LD b, C
LD hl, gc_wdim
gc_mlp LD de, 34
ADD hl, DE
DJNZ gc_mlp
ADD a, L
LD l, A
LD a, 0
ADC a, H
LD h, A
RET
! ASSM! Off
;-------------------- In HL-IL in DIM> Out DE-X, Y
gc_c_el LD de, 0
LD bc, 1934
AND a
gc_dlp SBC hl, BC
JR c, dv_end
JR z, dv_end +1
INC e
JR gc_dlp
dv_end ADD hl, BC
LD d, L
DEC e
DEC d
RET
! CONT
;----------------- Very Fatly array
gc_wdim DS 748, # 3030
go_way DS 1950, # ffff
gcb_way
Initially, map location is copied to
buffer equal to the size of her own so as not to
crap her darling its iterations.
As can be seen not armed glance
Massively yuzaetsya team CPIR, so I
I think soon it turns out, though I did not
then optimized, so the time to utrahal
itself percent. If you're still bothered
read that article, of which I mentioned earlier, then you, my
friend, everything will be clear ... Team CPIR looking iteration
check her for lice, then look for
another and so on until the end of the array. I think
everything is clear.
When the entire array of processed (blocked
crap), we begin the countdown until
not arrive at the goal.
Done it all pretty mudovo and
even evident in the game, although INT, of course,
no stopper>: ->
The maximum time for the calculation of the path
goes away when the path does not exist. Then an array of
elements completely 32h20 zamuzhivaetsya and I have it takes as
much 13 interruptions, I mean 900,000 strokes!
Mudovo ...
Since the items on location, I
are impassable, then when you specify
arrow on them is chosen in advance
score coordinates allowable space
which can be figured out a way for my
asshole. Ie If you choose, for example,
Bouncer, the program determines that you
chosen subject, and take from the table
coordinates for him (or rather for the hero,
To approach the subject) and their podsunet
rhyme path rather than real coordinates
cursor, as it would otherwise.
And more ... Percent of walking the hero I was done so
that it works through the command (of course only to her, of
couse). A special buffer is passed a string of commands to be
executed, such as: "L, 1910, U, 2, R, 3, ff" - this is oznachet
what to do 10 steps to the left, then 2 up and then 3 steps to
the right, # FF - end.Vse simple, but very comfortable now
tinkering with the engine under the colored sprites, I just
change my values in the Table size of sprites and everything!
So do not be surprised values of type "L" or "D"
In the above code.
Also made by the other procedure, so
I beg to wind on the condition ...
*
So, People, took a lot of
time after writing the first part of
article, so you can take This is all complete crap.
"What does not suck?" - You ask
me, my dear friend. All that I or
you've done is suck, because Is that what you
Already did! And that's something we have not yet
can - in essence, is for us an end in itself.
There have recently learned that sort of thing - a mask
for a sprite can be stored interlaced, ie, two times less! Such
a simple and cool idea, but by itself nor hell no come, oh, if
earlier know ... For those who do not understand: the masks
containing only even or, conversely,
odd lines. In the derivation of the shortage is compensated by
overlapping lines. All simple - but the sprite is already in
1,5 times less space (eg, $ 16
pounds have cost as much at 6 pounds!). I think this feature
has long been known geymmeykeram, but I am personally grateful
for VOLGASOFT this idea. You can with this garbage
ekskrementirovat somehow, can anyone generally get some rulez.
SchA ANT and Vaniac redrawn by
zero graphics for FULL SHIT, and because it
all the color, then respectively, I encountered some problems
...
Sprites are drawn by special-
(Fig. 1):
Cprayt draws what you want by the Incas and paperami, but
the border is not dolzhon use of colors, this allows a square
bezel attribute graph turn into a completely acceptable now.
Figure 2 shows the essence of the method:
All the crap (from the word "crap") begin when all this
Tuyev Khuchua pieces should display as a single whole. Here, in
principle, there is nothing particularly complicated, but I did
this:
1. All the "pieces" are
individual sprites, color output as
sprite attributes (PUT), and edging
as simple (LINES) on OR-method (although
and can mask, but the bars to eat more
will be, but the effect is not much different).
2. Accordingly, we write (or take)
files of sprites to display this
format (store and show how can
anything, I have this method: 8 lines
data, and then for them to line attributes
Peninsula 8 lines ...) and make the manager ...
3. "Pieces" of a sprite is stored
I have the same array:
+0 X, Y (the first one is always on zero)
+2 X_SIZE, Y_SIZE
4 The shift to the next MTEF. (NN)
6 Type sprite (ATTR, LINES ... = 0,1 ,...)
7 Sam sprite
.. ............
.. ............
+ NN X, Y with respect to the very first AB.
NN +1 X_SIZE, Y_SIZE ...
... and so on ...
Now write the manager who
will all of these arrays to rake and decide what to paint next
vyvodilkoy "piece" and where to draw it (the translation
relative to the actual cords).
Clearly, all this will work
not as fast as we would like, but first, it is normal, and
secondly - optimization knows no bounds! One can already
ready-made arrays of the envelope in a different format in a
way that does not count for each "piece" cords. I am personally
in the scrap, but if tozit, of course, I'll do it ... And so,
you can convertible ready arrays of "pieces" so as not to
recalculate every time cords for them, okay?
Well, to hell, got to write,
Bye!
Other articles:
Similar articles:
В этот день... 29 April