need for speed!

Thu, 26 Jan 2006 14:56:01 -0700

I have a feeling this will be a dificult question.
my problem is that my pixel setting routine is too slow:

PSET is the original pixel setting routine, but of course, we all know how slow it is.
then there's GET & PUT, quite fast, but no way to manipulate the sprites well for scrolling, special effects, etc.

then there's the POKEing method, where you POKE a color value into the VRAM (SEG = A000), at an offset that represents the pixel coordinates
(POKE 320& * Y% + X%)
this is very fast, (at least two times faster than PSET.)
and even faster in compilation.
and is what I have been using so far.
but... it seems it's STILL too slow, is there a faster way of plotting pixels?
should I do something different than plotting for drawing large scrolling backgrounds?
I want to do this without libraries. but ASM in the code is ok with me.
I've been trying to figure out the assembly equivelent to this code.
it seems easy, but I don't have TASM or NASM, so DEBUG:A is what I got.
and MOV [A000]:AX, BX doesn't work in debug, (what some one suggested to me).
I'm sorry for the long post, but if anyone can help me, then I would appreciate it greatly

Re: need for speed!

Sat, 04 Mar 2006 13:27:49 -0700

try
MOV [A000+AX], BX
I would use some REP STOSB or something similar.
REP prefixed instructions are really fast.

This code sets up for the rep movsb and executes it.

MOV CX, 10 ;how many bytes to be copied
MOV AX, yourSegment ;wherever you want to move from
MOV BX, yourAddress
MOV ES, AX
MOV DI, BX
MOV AX, 0A000h
MOV BX, 0           ;wherever you want to move to
MOV DS, AX
MOV SI, BX
REP MOVSB

This is 25 byte compiled, but the REP MOVSB is only 2...
This takes CX clock cycles so it's fast as hell.

I downloaded this program some years ago.
The memcopy func is almost as mine above.
Check it out:

' Bmpload Version 2 By Doug Barry (PD Computers)
'  This is the fastest BMP Loader I have ever seen, and I wrote it !!!!
'  God it's fast, anyway it uses assembler to copy a variable to the another
'  part of the memory, here I have used it to put the image data into the
'  bit of the Physical ram that overlaps the video memory, hence loading
'  the picture in one vertical blank space (one 50Hz cycle)
'
' Thanks load to Dan Holmes for finding out the memory copying routine, and
'  Andrew Griffin/Jon Sutton for info on the BMP structure.
'  Also to the guy/gal that posted ShowBMP9.bas on the net and to the
'  guy/gal that wrote CPLASMA.BAS for the "OUT" command for palette setting.

' Enjoy, even though it's uncommented it should be easy to understand,
'  being only 65, yes count 'em 65 lines of code (WOW!!!!!!!!).

' My E-Mail is pdcomputers@iname.com
' or s3dougla@doreen.rainhammark.kent.sch.uk if this don't work.
' i will try to reply and help with any of your problems.

DECLARE SUB SetPALETTE (Slot%, R%, G%, B%)
DECLARE SUB memcopy (fromseg%, fromoffset%, toseg%, tooffset%, bytes%)
DIM SHARED Buffer(319, 199) AS STRING * 1
DIM SHARED Pointer AS STRING * 1
DIM SHARED ImageDataSegment(200) AS STRING * 320
DEFINT A-Z
FileName$ = LCASE$(LTRIM$(RTRIM$(COMMAND$)))
'FileName$ = "Test.Bmp"
OPEN FileName$ FOR BINARY AS #1
IF LOF(1) < 2 THEN PRINT "File does not exist": KILL FileName$: SYSTEM
GET #1, 54, Pointer
SCREEN 13
FOR Slot% = 0 TO 255
 GET #1, , Pointer
 B% = INT(ASC(Pointer) / 4)
 GET #1, , Pointer
 G% = INT(ASC(Pointer) / 4)
 GET #1, , Pointer
 R% = INT(ASC(Pointer) / 4)
 SetPALETTE Slot%, R%, B%, G%
 GET #1, , Pointer
NEXT Slot%
FOR Y = 199 TO 0 STEP -1
  GET #1, , ImageDataSegment(Y)
NEXT
CLOSE
memcopy VARSEG(ImageDataSegment(0)), VARPTR(ImageDataSegment(0)), &HA000, 0, &HFA00
SYSTEM

SUB memcopy (fromseg%, fromoffset%, toseg%, tooffset%, bytes%)
  asm$ = ""
  asm$ = asm$ + CHR$(85)
  asm$ = asm$ + CHR$(137) + CHR$(229)
  asm$ = asm$ + CHR$(30)
  asm$ = asm$ + CHR$(139) + CHR$(70) + CHR$(10)
  asm$ = asm$ + CHR$(142) + CHR$(192)
  asm$ = asm$ + CHR$(139) + CHR$(70) + CHR$(14)
  asm$ = asm$ + CHR$(142) + CHR$(216)
  asm$ = asm$ + CHR$(139) + CHR$(118) + CHR$(8)
  asm$ = asm$ + CHR$(139) + CHR$(126) + CHR$(12)
  asm$ = asm$ + CHR$(139) + CHR$(78) + CHR$(6)
  asm$ = asm$ + CHR$(243)
  asm$ = asm$ + CHR$(164)
  asm$ = asm$ + CHR$(31)
  asm$ = asm$ + CHR$(93)
  asm$ = asm$ + CHR$(203)
  WAIT &H3DA, 8
  DEF SEG = VARSEG(asm$)
    CALL Absolute(BYVAL fromseg%, BYVAL fromoffset%, BYVAL toseg%, BYVAL tooffset%, BYVAL bytes%, SADD(asm$))
  DEF SEG
END SUB

SUB SetPALETTE (Slot, R, G, B)
  OUT &H3C8, Slot
  OUT &H3C9, R
  OUT &H3C9, B
  OUT &H3C9, G
END SUB

I don't know what you want to do, if you give me the basic code I can convert it to (f)asm

Re: need for speed!

Tue, 25 Apr 2006 09:34:51 -0700

hi ,
have you tried to use assembly language ?
it is very fast .
contact me to have the dma-straight-to-screen-kill-ya-momma-with-an-axe
pixel routine (it is 48 microprocessor clock cycles)
hoping you know how to include assembly in qbasic.

'need for speed!' Thread RSS Feed

need for speed!

Re: need for speed!

Re: need for speed!