Buffer scaling Sub

[SotSvart]

I'v made a sub for copying a 320x240-16bit buffer to a 640x480-16bit screen. I want to use the 320x240x16 screen mode, but the mode is not supported any all computers, so I need a backup plan

I'v tried to optimize it, but its to slow. Can somebody help me speed it up?

This code is made for the sdl lib, but I havent changed it to work with gfxlib(wich I'm gonna use)

Code:

'buffer and video are Dim'ed as SDL_Surface ptr'rs

sub xCopyBuffer
   dim srcOfset as unsigned integer,  srcPixel as ushort ptr
   dim dstOfset1 as unsigned integer, dstOfset2 as unsigned integer
   dim dstPixel as unsigned integer ptr
   dim x as integer, y as integer, doublepixel as unsigned integer
  
   srcOfset = 0
   dstOfset1 = 0
   dstOfset2 = video->pitch
  
   SDL_LockSurface(video)
   for y = 0 to 239
      for x = 0 to 639 step 2
         srcPixel = buffer->pixels + SrcOfset + x
        
         doublepixel = *srcPixel
         doublepixel = doublepixel shl 16
         doublepixel = doublepixel or *srcPixel
        
         dstPixel = video->pixels + DstOfset1 + (x*2)
         *dstPixel = doublepixel
         dstPixel = video->pixels + DstOfset2 + (x*2)
         *dstPixel = doublepixel
      next
      
      srcOfset = srcOfset + buffer->pitch
      dstOfset1 = dstOfset1 + (video->pitch*2)
      dstOfset2 = dstOfset2 + (video->pitch*2)
   next
  
   SDL_UnlockSurface(video)
   SDL_Updaterect Video,0,0,0,0
end sub

[VonGodric]

Search for Rel's scale2x algo he posted a while ago. I think you could also check his website

[SotSvart]

Thanx, I'll look into it to night

Heh 2min and someone helps out

[SotSvart]

scale2x is not what Im looking for, I want the blocky scaling not rounding the edges. I want the code to make the 320x240 buffer look the same in a 640x480 mode as it would look in a 320x240 mode, but my algo is to slow..

[Antoni Gual]

Rel's demo scaled the bitmap twice, with the "blocky" method and with the "smart method"

[v3cz0r]

Are you reading pixels from the video surface? Try allocating a software surface, because a hardware one will be a magnitude slower, as reading for VRAM is painful slow.

[SotSvart]

I have ported my xCopyBuffer from sdl to gfxlib, but I have come across a problem. Heres a working example:

Code:

declare sub xCopyBuffer()

TYPE SCREENINFOTYPE
    driver_name AS STRING
    w AS INTEGER
    h AS INTEGER
    depth AS INTEGER
    pitch AS INTEGER
    bpp AS INTEGER
    mask_color AS INTEGER
    num_pages AS INTEGER
    flags AS INTEGER
END TYPE

dim shared buffer(76801) as ushort, info as screeninfotype ptr

buffer(0) = 320 shl 3
buffer(1) = 240

screen 18, 16, 0
info = screeninfo
PRINT STR$(info->w) + "x" + STR$(info->h) + "x" + STR$(info->depth);
PRINT " using " + info->driver_name + " driver"
PRINT "pitch:"; STR$(info->pitch)
do:loop until inkey$ <> ""

for i = 2 to 76801
   buffer(i) = rnd*65000
next

'line buffer, (0,0) - (319,239), RGB(255,255,255)
xCopyBuffer
do:loop until inkey$ <> ""

end


sub xCopyBuffer
   dim srcOfset as unsigned integer,  srcPixel as ushort ptr
   dim dstOfset1 as unsigned integer, dstOfset2 as unsigned integer
   dim dstPixel as unsigned integer ptr
   dim x as integer, y as integer, doublepixel as unsigned integer
  
   srcOfset = 0
   dstOfset1 = 0
   dstOfset2 = info->pitch
  
   screenlock
   for y = 0 to 239
      for x = 0 to 639 step 2
         srcPixel = varptr(buffer(2)) + SrcOfset + x
        
         doublepixel = *srcPixel
         doublepixel = doublepixel shl 16
         doublepixel = doublepixel or *srcPixel
        
         dstPixel = screenptr + DstOfset1 + (x*2)
         *dstPixel = doublepixel
         dstPixel = screenptr + DstOfset2 + (x*2)
         *dstPixel = doublepixel
      next
      srcOfset = srcOfset + 640
      dstOfset1 = dstOfset1 + (info->pitch*2)
      dstOfset2 = dstOfset2 + (info->pitch*2)
   next
   screenunlock
end sub

If you try to add the Line command that I have commented out, it crashes.. and I cant understand why. Could someone please help, coz this has troubled me for a couple of days now :barf: