void Blit(LPDIRECTDRAWSURFACE7 dest, LPDIRECTDRAWSURFACE7 src, int xpos, int ypos, RECT area)
{
int nHeight = area.bottom - area.top,
nWidth = area.right - area.left;
DDSURFACEDESC2 ddsd_dest, ddsd_src;
// LOCK BACKBUFFER
ZeroMemory(&ddsd_dest, sizeof(ddsd_dest));
ddsd_dest.dwSize = sizeof(ddsd_dest);
dest->Lock(NULL, &ddsd_dest, DDLOCK_WAIT, NULL);
//LOCK MAINGFX
ZeroMemory(&ddsd_src, sizeof(ddsd_src));
ddsd_src.dwSize = sizeof(ddsd_src);
src->Lock(NULL, &ddsd_src, DDLOCK_WAIT, NULL);
BYTE *lp_dest = (BYTE *)ddsd_dest.lpSurface;
BYTE *lp_src = (BYTE *)ddsd_src.lpSurface;
//Move down to the correct line
lp_src += area.top * ddsd_src.lPitch;
//Move over to correct pixel
lp_src += area.left;
//Move down to the correct block frame
//Don''t multiply this value by 2!
lp_src += area.left * ddsd_src.lPitch;
//Move over to correct block
//(Times 2 cause we are in 16 bit color)
lp_src += area.left * 2;
//Move down to correct y co-ord
lp_dest += ypos * ddsd_dest.lPitch;
//Move over to correct x co-ord
lp_dest += xpos * 2;
for (int CurrentLine = 0; CurrentLine < nHeight; CurrentLine++)
{
memcpy(lp_dest, lp_src, nWidth);
//Move to next line..
lp_src += ddsd_src.lPitch;
//Move the destination to the next line..
lp_dest += ddsd_dest.lPitch;
}
dest->Unlock(NULL);
src->Unlock(NULL);
}
Again, I think it''s when I access the surface by using a BYTE. Using a WORD or DWORD in replacement has given me errors and such. Please help.
alternative to directx blit...help!
Author
I read this article at the directx developers page (or whatever it''s called now) and I followed the code to an alternative blitting function code for code. The problem is that the function only blits half the given width. For example, I passed a 64x64 rect to blit but only recieve a 32x64 rect. I believe it''s how I access the surface but not sure what to do.
This function will copy only half of width becouse of "BYTE".
When you are calculating width you must do *4 not *2 becouse you are in 32bpp (32bit=4 bytes, 16bit=2 bytes,...)
When you are calculating width you must do *4 not *2 becouse you are in 32bpp (32bit=4 bytes, 16bit=2 bytes,...)
Good GOD!
Who put multiplications in a time critical routine?????
NEVER use a multiplication in a time critical routine! It''ll DESTROY your speed!
Use bit shifts instead... for instance, saying
18*2
is done in a SINGLE CYCLE (sometimes less) by saying:
18<<1
And 18*4 would be 18<<2.
Holy crow, I wouldn''t trust that routine. You''re already talking about a 64 cycle speedup (maybe more) on that one notion alone!! It doesn''t seem like much, but I''ve found that those 64 cycles are better used ANYWHERE else but in a main rendering loop!
Who put multiplications in a time critical routine?????
NEVER use a multiplication in a time critical routine! It''ll DESTROY your speed!
Use bit shifts instead... for instance, saying
18*2
is done in a SINGLE CYCLE (sometimes less) by saying:
18<<1
And 18*4 would be 18<<2.
Holy crow, I wouldn''t trust that routine. You''re already talking about a 64 cycle speedup (maybe more) on that one notion alone!! It doesn''t seem like much, but I''ve found that those 64 cycles are better used ANYWHERE else but in a main rendering loop!
Of course, most compilers these days will spot a multiplication by a constant and optimise it out into a shift if applicable, maybe into a few additions if not. Of course, the principle is correct though - if you change it to a shift yourself, you''re leaving nothing to chance.
I think what might be causing the problem is
nWidth = area.right - area.left;
because when you call your memcpy like this:
memcpy(lp_dest, lp_src, nWidth);
you are copying nWidth 8-bit bytes, instead of nWidth 16-bit pixels... So I think if you change the nWidth line to
nWidth = (area.right - area.left) << 1;
you should get the full width of the src getting copied. Thats all that I could notice, of course I might be wrong cause I''ve been working in 8-bit mode a lot more than 16-bit...
Hope that helps.
Where's my Golden Fleece?
nWidth = area.right - area.left;
because when you call your memcpy like this:
memcpy(lp_dest, lp_src, nWidth);
you are copying nWidth 8-bit bytes, instead of nWidth 16-bit pixels... So I think if you change the nWidth line to
nWidth = (area.right - area.left) << 1;
you should get the full width of the src getting copied. Thats all that I could notice, of course I might be wrong cause I''ve been working in 8-bit mode a lot more than 16-bit...
Hope that helps.
Where's my Golden Fleece?
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
