My scaler...

[Tinnus]

Know what, I've just found the scaler I wrote can scale an image to an arbitrary dimension and not just multiples... Which means it can be used to scale NES, SNES and such as well.

Here's the code:

void BlitScaled2(UInt16* src, UInt16* dst, UInt16 srcwidth, UInt16 srcheight, UInt16 dstwidth, UInt16 dstheight, UInt16 xoffset, UInt16 yoffset, float factor)
{
    if ((srcheight * factor + yoffset) > dstheight) return;
    if ((srcwidth * factor + xoffset) > dstwidth) return;
   
    UInt16 i, j, yofs;
    yofs = yoffset * dstwidth + xoffset;
   
    UInt16 ifactor, dstyofs;
    const u32 fach = (UInt16)(factor * srcheight);
    const u32 facw = (UInt16)(factor * srcwidth);

    for (i = 0; i < fach; i++)
    {
        ifactor = ((UInt16)((float)i/factor)) * srcwidth;
        dstyofs = (UInt16)(dst + yofs);
       
        for (j = 0; j < facw; j++)
        {
            *((UInt16*)dstyofs + j) = *(src + ifactor + (UInt16 )((float)j/factor));
        }
       
        yofs += dstwidth;
    }
}

...or to scale an image with an arbitrary separate factor for x and y:

void BlitScaled3(UInt16* src, UInt16* dst, UInt16 srcwidth, UInt16 srcheight, UInt16 dstwidth, UInt16 dstheight, UInt16 xoffset, UInt16 yoffset, float factorx, float factory)
{
    if ((srcheight * factory + yoffset) > dstheight) return;
    if ((srcwidth * factorx + xoffset) > dstwidth) return;
   
    UInt16 i, j, yofs;
    yofs = yoffset * dstwidth + xoffset;
   
    UInt16 ifactor, dstyofs;
    const UInt16 fach = (UInt16)(factory * srcheight);
    const UInt16 facw = (UInt16)(factorx * srcwidth);

    for (i = 0; i < fach; i++)
    {
        ifactor = ((UInt16)((float)i/factory)) * srcwidth;
        dstyofs = (UInt16)(dst + yofs);
       
        for (j = 0; j < facw; j++)
        {
            *((UInt16*)dstyofs + j) = *(src + ifactor + (UInt16)((float)j/factorx));
        }
       
        yofs += dstwidth;
    }
}

[yoyofr]

thx for your help, but -no offense- LJP need more optimized code ;D
Using float and 16 bits access isn't very good with arm processor.
I'll do some blitter for different zoom/orientation in arm assembly later...

If you want some tips :
1/ think fixed floating point coding when working
with arm cpu.
2/ think 32 bits access when reading/writing
3/ think assembly!
here is an example of the current blitter for snes :

void os9x_asmfastrender(uint8 *screen,uint8 *source,int height,int pitch)
{
__asm{
cmp r2,#0
beq fin
stmfd r13!,{r3-r12,r14}
mov r14,r3
loopY:
ldmia r1!,{r3-r12} //r3-r12 : 10regs -> 40bytes <=> 20pixels
stmia r0!,{r3-r12} //20
ldmia r1!,{r3-r12}
stmia r0!,{r3-r12} //40
ldmia r1!,{r3-r12}
stmia r0!,{r3-r12} //60
ldmia r1!,{r3-r12}
stmia r0!,{r3-r12} //80
ldmia r1!,{r3-r12}
stmia r0!,{r3-r12} //100
ldmia r1!,{r3-r12}
stmia r0!,{r3-r12} //120
ldmia r1!,{r3-r12}
stmia r0!,{r3-r12} //140
ldmia r1!,{r3-r12}
stmia r0!,{r3-r12} //160
ldmia r1!,{r3-r12}
stmia r0!,{r3-r12} //180
ldmia r1!,{r3-r12}
stmia r0!,{r3-r12} //200
ldmia r1!,{r3-r12}
stmia r0!,{r3-r12} //220
ldmia r1!,{r3-r12}
stmia r0!,{r3-r12} //240
ldmia r1!,{r3-r10}
stmia r0!,{r3-r10} //256
add r0,r0,r14
subs r2,r2,#1
bne loopY


ldmfd r13!,{r3-r12,r14}
fin:
}
}

[Tinnus]

Hm... no problem, I know I'm not even near the level of optimization you guys can do ;D

ASM is too much for me... ar least for now.