Here is the code for a 1.5x scaler Ayla wrote for Ohboy. You can use it in Gambatte if you want.
/*Ayla's 1.5x Upscaler - 160x144 to 240x216 */
/* Before:
* a b c d
* e f g h
*
* After (parenthesis = average):
* a (a,b) b c (c,d) d
* (a,e) (a,b,e,f) (b,f) (c,g) (c,d,g,h) (d,h)
* e (e,f) f g (g,h) h
*/
void ayla_scale15x(uint32_t *to, uint32_t *from){
uint32_t reg1, reg2, reg3, reg4, reg5;
size_t x, y;
for (y=0; y<216/3; y++) {
for (x=0; x<240/6; x++) {
__builtin_prefetch(to+4, 1);
// Read b-a
reg1 = *from;
reg5 = reg1 >> 16;
reg2 = (reg1 & 0xf7de0000) >> 1;
reg1 &= 0xffff;
reg1 |= reg2 + ((reg1 & 0xf7de) << 15);
// Write (a,b)-a
*to = reg1;
reg1 = (reg1 & 0xf7def7de) >> 1;
// Read f-e
reg3 = *(from++ + 160/2);
reg2 = reg3 >> 16;
reg4 = (reg3 & 0xf7de0000) >> 1;
reg3 &= 0xffff;
reg3 |= reg4 + ((reg3 & 0xf7de) << 15);
// Write (e,f)-e
*(to + 2*320/2) = reg3;
reg3 = (reg3 & 0xf7def7de) >> 1;
// Write (a,b,e,f)-(a,e)
*(to++ + 320/2) = reg1 + reg3;
// Read d-c
reg1 = *from;
// Write c-b
reg5 |= (reg1 << 16);
*to = reg5;
reg5 = (reg5 & 0xf7def7de) >> 1;
// Read h-g
reg3 = *(from++ + 160/2);
// Write g-f
reg2 |= (reg3 << 16);
*(to + 2*320/2) = reg2;
reg2 = (reg2 & 0xf7def7de) >> 1;
// Write (c,g)-(b,f)
*(to++ + 320/2) = reg2 + reg5;
// Write d-(c,d)
reg2 = (reg1 & 0xf7def7de) >> 1;
reg1 = (reg1 & 0xffff0000) | ((reg2 + (reg2 >> 16)) & 0xffff);
*to = reg1;
reg1 = (reg1 & 0xf7def7de) >> 1;
// Write h-(g,h)
reg2 = (reg3 & 0xf7def7de) >> 1;
reg3 = (reg3 & 0xffff0000) | ((reg2 + (reg2 >> 16)) & 0xffff);
*(to + 2*320/2) = reg3;
reg3 = ((reg3 & 0xf7def7de) >> 1);
// Write (d,h)-(c,d,g,h)
*(to++ + 320/2) = reg1 + reg3;
}
to += 2*360/2;
from += 160/2;
}
}