Hey guys,

These last days I've been working on a textured triangle routine, mostly for the sake of having done one. For now it's Axe, I think it's rather well optimized (not too much) but it's still way too slow for my taste, so I thought of porting it to ASM.

So far I have this. It doesn't do any clipping on any side, so trying to draw out-screen will mess up your calc real good :

:_E8000→°DX1+2→°DX2+2→°DX3+2→°X1+2→°X2+2→°DU1+2→°DU2+2→°DU3+2→°DV1+2→°DV2+2→°DV3+2→°DU+2→°DV+2→°U1+2→°U2+2→°V1+2→°V2+2→°SFlag+2→°SMid+2→°SEnd+2→°SX1+2→°SX2+2→°SU+2→°SV
:.Tri(°X1Y1,°X2Y2,°X3Y3,°U1V1,°U2V2,°U3V3)
:.Pic1 is a 16*16 image
:Lbl Tri
:Data(128,64,32,16,8,4,2)→GDB1MASK
:If {r1+1}>{r2+1}
:Exch(°r1,°r2,2
:Exch(°r3,°r4,2
:End
:If {r2+1}>{r3+1}
:Exch(°r2,°r3,2
:Exch(°r4,°r5,2
:End
:If {r1+1}>{r2+1}
:Exch(°r1,°r2,2
:Exch(°r3,°r4,2
:End
:
:{r2+1}-{r1+1}?,+1
:→Θ
:{r2}-{r1}*256//Θ→DX1
:{r5}-{r4}*256//Θ→DU1
:{r5+1}-{r4+1}*256//Θ→DV1
:
:{r3+1}-{r1+1}?,+1
:→Θ
:{r3}-{r1}*256//Θ→DX2
:{r6}-{r4}*256//Θ→DU2
:{r6+1}-{r4+1}*256//Θ→DV2
:
:{r3+1}-{r2+1}?,+1
:→Θ
:{r3}-{r2}*256//Θ→DX3
:{r6}-{r5}*256//Θ→DU3
:{r6+1}-{r5+1}*256//Θ→DV3
:
:DX2-DX1?,1.0
:→Θ
:DU2-DU1/*Θ→DU
:DV2-DV1/*Θ→DV
:0→SFlag
:
:{r1}*256→X1→X2
:{r1+1}→Y
:{r2+1}→SMid
:{r3+1}→SEnd
:{r4}*256→U1→U2
:{r4+1}*256→V1→V2
:
:If DX1-DX2/32768
:°X1+1→SX1
:°X2+1→SX2
:°U1→SU
:°V1→SV
:Else
:°X2+1→SX1
:°X1+1→SX2
:°U2→SU
:°V2→SV
:End
:
:While 1
:.Draw texture slice
:{SX1}→X
:{SX2}→Θ
:{SU}^r→U
:{SV}^r→V
:X and 7→r5
:X/2/2/2+(Y*12)+L6→r6
:!If r5
:Else!If -1
:{r6}
:./128
:Asm(3E80A5076F
:Else!If -1
:{r6}
:./64
:Asm(3EC0A507076F
:Else!If -1
:{r6}
:./32
:Asm(3EE0A50707076F
:Else!If -1
:nib{r6*2}
:Else!If -1
:{r6}
:./8
:Asm(CB3DCB3DCB3D
:Else!If -1
:{r6}
:./4
:Asm(CB3DCB3D
:Else!If -1
:{r6}
:./2
:Asm(CB3D
:End
:→r4
:While 1
:r4*2
:+({V/256*2+Pic1}^rree(U/256))→r4
:!If r5+1 and 7→r5
:r4→{r6}
:r6++
:0→r4
:End
:DU+U→U
:DV+V→V
:EndIf X++>Θ
:r5?r4*{r5-1+GDB1MASK} or {r6}→{r6
:
:If SFlag
:DX3+X1→X1
:DU3+U1→U1
:DV3+V1→V1
:Else
:DX1+X1→X1
:DU1+U1→U1
:DV1+V1→V1
:End
:DX2+X2→X2
:DU2+U2→U2
:DV2+V2→V2
:Select(Y++,≥SMid?→SFlag)
:EndIf >SEnd
:Return

It does work, but it's reaaally slow (drawing 2 triangles is 6 FPS at 6 MHz) :



So before starting porting it to ASM (after I've got clipping working, but that is easy with the current code), I wanted to know if there was a better technique to do that. Basically, I'm doing it on a per-pixel logic (of course) but I interpolate every X and every Y, which I think is too slow.

Then I came around that 10-years old ASM scaled sprite routine by Badja and looking at the source and at how fast it was I felt like that technique could be used, although it can only shrink sprites. Basically, instead of interpolating every X, it calculates how many pixels has to be skipped and does the corresponding shifts according to it.

Would that be something feasible ? I didn't give it much thought by now, but seeing how slow my routine is, I'm pretty confident that there is a faster way. If it's not feasible in Axe but only ASM, it's fine, since I ultimately want to have an ASM textured triangle filler.