Hi Thomas,
> The known 'elegant' DCP-Solution needs 26 cycles and 11 of those
cycles
> are overhead for just determining wether to draw or not. The new
> solution reduces that overhead down to only 5 cycles!
I changed the skipDraw routine slightly for my prototype release of
Climber 5 too. It's not as elegant but it saved me a few cycles. During
VBLANK I set the x-register as an offset...
sec
sbc verPosP0
adc #PLAYER_HEIGHT
tax
Then in the kernel for each new scanline I would do...
inx ;2
bmi .skipDraw ;2³
cpx #PLAYER_HEIGHT+1 ;2
bcs .skipDraw ;2³
lda playerGraphics,x ;3
.
.
.
This gave me roughly 7-8 cycles for overhead vs. 9-10 cycles used in the
non-illegal opcode version of skipDraw. I'm not counting the increment
of the x-register because that's a given. It takes more bytes but it
saves those precious CPU cycles.
What do you think? I'd love to hear comments on how it could've been
improved.
I will study this optimization too. It will help me get more familiar
with your tricks :)
Take care,
Dennis
----------------------------------------------------------------------------------------------
Archives (includes files) at http://www.biglist.com/lists/stella/archives/
Unsub & more at http://www.biglist.com/lists/stella/
