[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Second Sight again...



jlee@ics.uci.edu (Orion Pax) writes:

>: Decrement-and-branch-if-non-zero instructions. gotta love em.

>Which is no loss of speed if scanned the other way and checked to some
>explicit number.  What are we talking about here, 0.001 MHz for the dec/beq
>to be a speed factor?

More than that. If Jawaid's loop is tight enough, it's going to be very
significant. One reprieve is that different code can be used for filled
vs. non-filled scanlines.

I think the real question is how much extra code does it take to propagate
the fill value. At the very least it would add a pair of BIT/BNE instructions
to isolate the 320-mode pixels, on top of the loop order reversal.

Remember that the SS is not using a 65816 but a 64180 which is Z80-like
in its wastage of cycles. The CPU power of the SS is more on par with a
stock GS.

I agree that it should be investigated further though. Having the option to
get slower but accurate fill mode would be nice.

Todd Whitesel
toddpw @ ugcs.caltech.edu