Posted By
indi
on 2006-01-18
09:39:23
|  Re: Xply routines...
Actually... ROR shifts the bit into the TOP of "A". Also, since A starts out as 0, the resulting carry from this is always 0.
C -> [76543210] -> C
This routine was lifted directly from the source to Blood Money where it was used in circle calculations. It was then used in xeo3 for exactly the same thing (along with some clipping code) before I posted here to try and speed things up. It works fine - honest. Id say try it and see, but the second one posted is much quicker... so use that instead -particually when you unroll it. 
Actually...I just thought.... if you pre-decriment "AUX" you can ignore the CLC (since it will now add 1 extra ALWAYS) and that'll save 18 cycles more - worst case. But adds 5 best case... Still its more consistant then... And if its a constant your multiplying by, you can "pre-decrement" as well, saving you those 5 cycles...
It was 134 worst, and 98 best... Which now becomes 121 worst, 103 best. Thats "almost" a whole scan line saved from the first version posted.
|
