Algo wrote:
> A rough cycle count comparison is:
> add mul fld fxch total
> NEWER_mandelx_pattern: 5*3 3*3 3*1 0 27
> mandel_pattern: 4*3 3*3 2*1 6*1 29
So just that I get you right: Are you suggesting because your code can trick
a specific compiler on a specific architecture to produce slightly better
code for some processors of that architecture, the current code in POV-Ray
should be changed? Or is your code universally better on an algorithmic level?
Thorsten
Post a reply to this message
|