|
![](/i/fill.gif) |
Assuming gcc-3.4.0:
> -O3
Correct,
> -DCPU=athlon
not useful at all for the povray-3.50c Unix source,
> -march=athlon
probably "-march=athlon-xp -mtune=athlon-xp" could be
more suited, and also maybe "-msse" and, as reported elsewhere,
"-mfpmath=sse,387" (but use this one with care).
> check out "-finline-functions"
not useful since automatically turned on by -O3,
> "-funroll-loops"
do _not_ use this one, as it will most likely make the
code slower,
> maybe more.
eventually "-malign-double" "-minline-all-stringops" but there is
no evidence yet that they would actually improve efficiency.
- NC
Post a reply to this message
|
![](/i/fill.gif) |