Quote:
Originally Posted by jtshaw
THey aren't that much different honestly.
|
if night and day are not much different.
the trick is in doing it well.
check it out
with gcc -O2 my system
Dhrystone 2 using register variables 3900355.7 lps (10 secs, 6 samples)
with gcc -march=pentium4 -O3 -funroll-loops -ftracer -momit-leaf-frame-pointer -fprefetch-loop-arrays
Dhrystone 2 using register variables 4453259.6 lps (10 secs, 6 samples)
// about 14% increase over -O2
multiply that times say at least two layers of software like the kernel and you app and its underlying libs and you got a 25% ovrall performance increase
thats = to like what
2 grand for a new box ?