[XviD-devel] Linux (gcc) optimizations
chl at math.uni-bonn.de
Wed Dec 10 20:21:49 CET 2003
I, too, had problems with -fgcse gcc option in plattform.inc
Maybe, we should better remove it.
However, I never had problems with -O2 or -O3 and -funroll-loops
Speedup is not big, but constant 1-2% is better than nothing.
Are there specific reasons against those as default on gcc plattforms?
P.S. Intel C for Linux v8.0 is out and I indeed get 4-6% speedup compared
to gcc 3.4(pre).
More information about the XviD-devel