I currently use -O3 -march=core2 -fno-tree-vectorize cause it crashes if tree-vectorize is enabled (by -O3) (in relation to SEE being enabled (by core2)). Mysteriously, -ffast-math makes it slower after those options as well as unroll-loops and the other common default stuff. -mfpmath=sse also crashes it unless its invoked on its own IIRC.
Has anyone found any better ones?
Advertisement
Page 1 of 1
Optimisation options for gcc on mingw building ioquake
#2
Posted 07 March 2009 - 02:31 AM
found "-O3 -march=i686 -fno-math-errno -ffinite-math-only -fno-signaling-nans -fcx-limited-range -fno-trapping-math" in my Makefile.local. that's been there for ages though, but i remember trying various stuff...
i also remember the intel compiler (http://www.intel.com...lers/284132.htm) giving me even more performance. "-O3 -static-intel -xP " is what i found in my Makefile.local for that.
i also remember the intel compiler (http://www.intel.com...lers/284132.htm) giving me even more performance. "-O3 -static-intel -xP " is what i found in my Makefile.local for that.
Page 1 of 1
1 User(s) are reading this topic
0 members, 1 guests, 0 anonymous users
Advertisement