I would like to be able to build optimized blender with cmake but find that it is up to 6 times slower at rendering than my scons builds.
I have an Amd athlon64 (san diego) chip running linux slackware 64bit.
Cmake detects "sse2" ok and builds correctly but renders very slowly.
The scon flags I use are:

Code:

CCFLAGS = ['-pipe','-fPIC','-march=athlon64','-msse','-msse2','-mmmx','-mfpmath=sse','-funsigned-char','-fno-strict-aliasing'] CXXFLAGS = ['-pipe','-fPIC','-march=athlon64','-msse','-msse2','-mmmx','-mfpmath=sse','-funsigned-char','-fno-strict-aliasing'] REL_CFLAGS = ['-O2','-fomit-frame-pointer','-funroll-loops'] REL_CCFLAGS = ['-O2','-fomit-frame-pointer','-funroll-loops']

Any suggestions, or flags to include for cmake would be appreciated.
Thanks in advance.

There are still Makefiles in the source tree. I find I can get a usable executable just using Make.