Alain Toussaint wrote: > Hello everyone, > I am currently downloading a gentoo amd64 livedvd for my system as > well as taking notes regarding how I'll build my installation of LFS and I > am not sure about the CFLAGS I will use to build it. Here's the flags I've > chosen after reviewing the > http://gcc.gnu.org/onlinedocs/gcc-4.5.3/gcc/i386-and-x86_002d64-Options.html > web page: > > -O2 -march=native -mtune=native -mavx -mfpmath=sse -msse2avx -pipe > > The rationale behind the avx flags (as well as the mfpmath=sse flags) is > because, on amd64, 64bit sse math (yes it make the mfpmath=sse keyword > redundant) is used by default by gcc while on 32bit x86 cpu, the default is > 80bit extended precision (64bit mantissa) and furthermore, I want to program > my nvidia gpu using cuda and the double precision math on it is 64bit (53bit > mantissa, 10bit exponent, 1bit sign, same as amd64) and would like the same > precision math in my programs. > > Should I be okay with these flags?
Maybe. But for a first build, we recommend not setting CFLAGS at all unless specified in the book. Use the developer's defaults. They understand the software better than you (or we) do. Optimization rarely makes a discernible difference. -- Bruce -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
