ifort: command line warning #10006: ignoring unknown option '-ffast-math'

[hamid_mosaddeghi]

hi

after compile make file with gfortran I get this error:

[hamid@localhost srcmod]$ make gfortran
make FC="mpif90" LD="mpif90 -o" \\
LDFLAGS="-O2 -ffast-math" \\
FFLAGS="-c -O2 -ffast-math" \\
EX=DLPOLY.X BINROOT=../execute par
make[1]: Entering directory `/home/hamid/paralel/dl_poly_2.18/srcmod'
mpif90 -c -O2 -ffast-math parse_module.f
ifort: command line warning #10006: ignoring unknown option '-ffast-math'
mpif90 -c -O2 -ffast-math setup_module.f
ifort: command line warning #10006: ignoring unknown option '-ffast-math'
mpif90 -c -O2 -ffast-math error_module.f
ifort: command line warning #10006: ignoring unknown option '-ffast-math'
mpif90 -c -O2 -ffast-math site_module.f
ifort: command line warning #10006: ignoring unknown option '-ffast-math'
mpif90 -c -O2 -ffast-math config_module.f
ifort: command line warning #10006: ignoring unknown option '-ffast-math'
mpif90 -c -O2 -ffast-math pair_module.f
ifort: command line warning #10006: ignoring unknown option '-ffast-math'

------------------

please help me.

[fercook]

As far as I know, -ffast-math is an option for the gcc compiler, not the Intel Fortran compiler.

It looks like your MPI installation is configured to use ifort instead of gfortran, you should either change that or switch on the appropriate flags for the ifort compiler (beware though, -ffast-math turns off checks that can lead to unreliable results. Check out the gcc page to see what happens with -ffast-math:http://gcc.gnu.org/onlinedocs/gcc-4.1.1/gcc/Optimize-Options.html)

I would just change your LD and F FLAGS in the make file to "-O2" and "-c -O2" respectively, and then to O3 when you have tested your program enough.

[TimP]

I agree with the recommendation to drop -ffast-math. I would suggest also adding -assume protect_parens, at least until you have tested to see whether dropping it will be advantageous. ifort -O2 -fast=2 is roughly equivalent to gfortran -O3 -ffast-math. I doubt you would see an advantage in -fast=2 over the default, unless you are using complex arithmetic and intend to use limited range.