Issue on compiling IOAPI 3.1 with ifort 13.1.1 on scientific linux 6.4 x64

[lee_icc]

Dear All,

I tried to compile the IOAPI 3.1 with intel fortran compiler 13.1.1 on scientific linux 6.4 x64 platform. The terminal reported several error/warnings as follows:

[sl@localhost ioapi-3.1]$ make
Makefile:200: warning: overriding commands for target `/home/sl/ioapi-3.1/Linux2_x86_64ifort'
Makefile:197: warning: ignoring old commands for target `/home/sl/ioapi-3.1/Linux2_x86_64ifort'
(cd /home/sl/Software/ioapi-3.1/ioapi   ; make fixed_src)
make[1]: Entering directory `/home/sl/Software/ioapi-3.1/ioapi'
make[1]: Nothing to be done for `fixed_src'.
make[1]: Leaving directory `/home/sl/Software/ioapi-3.1/ioapi'
(cd /home/sl/Software/ioapi-3.1/ioapi   ; make all)
make[1]: Entering directory `/home/sl/Software/ioapi-3.1/ioapi'
cd /home/sl/Software/ioapi-3.1/Linux2_x86_64ifort; ifort -auto -cm -w90 -w95 -warn notruncated_source -Bstatic -c -DAUTO_ARRAYS=1 -DF90=1 -DFLDMN=1 -DFSTR_L=int -DIOAPI_NO_STDOUT=1 -DAVOID_FLUSH=1 -DBIT32=1  -O3 -no-fp-port -unroll -stack_temps -safe_cray_ptr -mtune core2 -traceback -openmp -I/home/sl/Software/ioapi-3.1/ioapi /home/sl/Software/ioapi-3.1/ioapi/bilin.f
ifort: command line remark #10010: option '-cm' is deprecated and will be removed in a future release. See '-help deprecated'
ifort: command line remark #10010: option '-w90' is deprecated and will be removed in a future release. See '-help deprecated'
ifort: command line remark #10010: option '-w95' is deprecated and will be removed in a future release. See '-help deprecated'
ifort: command line warning #10159: invalid argument for option '-m'
ifort: error #10236: File not found:  'core2'
make[1]: *** [bilin.o] Error 1
make[1]: Leaving directory `/home/sl/Software/ioapi-3.1/ioapi'
make: *** [all] Error 2

It looks like that the compile option -cm -w90 -w95 were deprecated and -m core2 was invalid. Are there any replacement for these options?

Thanks,

Lee

[Casey]

If you are only going to be running the binary on the system you are compiling on (or with equal or greate cpu capabilities) then you can use "-xHost" instead of "-m".  If you stick with "-m", the options it recognizes per the composer documentations are avx, sse4.2, sse4.1, ssse3, sse3, sse2, sse, ia32.  The -xHost option will pick the highest instruction set your processor can execute and compile for that.

[TimP]

Intel compilers don't recognize the -mtune option; in fact there is no supported option to optimize for a different CPU than the one used in the instruction set selection. 

[Lorri_M_Intel]

-w95 and -w90 don't do anything with ifort; they are only accepted to be compatible with a long-ago version of the Intel Fortran compiler.  In fact, they may disappear in the future.

-cm is equivalent to -warn nousage, which is the default unless you've someone specified "warn all".

                --Lorri

[lee_icc]

Thank you all for your help.

'-cm -w90 -w95 -mtune core2' were included in the original compile options. I checked the options with 'ifort -help deprecated', and replaced -cm with -warn nousage, removed -w90 -w95. Also, I look up the Intel Fortran Compiler XE 13.1 User and Reference Guides, and used the corrct usage '-mtune=core2'.

Finally, ifort only reported a remark for the deprecated option '-Vaxlib'. I did not find replacement for this option. Does this option can be ignored or removed safely from the compile options?

[Lorri_M_Intel]

Yes, you can safely remove or ignore -Vaxlib

That too, was for compatibility with a long-ago implementation, and has no effect on the current Intel Fortran compiler.

                  --Lorri

[lee_icc]

Lorri Menard (Intel) wrote:

Yes, you can safely remove or ignore -Vaxlib

That too, was for compatibility with a long-ago implementation, and has no effect on the current Intel Fortran compiler.

                  --Lorri

Thanks.

[Raymond_W_]

Lee,

I also am trying to build IOAPI on Scientifi Linux 6 (64-bit) on a HP Z800 using ifort (v 14). I got different error messages from you. However, may I suggest that you put your question to the "m3user" Models-3 Technical Support Forum? Carlie Coats will see your message and he has been supporting IOAPI for twenty years or so. He solved my problem for me very quickly (I'm just about to try out some tests with IOAPI before attempting to build CMAQ, which why I need it)

Ray