compiling

[grann]

This message has been moved to the Intel Math Kernel Library forum which launched on Dec. 18, 2003.

[TimP]

As I understood it, mkl no longer supports static linking, but:

I don't know either why you list all those libraries by name, including supplying possibly conflicting libraries.

-openmp -Vaxlib would cause the compiler to choose the libraries required for those options.

Are you trying to avoid using p4 mkl for the subroutines you call directly, while using p4 code for those you call only via lapack calls? I don't think that's supported either.

I suspect red hat 8 wasn't around long enough for it to gain full support, but it's a good shot. If your build works fine for you when you don't depart so far from supported options, that may be all you can ask.

[grann]

I forgot to say that I am using RH8 (glibc-2.2.93-5).

[grann]

The only reason for doing it the obscure way was because of an unknowledgeable person trying weird things to get it to work. Mysteriously, it did.

But now, doing it the right way, I'm having problems. I did this:

FLAGS = -r8 -i8 -g -132 -save -Vaxlib

rtmb_v7.3.1_linux: $(OBJS)
/opt/intel/compiler70/ia32/bin/ifc $(OBJS) $(FLAGS)
-L/opt/intel/mkl/lib/32 -L/opt/intel/compiler70/ia32/lib
-o $(TARGET)

AND I GET:

/home/grann/rtm/src/rtmb1_v7.3.1_linux.f:3774: undefined reference to `zgetrf_'
/home/grann/rtm/src/rtmb1_v7.3.1_linux.f:4187: undefined reference to `zgetrs_'

I DID THIS ON RH7.3, RH8 and RH9 - ALL WITH EXACTLY THIS RESULT.

ALSO, IF WE CAN'T STATICALLY LINK TO THE INTEL MATH LIBRARY THIS WILL CAUSE US ALL KINDS OF PROBLEMS AS WE TRY TO MOVE OUR ENGINEERING TOOLS TO LINUX ACROSS A COMPANY DISPERSED AT SITES IN SEVERAL STATES. TOO MANY VERSIONS OF LINUX, INTEL MATH LIBRARIES AND GCC.