Compiler detection
Posted: Thu Aug 21, 2014 7:39 pm
Hi,
On NERSC Hopper supercomputer, they ask to use a wrapper "ftn" as for a fortran compiler, no matter the compiler we use (intel,gcc,cray).
But when typing
return a first line that is useless and then the compiler vendor/version.
The problem is that in the configure script (or more precisely m4/lang-fortran.m4) the detection of the compiler is done using
which will fail for all compiler since the first line is not related to the compiler.
I could bypass this behavior for my cas (Intel) modifying the line with
which worked.
Abinit now detects "intel 13.1" instead of "generic 0.0".
This issue caused the compilation line not to use the option "-module" which messed up the module compilation and crashed at the very beginning.
Maybe the "head -n 1" should be replaced by a grep line the next line ?
Anyways, I had the problem and solved it myself but I wanted to report it here for any other user in the same situation.
Cheers
Jordan
On NERSC Hopper supercomputer, they ask to use a wrapper "ftn" as for a fortran compiler, no matter the compiler we use (intel,gcc,cray).
But when typing
Code: Select all
ftn -V
ftn --version
return a first line that is useless and then the compiler vendor/version.
The problem is that in the configure script (or more precisely m4/lang-fortran.m4) the detection of the compiler is done using
Code: Select all
$1 -V | head -n 1
which will fail for all compiler since the first line is not related to the compiler.
I could bypass this behavior for my cas (Intel) modifying the line with
Code: Select all
sed -n "2p"
which worked.
Abinit now detects "intel 13.1" instead of "generic 0.0".
This issue caused the compilation line not to use the option "-module" which messed up the module compilation and crashed at the very beginning.
Maybe the "head -n 1" should be replaced by a grep line the next line ?
Anyways, I had the problem and solved it myself but I wanted to report it here for any other user in the same situation.
Cheers
Jordan