Intel® Fortran Compiler for Linux* and Mac OS X*

poor parallel speedup of sample program: jacobi.f


I am trying to run the "jacobi.f" sample program provided on the website:

have not been able to achieve any kind of parallel speedup when I
compile and run this program with intel fortran compiler for linux.

Has anyone had sucess with this program using intel fortran linux compiler ? Any explicit ode solver ?
What is necessary to achieve parallel speedup with the openmp "jacobi.f" sample program ?

REAL(4) to REAL(8) conversion

Dear All,

I have a problem with a real number conversion. I would like to convert a single precision real number into a double precision. I wrote a simple program which uses DBLE() function.

% cat conversion.f
program conversion

real(4) sx1
real(8) dx1

write(*,*) sx1,dx1


When I run the program I get the following result:

% ./conversion
1.000010 1.00001001358032

I wish that the result of conversion is 1.00001000000000 (or perhaps 1.00001000000001 or. 1.00000999999999).

-r8 option makes deallocate fail

hi i am using fortran for linux 9.0
i have a bunch of legacy code written in f77 and there all real declaration are real*8 ,

i have been coding in f90 free format , and i wrote my modules and libraries and and have other people's libraries also in f90, here the declararation are simply real , and use some allocateble arrays

i added in my makefile the -r8 option for my f90 files and it compiles with no warning any time i run i got 'error glibc detected double free or corruption '
when i try to deallocate the arrays ,

Getting line number info when debugging


this is a trivial question, and I am only posting since I simply couldn't find the (perhaps obvious) answer even after browsing several manuals. Any solution, however obvious, is greatly appreciated.

When a compiled binary of mine segfaults, it would be really nice to simply get the source file and line number where this occurred. All that I can see indicates that ifort should be able to do this, but even with -O0, -g, idb, and the source directory specified on the idb command line, I simply get:

Fortran won't install on a Mac

I'm trying to install the 9.1.029 fortran compiler (standard edition) on an intel iMac (24"). Running OS 10.4.8. Installation of XCode 2.4 went w/o incident.

When I ran the fortran installer, it asked me for the serial number - which I input and was accepted as a valid number. I then hit the compiler Install button, which dimmed but came back after only a couple of seconds. It didn't appear to do anything. There is no intel directory in /opt.

icc/ifort compile time increases after firewall install


a few days ago I declared new firewall rules on my desktop Linux machine in order to block most of the incoming/outgoing TCP/IP packets and now the compile time of icc/ifort has increased significatively. Do the compilers need access to a specific TCP/UDP port or do they "phone home"?

If so, what is the port number or IP address?

Thanks in advance for your response.

MPI taking so much VIRT memory

I have a fortran program compiled with ifort and openMPI and I experienced a strange behaviour - when runing at one processor, it takes about 105M of VIRT memory (which is fine) but when I run it on 2 processors ( mpirun -np 2 ...) each process takes about 550MB of virt memory. There is nothing in the code that would scale with the number of processes and the amount of SHR memory fits well ( 44 MB with one proc and cca 52MB altogether with 2 procs). Any ideas why does it need so much memory?

Trying to write 8bit data as big_endian but -convert big_endian doesn't seem to work

When i compile my code with -c -u -O -auto -convert big_endian -assume byterecl -align none and use the open statement OPEN(unit=23, file=testfile, status="replace", action="write", access="sequential", form="unformatted") to write out 16bit data, this writes out a data file that is big_endian. When i use the same compile options and the open statement OPEN(unit=23, FILE=Filename, FORM='UNFORMATTED', ACTION='WRITE', STATUS='UNKNOWN', RECL=RecordLength, ACCESS='DIRECT') to write an 8 bit data array as big_endian it doesn't work. The file is as it should be except that it is little_endian.

Subscribe to Intel® Fortran Compiler for Linux* and Mac OS X*