On Nov 23, 2011, at 1:28 PM, Tim Gallagher wrote:
I couldn't agree more about using the modern versions of Fortran! Oh the fights we've had here about that one...
I thought I read somewhere that petsc F90 wasn't working correctly/had limited support, but perhaps that was old or incorrect. I think I got that impression here:
http://www.mcs.anl.gov/petsc/petsc-3.1/docs/manualpages/Sys/UsingFortran.htm...
"We recommend either 2 or 3." Only 1 is pure F77 style and there is probably no reason to use it.
I will recompile it with the F90 turned on instead.
You don't really have to change anything to use 2 or 3 , the same libraries, no different configuration. Just other examples Barry
Thanks,
Tim
----- Original Message ----- From: "Barry Smith" <[email protected]> To: [email protected], "For users of the development version of PETSc" <[email protected]> Sent: Wednesday, November 23, 2011 2:16:59 PM Subject: Re: [petsc-dev] Bug in example code
Tim,
You must have array bounds checking turned on for Fortran with the Intel debug version. The lines
subroutine FormInitialGuess(X,ierr) implicit none
#include "ex5f.h"
! Input/output variables: Vec X PetscErrorCode ierr
! Declarations for use with local arrays: PetscScalar lx_v(0:1) PetscOffset lx_i Vec localX
will break that because we access outside th lx_f(0:1) array. Our Fortran examples that use F77 have to violate the array bounds so must be compiled without that checking.
The Fortran 90 examples like ex5f90.F use F90 constructs and thus don't violate bounds. If you plan to use Fortran it makes sense to use the F90 version; after all 2011 >> 1977 :-)
Barry
On Nov 23, 2011, at 12:04 PM, Tim Gallagher wrote:
Hi,
I just cloned the petsc-dev branch for the first time and built it with GCC with both release and debug options and the tests during make test worked fine. I then built it with Intel with release options and the tests worked. Then I compiled with debug options and one of the tests generates an error:
Possible error running Fortran example src/snes/examples/tutorials/ex5f with 1 MPI process See http://www.mcs.anl.gov/petsc/documentation/faq.html forrtl: severe (408): fort: (2): Subscript #1 of the array LX_V has value 15092 which is greater than the upper bound of 1
Image PC Routine Line Source libirc.so 00007F4245C589AA Unknown Unknown Unknown libirc.so 00007F4245C574A6 Unknown Unknown Unknown libifcore.so.5 00007F4240A5B6FC Unknown Unknown Unknown libifcore.so.5 00007F42409CEDA2 Unknown Unknown Unknown libifcore.so.5 00007F42409CF323 Unknown Unknown Unknown ex5f 0000000000403F58 forminitialguess_ 294 ex5f.F ex5f 0000000000403719 MAIN__ 224 ex5f.F ex5f 00000000004028AC Unknown Unknown Unknown libc.so.6 00007F42401AB23D Unknown Unknown Unknown ex5f 00000000004027A9 Unknown Unknown Unknown
I am using:
Intel(R) Fortran Intel(R) 64 Compiler XE for applications running on Intel(R) 64, Version 12.1.1.256 Build 20111011 Copyright (C) 1985-2011 Intel Corporation. All rights reserved.
and petsc was configured with:
./configure PETSC_ARCH=linux-intel-debug --prefix=/opt/petsc/intel/2011.sp1.1/debug --with-make-np=12 --with-clanguage=C++ --with-c-support=1--with-fortran-kernels=1 --with-shared-libraries=1--download-scalapack=yes --download-parmetis=yes --download-metis=yes --with-imagemagick=1 --download-sundials=yes --download-mpi4py=yes --download-petsc4py=yes --download-boost=yes --with-hdf5=1 --with-hdf5-dir=$HDF5_DIR --with-imagemagick=0 --with-shared-libraries --with-debugging=no COPTFLAGS='-g -C -traceback -fpe0 -check noarg_temp_created -fp-stack-check -check uninit' FOPTFLAGS='-g -C -traceback -fpe0 -check noarg_temp_created -fp-stack-check -check uninit'
I am going to try looking for the problem, but since this is the first 10 minutes I've looked at the code, it will probably take me a lot longer than somebody else to find the problem. Let me know if there's anything I can do to help track down the problem.
Thanks,
Tim