[petsc-users] reproduced the problem

[Balay, Satish]

>>>>>>>>>
================================================================================
TEST checkRuntimeIssues from config.packages.BlasLapack(/depot/kildisha/apps/brown/nemo5/libs/petsc/build-real3.11/config/BuildSystem/config/packages/BlasLapack.py:579)
TESTING: checkRuntimeIssues from config.packages.BlasLapack(config/BuildSystem/config/packages/BlasLapack.py:579)
  Determines if BLAS/LAPACK routines use 32 or 64 bit integers
Checking if BLAS/LAPACK routines use 32 or 64 bit integersExecuting: mpicc -c -o /tmp/petsc-wf99X2/config.packages.BlasLapack/conftest.o -I/tmp/petsc-wf99X2/config.setCompilers -I/tmp/petsc-wf99X2/config.compilers -I/tmp/petsc-wf99X2/config.utilities.closure -I/tmp/petsc-wf99X2/config.headers -I/tmp/petsc-wf99X2/config.utilities.cacheDetails -I/tmp/petsc-wf99X2/config.atomics -I/tmp/petsc-wf99X2/config.libraries -I/tmp/petsc-wf99X2/config.functions -I/tmp/petsc-wf99X2/config.utilities.featureTestMacros -I/tmp/petsc-wf99X2/config.utilities.missing -I/tmp/petsc-wf99X2/config.types -I/tmp/petsc-wf99X2/config.packages.MPI -I/tmp/petsc-wf99X2/config.packages.valgrind -I/tmp/petsc-wf99X2/config.packages.pthread -I/tmp/petsc-wf99X2/config.packages.metis -I/tmp/petsc-wf99X2/config.packages.hdf5 -I/tmp/petsc-wf99X2/config.packages.BlasLapack -fopenmp -fPIC  /tmp/petsc-wf99X2/config.packages.BlasLapack/conftest.c 
Successful compile:
Source:
#include "confdefs.h"
#include "conffix.h"
#include <sys/types.h>
#if STDC_HEADERS
#include <stdlib.h>
#include <stdio.h>
#include <stddef.h>
#endif

int main() {
FILE *output = fopen("runtimetestoutput","w");
extern double ddot_(const int*,const double*,const int *,const double*,const int*);
                  double x1mkl[4] = {3.0,5.0,7.0,9.0};
                  int one1mkl = 1,nmkl = 2;
                  double dotresultmkl = 0;
                  dotresultmkl = ddot_(&nmkl,x1mkl,&one1mkl,x1mkl,&one1mkl);
                  fprintf(output, "-known-64-bit-blas-indices=%d",dotresultmkl != 34);;
  return 0;
}
Executing: mpicc  -o /tmp/petsc-wf99X2/config.packages.BlasLapack/conftest   -fopenmp -fPIC /tmp/petsc-wf99X2/config.packages.BlasLapack/conftest.o -L/apps/cent7/intel/compilers_and_libraries_2017.1.132/linux/mkl/lib/intel64 -lmkl_intel_lp64 -lmkl_gnu_thread -lmkl_core -lm -lstdc++ -ldl -L/apps/brown/openmpi.20190215/2.1.6_gcc-5.2.0/lib -lmpi_usempif08 -lmpi_usempi_ignore_tkr -lmpi_mpifh -lmpi -lgfortran -lm -L/apps/cent7/gcc/5.2.0/lib/gcc/x86_64-unknown-linux-gnu/5.2.0 -L/apps/cent7/gcc/5.2.0/lib64 -L/apps/cent7/gcc/5.2.0/lib -Wl,-rpath,/apps/brown/openmpi.20190215/2.1.6_gcc-5.2.0/lib -lgfortran -lm -lgomp -lgcc_s -lquadmath -lpthread -lstdc++ -ldl 
Testing executable /tmp/petsc-wf99X2/config.packages.BlasLapack/conftest to see if it can be run
Executing: /tmp/petsc-wf99X2/config.packages.BlasLapack/conftest
Executing: /tmp/petsc-wf99X2/config.packages.BlasLapack/conftest
ERROR while running executable: Could not execute "['/tmp/petsc-wf99X2/config.packages.BlasLapack/conftest']":
/tmp/petsc-wf99X2/config.packages.BlasLapack/conftest: error while loading shared libraries: libmkl_intel_lp64.so: cannot open shared object file: No such file or directory

                Defined "HAVE_64BIT_BLAS_INDICES" to "1"
Checking for 64 bit blas indices: program did not return therefor assuming 64 bit blas indices
            Defined "HAVE_LIBMKL_INTEL_ILP64" to "1"

<<<<<<<<

So this test has an error but yet the flag HAVE_64BIT_BLAS_INDICES is set.

Is your compiler not returning correct error codes?

Does it make a difference if you also specify -Wl,-rpath along with -L in --with-blaslapack-lib option?


Satish

On Fri, 20 Sep 2019, Povolotskyi, Mykhailo wrote:

> Dear Matthew and Satish,
> 
> I just wrote that the error disappeared, but it still exists (I had to 
> wait longer).
> 
> The configuration log can be accessed here:
> 
> https://www.dropbox.com/s/tmkksemu294j719/configure.log?dl=0
> 
> Sorry for the last e-mail.
> 
> Michael.
> 
> 
> On 09/20/2019 03:53 PM, Balay, Satish wrote:
> > --with-64-bit-indices=1 => PetscInt = int64_t
> > --known-64-bit-blas-indices=1 => blas specified uses 64bit indices.
> >
> > What is your requirement (use case)?
> >
> > Satish
> >
> > On Fri, 20 Sep 2019, Povolotskyi, Mykhailo via petsc-users wrote:
> >
> >> Does it mean I have to configure petsc with --with-64-bit-indices=1 ?
> >>
> >> On 09/20/2019 03:41 PM, Matthew Knepley wrote:
> >> On Fri, Sep 20, 2019 at 1:55 PM Povolotskyi, Mykhailo via petsc-users <petsc-users at mcs.anl.gov<mailto:petsc-users at mcs.anl.gov>> wrote:
> >> Hello,
> >>
> >> I'm upgrading petsc from 3.8 to 3.11.
> >>
> >> In doing so, I see an error message:
> >>
> >>    UNABLE to CONFIGURE with GIVEN OPTIONS    (see configure.log for details):
> >> -------------------------------------------------------------------------------
> >> Cannot use SuperLU_DIST with 64 bit BLAS/Lapack indices
> >> *******************************************************************************
> >>
> >> I wonder why this configuration step worked well for 3.8?  I did not
> >> change anything else but version of petsc.
> >>
> >> This never worked. We are just checking now.
> >>
> >>    Thanks,
> >>
> >>      Matt
> >>
> >> Thank you,
> >>
> >> Michael.
> >>
> >>
> >>
> >> --
> >> What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.
> >> -- Norbert Wiener
> >>
> >> https://www.cse.buffalo.edu/~knepley/<http://www.cse.buffalo.edu/%7Eknepley/>
> >>
> >>
> 
>