Dalton 64-bits error

[mcasanova]

Dears,

I am trying to compile Dalton in serial (for now) but I am having errors when I try to do a calculation.

I am using a Mac mojave (10.14.3) with intel 2017 update 8 (the last update in 2017). My setup is the same as in the installation webpage
(https://dalton-installation.readthedocs ... neral.html)

./setup --fc=ifort --cc=icc --cxx=icpc --mkl=sequential --int64

this is the output

FC=ifort CC=icc CXX=icpc cmake -DENABLE_MPI=OFF -DENABLE_SGI_MPT=OFF -DENABLE_OPENMP=OFF -DENABLE_64BIT_INTEGERS=ON -DENABLE_GPU=OFF -DENABLE_CUBLAS=OFF -DENABLE_REAL_SP=OFF -DENABLE_STATIC_LINKING=OFF -DENABLE_SCALAPACK=OFF -DENABLE_AUTO_BLAS=ON -DENABLE_AUTO_LAPACK=ON -DMKL_FLAG="-mkl=sequential" -DENABLE_AUTO_BLAS=OFF -DENABLE_AUTO_LAPACK=OFF -DCMAKE_BUILD_TYPE=release /Users/student/Programs/dalton

--Switch off warnings due to incompatibility XCode 4 and Intel 11 on OsX 10.6
-- User set explicit MKL flag which is passed to the compiler and linker: -mkl=sequential
-- This disables math detection and builtin math libraries
-- Setting -DHAVE_MKL_BLAS and -DHAVE_MKL_LAPACK
-- GIT_BRANCH : master
-- System : Darwin
-- Processor type : x86_64
-- Fortran compiler flags: -fpp -assume byterecl -DVAR_IFORT -mkl=sequential -i8 -Qoption,ld,-w -O3 -ip -diag-disable 8290 -diag-disable 8291
-- C compiler flags : -g -wd981 -wd279 -wd383 -vec-report0 -wd1572 -wd1777 -restrict -DRESTRICT=restrict -mkl=sequential -O3 -ip
-- Libraries : /Users/student/Programs/dalton/build/external/lib/libqfitlib.a;/Users/student/Programs/dalton/build/external/lib/libpelib.a;-mkl=sequential
-- Definitions : SYS_DARWIN;SYS_UNIX;HAVE_NO_LSEEK64;VAR_IFORT;HAVE_MKL_BLAS;HAVE_MKL_LAPACK;BUILD_GEN1INT;BUILD_QFITLIB;VAR_MFDS;_FILE_OFFSET_BITS=64;IMPLICIT_NONE;BINARY_INFO_AVAILABLE;INSTALL_BASDIR="/Users/student/Programs/dalton/build/basis";VAR_INT64;VAR_64BITS;INSTALL_WRKMEM=64000000;INSTALL_MMWORK=1;BUILD_PELIB;BUILD_QFITLIB
-- The Fortran compiler identification is Intel 17.0.8.20180907
-- The C compiler identification is Intel 17.0.8.20180907
-- The CXX compiler identification is Intel 17.0.8.20180907
-- Check for working Fortran compiler: /Users/student/Programs/Intel/2017.8.059/bin/ifort
-- Check for working Fortran compiler: /Users/student/Programs/Intel/2017.8.059/bin/ifort -- works
-- Detecting Fortran compiler ABI info
-- Detecting Fortran compiler ABI info - done
-- Checking whether /Users/student/Programs/Intel/2017.8.059/bin/ifort supports Fortran 90
-- Checking whether /Users/student/Programs/Intel/2017.8.059/bin/ifort supports Fortran 90 -- yes
-- Check for working C compiler: /Users/student/Programs/Intel/2017.8.059/bin/icc
-- Check for working C compiler: /Users/student/Programs/Intel/2017.8.059/bin/icc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /Users/student/Programs/Intel/2017.8.059/bin/icpc
-- Check for working CXX compiler: /Users/student/Programs/Intel/2017.8.059/bin/icpc -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Found Git: /usr/bin/git
-- Polarizable Continuum Model via PCMSolver DISABLED
-- Configuring done
-- Generating done
-- Build files have been written to: /Users/student/Programs/dalton/build

configure step is done
now you need to compile the sources:
$ cd build
$ make

then I go to build and do: make -j 8

In the middle of the output I can read this messages

[ 8%] Building C object CMakeFiles/dalton.dir/DALTON/dft/cube-fast.c.o
icc: command line remark #10148: option '-vec-report0' not supported
[ 8%] Building Fortran object CMakeFiles/dalton.dir/DALTON/dft/dft_dftd.F.o
icc: command line remark #10148: option '-vec-report0' not supported
icc: command line remark #10148: option '-vec-report0' not supported
...more code
[ 99%] Built target dalton
Scanning dependencies of target dalton.x
[ 99%] Building Fortran object CMakeFiles/dalton.x.dir/DALTON/main/dalton.F.o
[100%] Linking Fortran executable dalton.x
[100%] Built target dalton.x

and then the error when I do a calculation:

Negative ICKITR in CCSD_INIT1
Calculation too large for 32-bit integersTry rebuilding Dalton using 64-bit integers

if I use otool -L (same as ldd)

otool -L /Users/student/Programs/dalton/build/dalton.x
/Users/student/Programs/dalton/build/dalton.x:
@rpath/libmkl_intel_ilp64.dylib (compatibility version 0.0.0, current version 0.0.0)
@rpath/libmkl_sequential.dylib (compatibility version 0.0.0, current version 0.0.0)
@rpath/libmkl_core.dylib (compatibility version 0.0.0, current version 0.0.0)
/usr/lib/libc++.1.dylib (compatibility version 1.0.0, current version 307.5.0)
/usr/lib/libSystem.B.dylib (compatibility version 1.0.0, current version 1238.60.2)


Any help would be appreciated

Kind regrards,

Marco

p.d I have used also install_name_tool -change for the @rpath libraries, but still the same error
rpath tutorial can be found here: https://software.intel.com/en-us/forums ... pic/698021

[magnus]

The reason for the error is most likely not related to 64-bit integers but rather that you are requesting to freeze 30 core orbitals which is difficult when you only have 8 occupied orbitals In this case it seems more reasonable to just freeze the two core orbitals. Try to change the relevant part of your input to

Code: Select all

.FREEZE
2 0

[mcasanova]

Oh Thanks a lot! not it's working fine!
Sorry all these silly questions, I am still new with this program
In another topic I will ask about compiling using openmpi 4.0.0 because I got an error almost in the end regarding openmpi

Thanks again!

[mcasanova]

Hello Again,

I was able to compile dalton using 64 bits and serial, but know I would like to compile it using intel MKL and openMPI... I have been using this command:

./setup --int64 --fc=ifort --cc=icc --cxx=icpc --mkl=parallel --mpi --blas auto --lapack auto parallel

but in the compilation I found the next error:

[ 0%] Built target FChk2HES
[ 1%] Built target labread
[ 1%] Built target peter_utils_blocks
[ 1%] Built target generate_binary_info
[ 1%] Built target check_external_timestamp_pelib
[ 1%] Built target check_external_timestamp_qfitlib
[ 1%] Built target check_external_timestamp_gen1int
[ 1%] Built target distances
[ 2%] Built target aces2dalton
[ 2%] Built target xyz2dalton
[ 3%] Built target gen1int
[ 3%] Built target gen1int_interface
[ 5%] Built target pelib
[ 5%] Built target qfitlib
[ 5%] Built target qfitlib_interface
[ 5%] Built target pelib_interface
Scanning dependencies of target dalton
[ 5%] Building Fortran object CMakeFiles/dalton.dir/binary_info.F90.o
[ 5%] Linking Fortran static library lib/libdalton.a
[ 99%] Built target dalton
[100%] Linking Fortran executable dalton.x
Undefined symbols for architecture x86_64:
"_MPI_Barrier", referenced from:
_grid_open in libdalton.a(grid-gen.c.o)
_opnqua_ in libdalton.a(grid-gen.c.o)
"_MPI_Bcast", referenced from:
_mpi_sync_data in libdalton.a(general.c.o)
_dft_wake_slaves in libdalton.a(general.c.o)
_dft_cslave_ in libdalton.a(general.c.o)
_dftfuncsync_ in libdalton.a(general.c.o)
_dft_lin_resp_ in libdalton.a(prop-eval.c.o)
_dft_kohn_shamab_slave in libdalton.a(prop-eval.c.o)
_dft_kohn_shamab_ in libdalton.a(prop-eval.c.o)
...

I tried with openmpi 3.1.2 and 4.0.0 and I got the same error...
Am I missing something?

thanks a lot again

[magnus]

Can you provide the output from the setup part?

Note that Dalton itself is only MPI parallel and NOT OpenMP threaded, so you have to be careful using parallel MKL together with MPI-enabled Dalton. Most often you get the best performance using just MPI, unless you are using non-parallel parts of Dalton, such as the CC module.

Also are you sure need 64-bit integers? Not all functionality will work with 64-bit integers and AFAIK you would only need it for large CC calculations and possibly a few other parts.

[mcasanova]

Hi,

This is the output of the setup part:

FC=ifort CC=icc CXX=icpc cmake -DENABLE_MPI=ON -DENABLE_SGI_MPT=OFF -DENABLE_OPENMP=OFF -DENABLE_64BIT_INTEGERS=ON -DENABLE_GPU=OFF -DENABLE_CUBLAS=OFF -DENABLE_REAL_SP=OFF -DENABLE_STATIC_LINKING=OFF -DENABLE_SCALAPACK=OFF -DENABLE_AUTO_BLAS=ON -DENABLE_AUTO_LAPACK=ON -DMKL_FLAG="-mkl=parallel" -DENABLE_AUTO_BLAS=OFF -DENABLE_AUTO_LAPACK=OFF -DCMAKE_BUILD_TYPE=release /Users/student/Programs/dalton

--Switch off warnings due to incompatibility XCode 4 and Intel 11 on OsX 10.6
-- User set explicit MKL flag which is passed to the compiler and linker: -mkl=parallel
-- This disables math detection and builtin math libraries
-- Setting -DHAVE_MKL_BLAS and -DHAVE_MKL_LAPACK
WARNING: Your compiler does not seem to be MPI compatible
-- WARNING: mpi module not found, will use mpif.h instead
-- WARNING: integer check not successful, assuming a 64bit mpif.h instead (if this is not true, please specify -DFORCE_32BIT_MPI_INTERFACE=ON)
-- GIT_BRANCH : master
-- System : Darwin
-- Processor type : x86_64
-- Fortran compiler flags: -fpp -assume byterecl -DVAR_IFORT -mkl=parallel -i8 -Qoption,ld,-w -O3 -ip -diag-disable 8290 -diag-disable 8291
-- C compiler flags : -g -wd981 -wd279 -wd383 -vec-report0 -wd1572 -wd1777 -restrict -DRESTRICT=restrict -mkl=parallel -O3 -ip
-- Libraries : /Users/student/Programs/dalton/parallel/external/lib/libqfitlib.a;/Users/student/Programs/dalton/parallel/external/lib/libpelib.a;-mkl=parallel
-- Definitions : SYS_DARWIN;SYS_UNIX;HAVE_NO_LSEEK64;VAR_IFORT;HAVE_MKL_BLAS;HAVE_MKL_LAPACK;VAR_MPI;BUILD_GEN1INT;BUILD_QFITLIB;VAR_MFDS;_FILE_OFFSET_BITS=64;IMPLICIT_NONE;BINARY_INFO_AVAILABLE;INSTALL_BASDIR="/Users/student/Programs/dalton/parallel/basis";VAR_INT64;VAR_64BITS;INSTALL_WRKMEM=64000000;INSTALL_MMWORK=1;BUILD_PELIB;BUILD_QFITLIB
-- The Fortran compiler identification is Intel 17.0.8.20180907
-- The C compiler identification is Intel 17.0.8.20180907
-- The CXX compiler identification is Intel 17.0.8.20180907
-- Check for working Fortran compiler: /Users/student/Programs/Intel/2017.8.059/bin/ifort
-- Check for working Fortran compiler: /Users/student/Programs/Intel/2017.8.059/bin/ifort -- works
-- Detecting Fortran compiler ABI info
-- Detecting Fortran compiler ABI info - done
-- Checking whether /Users/student/Programs/Intel/2017.8.059/bin/ifort supports Fortran 90
-- Checking whether /Users/student/Programs/Intel/2017.8.059/bin/ifort supports Fortran 90 -- yes
-- Check for working C compiler: /Users/student/Programs/Intel/2017.8.059/bin/icc
-- Check for working C compiler: /Users/student/Programs/Intel/2017.8.059/bin/icc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /Users/student/Programs/Intel/2017.8.059/bin/icpc
-- Check for working CXX compiler: /Users/student/Programs/Intel/2017.8.059/bin/icpc -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Found MPI_C: /Users/student/Programs/openmpi/4.0.0/lib/libmpi.dylib (found version "3.1")
-- Found MPI_CXX: /Users/student/Programs/openmpi/4.0.0/lib/libmpi.dylib (found version "3.1")
-- Found MPI_Fortran: /Users/student/Programs/openmpi/4.0.0/lib/libmpi_usempif08.dylib (found version "3.1")
-- Found MPI: TRUE (found version "3.1")
-- Performing Test MPI_COMPATIBLE
-- Performing Test MPI_COMPATIBLE - Failed
-- Performing Test MPI_F90_I4
-- Performing Test MPI_F90_I4 - Failed
-- Performing Test MPI_F90_I8
-- Performing Test MPI_F90_I8 - Failed
-- Performing Test ENABLE_MPI3_FEATURES
-- Performing Test ENABLE_MPI3_FEATURES - Failed
-- Found Git: /usr/bin/git
-- Polarizable Continuum Model via PCMSolver DISABLED
-- Configuring done
-- Generating done
-- Build files have been written to: /Users/student/Programs/dalton/parallel

and some library paths
echo $LD_LIBRARY_PATH
/Users/student/Programs/openmpi/4.0.0/lib
echo $DYLD_LIBRARY_PATH
/Users/student/Programs/openmpi/4.0.0/lib

Also are you sure need 64-bit integers? Not all functionality will work with 64-bit integers and AFAIK you would only need it for large CC calculations and possibly a few other parts.

I am not sure.. I need to calculate some CCSDR(3) excitation energies and I was using the 64-bits because the error I had my first message. Maybe I should remove it...

Thanks!

[magnus]

The error you had in your first message was (most likely) because of wrong input so that does not necessarily mean that you need 64-bit integers. Note that the CC module in Dalton is not parallelized so MPI will not give you anything. You may however gain some speedup by using a threaded math library such as MKL. In this case use the "-mkl=parallel" as you did but remove the "--mpi" (and I would recommend to first try without "--int64" as well).