|
| From: | Chen,Jinzhen |
| Subject: | Re: [EXT] Re: getfem installation |
| Date: | Tue, 30 Nov 2021 19:00:31 +0000 |
| User-agent: | Microsoft-MacOutlook/16.43.20110804 |
|
Dear Dr. Poulios, Thank you so much for your prompt response and information. They are very helpful. I cc’ed this email to
getfem-users@nongnu.org so that I don’t need to send to your personal email in next one. After reading your reference of the configure command, I started from scratch but still got compiling errors. I installed metis and mumps in the system from EPEL reporistory. [root@ ~]# rpm -qa |grep metis metis-devel-5.1.0-12.el7.x86_64 metis-5.1.0-12.el7.x86_64 [root@ ~]# rpm -qa |grep metis metis-devel-5.1.0-12.el7.x86_64 metis-5.1.0-12.el7.x86_64 [root@ ~]# rpm -qa |grep MUMPS MUMPS-srpm-macros-5.3.5-1.el7.noarch MUMPS-5.3.5-1.el7.x86_64 MUMPS-devel-5.3.5-1.el7.x86_64 MUMPS-common-5.3.5-1.el7.noarch I used the openmpi/4.1.1, gcc/3.9.0, blas/3.8.0 and zlib/1.2.11 modules for this build. Below is the module list command output. [ris_hpc_apps@r1drpswdev3 getfem-5.4.1]$ module list Currently Loaded Modules: 1) openmpi/4.1.1 2) python/3.7.3-anaconda 3) blas/3.8.0 4) zlib/1.2.11 5) gcc/9.3.0 Here is my configuration command running on getfem-5.4.1 directory ./configure CXX="/risapps/rhel7/openmpi/4.1.1/bin/mpic++" CC="/risapps/rhel7/openmpi/4.1.1/bin/mpicc" FC="/risapps/rhel7/openmpi/4.1.1/bin/mpifort" LIBS="-L/risapps/rhel7/gcc/9.3.0/lib64
-L/risapps/rhel7/openmpi/4.1.1/lib -L/risapps/rhel7/blas/3.8.0 -L/risapps/rhel7/zlib/1.2.11/lib -L/usr/lib64 -lmetis -lzmumps -ldmumps -lcmumps -lsmumps -lmumps_common" CXXFLAGES="-I/risapps/rhel7/gcc/9.3.0/include -I/risapps/rhel7/openmpi/4.1.1/include -I/risapps/rhel7/zlib/1.2.11/include
-I/usr/include -I/usr/include/MUMPS" CPPFLAGES="-I/risapps/rhel7/gcc/9.3.0/include -I/risapps/rhel7/openmpi/4.1.1/include -I/risapps/rhel7/zlib/1.2.11/include -I/usr/include -I/usr/include/MUMPS" CFLAGS="-I/risapps/rhel7/gcc/9.3.0/include -I/risapps/rhel7/openmpi/4.1.1/include
-I/risapps/rhel7/zlib/1.2.11/include -I/usr/include -I/usr/include/MUMPS" PYTHON="/risapps/rhel7/python/3.7.3/bin/python" PYTHON_VERSION=3.7.3 --with-mumps="-L/usr/lib64" --with-mumps-include-dir="-I/usr/include/MUMPS" --with-blas="-L/risapps/rhel7/blas/3.8.0"
--prefix=/risapps/rhel7/getfem-mpi/5.4.1 --enable-shared --enable-metis --enable-par-mumps -enable-paralevel=2 The current errors on the make command are the following: …. libtool: compile: /risapps/rhel7/openmpi/4.1.1/bin/mpic++ -DHAVE_CONFIG_H -I. -I.. -I../src -I../src -I.. -I/usr/local/include -DGETFEM_PARA_LEVEL=2
-DGMM_USES_MPI -DGMM_USES_BLAS -DGMM_U SES_BLAS_INTERFACE -I/usr/include/MUMPS -O3 -std=c++14 -MT dal_bit_vector.lo -MD -MP -MF .deps/dal_bit_vector.Tpo -c dal_bit_vector.cc -fPIC -DPIC -o .libs/dal_bit_vector.o In file included from ./gmm/gmm_kernel.h:49, from getfem/bgeot_config.h:50, from getfem/getfem_omp.h:46, from getfem/dal_basic.h:42, from getfem/dal_bit_vector.h:51, from dal_bit_vector.cc:23: ./gmm/gmm_matrix.h:956:32: error: ‘MPI_Datatype’ does not name a type 956 | template <typename T> inline MPI_Datatype mpi_type(T) | ^~~~~~~~~~~~ ./gmm/gmm_matrix.h:958:10: error: ‘MPI_Datatype’ does not name a type 958 | inline MPI_Datatype mpi_type(double) { return MPI_DOUBLE; } | ^~~~~~~~~~~~ ./gmm/gmm_matrix.h:959:10: error: ‘MPI_Datatype’ does not name a type 959 | inline MPI_Datatype mpi_type(float) { return MPI_FLOAT; } | ^~~~~~~~~~~~ ./gmm/gmm_matrix.h:960:10: error: ‘MPI_Datatype’ does not name a type 960 | inline MPI_Datatype mpi_type(long double) { return MPI_LONG_DOUBLE; } | ^~~~~~~~~~~~ ./gmm/gmm_matrix.h:962:10: error: ‘MPI_Datatype’ does not name a type 962 | inline MPI_Datatype mpi_type(std::complex<float>) { return MPI_COMPLEX; } | ^~~~~~~~~~~~ ./gmm/gmm_matrix.h:963:10: error: ‘MPI_Datatype’ does not name a type 963 | inline MPI_Datatype mpi_type(std::complex<double>) { return MPI_DOUBLE_COMPLEX; } | ^~~~~~~~~~~~ ./gmm/gmm_matrix.h:965:10: error: ‘MPI_Datatype’ does not name a type 965 | inline MPI_Datatype mpi_type(int) { return MPI_INT; } | ^~~~~~~~~~~~ ./gmm/gmm_matrix.h:966:10: error: ‘MPI_Datatype’ does not name a type 966 | inline MPI_Datatype mpi_type(unsigned int) { return MPI_UNSIGNED; } | ^~~~~~~~~~~~ ./gmm/gmm_matrix.h:967:10: error: ‘MPI_Datatype’ does not name a type 967 | inline MPI_Datatype mpi_type(long) { return MPI_LONG; } | ^~~~~~~~~~~~ ./gmm/gmm_matrix.h:968:10: error: ‘MPI_Datatype’ does not name a type 968 | inline MPI_Datatype mpi_type(unsigned long) { return MPI_UNSIGNED_LONG; } | ^~~~~~~~~~~~ ./gmm/gmm_matrix.h: In function ‘typename gmm::strongest_value_type3<V1, V2, MATSP>::value_type gmm::vect_sp(const gmm::mpi_distributed_matrix<MAT>&,
const V1&, const V2&)’: ./gmm/gmm_matrix.h:1012:50: error: ‘MPI_SUM’ was not declared in this scope 1012 | MPI_Allreduce(&res, &rest, 1, mpi_type(T()), MPI_SUM,MPI_COMM_WORLD); | ^~~~~~~ ./gmm/gmm_matrix.h: In function ‘void gmm::mult_add(const gmm::mpi_distributed_matrix<MAT>&, const V1&, V2&)’: ./gmm/gmm_matrix.h:1023:20: error: there are no arguments to ‘MPI_Wtime’ that depend on a template parameter, so a declaration of ‘MPI_Wtime’ must be
available [-fpermissive] 1023 | double t_ref = MPI_Wtime(); | ^~~~~~~~~ ./gmm/gmm_matrix.h:1023:20: note: (if you use ‘-fpermissive’, G++ will accept your code, but allowing the use of an undeclared name is deprecated) ./gmm/gmm_matrix.h:1026:21: error: there are no arguments to ‘MPI_Wtime’ that depend on a template parameter, so a declaration of ‘MPI_Wtime’ must be
available [-fpermissive] 1026 | double t_ref2 = MPI_Wtime(); | ^~~~~~~~~ ./gmm/gmm_matrix.h:1028:19: error: ‘MPI_SUM’ was not declared in this scope 1028 | MPI_SUM,MPI_COMM_WORLD); | ^~~~~~~ ./gmm/gmm_matrix.h:1029:18: error: there are no arguments to ‘MPI_Wtime’ that depend on a template parameter, so a declaration of ‘MPI_Wtime’ must be
available [-fpermissive] 1029 | tmult_tot2 = MPI_Wtime()-t_ref2; | ^~~~~~~~~ ./gmm/gmm_matrix.h:1032:17: error: there are no arguments to ‘MPI_Wtime’ that depend on a template parameter, so a declaration of ‘MPI_Wtime’ must be
available [-fpermissive] 1032 | tmult_tot = MPI_Wtime()-t_ref; | ^~~~~~~~~ make[2]: *** [Makefile:941: dal_bit_vector.lo] Error 1 make[2]: Leaving directory '/risapps/build7/getfem-5.4.1/src' make[1]: *** [Makefile:577: all-recursive] Error 1 make[1]: Leaving directory '/risapps/build7/getfem-5.4.1' make: *** [Makefile:466: all] Error 2 I really appreciate your help. Thank you again ! Best Regards Jinzhen Chen From: Konstantinos Poulios <kopo@mek.dtu.dk>
Dear Jinzhen Chen, Thanks for your question. Yes you should be able to compile GetFEM, also the parallel version of it on Redhat. I haven't tried it but there isn't anything distribution specific in the GetFEM code. Having said that, the parallel version has not been tested very recently, and it might need some performance fixes from our side, to get a good scaling for Anne Ceciles problem. Having compiled the parallel
version of GetFEM on your cluster is a good starting point in any case to detect bottlenecks. The tricky parts for building GetFEM are normally how to link to mumps, metis and other dependencies. If you could send me the compilation errors that you get, either on this address or on the getfem mailing
list getfem-users@nongnu.org I can try to help you to resolve the issues. As an additional reference, some time ago, when I asked our cluster administrators to compile GetFEM they used the following configure options: $ ../configure CXX=mpicxx CC=mpicc FC=mpifort LIBS= -L/zdata/groups/common/nicpa/2018-feb/generic/build-tools/1.0/lib -L/zdata/groups/common/nicpa/2018-feb/generic/build-tools/1.0/lib64
-L/zdata/groups/common/nicpa/2018-feb/generic/gcc/7.3.0/lib -L/zdata/groups/common/nicpa/2018-feb/generic/gcc/7.3.0/lib64 -L/zdata/groups/common/nicpa/2018-feb/XeonX5550/zlib/1.2.11/gnu-7.3.0/lib -L/zdata/groups/common/nicpa/2018-feb/generic/numactl/2.0.11/lib
-L/zdata/groups/common/nicpa/2018-feb/XeonX5550/libxml2/2.9.7/gnu-7.3.0/lib -L/zdata/groups/common/nicpa/2018-feb/XeonX5550/hwloc/1.11.9/gnu-7.3.0/lib -L/zdata/groups/common/nicpa/2018-feb/XeonX5550/openmpi/3.0.0/gnu-7.3.0/lib -L/zdata/groups/common/nicpa/2018-feb/XeonX5550/parmetis/4.0.3/gnu-7.3.0/lib
-L/zdata/groups/common/nicpa/2018-feb/XeonX5550/scalapack/204/gnu-7.3.0/lib -L/zdata/groups/common/nicpa/2018-feb/XeonX5550/openblas/0.2.20/gnu-7.3.0/lib -L/zdata/groups/common/nicpa/2018-feb/XeonX5550/scotch/6.0.4/gnu-7.3.0/lib -L/zdata/g Best regards Konstantinos On Tue, 2021-11-30 at 00:24 +0000, Chen,Jinzhen wrote:
The information contained in this e-mail message may be privileged, confidential, and/or protected from disclosure. This e-mail message may contain protected health information (PHI); dissemination of PHI should comply with applicable federal and state laws. If you are not the intended recipient, or an authorized representative of the intended recipient, any further review, disclosure, use, dissemination, distribution, or copying of this message or any attachment (or the information contained therein) is strictly prohibited. If you think that you have received this e-mail message in error, please notify the sender by return e-mail and delete all references to it and its contents from your systems.
|
| [Prev in Thread] | Current Thread | [Next in Thread] |