Up-to-date build from source on IC HPC cluster
Hello, I was just wondering how people compile Nektar++ from source on the IC cluster these days and if there are any recommended adjustments compared with that this link says https://www.nektar.info/nektar-cx1/ It is 4 years old but the instructions still work, it seems I'm able to compile Nektar++ with module load mpi/intel-2019.6.166 module load intel-suite/2017.6 module load fftw/3.3.3-double module load flex/2.6.4 module load scotch/6.0.4 module load boost/1.66.0 CC=mpicc CXX=mpicxx cmake -DNEKTAR_USE_FFTW=ON -DNEKTAR_USE_MKL=ON DNEKTAR_USE_MPI=ON -DNEKTAR_USE_SYSTEM_BLAS_LAPACK=OFF -DNEKTAR_TEST_FORCEMPIEXEC=ON -DCMAKE_CXX_FLAGS="-std=c++11 -O3 -xSSE4.2 -axAVX,CORE-AVX-I,CORE-AVX2" -DCMAKE_C_FLAGS="-O3 -xSSE4.2 -axAVX,CORE-AVX-I,CORE-AVX2" .. Would you recommend any change with respect to newer packages or better practice for efficiency? Thanks for the support, Thibaut
Hi Thibaut, Although this blog was originally written a while ago we have been updating the information as the system has been updated and so it is good that the instructions still work. I am not sure we have tried playing with the compiler flags however so I am not aware of any other options to improve the efficiency of the compiled code. Best regards, Spencer. On 17 Jan 2020, at 13:21, Thibaut Appel <t.appel17@imperial.ac.uk<mailto:t.appel17@imperial.ac.uk>> wrote: Hello, I was just wondering how people compile Nektar++ from source on the IC cluster these days and if there are any recommended adjustments compared with that this link says https://www.nektar.info/nektar-cx1/ It is 4 years old but the instructions still work, it seems I'm able to compile Nektar++ with module load mpi/intel-2019.6.166 module load intel-suite/2017.6 module load fftw/3.3.3-double module load flex/2.6.4 module load scotch/6.0.4 module load boost/1.66.0 CC=mpicc CXX=mpicxx cmake -DNEKTAR_USE_FFTW=ON -DNEKTAR_USE_MKL=ON DNEKTAR_USE_MPI=ON -DNEKTAR_USE_SYSTEM_BLAS_LAPACK=OFF -DNEKTAR_TEST_FORCEMPIEXEC=ON -DCMAKE_CXX_FLAGS="-std=c++11 -O3 -xSSE4.2 -axAVX,CORE-AVX-I,CORE-AVX2" -DCMAKE_C_FLAGS="-O3 -xSSE4.2 -axAVX,CORE-AVX-I,CORE-AVX2" .. Would you recommend any change with respect to newer packages or better practice for efficiency? Thanks for the support, Thibaut _______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk<mailto:Nektar-users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/nektar-users Spencer Sherwin FREng, FRAeS Head of Aerodynamics Section, Director of Research Computing Service, Professor of Computational Fluid Mechanics, Department of Aeronautics, s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk> South Kensington Campus, Phone: +44 (0)20 7594 5052 Imperial College London, Fax: +44 (0)20 7594 1974 London, SW7 2AZ, UK http://www.imperial.ac.uk/people/s.sherwin/
Hello Spencer, Thank you for that piece of information. I requested a newer version of FFTW since it has been optimised for AVX,AVX2 instruction sets but I don't know what fraction of computing time it typically represents in a quasi-3D Nektar++ run. Anyway, good to know Have a nice day, Thibaut On 17/01/2020 17:09, Sherwin, Spencer J wrote:
Hi Thibaut,
Although this blog was originally written a while ago we have been updating the information as the system has been updated and so it is good that the instructions still work. I am not sure we have tried playing with the compiler flags however so I am not aware of any other options to improve the efficiency of the compiled code.
Best regards, Spencer.
On 17 Jan 2020, at 13:21, Thibaut Appel <t.appel17@imperial.ac.uk <mailto:t.appel17@imperial.ac.uk>> wrote:
Hello,
I was just wondering how people compile Nektar++ from source on the IC cluster these days and if there are any recommended adjustments compared with that this link says https://www.nektar.info/nektar-cx1/
It is 4 years old but the instructions still work, it seems I'm able to compile Nektar++ with
module load mpi/intel-2019.6.166 module load intel-suite/2017.6 module load fftw/3.3.3-double module load flex/2.6.4 module load scotch/6.0.4 module load boost/1.66.0
CC=mpicc CXX=mpicxx cmake -DNEKTAR_USE_FFTW=ON -DNEKTAR_USE_MKL=ON DNEKTAR_USE_MPI=ON -DNEKTAR_USE_SYSTEM_BLAS_LAPACK=OFF -DNEKTAR_TEST_FORCEMPIEXEC=ON -DCMAKE_CXX_FLAGS="-std=c++11 -O3 -xSSE4.2 -axAVX,CORE-AVX-I,CORE-AVX2" -DCMAKE_C_FLAGS="-O3 -xSSE4.2 -axAVX,CORE-AVX-I,CORE-AVX2" ..
Would you recommend any change with respect to newer packages or better practice for efficiency?
Thanks for the support,
Thibaut
_______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk <mailto:Nektar-users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/nektar-users
Spencer Sherwin FREng, FRAeS Head of Aerodynamics Section, Director of Research Computing Service, Professor of Computational Fluid Mechanics, Department of Aeronautics, s.sherwin@imperial.ac.uk <mailto:s.sherwin@imperial.ac.uk> South Kensington Campus, Phone: +44 (0)20 7594 5052 Imperial College London, Fax: +44 (0)20 7594 1974 London, SW7 2AZ, UK http://www.imperial.ac.uk/people/s.sherwin/
participants (2)
-
Sherwin, Spencer J
-
Thibaut Appel