User Tools

Site Tools


optimization

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
optimization [2023/03/09 22:07]
root T
optimization [2023/03/09 22:16] (current)
root
Line 8: Line 8:
  
 * Intel proprietary: icc/icpc/ifort * Intel proprietary: icc/icpc/ifort
- 
 * Intel oneAPI Clang/LLVM based: icx/icpx/ifx * Intel oneAPI Clang/LLVM based: icx/icpx/ifx
- 
 * AMD Clang/LLVM based: clang/clang++/flang * AMD Clang/LLVM based: clang/clang++/flang
- 
 * NVidia PGI based: pgcc/pgc++/pgf90 * NVidia PGI based: pgcc/pgc++/pgf90
- 
 * GNU: gcc/g++/gfortran * GNU: gcc/g++/gfortran
- 
 * Also base Clang/LLVM is available, but not necessary with two optimized versions * Also base Clang/LLVM is available, but not necessary with two optimized versions
  
Line 51: Line 46:
 == OpenMP == == OpenMP ==
  
-The automated parallelization is not usually very good, so it requires directives in the code for good performance+The automated parallelization is not usually very good, so it requires directives in the code for good performance.  But generally a compiler option is necessary to enable OpenMP.
  
 * icc -qopenmp -parallel * icc -qopenmp -parallel
Line 66: Line 61:
 These include These include
  
-* BLAS and LAPACK:  Intel MKL and OpenBLAS +* BLAS and LAPACK: Intel MKL, AMD AOCL, OpenBLAS 
-* FFT;              FFTW and MKL +* FFTFFTWMKL, AOCL 
-* Solvers:          MKL, Scalapack, Elsa, PetSC, and others+* Solvers: AOCL, MKL, Scalapack, Elpa, PetSC, and others 
 +* Random Numbers: AOCL, MKL
  
 ==MPI Versions== ==MPI Versions==
  
-* Intel MPI: usually the easiest as it has run-time interfaces with multiple compilers +* Intel MPI: usually the easiest as it has run-time interfaces for multiple compilers 
-* Open MPI: often the fastest, must be compiled with the compiler +* Open MPI: often the fastest, must be compiled with the compiler in use 
-* MVAPICH: (MPICH for Infiniband): sometimes the fastest, must be compiled with the compiler+* MVAPICH: (MPICH for Infiniband): sometimes the fastest, must be compiled with the compiler in use
  
  
  
optimization.1678399679.txt.gz · Last modified: 2023/03/09 22:07 by root