diff --git a/doc/src/Manual_build.txt b/doc/src/Manual_build.txt index 747a55a7296d0ab206c0119ecb7a85dd5e8e7a36..a6b881cb79708c37760dc482cac2a7017342d7df 100644 --- a/doc/src/Manual_build.txt +++ b/doc/src/Manual_build.txt @@ -81,7 +81,7 @@ sudo yum install python3-virtualenv :pre Fedora (since version 22) :h4 -sudo dnf install python3-virtualenv pre +sudo dnf install python3-virtualenv :pre MacOS X :h4 diff --git a/doc/src/Packages_details.txt b/doc/src/Packages_details.txt index 5ab85a80c842c8a0a1672af8140e27e0be33d0e8..892774be383f379ade01ca6265e69ea6b6f4bcfc 100644 --- a/doc/src/Packages_details.txt +++ b/doc/src/Packages_details.txt @@ -382,6 +382,11 @@ switches"_Run_options.html. Also see the "GPU"_#PKG-GPU, "OPT"_#PKG-OPT, have styles optimized for CPUs, KNLs, and GPUs. You must have a C++11 compatible compiler to use this package. +KOKKOS makes extensive use of advanced C++ features, which can +expose compiler bugs, especially when compiling for maximum +performance at high optimization levels. Please see the file +lib/kokkos/README for a list of compilers and their respective +platforms, that are known to work. [Authors:] The KOKKOS package was created primarily by Christian Trott and Stan Moore (Sandia), with contributions from other folks as well. diff --git a/doc/src/Packages_standard.txt b/doc/src/Packages_standard.txt index a79caae141c68a4a0f5bcd6383658e20c3a307a9..55d0d616f417a5511898b98f61d9b7b0b348edd9 100644 --- a/doc/src/Packages_standard.txt +++ b/doc/src/Packages_standard.txt @@ -25,42 +25,42 @@ refers to the examples/USER/atc directory. The "Library" column indicates whether an extra library is needed to build and use the package: -dash = no library +no = no library sys = system library: you likely have it on your machine int = internal library: provided with LAMMPS, but you may need to build it ext = external library: you will need to download and install it on your machine :ul Package, Description, Doc page, Example, Library -"ASPHERE"_Packages_details.html#PKG-ASPHERE, aspherical particle models, "Howto spherical"_Howto_spherical.html, ellipse, - -"BODY"_Packages_details.html#PKG-BODY, body-style particles, "Howto body"_Howto_body.html, body, - -"CLASS2"_Packages_details.html#PKG-CLASS2, class 2 force fields, "pair_style lj/class2"_pair_class2.html, -, - -"COLLOID"_Packages_details.html#PKG-COLLOID, colloidal particles, "atom_style colloid"_atom_style.html, colloid, - -"COMPRESS"_Packages_details.html#PKG-COMPRESS, I/O compression, "dump */gz"_dump.html, -, sys -"CORESHELL"_Packages_details.html#PKG-CORESHELL, adiabatic core/shell model, "Howto coreshell"_Howto_coreshell.html, coreshell, - -"DIPOLE"_Packages_details.html#PKG-DIPOLE, point dipole particles, "pair_style dipole/cut"_pair_dipole.html, dipole, - +"ASPHERE"_Packages_details.html#PKG-ASPHERE, aspherical particle models, "Howto spherical"_Howto_spherical.html, ellipse, no +"BODY"_Packages_details.html#PKG-BODY, body-style particles, "Howto body"_Howto_body.html, body, no +"CLASS2"_Packages_details.html#PKG-CLASS2, class 2 force fields, "pair_style lj/class2"_pair_class2.html, n/a, no +"COLLOID"_Packages_details.html#PKG-COLLOID, colloidal particles, "atom_style colloid"_atom_style.html, colloid, no +"COMPRESS"_Packages_details.html#PKG-COMPRESS, I/O compression, "dump */gz"_dump.html, n/a, sys +"CORESHELL"_Packages_details.html#PKG-CORESHELL, adiabatic core/shell model, "Howto coreshell"_Howto_coreshell.html, coreshell, no +"DIPOLE"_Packages_details.html#PKG-DIPOLE, point dipole particles, "pair_style dipole/cut"_pair_dipole.html, dipole, no "GPU"_Packages_details.html#PKG-GPU, GPU-enabled styles, "Section gpu"_Speed_gpu.html, "Benchmarks"_http://lammps.sandia.gov/bench.html, int -"GRANULAR"_Packages_details.html#PKG-GRANULAR, granular systems, "Howto granular"_Howto_granular.html, pour, - +"GRANULAR"_Packages_details.html#PKG-GRANULAR, granular systems, "Howto granular"_Howto_granular.html, pour, no "KIM"_Packages_details.html#PKG-KIM, OpenKIM wrapper, "pair_style kim"_pair_kim.html, kim, ext -"KOKKOS"_Packages_details.html#PKG-KOKKOS, Kokkos-enabled styles, "Speed kokkos"_Speed_kokkos.html, "Benchmarks"_http://lammps.sandia.gov/bench.html, - -"KSPACE"_Packages_details.html#PKG-KSPACE, long-range Coulombic solvers, "kspace_style"_kspace_style.html, peptide, - +"KOKKOS"_Packages_details.html#PKG-KOKKOS, Kokkos-enabled styles, "Speed kokkos"_Speed_kokkos.html, "Benchmarks"_http://lammps.sandia.gov/bench.html, no +"KSPACE"_Packages_details.html#PKG-KSPACE, long-range Coulombic solvers, "kspace_style"_kspace_style.html, peptide, no "LATTE"_Packages_details.html#PKG-LATTE, quantum DFTB forces via LATTE, "fix latte"_fix_latte.html, latte, ext -"MANYBODY"_Packages_details.html#PKG-MANYBODY, many-body potentials, "pair_style tersoff"_pair_tersoff.html, shear, - -"MC"_Packages_details.html#PKG-MC, Monte Carlo options, "fix gcmc"_fix_gcmc.html, -, - +"MANYBODY"_Packages_details.html#PKG-MANYBODY, many-body potentials, "pair_style tersoff"_pair_tersoff.html, shear, no +"MC"_Packages_details.html#PKG-MC, Monte Carlo options, "fix gcmc"_fix_gcmc.html, n/a, no "MEAM"_Packages_details.html#PKG-MEAM, modified EAM potential, "pair_style meam"_pair_meam.html, meam, int -"MISC"_Packages_details.html#PKG-MISC, miscellanous single-file commands, -, -, - -"MOLECULE"_Packages_details.html#PKG-MOLECULE, molecular system force fields, "Howto bioFF"_Howto_bioFF.html, peptide, - -"MPIIO"_Packages_details.html#PKG-MPIIO, MPI parallel I/O dump and restart, "dump"_dump.html, -, - +"MISC"_Packages_details.html#PKG-MISC, miscellanous single-file commands, n/a, no, no +"MOLECULE"_Packages_details.html#PKG-MOLECULE, molecular system force fields, "Howto bioFF"_Howto_bioFF.html, peptide, no +"MPIIO"_Packages_details.html#PKG-MPIIO, MPI parallel I/O dump and restart, "dump"_dump.html, n/a, no "MSCG"_Packages_details.html#PKG-MSCG, multi-scale coarse-graining wrapper, "fix mscg"_fix_mscg.html, mscg, ext -"OPT"_Packages_details.html#PKG-OPT, optimized pair styles, "Speed opt"_Speed_opt.html, "Benchmarks"_http://lammps.sandia.gov/bench.html, - -"PERI"_Packages_details.html#PKG-PERI, Peridynamics models, "pair_style peri"_pair_peri.html, peri, - +"OPT"_Packages_details.html#PKG-OPT, optimized pair styles, "Speed opt"_Speed_opt.html, "Benchmarks"_http://lammps.sandia.gov/bench.html, no +"PERI"_Packages_details.html#PKG-PERI, Peridynamics models, "pair_style peri"_pair_peri.html, peri, no "POEMS"_Packages_details.html#PKG-POEMS, coupled rigid body motion, "fix poems"_fix_poems.html, rigid, int "PYTHON"_Packages_details.html#PKG-PYTHON, embed Python code in an input script, "python"_python.html, python, sys -"QEQ"_Packages_details.html#PKG-QEQ, QEq charge equilibration, "fix qeq"_fix_qeq.html, qeq, - +"QEQ"_Packages_details.html#PKG-QEQ, QEq charge equilibration, "fix qeq"_fix_qeq.html, qeq, no "REAX"_Packages_details.html#PKG-REAX, ReaxFF potential (Fortran), "pair_style reax"_pair_reax.html, reax, int -"REPLICA"_Packages_details.html#PKG-REPLICA2, multi-replica methods, "Howto replica"_Howto_replica.html, tad, - -"RIGID"_Packages_details.html#PKG-RIGID, rigid bodies and constraints, "fix rigid"_fix_rigid.html, rigid, - -"SHOCK"_Packages_details.html#PKG-SHOCK, shock loading methods, "fix msst"_fix_msst.html, -, - -"SNAP"_Packages_details.html#PKG-SNAP, quantum-fitted potential, "pair_style snap"_pair_snap.html, snap, - -"SPIN"_Packages_details.html#PKG-SPIN, magnetic atomic spin dynamics, "Howto spins"_Howto_spins.html, SPIN, - -"SRD"_Packages_details.html#PKG-SRD, stochastic rotation dynamics, "fix srd"_fix_srd.html, srd, - -"VORONOI"_Packages_details.html#PKG-VORONOI, Voronoi tesselation, "compute voronoi/atom"_compute_voronoi_atom.html, -, ext :tb(ea=c,ca1=l) +"REPLICA"_Packages_details.html#PKG-REPLICA2, multi-replica methods, "Howto replica"_Howto_replica.html, tad, no +"RIGID"_Packages_details.html#PKG-RIGID, rigid bodies and constraints, "fix rigid"_fix_rigid.html, rigid, no +"SHOCK"_Packages_details.html#PKG-SHOCK, shock loading methods, "fix msst"_fix_msst.html, n/a, no +"SNAP"_Packages_details.html#PKG-SNAP, quantum-fitted potential, "pair_style snap"_pair_snap.html, snap, no +"SPIN"_Packages_details.html#PKG-SPIN, magnetic atomic spin dynamics, "Howto spins"_Howto_spins.html, SPIN, no +"SRD"_Packages_details.html#PKG-SRD, stochastic rotation dynamics, "fix srd"_fix_srd.html, srd, no +"VORONOI"_Packages_details.html#PKG-VORONOI, Voronoi tesselation, "compute voronoi/atom"_compute_voronoi_atom.html, n/a, ext :tb(ea=c,ca1=l) diff --git a/doc/src/Packages_user.txt b/doc/src/Packages_user.txt index 7157ebead0ffaef0ec03508272f4a58b698719da..c1a52fd0d0d864493d49f849aa5c531fcb1cd40f 100644 --- a/doc/src/Packages_user.txt +++ b/doc/src/Packages_user.txt @@ -32,7 +32,7 @@ refers to the examples/USER/atc directory. The "Library" column indicates whether an extra library is needed to build and use the package: -dash = no library +no = no library sys = system library: you likely have it on your machine int = internal library: provided with LAMMPS, but you may need to build it ext = external library: you will need to download and install it on your machine :ul @@ -40,35 +40,35 @@ ext = external library: you will need to download and install it on your machine Package, Description, Doc page, Example, Library "USER-ATC"_Packages_details.html#PKG-USER-ATC, atom-to-continuum coupling, "fix atc"_fix_atc.html, USER/atc, int "USER-AWPMD"_Packages_details.html#PKG-USER-AWPMD, wave-packet MD, "pair_style awpmd/cut"_pair_awpmd.html, USER/awpmd, int -"USER-BOCS"_Packages_details.html#PKG-USER-BOCS, BOCS bottom up coarse graining, "fix bocs"_fix_bocs.html, USER/bocs, - -"USER-CGDNA"_Packages_details.html#PKG-USER-CGDNA, coarse-grained DNA force fields, src/USER-CGDNA/README, USER/cgdna, - -"USER-CGSDK"_Packages_details.html#PKG-USER-CGSDK, SDK coarse-graining model, "pair_style lj/sdk"_pair_sdk.html, USER/cgsdk, - +"USER-BOCS"_Packages_details.html#PKG-USER-BOCS, BOCS bottom up coarse graining, "fix bocs"_fix_bocs.html, USER/bocs, no +"USER-CGDNA"_Packages_details.html#PKG-USER-CGDNA, coarse-grained DNA force fields, src/USER-CGDNA/README, USER/cgdna, no +"USER-CGSDK"_Packages_details.html#PKG-USER-CGSDK, SDK coarse-graining model, "pair_style lj/sdk"_pair_sdk.html, USER/cgsdk, no "USER-COLVARS"_Packages_details.html#PKG-USER-COLVARS, collective variables library, "fix colvars"_fix_colvars.html, USER/colvars, int -"USER-DIFFRACTION"_Packages_details.html#PKG-USER-DIFFRACTION, virtual x-ray and electron diffraction,"compute xrd"_compute_xrd.html, USER/diffraction, - -"USER-DPD"_Packages_details.html#PKG-USER-DPD, reactive dissipative particle dynamics, src/USER-DPD/README, USER/dpd, - -"USER-DRUDE"_Packages_details.html#PKG-USER-DRUDE, Drude oscillators, "Howto drude"_Howto_drude.html, USER/drude, - -"USER-EFF"_Packages_details.html#PKG-USER-EFF, electron force field,"pair_style eff/cut"_pair_eff.html, USER/eff, - -"USER-FEP"_Packages_details.html#PKG-USER-FEP, free energy perturbation,"compute fep"_compute_fep.html, USER/fep, - -"USER-H5MD"_Packages_details.html#PKG-USER-H5MD, dump output via HDF5,"dump h5md"_dump_h5md.html, -, ext -"USER-INTEL"_Packages_details.html#PKG-USER-INTEL, optimized Intel CPU and KNL styles,"Speed intel"_Speed_intel.html, "Benchmarks"_http://lammps.sandia.gov/bench.html, - -"USER-LB"_Packages_details.html#PKG-USER-LB, Lattice Boltzmann fluid,"fix lb/fluid"_fix_lb_fluid.html, USER/lb, - -"USER-MANIFOLD"_Packages_details.html#PKG-USER-MANIFOLD, motion on 2d surfaces,"fix manifoldforce"_fix_manifoldforce.html, USER/manifold, - -"USER-MEAMC"_Packages_details.html#PKG-USER-MEAMC, modified EAM potential (C++), "pair_style meam/c"_pair_meam.html, meam, - -"USER-MESO"_Packages_details.html#PKG-USER-MESO, mesoscale DPD models, "pair_style edpd"_pair_meso.html, USER/meso, - -"USER-MGPT"_Packages_details.html#PKG-USER-MGPT, fast MGPT multi-ion potentials, "pair_style mgpt"_pair_mgpt.html, USER/mgpt, - -"USER-MISC"_Packages_details.html#PKG-USER-MISC, single-file contributions, USER-MISC/README, USER/misc, - -"USER-MOFFF"_Packages_details.html#PKG-USER-MOFFF, styles for "MOF-FF"_MOFplus force field, "pair_style buck6d/coul/gauss"_pair_buck6d_coul_gauss.html, USER/mofff, - -"USER-MOLFILE"_Packages_details.html#PKG-USER-MOLFILE, "VMD"_vmd_home molfile plug-ins,"dump molfile"_dump_molfile.html, -, ext -"USER-NETCDF"_Packages_details.html#PKG-USER-NETCDF, dump output via NetCDF,"dump netcdf"_dump_netcdf.html, -, ext -"USER-OMP"_Packages_details.html#PKG-USER-OMP, OpenMP-enabled styles,"Speed omp"_Speed_omp.html, "Benchmarks"_http://lammps.sandia.gov/bench.html, - -"USER-PHONON"_Packages_details.html#PKG-USER-PHONON, phonon dynamical matrix,"fix phonon"_fix_phonon.html, USER/phonon, - +"USER-DIFFRACTION"_Packages_details.html#PKG-USER-DIFFRACTION, virtual x-ray and electron diffraction,"compute xrd"_compute_xrd.html, USER/diffraction, no +"USER-DPD"_Packages_details.html#PKG-USER-DPD, reactive dissipative particle dynamics, src/USER-DPD/README, USER/dpd, no +"USER-DRUDE"_Packages_details.html#PKG-USER-DRUDE, Drude oscillators, "Howto drude"_Howto_drude.html, USER/drude, no +"USER-EFF"_Packages_details.html#PKG-USER-EFF, electron force field,"pair_style eff/cut"_pair_eff.html, USER/eff, no +"USER-FEP"_Packages_details.html#PKG-USER-FEP, free energy perturbation,"compute fep"_compute_fep.html, USER/fep, no +"USER-H5MD"_Packages_details.html#PKG-USER-H5MD, dump output via HDF5,"dump h5md"_dump_h5md.html, n/a, ext +"USER-INTEL"_Packages_details.html#PKG-USER-INTEL, optimized Intel CPU and KNL styles,"Speed intel"_Speed_intel.html, "Benchmarks"_http://lammps.sandia.gov/bench.html, no +"USER-LB"_Packages_details.html#PKG-USER-LB, Lattice Boltzmann fluid,"fix lb/fluid"_fix_lb_fluid.html, USER/lb, no +"USER-MANIFOLD"_Packages_details.html#PKG-USER-MANIFOLD, motion on 2d surfaces,"fix manifoldforce"_fix_manifoldforce.html, USER/manifold, no +"USER-MEAMC"_Packages_details.html#PKG-USER-MEAMC, modified EAM potential (C++), "pair_style meam/c"_pair_meam.html, meam, no +"USER-MESO"_Packages_details.html#PKG-USER-MESO, mesoscale DPD models, "pair_style edpd"_pair_meso.html, USER/meso, no +"USER-MGPT"_Packages_details.html#PKG-USER-MGPT, fast MGPT multi-ion potentials, "pair_style mgpt"_pair_mgpt.html, USER/mgpt, no +"USER-MISC"_Packages_details.html#PKG-USER-MISC, single-file contributions, USER-MISC/README, USER/misc, no +"USER-MOFFF"_Packages_details.html#PKG-USER-MOFFF, styles for "MOF-FF"_MOFplus force field, "pair_style buck6d/coul/gauss"_pair_buck6d_coul_gauss.html, USER/mofff, no +"USER-MOLFILE"_Packages_details.html#PKG-USER-MOLFILE, "VMD"_vmd_home molfile plug-ins,"dump molfile"_dump_molfile.html, n/a, ext +"USER-NETCDF"_Packages_details.html#PKG-USER-NETCDF, dump output via NetCDF,"dump netcdf"_dump_netcdf.html, n/a, ext +"USER-OMP"_Packages_details.html#PKG-USER-OMP, OpenMP-enabled styles,"Speed omp"_Speed_omp.html, "Benchmarks"_http://lammps.sandia.gov/bench.html, no +"USER-PHONON"_Packages_details.html#PKG-USER-PHONON, phonon dynamical matrix,"fix phonon"_fix_phonon.html, USER/phonon, no "USER-QMMM"_Packages_details.html#PKG-USER-QMMM, QM/MM coupling,"fix qmmm"_fix_qmmm.html, USER/qmmm, ext -"USER-QTB"_Packages_details.html#PKG-USER-QTB, quantum nuclear effects,"fix qtb"_fix_qtb.html "fix qbmsst"_fix_qbmsst.html, qtb, - +"USER-QTB"_Packages_details.html#PKG-USER-QTB, quantum nuclear effects,"fix qtb"_fix_qtb.html "fix qbmsst"_fix_qbmsst.html, qtb, no "USER-QUIP"_Packages_details.html#PKG-USER-QUIP, QUIP/libatoms interface,"pair_style quip"_pair_quip.html, USER/quip, ext -"USER-REAXC"_Packages_details.html#PKG-USER-REAXC, ReaxFF potential (C/C++) ,"pair_style reaxc"_pair_reaxc.html, reax, - +"USER-REAXC"_Packages_details.html#PKG-USER-REAXC, ReaxFF potential (C/C++) ,"pair_style reaxc"_pair_reaxc.html, reax, no "USER-SMD"_Packages_details.html#PKG-USER-SMD, smoothed Mach dynamics,"SMD User Guide"_PDF/SMD_LAMMPS_userguide.pdf, USER/smd, ext -"USER-SMTBQ"_Packages_details.html#PKG-USER-SMTBQ, second moment tight binding QEq potential,"pair_style smtbq"_pair_smtbq.html, USER/smtbq, - -"USER-SPH"_Packages_details.html#PKG-USER-SPH, smoothed particle hydrodynamics,"SPH User Guide"_PDF/SPH_LAMMPS_userguide.pdf, USER/sph, - -"USER-TALLY"_Packages_details.html#PKG-USER-TALLY, pairwise tally computes,"compute XXX/tally"_compute_tally.html, USER/tally, - -"USER-UEF"_Packages_details.html#PKG-USER-UEF, extensional flow,"fix nvt/uef"_fix_nh_uef.html, USER/uef, - -"USER-VTK"_Packages_details.html#PKG-USER-VTK, dump output via VTK, "compute vtk"_dump_vtk.html, -, ext :tb(ea=c,ca1=l) +"USER-SMTBQ"_Packages_details.html#PKG-USER-SMTBQ, second moment tight binding QEq potential,"pair_style smtbq"_pair_smtbq.html, USER/smtbq, no +"USER-SPH"_Packages_details.html#PKG-USER-SPH, smoothed particle hydrodynamics,"SPH User Guide"_PDF/SPH_LAMMPS_userguide.pdf, USER/sph, no +"USER-TALLY"_Packages_details.html#PKG-USER-TALLY, pairwise tally computes,"compute XXX/tally"_compute_tally.html, USER/tally, no +"USER-UEF"_Packages_details.html#PKG-USER-UEF, extensional flow,"fix nvt/uef"_fix_nh_uef.html, USER/uef, no +"USER-VTK"_Packages_details.html#PKG-USER-VTK, dump output via VTK, "compute vtk"_dump_vtk.html, n/a, ext :tb(ea=c,ca1=l) diff --git a/doc/src/Speed_compare.txt b/doc/src/Speed_compare.txt index 1a17b39c79b672e43062770253ae5e82fc1f998d..c93407515e5cebe6b8e278c62ea36b731c2b121b 100644 --- a/doc/src/Speed_compare.txt +++ b/doc/src/Speed_compare.txt @@ -9,65 +9,108 @@ Documentation"_ld - "LAMMPS Commands"_lc :c Comparison of various accelerator packages :h3 -NOTE: this section still needs to be re-worked with additional KOKKOS -and USER-INTEL information. - The next section compares and contrasts the various accelerator options, since there are multiple ways to perform OpenMP threading, -run on GPUs, and run on Intel Xeon Phi coprocessors. +run on GPUs, optimize for vector units on CPUs and run on Intel +Xeon Phi (co-)processors. -All 3 of these packages accelerate a LAMMPS calculation using NVIDIA -hardware, but they do it in different ways. +All of these packages can accelerate a LAMMPS calculation taking +advantage of hardware features, but they do it in different ways +and acceleration is not always guaranteed. As a consequence, for a particular simulation on specific hardware, -one package may be faster than the other. We give guidelines below, -but the best way to determine which package is faster for your input -script is to try both of them on your machine. See the benchmarking +one package may be faster than the other. We give some guidelines +below, but the best way to determine which package is faster for your +input script is to try multiple of them on your machine and experiment +with available performance tuning settings. See the benchmarking section below for examples where this has been done. [Guidelines for using each package optimally:] -The GPU package allows you to assign multiple CPUs (cores) to a single -GPU (a common configuration for "hybrid" nodes that contain multicore -CPU(s) and GPU(s)) and works effectively in this mode. :ulb,l - -The GPU package moves per-atom data (coordinates, forces) -back-and-forth between the CPU and GPU every timestep. The -KOKKOS/CUDA package only does this on timesteps when a CPU calculation -is required (e.g. to invoke a fix or compute that is non-GPU-ized). -Hence, if you can formulate your input script to only use GPU-ized -fixes and computes, and avoid doing I/O too often (thermo output, dump -file snapshots, restart files), then the data transfer cost of the -KOKKOS/CUDA package can be very low, causing it to run faster than the -GPU package. :l - -The GPU package is often faster than the KOKKOS/CUDA package, if the -number of atoms per GPU is smaller. The crossover point, in terms of -atoms/GPU at which the KOKKOS/CUDA package becomes faster depends -strongly on the pair style. For example, for a simple Lennard Jones +Both, the GPU and the KOKKOS package allows you to assign multiple +MPI ranks (= CPU cores) to the same GPU. For the GPU package, this +can lead to a speedup through better utilization of the GPU (by +overlapping computation and data transfer) and more efficient +computation of the non-GPU accelerated parts of LAMMPS through MPI +parallelization, as all system data is maintained and updated on +the host. For KOKKOS, there is less to no benefit from this, due +to its different memory management model, which tries to retain +data on the GPU. + :ulb,l + +The GPU package moves per-atom data (coordinates, forces, and +(optionally) neighbor list data, if not computed on the GPU) between +the CPU and GPU at every timestep. The KOKKOS/CUDA package only does +this on timesteps when a CPU calculation is required (e.g. to invoke +a fix or compute that is non-GPU-ized). Hence, if you can formulate +your input script to only use GPU-ized fixes and computes, and avoid +doing I/O too often (thermo output, dump file snapshots, restart files), +then the data transfer cost of the KOKKOS/CUDA package can be very low, +causing it to run faster than the GPU package. :l + +The GPU package is often faster than the KOKKOS/CUDA package, when the +number of atoms per GPU is on the smaller side. The crossover point, +in terms of atoms/GPU at which the KOKKOS/CUDA package becomes faster +depends strongly on the pair style. For example, for a simple Lennard Jones system the crossover (in single precision) is often about 50K-100K atoms per GPU. When performing double precision calculations the crossover point can be significantly smaller. :l -Both packages compute bonded interactions (bonds, angles, etc) on the -CPU. If the GPU package is running with several MPI processes +Both KOKKOS and GPU package compute bonded interactions (bonds, angles, +etc) on the CPU. If the GPU package is running with several MPI processes assigned to one GPU, the cost of computing the bonded interactions is -spread across more CPUs and hence the GPU package can run faster. :l - -When using the GPU package with multiple CPUs assigned to one GPU, its -performance depends to some extent on high bandwidth between the CPUs -and the GPU. Hence its performance is affected if full 16 PCIe lanes -are not available for each GPU. In HPC environments this can be the -case if S2050/70 servers are used, where two devices generally share -one PCIe 2.0 16x slot. Also many multi-GPU mainboards do not provide -full 16 lanes to each of the PCIe 2.0 16x slots. :l +spread across more CPUs and hence the GPU package can run faster in these +cases. :l + +When using LAMMPS with multiple MPI ranks assigned to the same GPU, its +performance depends to some extent on the available bandwidth between +the CPUs and the GPU. This can differ significantly based on the +available bus technology, capability of the host CPU and mainboard, +the wiring of the buses and whether switches are used to increase the +number of available bus slots, or if GPUs are housed in an external +enclosure. This can become quite complex. :l + +To achieve significant acceleration through GPUs, both KOKKOS and GPU +package require capable GPUs with fast on-device memory and efficient +data transfer rates. This requests capable upper mid-level to high-end +(desktop) GPUs. Using lower performance GPUs (e.g. on laptops) may +result in a slowdown instead. :l + +For the GPU package, specifically when running in parallel with MPI, +if it often more efficient to exclude the PPPM kspace style from GPU +acceleration and instead run it - concurrently with a GPU accelerated +pair style - on the CPU. This can often be easily achieved with placing +a {suffix off} command before and a {suffix on} command after the +{kspace_style pppm} command. :l + +The KOKKOS/OpenMP and USER-OMP package have different thread management +strategies, which should result in USER-OMP being more efficient for a +small number of threads with increasing overhead as the number of threads +per MPI rank grows. The KOKKOS/OpenMP kernels have less overhead in that +case, but have lower performance with few threads. :l + +The USER-INTEL package contains many options and settings for achieving +additional performance on Intel hardware (CPU and accelerator cards), but +to unlock this potential, an Intel compiler is required. The package code +will compile with GNU gcc, but it will not be as efficient. :l :ule -[Differences between the two packages:] +[Differences between the GPU and KOKKOS packages:] -The GPU package accelerates only pair force, neighbor list, and PPPM -calculations. :ulb,l +The GPU package accelerates only pair force, neighbor list, and (parts +of) PPPM calculations. The KOKKOS package attempts to run most of the +calculation on the GPU, but can transparently support non-accelerated +code (with a performance penalty due to having data transfers between +host and GPU). :ulb,l The GPU package requires neighbor lists to be built on the CPU when using exclusion lists, hybrid pair styles, or a triclinic simulation box. :l + +The GPU package can be compiled for CUDA or OpenCL and thus supports +both, Nvidia and AMD GPUs well. On Nvidia hardware, using CUDA is typically +resulting in equal or better performance over OpenCL. :l + +OpenCL in the GPU package does theoretically also support Intel CPUs or +Intel Xeon Phi, but the native support for those in KOKKOS (or USER-INTEL) +is superior. :l :ule diff --git a/src/USER-COLVARS/fix_colvars.h b/src/USER-COLVARS/fix_colvars.h index 509eca5de35892b3094cf5749b7d6d088d7b0d18..3029ba9db541e095af265657d5f940998cce224f 100644 --- a/src/USER-COLVARS/fix_colvars.h +++ b/src/USER-COLVARS/fix_colvars.h @@ -34,7 +34,6 @@ FixStyle(colvars,FixColvars) #define LMP_FIX_COLVARS_H #include "fix.h" -#include <vector> // forward declaration class colvarproxy_lammps; @@ -77,13 +76,6 @@ class FixColvars : public Fix { int num_coords; // total number of atoms controlled by this fix tagint *taglist; // list of all atom IDs referenced by colvars. - // TODO get rid of these - // std::vector<cvm::atom_pos> *coords; // coordinates of colvar atoms - // std::vector<cvm::rvector> *forces; // received forces of colvar atoms - // std::vector<cvm::rvector> *oforce; // old total forces of colvar atoms - // std::vector<cvm::real> *masses; - // std::vector<cvm::real> *charges; - int nmax; // size of atom communication buffer. int size_one; // bytes per atom in communication buffer. struct commdata *comm_buf; // communication buffer