(28194a14) · Commits · 郑智淋 / lammps

lib/gpu/Makefile.linux

+6 −3

Original line number	Diff line number	Diff line
		@@ -17,13 +17,16 @@ CUDA_ARCH = -arch=sm_21
		# older CUDA
		#CUDA_ARCH = -arch=sm_10 -DCUDA_PRE_THREE

		CUDA_PRECISION = -D_SINGLE_SINGLE
		# system-specific settings, should match with LAMMPS_SMALLBIG (default), LAMMPS_BIGBIG or LAMMPS_SMALLSMALL
		LMP_INC = -DLAMMPS_BIGBIG

		CUDA_PRECISION = -D_SINGLE_DOUBLE
		CUDA_INCLUDE = -I$(CUDA_HOME)/include
		CUDA_LIB = -L$(CUDA_HOME)/lib64
		CUDA_OPTS = -DUNIX -O3 -Xptxas -v --use_fast_math
		CUDA_OPTS = -DUNIX -O3 -Xptxas -v --use_fast_math $(LMP_INC)

		CUDR_CPP = mpic++ -DMPI_GERYON -DUCL_NO_EXIT -DMPICH_IGNORE_CXX_SEEK
		CUDR_OPTS = -O2 # -xHost -no-prec-div -ansi-alias
		CUDR_OPTS = -O2 $(LMP_INC) # -xHost -no-prec-div -ansi-alias

		BIN_DIR = ./
		OBJ_DIR = ./

lib/gpu/Makefile.linux_opencl

+6 −2

Original line number	Diff line number	Diff line
		@@ -11,9 +11,13 @@ OCL_TUNE = -DFERMI_OCL # -- Uncomment for NVIDIA Fermi
		# OCL_TUNE = -DCYPRESS_OCL # -- Uncomment for AMD Cypress
		# OCL_TUNE = -DGENERIC_OCL # -- Uncomment for generic device

		OCL_CPP = mpic++ $(DEFAULT_DEVICE) -O3 -DMPI_GERYON -DUCL_NO_EXIT -DMPICH_IGNORE_CXX_SEEK
		# system-specific settings, should match with LAMMPS_SMALLBIG (default), LAMMPS_BIGBIG and LAMMPS_SMALLSMALL
		LMP_INC = #-DLAMMPS_BIGBIG

		OCL_INC = -I/usr/local/cuda/include # Path to CL directory
		OCL_CPP = mpic++ $(DEFAULT_DEVICE) -O3 -DMPI_GERYON -DUCL_NO_EXIT -DMPICH_IGNORE_CXX_SEEK $(LMP_INC) $(OCL_INC)
		OCL_LINK = -lOpenCL
		OCL_PREC = -D_SINGLE_SINGLE
		OCL_PREC = -D_SINGLE_DOUBLE

		BIN_DIR = ./
		OBJ_DIR = ./

lib/gpu/README

+32 −32

Original line number	Diff line number	Diff line
		@@ -82,7 +82,7 @@ and

		Brown, W.M., Masako, Y. Implementing Molecular Dynamics on Hybrid High
		Performance Computers - Three-Body Potentials. Computer Physics Communications.
		2013. In press.
		2013. 184: p. 2785–2793.

		----

		@@ -103,37 +103,33 @@ Current styles supporting GPU acceleration:
		11 eam/alloy
		12 eam/fs
		13 eam
		14 eam/lj
		15 gauss
		16 gayberne
		17 lj96/cut
		18 lj/charmm/coul/long
		19 lj/class2/coul/long
		20 lj/class2
		21 lj/cut/coul/cut
		22 lj/cut/coul/debye
		23 lj/cut/coul/dsf
		24 lj/cut/coul/long
		25 lj/cut/coul/msm
		26 lj/cut/coul/wolf/fsw
		27 lj/cut/dipole/cut
		28 lj/cut
		29 lj/cut/tgpu
		30 lj/expand
		31 lj/sdk/coul/long
		32 cg/cmm/coul/long
		33 lj/sdk
		34 cg/cmm
		35 lj/sf/dipole/sf
		36 mie/cut
		37 morse
		38 resquared
		39 soft
		40 sw
		41 table
		42 yukawa/colloid
		43 yukawa
		44 pppm
		14 gauss
		15 gayberne
		16 lj96/cut
		17 lj/charmm/coul/long
		18 lj/class2/coul/long
		19 lj/class2
		20 lj/cut/coul/cut
		21 lj/cut/coul/debye
		22 lj/cut/coul/dsf
		23 lj/cut/coul/long
		24 lj/cut/coul/msm
		25 lj/cut/dipole/cut
		26 lj/cut
		27 lj/expand
		28 lj/gromacs
		29 lj/sdk/coul/long
		30 lj/sdk
		31 lj/sf/dipole/sf
		32 mie/cut
		33 morse
		34 resquared
		35 soft
		36 sw
		37 table
		38 yukawa/colloid
		39 yukawa
		40 pppm


		MULTIPLE LAMMPS PROCESSES
		@@ -200,6 +196,10 @@ NOTE: The lj/cut/coul/long/gpu, cg/cmm/coul/long/gpu, coul/long/gpu,
		lj/charmm/coul/long/gpu and pppm/gpu styles will only be installed
		if the KSPACE package has been installed.

		NOTE: The system-specific setting LAMMPS_SMALLBIG (default), LAMMPS_BIGBIG,
		or LAMMPS_SMALLSMALL if specified when building LAMMPS (i.e. in
		src/MAKE/Makefile.foo) should be consistent with that specified
		when building libgpu.a (i.e. by LMP_INC in the lib/gpu/Makefile.bar).

		EXAMPLE BUILD PROCESS
		--------------------------------

lib/gpu/lal_atom.h

+1 −1

Original line number	Diff line number	Diff line
		@@ -373,7 +373,7 @@ class Atom {
		/// Cell list identifiers for device nbor builds
		UCL_D_Vec<int> dev_particle_id;
		/// Atom tag information for device nbor builds
		UCL_D_Vec<int> dev_tag;
		UCL_D_Vec<tagint> dev_tag;

		/// Cell list identifiers for hybrid nbor builds
		UCL_H_Vec<int> host_cell_id;

lib/gpu/lal_base_atomic.cpp

+4 −4

Original line number	Diff line number	Diff line
		@@ -150,8 +150,8 @@ template <class numtyp, class acctyp>
		inline void BaseAtomicT::build_nbor_list(const int inum, const int host_inum,
		const int nall, double **host_x,
		int host_type, double sublo,
		double subhi, int tag,
		int nspecial, int special,
		double subhi, tagint tag,
		int nspecial, tagint special,
		bool &success) {
		success=true;
		resize_atom(inum,nall,success);
		@@ -216,8 +216,8 @@ void BaseAtomicT::compute(const int f_ago, const int inum_full,
		template <class numtyp, class acctyp>
		int ** BaseAtomicT::compute(const int ago, const int inum_full,
		const int nall, double *host_x, int host_type,
		double sublo, double subhi, int *tag,
		int nspecial, int special, const bool eflag,
		double sublo, double subhi, tagint *tag,
		int nspecial, tagint special, const bool eflag,
		const bool vflag, const bool eatom,
		const bool vatom, int &host_start,
		int ilist, int jnum,

Admin message