Merge pull request #2167 from Vsevak/master (66271448) · Commits · 郑智淋 / lammps

cmake/Modules/Packages/GPU.cmake

+11 −4

Original line number	Original line	Diff line number	Diff line
	@@ -268,10 +268,17 @@ elseif(GPU_API STREQUAL "HIP")
	if(HIP_PLATFORM STREQUAL "hcc")		if(HIP_PLATFORM STREQUAL "hcc")
	configure_file(${CU_FILE} ${CU_CPP_FILE} COPYONLY)		configure_file(${CU_FILE} ${CU_CPP_FILE} COPYONLY)

			if(HIP_COMPILER STREQUAL "clang")
			add_custom_command(OUTPUT ${CUBIN_FILE}
			VERBATIM COMMAND ${HIP_HIPCC_EXECUTABLE} --genco --offload-arch=${HIP_ARCH} -O3 -ffast-math -DUSE_HIP -D_${GPU_PREC_SETTING} -I${LAMMPS_LIB_SOURCE_DIR}/gpu -o ${CUBIN_FILE} ${CU_CPP_FILE}
			DEPENDS ${CU_CPP_FILE}
			COMMENT "Generating ${CU_NAME}.cubin")
			else()
	add_custom_command(OUTPUT ${CUBIN_FILE}		add_custom_command(OUTPUT ${CUBIN_FILE}
	VERBATIM COMMAND ${HIP_HIPCC_EXECUTABLE} --genco -t="${HIP_ARCH}" -f=\"-O3 -ffast-math -DUSE_HIP -D_${GPU_PREC_SETTING} -I${LAMMPS_LIB_SOURCE_DIR}/gpu\" -o ${CUBIN_FILE} ${CU_CPP_FILE}		VERBATIM COMMAND ${HIP_HIPCC_EXECUTABLE} --genco -t="${HIP_ARCH}" -f=\"-O3 -ffast-math -DUSE_HIP -D_${GPU_PREC_SETTING} -I${LAMMPS_LIB_SOURCE_DIR}/gpu\" -o ${CUBIN_FILE} ${CU_CPP_FILE}
	DEPENDS ${CU_CPP_FILE}		DEPENDS ${CU_CPP_FILE}
	COMMENT "Generating ${CU_NAME}.cubin")		COMMENT "Generating ${CU_NAME}.cubin")
			endif()
	elseif(HIP_PLATFORM STREQUAL "nvcc")		elseif(HIP_PLATFORM STREQUAL "nvcc")
	add_custom_command(OUTPUT ${CUBIN_FILE}		add_custom_command(OUTPUT ${CUBIN_FILE}
	VERBATIM COMMAND ${HIP_HIPCC_EXECUTABLE} --fatbin --use_fast_math -DUSE_HIP -D_${GPU_PREC_SETTING} ${HIP_CUDA_GENCODE} -I${LAMMPS_LIB_SOURCE_DIR}/gpu -o ${CUBIN_FILE} ${CU_FILE}		VERBATIM COMMAND ${HIP_HIPCC_EXECUTABLE} --fatbin --use_fast_math -DUSE_HIP -D_${GPU_PREC_SETTING} ${HIP_CUDA_GENCODE} -I${LAMMPS_LIB_SOURCE_DIR}/gpu -o ${CUBIN_FILE} ${CU_FILE}

doc/src/Speed_gpu.rst

+1 −1

Original line number	Original line	Diff line number	Diff line
	@@ -88,7 +88,7 @@ GPUs/node to 1.

	Using the "-pk" switch explicitly allows for setting of the number of		Using the "-pk" switch explicitly allows for setting of the number of
	GPUs/node to use and additional options. Its syntax is the same as		GPUs/node to use and additional options. Its syntax is the same as
	same as the "package gpu" command. See the :doc:`package <package>`		the "package gpu" command. See the :doc:`package <package>`
	command doc page for details, including the default values used for		command doc page for details, including the default values used for
	all its options if it is not specified.		all its options if it is not specified.

lib/gpu/Makefile.hip

+11 −5

Original line number	Original line	Diff line number	Diff line
	@@ -33,10 +33,10 @@ HIP_HOST_OPTS += -DMPI_GERYON -DUCL_NO_EXIT
	# this settings should match LAMMPS Makefile		# this settings should match LAMMPS Makefile
	MPI_COMP_OPTS = $(shell mpicxx --showme:compile)		MPI_COMP_OPTS = $(shell mpicxx --showme:compile)
	MPI_LINK_OPTS = $(shell mpicxx --showme:link)		MPI_LINK_OPTS = $(shell mpicxx --showme:link)
	#MPI_COMP_OPTS += -I/usr/include/mpi -DMPICH_IGNORE_CXX_SEEK -DOMPI_SKIP_MPICXX=1

	HIP_PATH ?= $(wildcard /opt/rocm/hip)		HIP_PATH ?= $(wildcard /opt/rocm/hip)
	HIP_PLATFORM=$(shell $(HIP_PATH)/bin/hipconfig --compiler)		HIP_PLATFORM=$(shell $(HIP_PATH)/bin/hipconfig --platform)
			HIP_COMPILER=$(shell $(HIP_PATH)/bin/hipconfig --compiler)

	ifeq (hcc,$(HIP_PLATFORM))		ifeq (hcc,$(HIP_PLATFORM))
	HIP_OPTS += -ffast-math		HIP_OPTS += -ffast-math
	@@ -48,8 +48,6 @@ else ifeq (nvcc,$(HIP_PLATFORM))
	-gencode arch=compute_50,code=[sm_50,compute_50] -gencode arch=compute_52,code=[sm_52,compute_52] -gencode arch=compute_53,code=[sm_53,compute_53]\		-gencode arch=compute_50,code=[sm_50,compute_50] -gencode arch=compute_52,code=[sm_52,compute_52] -gencode arch=compute_53,code=[sm_53,compute_53]\
	-gencode arch=compute_60,code=[sm_60,compute_60] -gencode arch=compute_61,code=[sm_61,compute_61] -gencode arch=compute_62,code=[sm_62,compute_62]\		-gencode arch=compute_60,code=[sm_60,compute_60] -gencode arch=compute_61,code=[sm_61,compute_61] -gencode arch=compute_62,code=[sm_62,compute_62]\
	-gencode arch=compute_70,code=[sm_70,compute_70] -gencode arch=compute_72,code=[sm_72,compute_72] -gencode arch=compute_75,code=[sm_75,compute_75]		-gencode arch=compute_70,code=[sm_70,compute_70] -gencode arch=compute_72,code=[sm_72,compute_72] -gencode arch=compute_75,code=[sm_75,compute_75]
	else
	$(error Specify HIP platform using 'export HIP_PLATFORM=(hcc,nvcc)')
	endif		endif

	BIN_DIR = .		BIN_DIR = .
	@@ -66,7 +64,15 @@ BSH = /bin/sh
	HIP_OPTS += -DUSE_HIP $(HIP_PRECISION)		HIP_OPTS += -DUSE_HIP $(HIP_PRECISION)
	HIP_GPU_OPTS += $(HIP_OPTS) -I./		HIP_GPU_OPTS += $(HIP_OPTS) -I./

	ifeq (hcc,$(HIP_PLATFORM))		ifeq (clang,$(HIP_COMPILER))
			HIP_HOST_OPTS += -fPIC
			HIP_GPU_CC = $(HIP_PATH)/bin/hipcc --genco
			HIP_GPU_OPTS_S = --offload-arch=$(HIP_ARCH)
			HIP_GPU_OPTS_E =
			HIP_KERNEL_SUFFIX = .cpp
			HIP_LIBS_TARGET = export HCC_AMDGPU_TARGET := $(HIP_ARCH)
			export HCC_AMDGPU_TARGET := $(HIP_ARCH)
			else ifeq (hcc,$(HIP_COMPILER))
	HIP_HOST_OPTS += -fPIC		HIP_HOST_OPTS += -fPIC
	HIP_GPU_CC = $(HIP_PATH)/bin/hipcc --genco		HIP_GPU_CC = $(HIP_PATH)/bin/hipcc --genco
	HIP_GPU_OPTS_S = -t="$(HIP_ARCH)" -f=\"		HIP_GPU_OPTS_S = -t="$(HIP_ARCH)" -f=\"

lib/gpu/lal_lj_tip4p_long.cpp

+9 −9

Original line number	Original line	Diff line number	Diff line
	@@ -250,22 +250,22 @@ void LJTIP4PLongT::copy_relations_data(int n, tagint tag, int map_array,
	if (ago == 0) {		if (ago == 0) {
	hneight.zero();		hneight.zero();


	{		{
	UCL_H_Vec<tagint> host_tag_write(nall,*(this->ucl_device),UCL_WRITE_ONLY);		UCL_H_Vec<tagint> host_tag_write;
			host_tag_write.view(tag, nall, *(this->ucl_device));
	this->tag.resize_ib(nall);		this->tag.resize_ib(nall);
	for(int i=0; i<nall; ++i) host_tag_write[i] = tag[i];		ucl_copy(this->tag, host_tag_write, false);
	ucl_copy(this->tag, host_tag_write, nall, false);
	}		}

	UCL_H_Vec<int> host_write(max_same,*(this->ucl_device),UCL_WRITE_ONLY);		UCL_H_Vec<int> host_write;
			host_write.view(sametag, max_same, *(this->ucl_device));
	this->atom_sametag.resize_ib(max_same);		this->atom_sametag.resize_ib(max_same);
	for(int i=0; i<max_same; ++i) host_write[i] = sametag[i];		ucl_copy(this->atom_sametag, host_write, false);
	ucl_copy(this->atom_sametag, host_write, max_same, false);

	host_write.resize_ib(map_size);
	this->map_array.resize_ib(map_size);		this->map_array.resize_ib(map_size);
	for(int i=0; i<map_size; ++i) host_write[i] = map_array[i];		host_write.view(map_array, map_size, *(this->ucl_device));
	ucl_copy(this->map_array, host_write, map_size, false);		ucl_copy(this->map_array, host_write, false);
	}		}
	}		}

tools/singularity/ubuntu18.04_amd_rocm.def

+21 −4

Original line number	Original line	Diff line number	Diff line
	BootStrap: docker		BootStrap: docker
	From: rocm/dev-ubuntu-18.04		From: ubuntu:18.04

	%environment		%environment
	export PATH=$PATH:/opt/rocm/bin:/opt/rocm/profiler/bin:/opt/rocm/opencl/bin/x86_64		export PATH=$PATH:/opt/rocm/bin:/opt/rocm/profiler/bin:/opt/rocm/opencl/bin/x86_64
	@@ -8,7 +8,24 @@ From: rocm/dev-ubuntu-18.04
	export DEBIAN_FRONTEND=noninteractive		export DEBIAN_FRONTEND=noninteractive
	apt-get update		apt-get update
	apt-get upgrade --no-install-recommends -y		apt-get upgrade --no-install-recommends -y

			apt-get install -y --no-install-recommends curl libnuma-dev gnupg

			curl -sL http://repo.radeon.com/rocm/apt/debian/rocm.gpg.key \| apt-key add -
			printf "deb [arch=amd64] http://repo.radeon.com/rocm/apt/debian/ xenial main" > /etc/apt/sources.list.d/rocm.list

			apt-get update
			apt-get install --no-install-recommends -y \
			kmod \
			file \
			sudo \
			libelf1 \
			rocm-dev \
			rocm-libs \
			build-essential

	apt-get install --no-install-recommends -y software-properties-common		apt-get install --no-install-recommends -y software-properties-common

	apt-get install --no-install-recommends -y \		apt-get install --no-install-recommends -y \
	bc \		bc \
	build-essential \		build-essential \
	@@ -54,7 +71,6 @@ From: rocm/dev-ubuntu-18.04
	python3-pkg-resources \		python3-pkg-resources \
	python3-setuptools \		python3-setuptools \
	python3-virtualenv \		python3-virtualenv \
	rocm-libs \
	rsync \		rsync \
	ssh \		ssh \
	vim-nox \		vim-nox \
	@@ -65,11 +81,12 @@ From: rocm/dev-ubuntu-18.04
	valgrind \		valgrind \
	gdb		gdb


	export PATH=$PATH:/opt/rocm/bin:/opt/rocm/profiler/bin:/opt/rocm/opencl/bin/x86_64		export PATH=$PATH:/opt/rocm/bin:/opt/rocm/profiler/bin:/opt/rocm/opencl/bin/x86_64
	git clone -b master-rocm-3.3 https://github.com/ROCmSoftwarePlatform/hipCUB.git		git clone -b master-rocm-3.5 https://github.com/ROCmSoftwarePlatform/hipCUB.git
	mkdir hipCUB/build		mkdir hipCUB/build
	cd hipCUB/build		cd hipCUB/build
	CXX=hcc cmake -D BUILD_TEST=off ..		CXX=hipcc cmake -D BUILD_TEST=off ..
	make -j4		make -j4
	make package		make package
	make install		make install

Admin message