mirror of
https://github.com/NVIDIA/cuda-samples.git
synced 2026-10-23 21:28:17 +08:00
add and update samples for CUDA 11.6
This commit is contained in:
Rutwik Choughule
18
Samples/4_CUDA_Libraries/conjugateGradientMultiDeviceCG/.vscode/c_cpp_properties.json
vendored
Normal file
18
Samples/4_CUDA_Libraries/conjugateGradientMultiDeviceCG/.vscode/c_cpp_properties.json
vendored
Normal file
@@ -0,0 +1,18 @@
|
||||
{
|
||||
"configurations": [
|
||||
{
|
||||
"name": "Linux",
|
||||
"includePath": [
|
||||
"${workspaceFolder}/**",
|
||||
"${workspaceFolder}/../../../Common"
|
||||
],
|
||||
"defines": [],
|
||||
"compilerPath": "/usr/local/cuda/bin/nvcc",
|
||||
"cStandard": "gnu17",
|
||||
"cppStandard": "gnu++14",
|
||||
"intelliSenseMode": "linux-gcc-x64",
|
||||
"configurationProvider": "ms-vscode.makefile-tools"
|
||||
}
|
||||
],
|
||||
"version": 4
|
||||
}
|
||||
7
Samples/4_CUDA_Libraries/conjugateGradientMultiDeviceCG/.vscode/extensions.json
vendored
Normal file
7
Samples/4_CUDA_Libraries/conjugateGradientMultiDeviceCG/.vscode/extensions.json
vendored
Normal file
@@ -0,0 +1,7 @@
|
||||
{
|
||||
"recommendations": [
|
||||
"nvidia.nsight-vscode-edition",
|
||||
"ms-vscode.cpptools",
|
||||
"ms-vscode.makefile-tools"
|
||||
]
|
||||
}
|
||||
10
Samples/4_CUDA_Libraries/conjugateGradientMultiDeviceCG/.vscode/launch.json
vendored
Normal file
10
Samples/4_CUDA_Libraries/conjugateGradientMultiDeviceCG/.vscode/launch.json
vendored
Normal file
@@ -0,0 +1,10 @@
|
||||
{
|
||||
"configurations": [
|
||||
{
|
||||
"name": "CUDA C++: Launch",
|
||||
"type": "cuda-gdb",
|
||||
"request": "launch",
|
||||
"program": "${workspaceFolder}/conjugateGradientMultiDeviceCG"
|
||||
}
|
||||
]
|
||||
}
|
||||
15
Samples/4_CUDA_Libraries/conjugateGradientMultiDeviceCG/.vscode/tasks.json
vendored
Normal file
15
Samples/4_CUDA_Libraries/conjugateGradientMultiDeviceCG/.vscode/tasks.json
vendored
Normal file
@@ -0,0 +1,15 @@
|
||||
{
|
||||
"version": "2.0.0",
|
||||
"tasks": [
|
||||
{
|
||||
"label": "sample",
|
||||
"type": "shell",
|
||||
"command": "make dbg=1",
|
||||
"problemMatcher": ["$nvcc"],
|
||||
"group": {
|
||||
"kind": "build",
|
||||
"isDefault": true
|
||||
}
|
||||
}
|
||||
]
|
||||
}
|
||||
385
Samples/4_CUDA_Libraries/conjugateGradientMultiDeviceCG/Makefile
Normal file
385
Samples/4_CUDA_Libraries/conjugateGradientMultiDeviceCG/Makefile
Normal file
@@ -0,0 +1,385 @@
|
||||
################################################################################
|
||||
# Copyright (c) 2022, NVIDIA CORPORATION. All rights reserved.
|
||||
#
|
||||
# Redistribution and use in source and binary forms, with or without
|
||||
# modification, are permitted provided that the following conditions
|
||||
# are met:
|
||||
# * Redistributions of source code must retain the above copyright
|
||||
# notice, this list of conditions and the following disclaimer.
|
||||
# * Redistributions in binary form must reproduce the above copyright
|
||||
# notice, this list of conditions and the following disclaimer in the
|
||||
# documentation and/or other materials provided with the distribution.
|
||||
# * Neither the name of NVIDIA CORPORATION nor the names of its
|
||||
# contributors may be used to endorse or promote products derived
|
||||
# from this software without specific prior written permission.
|
||||
#
|
||||
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
||||
# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||||
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
# PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
||||
# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
||||
# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||||
# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||||
# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
||||
# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
#
|
||||
################################################################################
|
||||
#
|
||||
# Makefile project only supported on Mac OS X and Linux Platforms)
|
||||
#
|
||||
################################################################################
|
||||
|
||||
# Location of the CUDA Toolkit
|
||||
CUDA_PATH ?= /usr/local/cuda
|
||||
|
||||
##############################
|
||||
# start deprecated interface #
|
||||
##############################
|
||||
ifeq ($(x86_64),1)
|
||||
$(info WARNING - x86_64 variable has been deprecated)
|
||||
$(info WARNING - please use TARGET_ARCH=x86_64 instead)
|
||||
TARGET_ARCH ?= x86_64
|
||||
endif
|
||||
ifeq ($(ARMv7),1)
|
||||
$(info WARNING - ARMv7 variable has been deprecated)
|
||||
$(info WARNING - please use TARGET_ARCH=armv7l instead)
|
||||
TARGET_ARCH ?= armv7l
|
||||
endif
|
||||
ifeq ($(aarch64),1)
|
||||
$(info WARNING - aarch64 variable has been deprecated)
|
||||
$(info WARNING - please use TARGET_ARCH=aarch64 instead)
|
||||
TARGET_ARCH ?= aarch64
|
||||
endif
|
||||
ifeq ($(ppc64le),1)
|
||||
$(info WARNING - ppc64le variable has been deprecated)
|
||||
$(info WARNING - please use TARGET_ARCH=ppc64le instead)
|
||||
TARGET_ARCH ?= ppc64le
|
||||
endif
|
||||
ifneq ($(GCC),)
|
||||
$(info WARNING - GCC variable has been deprecated)
|
||||
$(info WARNING - please use HOST_COMPILER=$(GCC) instead)
|
||||
HOST_COMPILER ?= $(GCC)
|
||||
endif
|
||||
ifneq ($(abi),)
|
||||
$(error ERROR - abi variable has been removed)
|
||||
endif
|
||||
############################
|
||||
# end deprecated interface #
|
||||
############################
|
||||
|
||||
# architecture
|
||||
HOST_ARCH := $(shell uname -m)
|
||||
TARGET_ARCH ?= $(HOST_ARCH)
|
||||
ifneq (,$(filter $(TARGET_ARCH),x86_64 aarch64 sbsa ppc64le armv7l))
|
||||
ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
||||
ifneq (,$(filter $(TARGET_ARCH),x86_64 aarch64 sbsa ppc64le))
|
||||
TARGET_SIZE := 64
|
||||
else ifneq (,$(filter $(TARGET_ARCH),armv7l))
|
||||
TARGET_SIZE := 32
|
||||
endif
|
||||
else
|
||||
TARGET_SIZE := $(shell getconf LONG_BIT)
|
||||
endif
|
||||
else
|
||||
$(error ERROR - unsupported value $(TARGET_ARCH) for TARGET_ARCH!)
|
||||
endif
|
||||
|
||||
# sbsa and aarch64 systems look similar. Need to differentiate them at host level for now.
|
||||
ifeq ($(HOST_ARCH),aarch64)
|
||||
ifeq ($(CUDA_PATH)/targets/sbsa-linux,$(shell ls -1d $(CUDA_PATH)/targets/sbsa-linux 2>/dev/null))
|
||||
HOST_ARCH := sbsa
|
||||
TARGET_ARCH := sbsa
|
||||
endif
|
||||
endif
|
||||
|
||||
ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
||||
ifeq (,$(filter $(HOST_ARCH)-$(TARGET_ARCH),aarch64-armv7l x86_64-armv7l x86_64-aarch64 x86_64-sbsa x86_64-ppc64le))
|
||||
$(error ERROR - cross compiling from $(HOST_ARCH) to $(TARGET_ARCH) is not supported!)
|
||||
endif
|
||||
endif
|
||||
|
||||
# When on native aarch64 system with userspace of 32-bit, change TARGET_ARCH to armv7l
|
||||
ifeq ($(HOST_ARCH)-$(TARGET_ARCH)-$(TARGET_SIZE),aarch64-aarch64-32)
|
||||
TARGET_ARCH = armv7l
|
||||
endif
|
||||
|
||||
# operating system
|
||||
HOST_OS := $(shell uname -s 2>/dev/null | tr "[:upper:]" "[:lower:]")
|
||||
TARGET_OS ?= $(HOST_OS)
|
||||
ifeq (,$(filter $(TARGET_OS),linux darwin qnx android))
|
||||
$(error ERROR - unsupported value $(TARGET_OS) for TARGET_OS!)
|
||||
endif
|
||||
|
||||
# host compiler
|
||||
ifeq ($(TARGET_OS),darwin)
|
||||
ifeq ($(shell expr `xcodebuild -version | grep -i xcode | awk '{print $$2}' | cut -d'.' -f1` \>= 5),1)
|
||||
HOST_COMPILER ?= clang++
|
||||
endif
|
||||
else ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
||||
ifeq ($(HOST_ARCH)-$(TARGET_ARCH),x86_64-armv7l)
|
||||
ifeq ($(TARGET_OS),linux)
|
||||
HOST_COMPILER ?= arm-linux-gnueabihf-g++
|
||||
else ifeq ($(TARGET_OS),qnx)
|
||||
ifeq ($(QNX_HOST),)
|
||||
$(error ERROR - QNX_HOST must be passed to the QNX host toolchain)
|
||||
endif
|
||||
ifeq ($(QNX_TARGET),)
|
||||
$(error ERROR - QNX_TARGET must be passed to the QNX target toolchain)
|
||||
endif
|
||||
export QNX_HOST
|
||||
export QNX_TARGET
|
||||
HOST_COMPILER ?= $(QNX_HOST)/usr/bin/arm-unknown-nto-qnx6.6.0eabi-g++
|
||||
else ifeq ($(TARGET_OS),android)
|
||||
HOST_COMPILER ?= arm-linux-androideabi-g++
|
||||
endif
|
||||
else ifeq ($(TARGET_ARCH),aarch64)
|
||||
ifeq ($(TARGET_OS), linux)
|
||||
HOST_COMPILER ?= aarch64-linux-gnu-g++
|
||||
else ifeq ($(TARGET_OS),qnx)
|
||||
ifeq ($(QNX_HOST),)
|
||||
$(error ERROR - QNX_HOST must be passed to the QNX host toolchain)
|
||||
endif
|
||||
ifeq ($(QNX_TARGET),)
|
||||
$(error ERROR - QNX_TARGET must be passed to the QNX target toolchain)
|
||||
endif
|
||||
export QNX_HOST
|
||||
export QNX_TARGET
|
||||
HOST_COMPILER ?= $(QNX_HOST)/usr/bin/q++
|
||||
else ifeq ($(TARGET_OS), android)
|
||||
HOST_COMPILER ?= aarch64-linux-android-clang++
|
||||
endif
|
||||
else ifeq ($(TARGET_ARCH),sbsa)
|
||||
HOST_COMPILER ?= aarch64-linux-gnu-g++
|
||||
else ifeq ($(TARGET_ARCH),ppc64le)
|
||||
HOST_COMPILER ?= powerpc64le-linux-gnu-g++
|
||||
endif
|
||||
endif
|
||||
HOST_COMPILER ?= g++
|
||||
NVCC := $(CUDA_PATH)/bin/nvcc -ccbin $(HOST_COMPILER)
|
||||
|
||||
# internal flags
|
||||
NVCCFLAGS := -m${TARGET_SIZE}
|
||||
CCFLAGS :=
|
||||
LDFLAGS :=
|
||||
|
||||
# build flags
|
||||
ifeq ($(TARGET_OS),darwin)
|
||||
LDFLAGS += -rpath $(CUDA_PATH)/lib
|
||||
CCFLAGS += -arch $(HOST_ARCH)
|
||||
else ifeq ($(HOST_ARCH)-$(TARGET_ARCH)-$(TARGET_OS),x86_64-armv7l-linux)
|
||||
LDFLAGS += --dynamic-linker=/lib/ld-linux-armhf.so.3
|
||||
CCFLAGS += -mfloat-abi=hard
|
||||
else ifeq ($(TARGET_OS),android)
|
||||
LDFLAGS += -pie
|
||||
CCFLAGS += -fpie -fpic -fexceptions
|
||||
endif
|
||||
|
||||
ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-linux)
|
||||
ifneq ($(TARGET_FS),)
|
||||
GCCVERSIONLTEQ46 := $(shell expr `$(HOST_COMPILER) -dumpversion` \<= 4.6)
|
||||
ifeq ($(GCCVERSIONLTEQ46),1)
|
||||
CCFLAGS += --sysroot=$(TARGET_FS)
|
||||
endif
|
||||
LDFLAGS += --sysroot=$(TARGET_FS)
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/lib
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib/arm-linux-gnueabihf
|
||||
endif
|
||||
endif
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-linux)
|
||||
ifneq ($(TARGET_FS),)
|
||||
GCCVERSIONLTEQ46 := $(shell expr `$(HOST_COMPILER) -dumpversion` \<= 4.6)
|
||||
ifeq ($(GCCVERSIONLTEQ46),1)
|
||||
CCFLAGS += --sysroot=$(TARGET_FS)
|
||||
endif
|
||||
LDFLAGS += --sysroot=$(TARGET_FS)
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/lib -L$(TARGET_FS)/lib
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/lib/aarch64-linux-gnu -L$(TARGET_FS)/lib/aarch64-linux-gnu
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib -L$(TARGET_FS)/usr/lib
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib/aarch64-linux-gnu -L$(TARGET_FS)/usr/lib/aarch64-linux-gnu
|
||||
LDFLAGS += --unresolved-symbols=ignore-in-shared-libs
|
||||
CCFLAGS += -isystem=$(TARGET_FS)/usr/include -I$(TARGET_FS)/usr/include -I$(TARGET_FS)/usr/include/libdrm
|
||||
CCFLAGS += -isystem=$(TARGET_FS)/usr/include/aarch64-linux-gnu -I$(TARGET_FS)/usr/include/aarch64-linux-gnu
|
||||
endif
|
||||
endif
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-qnx)
|
||||
NVCCFLAGS += -D_QNX_SOURCE
|
||||
NVCCFLAGS += --qpp-config 8.3.0,gcc_ntoaarch64le
|
||||
CCFLAGS += -DWIN_INTERFACE_CUSTOM -I/usr/include/aarch64-qnx-gnu
|
||||
LDFLAGS += -lsocket
|
||||
LDFLAGS += -L/usr/lib/aarch64-qnx-gnu
|
||||
CCFLAGS += "-Wl\,-rpath-link\,/usr/lib/aarch64-qnx-gnu"
|
||||
ifdef TARGET_OVERRIDE
|
||||
LDFLAGS += -lslog2
|
||||
endif
|
||||
|
||||
ifneq ($(TARGET_FS),)
|
||||
LDFLAGS += -L$(TARGET_FS)/usr/lib
|
||||
CCFLAGS += "-Wl\,-rpath-link\,$(TARGET_FS)/usr/lib"
|
||||
LDFLAGS += -L$(TARGET_FS)/usr/libnvidia
|
||||
CCFLAGS += "-Wl\,-rpath-link\,$(TARGET_FS)/usr/libnvidia"
|
||||
CCFLAGS += -I$(TARGET_FS)/../include
|
||||
endif
|
||||
endif
|
||||
endif
|
||||
|
||||
ifdef TARGET_OVERRIDE # cuda toolkit targets override
|
||||
NVCCFLAGS += -target-dir $(TARGET_OVERRIDE)
|
||||
endif
|
||||
|
||||
# Install directory of different arch
|
||||
CUDA_INSTALL_TARGET_DIR :=
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-linux)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/armv7-linux-gnueabihf/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-linux)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/aarch64-linux/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),sbsa-linux)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/sbsa-linux/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-android)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/armv7-linux-androideabi/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-android)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/aarch64-linux-androideabi/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-qnx)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/ARMv7-linux-QNX/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-qnx)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/aarch64-qnx/
|
||||
else ifeq ($(TARGET_ARCH),ppc64le)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/ppc64le-linux/
|
||||
endif
|
||||
|
||||
# Debug build flags
|
||||
ifeq ($(dbg),1)
|
||||
NVCCFLAGS += -g -G
|
||||
BUILD_TYPE := debug
|
||||
else
|
||||
BUILD_TYPE := release
|
||||
endif
|
||||
|
||||
ALL_CCFLAGS :=
|
||||
ALL_CCFLAGS += $(NVCCFLAGS)
|
||||
ALL_CCFLAGS += $(EXTRA_NVCCFLAGS)
|
||||
ALL_CCFLAGS += $(addprefix -Xcompiler ,$(CCFLAGS))
|
||||
ALL_CCFLAGS += $(addprefix -Xcompiler ,$(EXTRA_CCFLAGS))
|
||||
|
||||
SAMPLE_ENABLED := 1
|
||||
|
||||
# This sample is not supported on Mac OSX
|
||||
ifeq ($(TARGET_OS),darwin)
|
||||
$(info >>> WARNING - conjugateGradientMultiDeviceCG is not supported on Mac OSX - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
# This sample is not supported on ARMv7
|
||||
ifeq ($(TARGET_ARCH),armv7l)
|
||||
$(info >>> WARNING - conjugateGradientMultiDeviceCG is not supported on ARMv7 - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
# This sample is not supported on QNX
|
||||
ifeq ($(TARGET_OS),qnx)
|
||||
$(info >>> WARNING - conjugateGradientMultiDeviceCG is not supported on QNX - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
ALL_LDFLAGS :=
|
||||
ALL_LDFLAGS += $(ALL_CCFLAGS)
|
||||
ALL_LDFLAGS += $(addprefix -Xlinker ,$(LDFLAGS))
|
||||
ALL_LDFLAGS += $(addprefix -Xlinker ,$(EXTRA_LDFLAGS))
|
||||
|
||||
# Common includes and paths for CUDA
|
||||
INCLUDES := -I../../../Common
|
||||
LIBRARIES :=
|
||||
|
||||
################################################################################
|
||||
|
||||
#Detect if installed version of GCC supports required C++11
|
||||
ifeq ($(TARGET_OS),linux)
|
||||
empty :=
|
||||
space := $(empty) $(empty)
|
||||
GCCVERSIONSTRING := $(shell expr `$(HOST_COMPILER) -dumpversion`)
|
||||
#Create version number without "."
|
||||
GCCVERSION := $(shell expr `echo $(GCCVERSIONSTRING)` | cut -f1 -d.)
|
||||
GCCVERSION += $(shell expr `echo $(GCCVERSIONSTRING)` | cut -f2 -d.)
|
||||
GCCVERSION += $(shell expr `echo $(GCCVERSIONSTRING)` | cut -f3 -d.)
|
||||
# Make sure the version number has at least 3 decimals
|
||||
GCCVERSION += 00
|
||||
# Remove spaces from the version number
|
||||
GCCVERSION := $(subst $(space),$(empty),$(GCCVERSION))
|
||||
#$(warning $(GCCVERSION))
|
||||
|
||||
IS_MIN_VERSION := $(shell expr `echo $(GCCVERSION)` \>= 47000)
|
||||
|
||||
ifeq ($(IS_MIN_VERSION), 1)
|
||||
$(info >>> GCC Version is greater or equal to 4.7.0 <<<)
|
||||
else
|
||||
$(info >>> Waiving build. Minimum GCC version required is 4.7.0<<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
endif
|
||||
|
||||
# Gencode arguments
|
||||
ifeq ($(TARGET_ARCH),$(filter $(TARGET_ARCH),armv7l aarch64 sbsa))
|
||||
SMS ?= 61 70 72 75 80 86 87
|
||||
else
|
||||
SMS ?= 60 61 70 75 80 86
|
||||
endif
|
||||
|
||||
ifeq ($(SMS),)
|
||||
$(info >>> WARNING - no SM architectures have been specified - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
ifeq ($(GENCODE_FLAGS),)
|
||||
# Generate SASS code for each SM architecture listed in $(SMS)
|
||||
$(foreach sm,$(SMS),$(eval GENCODE_FLAGS += -gencode arch=compute_$(sm),code=sm_$(sm)))
|
||||
|
||||
# Generate PTX code from the highest SM architecture in $(SMS) to guarantee forward-compatibility
|
||||
HIGHEST_SM := $(lastword $(sort $(SMS)))
|
||||
ifneq ($(HIGHEST_SM),)
|
||||
GENCODE_FLAGS += -gencode arch=compute_$(HIGHEST_SM),code=compute_$(HIGHEST_SM)
|
||||
endif
|
||||
endif
|
||||
|
||||
ALL_CCFLAGS += -ewp -maxrregcount=64 --std=c++11 --threads 0
|
||||
|
||||
LIBRARIES += -lcudadevrt
|
||||
|
||||
ifeq ($(SAMPLE_ENABLED),0)
|
||||
EXEC ?= @echo "[@]"
|
||||
endif
|
||||
|
||||
################################################################################
|
||||
|
||||
# Target rules
|
||||
all: build
|
||||
|
||||
build: conjugateGradientMultiDeviceCG
|
||||
|
||||
check.deps:
|
||||
ifeq ($(SAMPLE_ENABLED),0)
|
||||
@echo "Sample will be waived due to the above missing dependencies"
|
||||
else
|
||||
@echo "Sample is ready - all dependencies have been met"
|
||||
endif
|
||||
|
||||
conjugateGradientMultiDeviceCG.o:conjugateGradientMultiDeviceCG.cu
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
conjugateGradientMultiDeviceCG: conjugateGradientMultiDeviceCG.o
|
||||
$(EXEC) $(NVCC) $(ALL_LDFLAGS) $(GENCODE_FLAGS) -o $@ $+ $(LIBRARIES)
|
||||
$(EXEC) mkdir -p ../../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)
|
||||
$(EXEC) cp $@ ../../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)
|
||||
|
||||
run: build
|
||||
$(EXEC) ./conjugateGradientMultiDeviceCG
|
||||
|
||||
testrun: build
|
||||
|
||||
clean:
|
||||
rm -f conjugateGradientMultiDeviceCG conjugateGradientMultiDeviceCG.o
|
||||
rm -rf ../../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)/conjugateGradientMultiDeviceCG
|
||||
|
||||
clobber: clean
|
||||
@@ -0,0 +1,98 @@
|
||||
<?xml version="1.0" encoding="UTF-8"?>
|
||||
<!DOCTYPE entry SYSTEM "SamplesInfo.dtd">
|
||||
<entry>
|
||||
<name>conjugateGradientMultiDeviceCG</name>
|
||||
<cflags>
|
||||
<flag>-ewp</flag>
|
||||
<flag>-maxrregcount=64</flag>
|
||||
<flag>--std=c++11</flag>
|
||||
</cflags>
|
||||
<cuda_api_list>
|
||||
<toolkit>cudaDeviceEnablePeerAccess</toolkit>
|
||||
<toolkit>cudaMemset</toolkit>
|
||||
<toolkit>cudaFree</toolkit>
|
||||
<toolkit>cudaMallocManaged</toolkit>
|
||||
<toolkit>cudaMemPrefetchAsync</toolkit>
|
||||
<toolkit>cudaHostAlloc</toolkit>
|
||||
<toolkit>cudaOccupancyMaxActiveBlocksPerMultiprocessor</toolkit>
|
||||
<toolkit>cudaStreamCreate</toolkit>
|
||||
<toolkit>cudaGetDeviceCount</toolkit>
|
||||
<toolkit>cudaFreeHost</toolkit>
|
||||
<toolkit>cudaSetDevice</toolkit>
|
||||
<toolkit>cudaDeviceCanAccessPeer</toolkit>
|
||||
<toolkit>cudaLaunchCooperativeKernel</toolkit>
|
||||
<toolkit>cudaStreamSynchronize</toolkit>
|
||||
<toolkit>cudaMemAdvise</toolkit>
|
||||
<toolkit>cudaGetDeviceProperties</toolkit>
|
||||
</cuda_api_list>
|
||||
<description><![CDATA[This sample implements a conjugate gradient solver on multiple GPUs using Multi Device Cooperative Groups, also uses Unified Memory optimized using prefetching and usage hints.]]></description>
|
||||
<devicecompilation>whole</devicecompilation>
|
||||
<includepaths>
|
||||
<path>./</path>
|
||||
<path>../</path>
|
||||
<path>../../../Common</path>
|
||||
</includepaths>
|
||||
<keyconcepts>
|
||||
<concept level="basic">Unified Memory</concept>
|
||||
<concept level="advanced">Linear Algebra</concept>
|
||||
<concept level="basic">Cooperative Groups</concept>
|
||||
<concept level="advanced">MultiDevice Cooperative Groups</concept>
|
||||
<concept level="advanced">CUBLAS Library</concept>
|
||||
<concept level="advanced">CUSPARSE Library</concept>
|
||||
</keyconcepts>
|
||||
<keywords>
|
||||
<keyword>CUDA</keyword>
|
||||
<keyword>Sparse Matrix</keyword>
|
||||
<keyword>Unified Memory</keyword>
|
||||
<keyword>Multi-GPU</keyword>
|
||||
<keyword>CPP11</keyword>
|
||||
</keywords>
|
||||
<libraries>
|
||||
<library>cudadevrt</library>
|
||||
</libraries>
|
||||
<librarypaths>
|
||||
</librarypaths>
|
||||
<nsight_eclipse>true</nsight_eclipse>
|
||||
<primary_file>conjugateGradientMultiDeviceCG.cu</primary_file>
|
||||
<required_dependencies>
|
||||
<dependency>UVM</dependency>
|
||||
<dependency>MDCG</dependency>
|
||||
<dependency>CPP11</dependency>
|
||||
</required_dependencies>
|
||||
<scopes>
|
||||
<scope>1:CUDA Advanced Topics</scope>
|
||||
<scope>3:Linear Algebra</scope>
|
||||
</scopes>
|
||||
<sm-arch>sm60</sm-arch>
|
||||
<sm-arch>sm61</sm-arch>
|
||||
<sm-arch>sm70</sm-arch>
|
||||
<sm-arch>sm72</sm-arch>
|
||||
<sm-arch>sm75</sm-arch>
|
||||
<sm-arch>sm80</sm-arch>
|
||||
<sm-arch>sm86</sm-arch>
|
||||
<sm-arch>sm87</sm-arch>
|
||||
<supported_envs>
|
||||
<env>
|
||||
<arch>x86_64</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
<env>
|
||||
<arch>ppc64le</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
<env>
|
||||
<platform>windows</platform>
|
||||
</env>
|
||||
<env>
|
||||
<platform>aarch64</platform>
|
||||
</env>
|
||||
<env>
|
||||
<platform>sbsa</platform>
|
||||
</env>
|
||||
</supported_envs>
|
||||
<supported_sm_architectures>
|
||||
<from>6.0</from>
|
||||
</supported_sm_architectures>
|
||||
<title>conjugateGradient using MultiDevice Cooperative Groups</title>
|
||||
<type>exe</type>
|
||||
</entry>
|
||||
@@ -0,0 +1,74 @@
|
||||
# conjugateGradientMultiDeviceCG - conjugateGradient using MultiDevice Cooperative Groups
|
||||
|
||||
## Description
|
||||
|
||||
This sample implements a conjugate gradient solver on multiple GPUs using Multi Device Cooperative Groups, also uses Unified Memory optimized using prefetching and usage hints.
|
||||
|
||||
## Key Concepts
|
||||
|
||||
Unified Memory, Linear Algebra, Cooperative Groups, MultiDevice Cooperative Groups, CUBLAS Library, CUSPARSE Library
|
||||
|
||||
## Supported SM Architectures
|
||||
|
||||
[SM 6.0 ](https://developer.nvidia.com/cuda-gpus) [SM 6.1 ](https://developer.nvidia.com/cuda-gpus) [SM 7.0 ](https://developer.nvidia.com/cuda-gpus) [SM 7.2 ](https://developer.nvidia.com/cuda-gpus) [SM 7.5 ](https://developer.nvidia.com/cuda-gpus) [SM 8.0 ](https://developer.nvidia.com/cuda-gpus) [SM 8.6 ](https://developer.nvidia.com/cuda-gpus) [SM 8.7 ](https://developer.nvidia.com/cuda-gpus)
|
||||
|
||||
## Supported OSes
|
||||
|
||||
Linux, Windows
|
||||
|
||||
## Supported CPU Architecture
|
||||
|
||||
x86_64, ppc64le, aarch64
|
||||
|
||||
## CUDA APIs involved
|
||||
|
||||
### [CUDA Runtime API](http://docs.nvidia.com/cuda/cuda-runtime-api/index.html)
|
||||
cudaDeviceEnablePeerAccess, cudaMemset, cudaFree, cudaMallocManaged, cudaMemPrefetchAsync, cudaHostAlloc, cudaOccupancyMaxActiveBlocksPerMultiprocessor, cudaStreamCreate, cudaGetDeviceCount, cudaFreeHost, cudaSetDevice, cudaDeviceCanAccessPeer, cudaLaunchCooperativeKernel, cudaStreamSynchronize, cudaMemAdvise, cudaGetDeviceProperties
|
||||
|
||||
## Dependencies needed to build/run
|
||||
[UVM](../../README.md#uvm), [MDCG](../../README.md#mdcg), [CPP11](../../README.md#cpp11)
|
||||
|
||||
## Prerequisites
|
||||
|
||||
Download and install the [CUDA Toolkit 11.6](https://developer.nvidia.com/cuda-downloads) for your corresponding platform.
|
||||
Make sure the dependencies mentioned in [Dependencies]() section above are installed.
|
||||
|
||||
## Build and Run
|
||||
|
||||
### Windows
|
||||
The Windows samples are built using the Visual Studio IDE. Solution files (.sln) are provided for each supported version of Visual Studio, using the format:
|
||||
```
|
||||
*_vs<version>.sln - for Visual Studio <version>
|
||||
```
|
||||
Each individual sample has its own set of solution files in its directory:
|
||||
|
||||
To build/examine all the samples at once, the complete solution files should be used. To build/examine a single sample, the individual sample solution files should be used.
|
||||
> **Note:** Some samples require that the Microsoft DirectX SDK (June 2010 or newer) be installed and that the VC++ directory paths are properly set up (**Tools > Options...**). Check DirectX Dependencies section for details."
|
||||
|
||||
### Linux
|
||||
The Linux samples are built using makefiles. To use the makefiles, change the current directory to the sample directory you wish to build, and run make:
|
||||
```
|
||||
$ cd <sample_dir>
|
||||
$ make
|
||||
```
|
||||
The samples makefiles can take advantage of certain options:
|
||||
* **TARGET_ARCH=<arch>** - cross-compile targeting a specific architecture. Allowed architectures are x86_64, ppc64le, aarch64.
|
||||
By default, TARGET_ARCH is set to HOST_ARCH. On a x86_64 machine, not setting TARGET_ARCH is the equivalent of setting TARGET_ARCH=x86_64.<br/>
|
||||
`$ make TARGET_ARCH=x86_64` <br/> `$ make TARGET_ARCH=ppc64le` <br/> `$ make TARGET_ARCH=aarch64` <br/>
|
||||
See [here](http://docs.nvidia.com/cuda/cuda-samples/index.html#cross-samples) for more details.
|
||||
* **dbg=1** - build with debug symbols
|
||||
```
|
||||
$ make dbg=1
|
||||
```
|
||||
* **SMS="A B ..."** - override the SM architectures for which the sample will be built, where `"A B ..."` is a space-delimited list of SM architectures. For example, to generate SASS for SM 50 and SM 60, use `SMS="50 60"`.
|
||||
```
|
||||
$ make SMS="50 60"
|
||||
```
|
||||
|
||||
* **HOST_COMPILER=<host_compiler>** - override the default g++ host compiler. See the [Linux Installation Guide](http://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#system-requirements) for a list of supported host compilers.
|
||||
```
|
||||
$ make HOST_COMPILER=g++
|
||||
```
|
||||
|
||||
## References (for more details)
|
||||
|
||||
@@ -0,0 +1,793 @@
|
||||
/* Copyright (c) 2022, NVIDIA CORPORATION. All rights reserved.
|
||||
*
|
||||
* Redistribution and use in source and binary forms, with or without
|
||||
* modification, are permitted provided that the following conditions
|
||||
* are met:
|
||||
* * Redistributions of source code must retain the above copyright
|
||||
* notice, this list of conditions and the following disclaimer.
|
||||
* * Redistributions in binary form must reproduce the above copyright
|
||||
* notice, this list of conditions and the following disclaimer in the
|
||||
* documentation and/or other materials provided with the distribution.
|
||||
* * Neither the name of NVIDIA CORPORATION nor the names of its
|
||||
* contributors may be used to endorse or promote products derived
|
||||
* from this software without specific prior written permission.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
||||
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||||
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
||||
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
||||
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||||
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||||
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
||||
* OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
*/
|
||||
|
||||
/*
|
||||
* This sample implements a conjugate gradient solver on multiple GPU using
|
||||
* Unified Memory optimized prefetching and usage hints.
|
||||
*
|
||||
*/
|
||||
|
||||
// includes, system
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
#include <map>
|
||||
#include <iostream>
|
||||
#include <set>
|
||||
#include <utility>
|
||||
|
||||
#include <cuda_runtime.h>
|
||||
|
||||
// Utilities and system includes
|
||||
#include <helper_cuda.h> // helper function CUDA error checking and initialization
|
||||
#include <helper_functions.h> // helper for shared functions common to CUDA Samples
|
||||
|
||||
#include <cooperative_groups.h>
|
||||
#include <cooperative_groups/reduce.h>
|
||||
|
||||
namespace cg = cooperative_groups;
|
||||
|
||||
const char *sSDKname = "conjugateGradientMultiDeviceCG";
|
||||
|
||||
#define ENABLE_CPU_DEBUG_CODE 0
|
||||
#define THREADS_PER_BLOCK 512
|
||||
|
||||
__device__ double grid_dot_result = 0.0;
|
||||
|
||||
/* genTridiag: generate a random tridiagonal symmetric matrix */
|
||||
void genTridiag(int *I, int *J, float *val, int N, int nz) {
|
||||
I[0] = 0, J[0] = 0, J[1] = 1;
|
||||
val[0] = (float)rand() / RAND_MAX + 10.0f;
|
||||
val[1] = (float)rand() / RAND_MAX;
|
||||
int start;
|
||||
|
||||
for (int i = 1; i < N; i++) {
|
||||
if (i > 1) {
|
||||
I[i] = I[i - 1] + 3;
|
||||
} else {
|
||||
I[1] = 2;
|
||||
}
|
||||
|
||||
start = (i - 1) * 3 + 2;
|
||||
J[start] = i - 1;
|
||||
J[start + 1] = i;
|
||||
|
||||
if (i < N - 1) {
|
||||
J[start + 2] = i + 1;
|
||||
}
|
||||
|
||||
val[start] = val[start - 1];
|
||||
val[start + 1] = (float)rand() / RAND_MAX + 10.0f;
|
||||
|
||||
if (i < N - 1) {
|
||||
val[start + 2] = (float)rand() / RAND_MAX;
|
||||
}
|
||||
}
|
||||
|
||||
I[N] = nz;
|
||||
}
|
||||
|
||||
// I - contains location of the given non-zero element in the row of the matrix
|
||||
// J - contains location of the given non-zero element in the column of the
|
||||
// matrix val - contains values of the given non-zero elements of the matrix
|
||||
// inputVecX - input vector to be multiplied
|
||||
// outputVecY - resultant vector
|
||||
void cpuSpMV(int *I, int *J, float *val, int nnz, int num_rows, float alpha,
|
||||
float *inputVecX, float *outputVecY) {
|
||||
for (int i = 0; i < num_rows; i++) {
|
||||
int num_elems_this_row = I[i + 1] - I[i];
|
||||
|
||||
float output = 0.0;
|
||||
for (int j = 0; j < num_elems_this_row; j++) {
|
||||
output += alpha * val[I[i] + j] * inputVecX[J[I[i] + j]];
|
||||
}
|
||||
outputVecY[i] = output;
|
||||
}
|
||||
|
||||
return;
|
||||
}
|
||||
|
||||
float dotProduct(float *vecA, float *vecB, int size) {
|
||||
float result = 0.0;
|
||||
|
||||
for (int i = 0; i < size; i++) {
|
||||
result = result + (vecA[i] * vecB[i]);
|
||||
}
|
||||
|
||||
return result;
|
||||
}
|
||||
|
||||
void scaleVector(float *vec, float alpha, int size) {
|
||||
for (int i = 0; i < size; i++) {
|
||||
vec[i] = alpha * vec[i];
|
||||
}
|
||||
}
|
||||
|
||||
void saxpy(float *x, float *y, float a, int size) {
|
||||
for (int i = 0; i < size; i++) {
|
||||
y[i] = a * x[i] + y[i];
|
||||
}
|
||||
}
|
||||
|
||||
void cpuConjugateGrad(int *I, int *J, float *val, float *x, float *Ax, float *p,
|
||||
float *r, int nnz, int N, float tol) {
|
||||
int max_iter = 10000;
|
||||
|
||||
float alpha = 1.0;
|
||||
float alpham1 = -1.0;
|
||||
float r0 = 0.0, b, a, na;
|
||||
|
||||
cpuSpMV(I, J, val, nnz, N, alpha, x, Ax);
|
||||
saxpy(Ax, r, alpham1, N);
|
||||
|
||||
float r1 = dotProduct(r, r, N);
|
||||
|
||||
int k = 1;
|
||||
|
||||
while (r1 > tol * tol && k <= max_iter) {
|
||||
if (k > 1) {
|
||||
b = r1 / r0;
|
||||
scaleVector(p, b, N);
|
||||
|
||||
saxpy(r, p, alpha, N);
|
||||
} else {
|
||||
for (int i = 0; i < N; i++) p[i] = r[i];
|
||||
}
|
||||
|
||||
cpuSpMV(I, J, val, nnz, N, alpha, p, Ax);
|
||||
|
||||
float dot = dotProduct(p, Ax, N);
|
||||
a = r1 / dot;
|
||||
|
||||
saxpy(p, x, a, N);
|
||||
na = -a;
|
||||
saxpy(Ax, r, na, N);
|
||||
|
||||
r0 = r1;
|
||||
r1 = dotProduct(r, r, N);
|
||||
|
||||
printf("\nCPU code iteration = %3d, residual = %e\n", k, sqrt(r1));
|
||||
k++;
|
||||
}
|
||||
}
|
||||
|
||||
// Data filled on CPU needed for MultiGPU operations.
|
||||
struct MultiDeviceData {
|
||||
unsigned char *hostMemoryArrivedList;
|
||||
unsigned int numDevices;
|
||||
unsigned int deviceRank;
|
||||
};
|
||||
|
||||
// Class used for coordination of multiple devices.
|
||||
class PeerGroup {
|
||||
const MultiDeviceData &data;
|
||||
const cg::grid_group &grid;
|
||||
|
||||
__device__ unsigned char load_arrived(unsigned char *arrived) const {
|
||||
#if __CUDA_ARCH__ < 700
|
||||
return *(volatile unsigned char *)arrived;
|
||||
#else
|
||||
unsigned int result;
|
||||
asm volatile("ld.acquire.sys.global.u8 %0, [%1];"
|
||||
: "=r"(result)
|
||||
: "l"(arrived)
|
||||
: "memory");
|
||||
return result;
|
||||
#endif
|
||||
}
|
||||
|
||||
__device__ void store_arrived(unsigned char *arrived,
|
||||
unsigned char val) const {
|
||||
#if __CUDA_ARCH__ < 700
|
||||
*(volatile unsigned char *)arrived = val;
|
||||
#else
|
||||
unsigned int reg_val = val;
|
||||
asm volatile(
|
||||
"st.release.sys.global.u8 [%1], %0;" ::"r"(reg_val) "l"(arrived)
|
||||
: "memory");
|
||||
|
||||
// Avoids compiler warnings from unused variable val.
|
||||
(void)(reg_val = reg_val);
|
||||
#endif
|
||||
}
|
||||
|
||||
public:
|
||||
__device__ PeerGroup(const MultiDeviceData &data, const cg::grid_group &grid)
|
||||
: data(data), grid(grid){};
|
||||
|
||||
__device__ unsigned int size() const { return data.numDevices * grid.size(); }
|
||||
|
||||
__device__ unsigned int thread_rank() const {
|
||||
return data.deviceRank * grid.size() + grid.thread_rank();
|
||||
}
|
||||
|
||||
__device__ void sync() const {
|
||||
grid.sync();
|
||||
|
||||
// One thread from each grid participates in the sync.
|
||||
if (grid.thread_rank() == 0) {
|
||||
if (data.deviceRank == 0) {
|
||||
// Leader grid waits for others to join and then releases them.
|
||||
// Other GPUs can arrive in any order, so the leader have to wait for
|
||||
// all others.
|
||||
for (int i = 0; i < data.numDevices - 1; i++) {
|
||||
while (load_arrived(&data.hostMemoryArrivedList[i]) == 0)
|
||||
;
|
||||
}
|
||||
for (int i = 0; i < data.numDevices - 1; i++) {
|
||||
store_arrived(&data.hostMemoryArrivedList[i], 0);
|
||||
}
|
||||
__threadfence_system();
|
||||
} else {
|
||||
// Other grids note their arrival and wait to be released.
|
||||
store_arrived(&data.hostMemoryArrivedList[data.deviceRank - 1], 1);
|
||||
while (load_arrived(&data.hostMemoryArrivedList[data.deviceRank - 1]) ==
|
||||
1)
|
||||
;
|
||||
}
|
||||
}
|
||||
|
||||
grid.sync();
|
||||
}
|
||||
};
|
||||
|
||||
__device__ void gpuSpMV(int *I, int *J, float *val, int nnz, int num_rows,
|
||||
float alpha, float *inputVecX, float *outputVecY,
|
||||
const PeerGroup &peer_group) {
|
||||
for (int i = peer_group.thread_rank(); i < num_rows; i += peer_group.size()) {
|
||||
int row_elem = I[i];
|
||||
int next_row_elem = I[i + 1];
|
||||
int num_elems_this_row = next_row_elem - row_elem;
|
||||
|
||||
float output = 0.0;
|
||||
for (int j = 0; j < num_elems_this_row; j++) {
|
||||
output += alpha * val[row_elem + j] * inputVecX[J[row_elem + j]];
|
||||
}
|
||||
|
||||
outputVecY[i] = output;
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void gpuSaxpy(float *x, float *y, float a, int size,
|
||||
const PeerGroup &peer_group) {
|
||||
for (int i = peer_group.thread_rank(); i < size; i += peer_group.size()) {
|
||||
y[i] = a * x[i] + y[i];
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void gpuDotProduct(float *vecA, float *vecB, int size,
|
||||
const cg::thread_block &cta,
|
||||
const PeerGroup &peer_group) {
|
||||
extern __shared__ double tmp[];
|
||||
|
||||
double temp_sum = 0.0;
|
||||
|
||||
for (int i = peer_group.thread_rank(); i < size; i += peer_group.size()) {
|
||||
temp_sum += (double)(vecA[i] * vecB[i]);
|
||||
}
|
||||
|
||||
cg::thread_block_tile<32> tile32 = cg::tiled_partition<32>(cta);
|
||||
|
||||
temp_sum = cg::reduce(tile32, temp_sum, cg::plus<double>());
|
||||
|
||||
if (tile32.thread_rank() == 0) {
|
||||
tmp[tile32.meta_group_rank()] = temp_sum;
|
||||
}
|
||||
|
||||
cg::sync(cta);
|
||||
|
||||
if (tile32.meta_group_rank() == 0) {
|
||||
temp_sum = tile32.thread_rank() < tile32.meta_group_size()
|
||||
? tmp[tile32.thread_rank()]
|
||||
: 0.0;
|
||||
temp_sum = cg::reduce(tile32, temp_sum, cg::plus<double>());
|
||||
|
||||
if (tile32.thread_rank() == 0) {
|
||||
atomicAdd(&grid_dot_result, temp_sum);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void gpuCopyVector(float *srcA, float *destB, int size,
|
||||
const PeerGroup &peer_group) {
|
||||
for (int i = peer_group.thread_rank(); i < size; i += peer_group.size()) {
|
||||
destB[i] = srcA[i];
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void gpuScaleVectorAndSaxpy(float *x, float *y, float a, float scale,
|
||||
int size, const PeerGroup &peer_group) {
|
||||
for (int i = peer_group.thread_rank(); i < size; i += peer_group.size()) {
|
||||
y[i] = a * x[i] + scale * y[i];
|
||||
}
|
||||
}
|
||||
|
||||
extern "C" __global__ void multiGpuConjugateGradient(
|
||||
int *I, int *J, float *val, float *x, float *Ax, float *p, float *r,
|
||||
double *dot_result, int nnz, int N, float tol,
|
||||
MultiDeviceData multi_device_data) {
|
||||
cg::thread_block cta = cg::this_thread_block();
|
||||
cg::grid_group grid = cg::this_grid();
|
||||
PeerGroup peer_group(multi_device_data, grid);
|
||||
|
||||
const int max_iter = 10000;
|
||||
|
||||
float alpha = 1.0;
|
||||
float alpham1 = -1.0;
|
||||
float r0 = 0.0, r1, b, a, na;
|
||||
|
||||
for (int i = peer_group.thread_rank(); i < N; i += peer_group.size()) {
|
||||
r[i] = 1.0;
|
||||
x[i] = 0.0;
|
||||
}
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
gpuSpMV(I, J, val, nnz, N, alpha, x, Ax, peer_group);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
gpuSaxpy(Ax, r, alpham1, N, peer_group);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
gpuDotProduct(r, r, N, cta, peer_group);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
if (grid.thread_rank() == 0) {
|
||||
atomicAdd_system(dot_result, grid_dot_result);
|
||||
grid_dot_result = 0.0;
|
||||
}
|
||||
peer_group.sync();
|
||||
|
||||
r1 = *dot_result;
|
||||
|
||||
int k = 1;
|
||||
while (r1 > tol * tol && k <= max_iter) {
|
||||
if (k > 1) {
|
||||
b = r1 / r0;
|
||||
gpuScaleVectorAndSaxpy(r, p, alpha, b, N, peer_group);
|
||||
} else {
|
||||
gpuCopyVector(r, p, N, peer_group);
|
||||
}
|
||||
|
||||
peer_group.sync();
|
||||
|
||||
gpuSpMV(I, J, val, nnz, N, alpha, p, Ax, peer_group);
|
||||
|
||||
if (peer_group.thread_rank() == 0) {
|
||||
*dot_result = 0.0;
|
||||
}
|
||||
peer_group.sync();
|
||||
|
||||
gpuDotProduct(p, Ax, N, cta, peer_group);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
if (grid.thread_rank() == 0) {
|
||||
atomicAdd_system(dot_result, grid_dot_result);
|
||||
grid_dot_result = 0.0;
|
||||
}
|
||||
peer_group.sync();
|
||||
|
||||
a = r1 / *dot_result;
|
||||
|
||||
gpuSaxpy(p, x, a, N, peer_group);
|
||||
|
||||
na = -a;
|
||||
|
||||
gpuSaxpy(Ax, r, na, N, peer_group);
|
||||
|
||||
r0 = r1;
|
||||
|
||||
peer_group.sync();
|
||||
|
||||
if (peer_group.thread_rank() == 0) {
|
||||
*dot_result = 0.0;
|
||||
}
|
||||
|
||||
peer_group.sync();
|
||||
|
||||
gpuDotProduct(r, r, N, cta, peer_group);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
if (grid.thread_rank() == 0) {
|
||||
atomicAdd_system(dot_result, grid_dot_result);
|
||||
grid_dot_result = 0.0;
|
||||
}
|
||||
peer_group.sync();
|
||||
|
||||
r1 = *dot_result;
|
||||
k++;
|
||||
}
|
||||
}
|
||||
|
||||
// Map of device version to device number
|
||||
std::multimap<std::pair<int, int>, int> getIdenticalGPUs() {
|
||||
int numGpus = 0;
|
||||
checkCudaErrors(cudaGetDeviceCount(&numGpus));
|
||||
|
||||
std::multimap<std::pair<int, int>, int> identicalGpus;
|
||||
|
||||
for (int i = 0; i < numGpus; i++) {
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, i));
|
||||
|
||||
// Filter unsupported devices
|
||||
if (deviceProp.cooperativeLaunch && deviceProp.concurrentManagedAccess) {
|
||||
identicalGpus.emplace(std::make_pair(deviceProp.major, deviceProp.minor),
|
||||
i);
|
||||
}
|
||||
printf("GPU Device %d: \"%s\" with compute capability %d.%d\n", i,
|
||||
deviceProp.name, deviceProp.major, deviceProp.minor);
|
||||
}
|
||||
|
||||
return identicalGpus;
|
||||
}
|
||||
|
||||
int main(int argc, char **argv) {
|
||||
constexpr size_t kNumGpusRequired = 2;
|
||||
int N = 0, nz = 0, *I = NULL, *J = NULL;
|
||||
float *val = NULL;
|
||||
const float tol = 1e-5f;
|
||||
float *x;
|
||||
float rhs = 1.0;
|
||||
float r1;
|
||||
float *r, *p, *Ax;
|
||||
|
||||
printf("Starting [%s]...\n", sSDKname);
|
||||
auto gpusByArch = getIdenticalGPUs();
|
||||
|
||||
auto it = gpusByArch.begin();
|
||||
auto end = gpusByArch.end();
|
||||
|
||||
auto bestFit = std::make_pair(it, it);
|
||||
// use std::distance to find the largest number of GPUs amongst architectures
|
||||
auto distance = [](decltype(bestFit) p) {
|
||||
return std::distance(p.first, p.second);
|
||||
};
|
||||
|
||||
// Read each unique key/pair element in order
|
||||
for (; it != end; it = gpusByArch.upper_bound(it->first)) {
|
||||
// first and second are iterators bounded within the architecture group
|
||||
auto testFit = gpusByArch.equal_range(it->first);
|
||||
// Always use devices with highest architecture version or whichever has the
|
||||
// most devices available
|
||||
if (distance(bestFit) <= distance(testFit)) bestFit = testFit;
|
||||
}
|
||||
|
||||
if (distance(bestFit) < kNumGpusRequired) {
|
||||
printf(
|
||||
"No two or more GPUs with same architecture capable of "
|
||||
"concurrentManagedAccess found. "
|
||||
"\nWaiving the sample\n");
|
||||
exit(EXIT_WAIVED);
|
||||
}
|
||||
|
||||
std::set<int> bestFitDeviceIds;
|
||||
|
||||
// Check & select peer-to-peer access capable GPU devices as enabling p2p
|
||||
// access between participating GPUs gives better performance.
|
||||
for (auto itr = bestFit.first; itr != bestFit.second; itr++) {
|
||||
int deviceId = itr->second;
|
||||
checkCudaErrors(cudaSetDevice(deviceId));
|
||||
|
||||
std::for_each(
|
||||
itr, bestFit.second,
|
||||
[&deviceId, &bestFitDeviceIds,
|
||||
&kNumGpusRequired](decltype(*itr) mapPair) {
|
||||
if (deviceId != mapPair.second) {
|
||||
int access = 0;
|
||||
checkCudaErrors(
|
||||
cudaDeviceCanAccessPeer(&access, deviceId, mapPair.second));
|
||||
printf("Device=%d %s Access Peer Device=%d\n", deviceId,
|
||||
access ? "CAN" : "CANNOT", mapPair.second);
|
||||
if (access && bestFitDeviceIds.size() < kNumGpusRequired) {
|
||||
bestFitDeviceIds.emplace(deviceId);
|
||||
bestFitDeviceIds.emplace(mapPair.second);
|
||||
} else {
|
||||
printf("Ignoring device %i (max devices exceeded)\n",
|
||||
mapPair.second);
|
||||
}
|
||||
}
|
||||
});
|
||||
|
||||
if (bestFitDeviceIds.size() >= kNumGpusRequired) {
|
||||
printf("Selected p2p capable devices - ");
|
||||
for (auto devicesItr = bestFitDeviceIds.begin();
|
||||
devicesItr != bestFitDeviceIds.end(); devicesItr++) {
|
||||
printf("deviceId = %d ", *devicesItr);
|
||||
}
|
||||
printf("\n");
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
// if bestFitDeviceIds.size() == 0 it means the GPUs in system are not p2p
|
||||
// capable, hence we add it without p2p capability check.
|
||||
if (!bestFitDeviceIds.size()) {
|
||||
printf("Devices involved are not p2p capable.. selecting %zu of them\n",
|
||||
kNumGpusRequired);
|
||||
std::for_each(bestFit.first, bestFit.second,
|
||||
[&bestFitDeviceIds,
|
||||
&kNumGpusRequired](decltype(*bestFit.first) mapPair) {
|
||||
if (bestFitDeviceIds.size() < kNumGpusRequired) {
|
||||
bestFitDeviceIds.emplace(mapPair.second);
|
||||
} else {
|
||||
printf("Ignoring device %i (max devices exceeded)\n",
|
||||
mapPair.second);
|
||||
}
|
||||
// Insert the sequence into the deviceIds set
|
||||
});
|
||||
} else {
|
||||
// perform cudaDeviceEnablePeerAccess in both directions for all
|
||||
// participating devices.
|
||||
for (auto p1_itr = bestFitDeviceIds.begin();
|
||||
p1_itr != bestFitDeviceIds.end(); p1_itr++) {
|
||||
checkCudaErrors(cudaSetDevice(*p1_itr));
|
||||
for (auto p2_itr = bestFitDeviceIds.begin();
|
||||
p2_itr != bestFitDeviceIds.end(); p2_itr++) {
|
||||
if (*p1_itr != *p2_itr) {
|
||||
checkCudaErrors(cudaDeviceEnablePeerAccess(*p2_itr, 0));
|
||||
checkCudaErrors(cudaSetDevice(*p1_itr));
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/* Generate a random tridiagonal symmetric matrix in CSR format */
|
||||
N = 10485760 * 2;
|
||||
nz = (N - 2) * 3 + 4;
|
||||
|
||||
checkCudaErrors(cudaMallocManaged((void **)&I, sizeof(int) * (N + 1)));
|
||||
checkCudaErrors(cudaMallocManaged((void **)&J, sizeof(int) * nz));
|
||||
checkCudaErrors(cudaMallocManaged((void **)&val, sizeof(float) * nz));
|
||||
|
||||
float *val_cpu = (float *)malloc(sizeof(float) * nz);
|
||||
|
||||
genTridiag(I, J, val_cpu, N, nz);
|
||||
|
||||
memcpy(val, val_cpu, sizeof(float) * nz);
|
||||
checkCudaErrors(
|
||||
cudaMemAdvise(I, sizeof(int) * (N + 1), cudaMemAdviseSetReadMostly, 0));
|
||||
checkCudaErrors(
|
||||
cudaMemAdvise(J, sizeof(int) * nz, cudaMemAdviseSetReadMostly, 0));
|
||||
checkCudaErrors(
|
||||
cudaMemAdvise(val, sizeof(float) * nz, cudaMemAdviseSetReadMostly, 0));
|
||||
|
||||
checkCudaErrors(cudaMallocManaged((void **)&x, sizeof(float) * N));
|
||||
|
||||
double *dot_result;
|
||||
checkCudaErrors(cudaMallocManaged((void **)&dot_result, sizeof(double)));
|
||||
|
||||
checkCudaErrors(cudaMemset(dot_result, 0, sizeof(double)));
|
||||
|
||||
// temp memory for ConjugateGradient
|
||||
checkCudaErrors(cudaMallocManaged((void **)&r, N * sizeof(float)));
|
||||
checkCudaErrors(cudaMallocManaged((void **)&p, N * sizeof(float)));
|
||||
checkCudaErrors(cudaMallocManaged((void **)&Ax, N * sizeof(float)));
|
||||
|
||||
std::cout << "\nRunning on GPUs = " << kNumGpusRequired << std::endl;
|
||||
cudaStream_t nStreams[kNumGpusRequired];
|
||||
|
||||
int sMemSize = sizeof(double) * ((THREADS_PER_BLOCK / 32) + 1);
|
||||
int numBlocksPerSm = INT_MAX;
|
||||
int numThreads = THREADS_PER_BLOCK;
|
||||
int numSms = INT_MAX;
|
||||
auto deviceId = bestFitDeviceIds.begin();
|
||||
|
||||
// set numSms & numBlocksPerSm to be lowest of 2 devices
|
||||
while (deviceId != bestFitDeviceIds.end()) {
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(cudaSetDevice(*deviceId));
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, *deviceId));
|
||||
|
||||
int numBlocksPerSm_current = 0;
|
||||
checkCudaErrors(cudaOccupancyMaxActiveBlocksPerMultiprocessor(
|
||||
&numBlocksPerSm_current, multiGpuConjugateGradient, numThreads,
|
||||
sMemSize));
|
||||
|
||||
if (numBlocksPerSm > numBlocksPerSm_current) {
|
||||
numBlocksPerSm = numBlocksPerSm_current;
|
||||
}
|
||||
if (numSms > deviceProp.multiProcessorCount) {
|
||||
numSms = deviceProp.multiProcessorCount;
|
||||
}
|
||||
deviceId++;
|
||||
}
|
||||
|
||||
if (!numBlocksPerSm) {
|
||||
printf(
|
||||
"Max active blocks per SM is returned as 0.\n Hence, Waiving the "
|
||||
"sample\n");
|
||||
exit(EXIT_WAIVED);
|
||||
}
|
||||
|
||||
int device_count = 0;
|
||||
int totalThreadsPerGPU = numSms * numBlocksPerSm * THREADS_PER_BLOCK;
|
||||
deviceId = bestFitDeviceIds.begin();
|
||||
while (deviceId != bestFitDeviceIds.end()) {
|
||||
checkCudaErrors(cudaSetDevice(*deviceId));
|
||||
checkCudaErrors(cudaStreamCreate(&nStreams[device_count]));
|
||||
|
||||
int perGPUIter = N / (totalThreadsPerGPU * kNumGpusRequired);
|
||||
int offset_Ax = device_count * totalThreadsPerGPU;
|
||||
int offset_r = device_count * totalThreadsPerGPU;
|
||||
int offset_p = device_count * totalThreadsPerGPU;
|
||||
int offset_x = device_count * totalThreadsPerGPU;
|
||||
|
||||
checkCudaErrors(cudaMemPrefetchAsync(I, sizeof(int) * N, *deviceId,
|
||||
nStreams[device_count]));
|
||||
checkCudaErrors(cudaMemPrefetchAsync(val, sizeof(float) * nz, *deviceId,
|
||||
nStreams[device_count]));
|
||||
checkCudaErrors(cudaMemPrefetchAsync(J, sizeof(float) * nz, *deviceId,
|
||||
nStreams[device_count]));
|
||||
|
||||
if (offset_Ax <= N) {
|
||||
for (int i = 0; i < perGPUIter; i++) {
|
||||
cudaMemAdvise(Ax + offset_Ax, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetPreferredLocation, *deviceId);
|
||||
cudaMemAdvise(r + offset_r, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetPreferredLocation, *deviceId);
|
||||
cudaMemAdvise(x + offset_x, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetPreferredLocation, *deviceId);
|
||||
cudaMemAdvise(p + offset_p, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetPreferredLocation, *deviceId);
|
||||
|
||||
cudaMemAdvise(Ax + offset_Ax, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetAccessedBy, *deviceId);
|
||||
cudaMemAdvise(r + offset_r, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetAccessedBy, *deviceId);
|
||||
cudaMemAdvise(p + offset_p, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetAccessedBy, *deviceId);
|
||||
cudaMemAdvise(x + offset_x, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetAccessedBy, *deviceId);
|
||||
|
||||
offset_Ax += totalThreadsPerGPU * kNumGpusRequired;
|
||||
offset_r += totalThreadsPerGPU * kNumGpusRequired;
|
||||
offset_p += totalThreadsPerGPU * kNumGpusRequired;
|
||||
offset_x += totalThreadsPerGPU * kNumGpusRequired;
|
||||
|
||||
if (offset_Ax >= N) {
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
device_count++;
|
||||
deviceId++;
|
||||
}
|
||||
|
||||
#if ENABLE_CPU_DEBUG_CODE
|
||||
float *Ax_cpu = (float *)malloc(sizeof(float) * N);
|
||||
float *r_cpu = (float *)malloc(sizeof(float) * N);
|
||||
float *p_cpu = (float *)malloc(sizeof(float) * N);
|
||||
float *x_cpu = (float *)malloc(sizeof(float) * N);
|
||||
|
||||
for (int i = 0; i < N; i++) {
|
||||
r_cpu[i] = 1.0;
|
||||
Ax_cpu[i] = x_cpu[i] = 0.0;
|
||||
}
|
||||
#endif
|
||||
|
||||
printf("Total threads per GPU = %d numBlocksPerSm = %d\n",
|
||||
numSms * numBlocksPerSm * THREADS_PER_BLOCK, numBlocksPerSm);
|
||||
dim3 dimGrid(numSms * numBlocksPerSm, 1, 1),
|
||||
dimBlock(THREADS_PER_BLOCK, 1, 1);
|
||||
|
||||
// Structure used for cross-grid synchronization.
|
||||
MultiDeviceData multi_device_data;
|
||||
checkCudaErrors(cudaHostAlloc(
|
||||
&multi_device_data.hostMemoryArrivedList,
|
||||
(kNumGpusRequired - 1) * sizeof(*multi_device_data.hostMemoryArrivedList),
|
||||
cudaHostAllocPortable));
|
||||
memset(multi_device_data.hostMemoryArrivedList, 0,
|
||||
(kNumGpusRequired - 1) *
|
||||
sizeof(*multi_device_data.hostMemoryArrivedList));
|
||||
multi_device_data.numDevices = kNumGpusRequired;
|
||||
multi_device_data.deviceRank = 0;
|
||||
|
||||
void *kernelArgs[] = {
|
||||
(void *)&I, (void *)&J, (void *)&val, (void *)&x,
|
||||
(void *)&Ax, (void *)&p, (void *)&r, (void *)&dot_result,
|
||||
(void *)&nz, (void *)&N, (void *)&tol, (void *)&multi_device_data,
|
||||
};
|
||||
|
||||
printf("Launching kernel\n");
|
||||
|
||||
deviceId = bestFitDeviceIds.begin();
|
||||
device_count = 0;
|
||||
while (deviceId != bestFitDeviceIds.end()) {
|
||||
checkCudaErrors(cudaSetDevice(*deviceId));
|
||||
checkCudaErrors(cudaLaunchCooperativeKernel(
|
||||
(void *)multiGpuConjugateGradient, dimGrid, dimBlock, kernelArgs,
|
||||
sMemSize, nStreams[device_count++]));
|
||||
multi_device_data.deviceRank++;
|
||||
deviceId++;
|
||||
}
|
||||
|
||||
checkCudaErrors(cudaMemPrefetchAsync(x, sizeof(float) * N, cudaCpuDeviceId));
|
||||
checkCudaErrors(
|
||||
cudaMemPrefetchAsync(dot_result, sizeof(double), cudaCpuDeviceId));
|
||||
|
||||
deviceId = bestFitDeviceIds.begin();
|
||||
device_count = 0;
|
||||
while (deviceId != bestFitDeviceIds.end()) {
|
||||
checkCudaErrors(cudaSetDevice(*deviceId));
|
||||
checkCudaErrors(cudaStreamSynchronize(nStreams[device_count++]));
|
||||
deviceId++;
|
||||
}
|
||||
|
||||
r1 = (float)*dot_result;
|
||||
|
||||
printf("GPU Final, residual = %e \n ", sqrt(r1));
|
||||
|
||||
#if ENABLE_CPU_DEBUG_CODE
|
||||
cpuConjugateGrad(I, J, val, x_cpu, Ax_cpu, p_cpu, r_cpu, nz, N, tol);
|
||||
#endif
|
||||
|
||||
float rsum, diff, err = 0.0;
|
||||
|
||||
for (int i = 0; i < N; i++) {
|
||||
rsum = 0.0;
|
||||
|
||||
for (int j = I[i]; j < I[i + 1]; j++) {
|
||||
rsum += val_cpu[j] * x[J[j]];
|
||||
}
|
||||
|
||||
diff = fabs(rsum - rhs);
|
||||
|
||||
if (diff > err) {
|
||||
err = diff;
|
||||
}
|
||||
}
|
||||
|
||||
checkCudaErrors(cudaFreeHost(multi_device_data.hostMemoryArrivedList));
|
||||
checkCudaErrors(cudaFree(I));
|
||||
checkCudaErrors(cudaFree(J));
|
||||
checkCudaErrors(cudaFree(val));
|
||||
checkCudaErrors(cudaFree(x));
|
||||
checkCudaErrors(cudaFree(r));
|
||||
checkCudaErrors(cudaFree(p));
|
||||
checkCudaErrors(cudaFree(Ax));
|
||||
checkCudaErrors(cudaFree(dot_result));
|
||||
free(val_cpu);
|
||||
|
||||
#if ENABLE_CPU_DEBUG_CODE
|
||||
free(Ax_cpu);
|
||||
free(r_cpu);
|
||||
free(p_cpu);
|
||||
free(x_cpu);
|
||||
#endif
|
||||
|
||||
printf("Test Summary: Error amount = %f \n", err);
|
||||
fprintf(stdout, "&&&& conjugateGradientMultiDeviceCG %s\n",
|
||||
(sqrt(r1) < tol) ? "PASSED" : "FAILED");
|
||||
exit((sqrt(r1) < tol) ? EXIT_SUCCESS : EXIT_FAILURE);
|
||||
}
|
||||
@@ -0,0 +1,20 @@
|
||||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2017
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientMultiDeviceCG", "conjugateGradientMultiDeviceCG_vs2017.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
@@ -0,0 +1,113 @@
|
||||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>conjugateGradientMultiDeviceCG_vs2017</RootNamespace>
|
||||
<ProjectName>conjugateGradientMultiDeviceCG</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(WindowsTargetPlatformVersion)'==''">
|
||||
<LatestTargetPlatformVersion>$([Microsoft.Build.Utilities.ToolLocationHelper]::GetLatestSDKTargetPlatformVersion('Windows', '10.0'))</LatestTargetPlatformVersion>
|
||||
<WindowsTargetPlatformVersion Condition="'$(WindowsTargetPlatformVersion)' == ''">$(LatestTargetPlatformVersion)</WindowsTargetPlatformVersion>
|
||||
<TargetPlatformVersion>$(WindowsTargetPlatformVersion)</TargetPlatformVersion>
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v141</PlatformToolset>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<UseDebugLibraries>true</UseDebugLibraries>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Release'">
|
||||
<WholeProgramOptimization>true</WholeProgramOptimization>
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
|
||||
<ImportGroup Label="ExtensionSettings">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientMultiDeviceCG.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;compute_80,sm_80;compute_86,sm_86;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" --threads 0 </AdditionalOptions>
|
||||
<Include>./;../../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<ExtensibleWholeProgramCompilation>true</ExtensibleWholeProgramCompilation>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="conjugateGradientMultiDeviceCG.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
@@ -0,0 +1,20 @@
|
||||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2019
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientMultiDeviceCG", "conjugateGradientMultiDeviceCG_vs2019.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
@@ -0,0 +1,109 @@
|
||||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>conjugateGradientMultiDeviceCG_vs2019</RootNamespace>
|
||||
<ProjectName>conjugateGradientMultiDeviceCG</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v142</PlatformToolset>
|
||||
<WindowsTargetPlatformVersion>10.0</WindowsTargetPlatformVersion>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<UseDebugLibraries>true</UseDebugLibraries>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Release'">
|
||||
<WholeProgramOptimization>true</WholeProgramOptimization>
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
|
||||
<ImportGroup Label="ExtensionSettings">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientMultiDeviceCG.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;compute_80,sm_80;compute_86,sm_86;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" --threads 0 </AdditionalOptions>
|
||||
<Include>./;../../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<ExtensibleWholeProgramCompilation>true</ExtensibleWholeProgramCompilation>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="conjugateGradientMultiDeviceCG.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
@@ -0,0 +1,20 @@
|
||||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2022
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientMultiDeviceCG", "conjugateGradientMultiDeviceCG_vs2022.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
@@ -0,0 +1,109 @@
|
||||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>conjugateGradientMultiDeviceCG_vs2022</RootNamespace>
|
||||
<ProjectName>conjugateGradientMultiDeviceCG</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v143</PlatformToolset>
|
||||
<WindowsTargetPlatformVersion>10.0</WindowsTargetPlatformVersion>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<UseDebugLibraries>true</UseDebugLibraries>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Release'">
|
||||
<WholeProgramOptimization>true</WholeProgramOptimization>
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
|
||||
<ImportGroup Label="ExtensionSettings">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientMultiDeviceCG.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;compute_80,sm_80;compute_86,sm_86;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" --threads 0 </AdditionalOptions>
|
||||
<Include>./;../../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<ExtensibleWholeProgramCompilation>true</ExtensibleWholeProgramCompilation>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="conjugateGradientMultiDeviceCG.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
Reference in New Issue
Block a user