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/conjugateGradientPrecond/.vscode/c_cpp_properties.json
vendored
Normal file
18
Samples/4_CUDA_Libraries/conjugateGradientPrecond/.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/conjugateGradientPrecond/.vscode/extensions.json
vendored
Normal file
7
Samples/4_CUDA_Libraries/conjugateGradientPrecond/.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/conjugateGradientPrecond/.vscode/launch.json
vendored
Normal file
10
Samples/4_CUDA_Libraries/conjugateGradientPrecond/.vscode/launch.json
vendored
Normal file
@@ -0,0 +1,10 @@
|
||||
{
|
||||
"configurations": [
|
||||
{
|
||||
"name": "CUDA C++: Launch",
|
||||
"type": "cuda-gdb",
|
||||
"request": "launch",
|
||||
"program": "${workspaceFolder}/conjugateGradientPrecond"
|
||||
}
|
||||
]
|
||||
}
|
||||
15
Samples/4_CUDA_Libraries/conjugateGradientPrecond/.vscode/tasks.json
vendored
Normal file
15
Samples/4_CUDA_Libraries/conjugateGradientPrecond/.vscode/tasks.json
vendored
Normal file
@@ -0,0 +1,15 @@
|
||||
{
|
||||
"version": "2.0.0",
|
||||
"tasks": [
|
||||
{
|
||||
"label": "sample",
|
||||
"type": "shell",
|
||||
"command": "make dbg=1",
|
||||
"problemMatcher": ["$nvcc"],
|
||||
"group": {
|
||||
"kind": "build",
|
||||
"isDefault": true
|
||||
}
|
||||
}
|
||||
]
|
||||
}
|
||||
330
Samples/4_CUDA_Libraries/conjugateGradientPrecond/Makefile
Normal file
330
Samples/4_CUDA_Libraries/conjugateGradientPrecond/Makefile
Normal file
@@ -0,0 +1,330 @@
|
||||
################################################################################
|
||||
# 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))
|
||||
|
||||
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 :=
|
||||
|
||||
################################################################################
|
||||
|
||||
# Gencode arguments
|
||||
SMS ?=
|
||||
|
||||
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_$(sm)))
|
||||
|
||||
ifeq ($(SMS),)
|
||||
ifeq ($(TARGET_ARCH),$(filter $(TARGET_ARCH),armv7l aarch64 sbsa))
|
||||
# Generate PTX code from SM 53
|
||||
GENCODE_FLAGS += -gencode arch=compute_53,code=compute_53
|
||||
else
|
||||
# Generate PTX code from SM 35
|
||||
GENCODE_FLAGS += -gencode arch=compute_35,code=compute_35
|
||||
endif
|
||||
endif
|
||||
|
||||
# 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 += --threads 0 --std=c++11
|
||||
|
||||
LIBRARIES += -lcublas -lcusparse
|
||||
|
||||
################################################################################
|
||||
|
||||
# Target rules
|
||||
all: build
|
||||
|
||||
build: conjugateGradientPrecond
|
||||
|
||||
main.o:main.cpp
|
||||
$(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
conjugateGradientPrecond: main.o
|
||||
$(NVCC) $(ALL_LDFLAGS) $(GENCODE_FLAGS) -o $@ $+ $(LIBRARIES)
|
||||
mkdir -p ../../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)
|
||||
cp $@ ../../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)
|
||||
|
||||
run: build
|
||||
./conjugateGradientPrecond
|
||||
|
||||
testrun: build
|
||||
|
||||
clean:
|
||||
rm -f conjugateGradientPrecond main.o
|
||||
rm -rf ../../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)/conjugateGradientPrecond
|
||||
|
||||
clobber: clean
|
||||
@@ -0,0 +1,88 @@
|
||||
<?xml version="1.0" encoding="UTF-8"?>
|
||||
<!DOCTYPE entry SYSTEM "SamplesInfo.dtd">
|
||||
<entry>
|
||||
<name>conjugateGradientPrecond</name>
|
||||
<cuda_api_list>
|
||||
<toolkit>cudaMemset</toolkit>
|
||||
<toolkit>cudaFree</toolkit>
|
||||
<toolkit>cudaMalloc</toolkit>
|
||||
<toolkit>cudaMemcpy</toolkit>
|
||||
<toolkit>cudaGetDeviceProperties</toolkit>
|
||||
</cuda_api_list>
|
||||
<description><![CDATA[This sample implements a preconditioned conjugate gradient solver on GPU using CUBLAS and CUSPARSE library.]]></description>
|
||||
<devicecompilation>whole</devicecompilation>
|
||||
<fallback_min_ptx>true</fallback_min_ptx>
|
||||
<includepaths>
|
||||
<path>./</path>
|
||||
<path>../</path>
|
||||
<path>../../../Common</path>
|
||||
</includepaths>
|
||||
<keyconcepts>
|
||||
<concept level="advanced">Linear Algebra</concept>
|
||||
<concept level="advanced">CUBLAS Library</concept>
|
||||
<concept level="advanced">CUSPARSE Library</concept>
|
||||
</keyconcepts>
|
||||
<keywords>
|
||||
<keyword>CUDA</keyword>
|
||||
<keyword>CUBLAS</keyword>
|
||||
<keyword>CUSPARSE</keyword>
|
||||
<keyword>Sparse Matrix</keyword>
|
||||
</keywords>
|
||||
<libraries>
|
||||
<library>cublas</library>
|
||||
<library>cusparse</library>
|
||||
</libraries>
|
||||
<librarypaths>
|
||||
</librarypaths>
|
||||
<nsight_eclipse>true</nsight_eclipse>
|
||||
<primary_file>main.cpp</primary_file>
|
||||
<required_dependencies>
|
||||
<dependency>CUBLAS</dependency>
|
||||
<dependency>CUSPARSE</dependency>
|
||||
</required_dependencies>
|
||||
<scopes>
|
||||
<scope>1:CUDA Advanced Topics</scope>
|
||||
<scope>3:Linear Algebra</scope>
|
||||
</scopes>
|
||||
<sm-arch>sm35</sm-arch>
|
||||
<sm-arch>sm37</sm-arch>
|
||||
<sm-arch>sm50</sm-arch>
|
||||
<sm-arch>sm52</sm-arch>
|
||||
<sm-arch>sm53</sm-arch>
|
||||
<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>
|
||||
<platform>windows7</platform>
|
||||
</env>
|
||||
<env>
|
||||
<arch>x86_64</arch>
|
||||
<platform>macosx</platform>
|
||||
</env>
|
||||
<env>
|
||||
<arch>arm</arch>
|
||||
</env>
|
||||
<env>
|
||||
<arch>sbsa</arch>
|
||||
</env>
|
||||
<env>
|
||||
<arch>ppc64le</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
</supported_envs>
|
||||
<supported_sm_architectures>
|
||||
<include>all</include>
|
||||
</supported_sm_architectures>
|
||||
<title>Preconditioned Conjugate Gradient</title>
|
||||
<type>exe</type>
|
||||
</entry>
|
||||
74
Samples/4_CUDA_Libraries/conjugateGradientPrecond/README.md
Normal file
74
Samples/4_CUDA_Libraries/conjugateGradientPrecond/README.md
Normal file
@@ -0,0 +1,74 @@
|
||||
# conjugateGradientPrecond - Preconditioned Conjugate Gradient
|
||||
|
||||
## Description
|
||||
|
||||
This sample implements a preconditioned conjugate gradient solver on GPU using CUBLAS and CUSPARSE library.
|
||||
|
||||
## Key Concepts
|
||||
|
||||
Linear Algebra, CUBLAS Library, CUSPARSE Library
|
||||
|
||||
## Supported SM Architectures
|
||||
|
||||
[SM 3.5 ](https://developer.nvidia.com/cuda-gpus) [SM 3.7 ](https://developer.nvidia.com/cuda-gpus) [SM 5.0 ](https://developer.nvidia.com/cuda-gpus) [SM 5.2 ](https://developer.nvidia.com/cuda-gpus) [SM 5.3 ](https://developer.nvidia.com/cuda-gpus) [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, armv7l
|
||||
|
||||
## CUDA APIs involved
|
||||
|
||||
### [CUDA Runtime API](http://docs.nvidia.com/cuda/cuda-runtime-api/index.html)
|
||||
cudaMemset, cudaFree, cudaMalloc, cudaMemcpy, cudaGetDeviceProperties
|
||||
|
||||
## Dependencies needed to build/run
|
||||
[CUBLAS](../../README.md#cublas), [CUSPARSE](../../README.md#cusparse)
|
||||
|
||||
## 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, armv7l.
|
||||
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=armv7l` <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 +1,20 @@
|
||||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2017
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientPrecond", "conjugateGradientPrecond_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 +1,112 @@
|
||||
<?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>conjugateGradientPrecond_vs2017</RootNamespace>
|
||||
<ProjectName>conjugateGradientPrecond</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;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cublas.lib;cusparse.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientPrecond.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_35,compute_35;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" --threads 0 </AdditionalOptions>
|
||||
<Include>./;../../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(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="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<ClCompile Include="main.cpp" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
@@ -0,0 +1,20 @@
|
||||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2019
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientPrecond", "conjugateGradientPrecond_vs2019.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
@@ -0,0 +1,108 @@
|
||||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>conjugateGradientPrecond_vs2019</RootNamespace>
|
||||
<ProjectName>conjugateGradientPrecond</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(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;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cublas.lib;cusparse.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientPrecond.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_35,compute_35;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" --threads 0 </AdditionalOptions>
|
||||
<Include>./;../../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(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="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<ClCompile Include="main.cpp" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
@@ -0,0 +1,20 @@
|
||||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2022
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientPrecond", "conjugateGradientPrecond_vs2022.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
@@ -0,0 +1,108 @@
|
||||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>conjugateGradientPrecond_vs2022</RootNamespace>
|
||||
<ProjectName>conjugateGradientPrecond</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(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;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cublas.lib;cusparse.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientPrecond.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_35,compute_35;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" --threads 0 </AdditionalOptions>
|
||||
<Include>./;../../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(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="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<ClCompile Include="main.cpp" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
531
Samples/4_CUDA_Libraries/conjugateGradientPrecond/main.cpp
Normal file
531
Samples/4_CUDA_Libraries/conjugateGradientPrecond/main.cpp
Normal file
@@ -0,0 +1,531 @@
|
||||
/* 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 preconditioned conjugate gradient solver on
|
||||
* the GPU using CUBLAS and CUSPARSE. Relative to the conjugateGradient
|
||||
* SDK example, this demonstrates the use of cusparseScsrilu02() for
|
||||
* computing the incompute-LU preconditioner and cusparseScsrsv2_solve()
|
||||
* for solving triangular systems. Specifically, the preconditioned
|
||||
* conjugate gradient method with an incomplete LU preconditioner is
|
||||
* used to solve the Laplacian operator in 2D on a uniform mesh.
|
||||
*
|
||||
* Note that the code in this example and the specific matrices used here
|
||||
* were chosen to demonstrate the use of the CUSPARSE library as simply
|
||||
* and as clearly as possible. This is not optimized code and the input
|
||||
* matrices have been chosen for simplicity rather than performance.
|
||||
* These should not be used either as a performance guide or for
|
||||
* benchmarking purposes.
|
||||
*/
|
||||
|
||||
// includes, system
|
||||
#include <math.h>
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
|
||||
// CUDA Runtime
|
||||
#include <cuda_runtime.h>
|
||||
|
||||
// Using updated (v2) interfaces for CUBLAS and CUSPARSE
|
||||
#include <cublas_v2.h>
|
||||
#include <cusparse.h>
|
||||
|
||||
// Utilities and system includes
|
||||
#include <helper_cuda.h> // CUDA error checking
|
||||
#include <helper_functions.h> // shared functions common to CUDA Samples
|
||||
|
||||
const char *sSDKname = "conjugateGradientPrecond";
|
||||
|
||||
/*
|
||||
* Generate a matrix representing a second order regular Laplacian operator
|
||||
* on a 2D domain in Compressed Sparse Row format.
|
||||
*/
|
||||
void genLaplace(int *row_ptr, int *col_ind, float *val, int M, int N, int nz,
|
||||
float *rhs) {
|
||||
assert(M == N);
|
||||
int n = (int)sqrt((double)N);
|
||||
assert(n * n == N);
|
||||
printf("laplace dimension = %d\n", n);
|
||||
int idx = 0;
|
||||
|
||||
// loop over degrees of freedom
|
||||
for (int i = 0; i < N; i++) {
|
||||
int ix = i % n;
|
||||
int iy = i / n;
|
||||
|
||||
row_ptr[i] = idx;
|
||||
|
||||
// up
|
||||
if (iy > 0) {
|
||||
val[idx] = 1.0;
|
||||
col_ind[idx] = i - n;
|
||||
idx++;
|
||||
} else {
|
||||
rhs[i] -= 1.0;
|
||||
}
|
||||
|
||||
// left
|
||||
if (ix > 0) {
|
||||
val[idx] = 1.0;
|
||||
col_ind[idx] = i - 1;
|
||||
idx++;
|
||||
} else {
|
||||
rhs[i] -= 0.0;
|
||||
}
|
||||
|
||||
// center
|
||||
val[idx] = -4.0;
|
||||
col_ind[idx] = i;
|
||||
idx++;
|
||||
|
||||
// right
|
||||
if (ix < n - 1) {
|
||||
val[idx] = 1.0;
|
||||
col_ind[idx] = i + 1;
|
||||
idx++;
|
||||
} else {
|
||||
rhs[i] -= 0.0;
|
||||
}
|
||||
|
||||
// down
|
||||
if (iy < n - 1) {
|
||||
val[idx] = 1.0;
|
||||
col_ind[idx] = i + n;
|
||||
idx++;
|
||||
} else {
|
||||
rhs[i] -= 0.0;
|
||||
}
|
||||
}
|
||||
|
||||
row_ptr[N] = idx;
|
||||
}
|
||||
|
||||
/*
|
||||
* Solve Ax=b using the conjugate gradient method
|
||||
* a) without any preconditioning,
|
||||
* b) using an Incomplete Cholesky preconditioner, and
|
||||
* c) using an ILU0 preconditioner.
|
||||
*/
|
||||
int main(int argc, char **argv) {
|
||||
const int max_iter = 1000;
|
||||
int k, M = 0, N = 0, nz = 0, *I = NULL, *J = NULL;
|
||||
int *d_col, *d_row;
|
||||
int qatest = 0;
|
||||
const float tol = 1e-12f;
|
||||
float *x, *rhs;
|
||||
float r0, r1, alpha, beta;
|
||||
float *d_val, *d_x;
|
||||
float *d_zm1, *d_zm2, *d_rm2;
|
||||
float *d_r, *d_p, *d_omega, *d_y;
|
||||
float *val = NULL;
|
||||
float *d_valsILU0;
|
||||
void *buffer = NULL;
|
||||
float rsum, diff, err = 0.0;
|
||||
float qaerr1, qaerr2 = 0.0;
|
||||
float dot, numerator, denominator, nalpha;
|
||||
const float floatone = 1.0;
|
||||
const float floatzero = 0.0;
|
||||
|
||||
int nErrors = 0;
|
||||
|
||||
printf("conjugateGradientPrecond starting...\n");
|
||||
|
||||
/* QA testing mode */
|
||||
if (checkCmdLineFlag(argc, (const char **)argv, "qatest")) {
|
||||
qatest = 1;
|
||||
}
|
||||
|
||||
/* This will pick the best possible CUDA capable device */
|
||||
cudaDeviceProp deviceProp;
|
||||
int devID = findCudaDevice(argc, (const char **)argv);
|
||||
printf("GPU selected Device ID = %d \n", devID);
|
||||
|
||||
if (devID < 0) {
|
||||
printf("Invalid GPU device %d selected, exiting...\n", devID);
|
||||
exit(EXIT_SUCCESS);
|
||||
}
|
||||
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, devID));
|
||||
|
||||
/* Statistics about the GPU device */
|
||||
printf(
|
||||
"> GPU device has %d Multi-Processors, "
|
||||
"SM %d.%d compute capabilities\n\n",
|
||||
deviceProp.multiProcessorCount, deviceProp.major, deviceProp.minor);
|
||||
|
||||
/* Generate a Laplace matrix in CSR (Compressed Sparse Row) format */
|
||||
M = N = 16384;
|
||||
nz = 5 * N - 4 * (int)sqrt((double)N);
|
||||
I = (int *)malloc(sizeof(int) * (N + 1)); // csr row pointers for matrix A
|
||||
J = (int *)malloc(sizeof(int) * nz); // csr column indices for matrix A
|
||||
val = (float *)malloc(sizeof(float) * nz); // csr values for matrix A
|
||||
x = (float *)malloc(sizeof(float) * N);
|
||||
rhs = (float *)malloc(sizeof(float) * N);
|
||||
|
||||
for (int i = 0; i < N; i++) {
|
||||
rhs[i] = 0.0; // Initialize RHS
|
||||
x[i] = 0.0; // Initial solution approximation
|
||||
}
|
||||
|
||||
genLaplace(I, J, val, M, N, nz, rhs);
|
||||
|
||||
/* Create CUBLAS context */
|
||||
cublasHandle_t cublasHandle = NULL;
|
||||
checkCudaErrors(cublasCreate(&cublasHandle));
|
||||
|
||||
/* Create CUSPARSE context */
|
||||
cusparseHandle_t cusparseHandle = NULL;
|
||||
checkCudaErrors(cusparseCreate(&cusparseHandle));
|
||||
|
||||
/* Description of the A matrix */
|
||||
cusparseMatDescr_t descr = 0;
|
||||
checkCudaErrors(cusparseCreateMatDescr(&descr));
|
||||
checkCudaErrors(cusparseSetMatType(descr, CUSPARSE_MATRIX_TYPE_GENERAL));
|
||||
checkCudaErrors(cusparseSetMatIndexBase(descr, CUSPARSE_INDEX_BASE_ZERO));
|
||||
|
||||
/* Allocate required memory */
|
||||
checkCudaErrors(cudaMalloc((void **)&d_col, nz * sizeof(int)));
|
||||
checkCudaErrors(cudaMalloc((void **)&d_row, (N + 1) * sizeof(int)));
|
||||
checkCudaErrors(cudaMalloc((void **)&d_val, nz * sizeof(float)));
|
||||
checkCudaErrors(cudaMalloc((void **)&d_x, N * sizeof(float)));
|
||||
checkCudaErrors(cudaMalloc((void **)&d_y, N * sizeof(float)));
|
||||
checkCudaErrors(cudaMalloc((void **)&d_r, N * sizeof(float)));
|
||||
checkCudaErrors(cudaMalloc((void **)&d_p, N * sizeof(float)));
|
||||
checkCudaErrors(cudaMalloc((void **)&d_omega, N * sizeof(float)));
|
||||
checkCudaErrors(cudaMalloc((void **)&d_valsILU0, nz * sizeof(float)));
|
||||
checkCudaErrors(cudaMalloc((void **)&d_zm1, (N) * sizeof(float)));
|
||||
checkCudaErrors(cudaMalloc((void **)&d_zm2, (N) * sizeof(float)));
|
||||
checkCudaErrors(cudaMalloc((void **)&d_rm2, (N) * sizeof(float)));
|
||||
|
||||
/* Wrap raw data into cuSPARSE generic API objects */
|
||||
cusparseSpMatDescr_t matA = NULL;
|
||||
checkCudaErrors(cusparseCreateCsr(&matA, N, N, nz, d_row, d_col, d_val,
|
||||
CUSPARSE_INDEX_32I, CUSPARSE_INDEX_32I,
|
||||
CUSPARSE_INDEX_BASE_ZERO, CUDA_R_32F));
|
||||
cusparseDnVecDescr_t vecp = NULL;
|
||||
checkCudaErrors(cusparseCreateDnVec(&vecp, N, d_p, CUDA_R_32F));
|
||||
cusparseDnVecDescr_t vecomega = NULL;
|
||||
checkCudaErrors(cusparseCreateDnVec(&vecomega, N, d_omega, CUDA_R_32F));
|
||||
|
||||
/* Initialize problem data */
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(d_col, J, nz * sizeof(int), cudaMemcpyHostToDevice));
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(d_row, I, (N + 1) * sizeof(int), cudaMemcpyHostToDevice));
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(d_val, val, nz * sizeof(float), cudaMemcpyHostToDevice));
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(d_x, x, N * sizeof(float), cudaMemcpyHostToDevice));
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(d_r, rhs, N * sizeof(float), cudaMemcpyHostToDevice));
|
||||
checkCudaErrors(cudaMemset(d_y, 0, sizeof(float) * N));
|
||||
|
||||
/* Create ILU(0) info object */
|
||||
csrilu02Info_t infoILU = NULL;
|
||||
checkCudaErrors(cusparseCreateCsrilu02Info(&infoILU));
|
||||
|
||||
/* Create L factor descriptor and triangular solve info */
|
||||
cusparseMatDescr_t descrL = NULL;
|
||||
checkCudaErrors(cusparseCreateMatDescr(&descrL));
|
||||
checkCudaErrors(cusparseSetMatType(descrL, CUSPARSE_MATRIX_TYPE_GENERAL));
|
||||
checkCudaErrors(cusparseSetMatIndexBase(descrL, CUSPARSE_INDEX_BASE_ZERO));
|
||||
checkCudaErrors(cusparseSetMatFillMode(descrL, CUSPARSE_FILL_MODE_LOWER));
|
||||
checkCudaErrors(cusparseSetMatDiagType(descrL, CUSPARSE_DIAG_TYPE_UNIT));
|
||||
csrsv2Info_t infoL = NULL;
|
||||
checkCudaErrors(cusparseCreateCsrsv2Info(&infoL));
|
||||
|
||||
/* Create U factor descriptor and triangular solve info */
|
||||
cusparseMatDescr_t descrU = NULL;
|
||||
checkCudaErrors(cusparseCreateMatDescr(&descrU));
|
||||
checkCudaErrors(cusparseSetMatType(descrU, CUSPARSE_MATRIX_TYPE_GENERAL));
|
||||
checkCudaErrors(cusparseSetMatIndexBase(descrU, CUSPARSE_INDEX_BASE_ZERO));
|
||||
checkCudaErrors(cusparseSetMatFillMode(descrU, CUSPARSE_FILL_MODE_UPPER));
|
||||
checkCudaErrors(cusparseSetMatDiagType(descrU, CUSPARSE_DIAG_TYPE_NON_UNIT));
|
||||
csrsv2Info_t infoU = NULL;
|
||||
checkCudaErrors(cusparseCreateCsrsv2Info(&infoU));
|
||||
|
||||
/* Allocate workspace for cuSPARSE */
|
||||
size_t bufferSize = 0;
|
||||
size_t tmp = 0;
|
||||
int stmp = 0;
|
||||
checkCudaErrors(cusparseSpMV_bufferSize(
|
||||
cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE, &floatone, matA, vecp,
|
||||
&floatzero, vecomega, CUDA_R_32F, CUSPARSE_SPMV_ALG_DEFAULT, &tmp));
|
||||
if (tmp > bufferSize) {
|
||||
bufferSize = stmp;
|
||||
}
|
||||
checkCudaErrors(cusparseScsrilu02_bufferSize(
|
||||
cusparseHandle, N, nz, descr, d_val, d_row, d_col, infoILU, &stmp));
|
||||
if (stmp > bufferSize) {
|
||||
bufferSize = stmp;
|
||||
}
|
||||
checkCudaErrors(cusparseScsrsv2_bufferSize(
|
||||
cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE, N, nz, descrL, d_val,
|
||||
d_row, d_col, infoL, &stmp));
|
||||
if (stmp > bufferSize) {
|
||||
bufferSize = stmp;
|
||||
}
|
||||
checkCudaErrors(cusparseScsrsv2_bufferSize(
|
||||
cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE, N, nz, descrU, d_val,
|
||||
d_row, d_col, infoU, &stmp));
|
||||
if (stmp > bufferSize) {
|
||||
bufferSize = stmp;
|
||||
}
|
||||
checkCudaErrors(cudaMalloc(&buffer, bufferSize));
|
||||
|
||||
/* Conjugate gradient without preconditioning.
|
||||
------------------------------------------
|
||||
|
||||
Follows the description by Golub & Van Loan,
|
||||
"Matrix Computations 3rd ed.", Section 10.2.6 */
|
||||
|
||||
printf("Convergence of CG without preconditioning: \n");
|
||||
k = 0;
|
||||
r0 = 0;
|
||||
checkCudaErrors(cublasSdot(cublasHandle, N, d_r, 1, d_r, 1, &r1));
|
||||
|
||||
while (r1 > tol * tol && k <= max_iter) {
|
||||
k++;
|
||||
|
||||
if (k == 1) {
|
||||
checkCudaErrors(cublasScopy(cublasHandle, N, d_r, 1, d_p, 1));
|
||||
} else {
|
||||
beta = r1 / r0;
|
||||
checkCudaErrors(cublasSscal(cublasHandle, N, &beta, d_p, 1));
|
||||
checkCudaErrors(cublasSaxpy(cublasHandle, N, &floatone, d_r, 1, d_p, 1));
|
||||
}
|
||||
|
||||
checkCudaErrors(cusparseSpMV(
|
||||
cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE, &floatone, matA, vecp,
|
||||
&floatzero, vecomega, CUDA_R_32F, CUSPARSE_SPMV_ALG_DEFAULT, buffer));
|
||||
checkCudaErrors(cublasSdot(cublasHandle, N, d_p, 1, d_omega, 1, &dot));
|
||||
alpha = r1 / dot;
|
||||
checkCudaErrors(cublasSaxpy(cublasHandle, N, &alpha, d_p, 1, d_x, 1));
|
||||
nalpha = -alpha;
|
||||
checkCudaErrors(cublasSaxpy(cublasHandle, N, &nalpha, d_omega, 1, d_r, 1));
|
||||
r0 = r1;
|
||||
checkCudaErrors(cublasSdot(cublasHandle, N, d_r, 1, d_r, 1, &r1));
|
||||
}
|
||||
|
||||
printf(" iteration = %3d, residual = %e \n", k, sqrt(r1));
|
||||
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(x, d_x, N * sizeof(float), cudaMemcpyDeviceToHost));
|
||||
|
||||
/* check result */
|
||||
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[j] * x[J[j]];
|
||||
}
|
||||
|
||||
diff = fabs(rsum - rhs[i]);
|
||||
|
||||
if (diff > err) {
|
||||
err = diff;
|
||||
}
|
||||
}
|
||||
|
||||
printf(" Convergence Test: %s \n", (k <= max_iter) ? "OK" : "FAIL");
|
||||
nErrors += (k > max_iter) ? 1 : 0;
|
||||
qaerr1 = err;
|
||||
|
||||
if (0) {
|
||||
// output result in matlab-style array
|
||||
int n = (int)sqrt((double)N);
|
||||
printf("a = [ ");
|
||||
|
||||
for (int iy = 0; iy < n; iy++) {
|
||||
for (int ix = 0; ix < n; ix++) {
|
||||
printf(" %f ", x[iy * n + ix]);
|
||||
}
|
||||
|
||||
if (iy == n - 1) {
|
||||
printf(" ]");
|
||||
}
|
||||
|
||||
printf("\n");
|
||||
}
|
||||
}
|
||||
|
||||
/* Preconditioned Conjugate Gradient using ILU.
|
||||
--------------------------------------------
|
||||
Follows the description by Golub & Van Loan,
|
||||
"Matrix Computations 3rd ed.", Algorithm 10.3.1 */
|
||||
|
||||
printf("\nConvergence of CG using ILU(0) preconditioning: \n");
|
||||
|
||||
/* Perform analysis for ILU(0) */
|
||||
checkCudaErrors(cusparseScsrilu02_analysis(
|
||||
cusparseHandle, N, nz, descr, d_val, d_row, d_col, infoILU,
|
||||
CUSPARSE_SOLVE_POLICY_USE_LEVEL, buffer));
|
||||
|
||||
/* Copy A data to ILU(0) vals as input*/
|
||||
checkCudaErrors(cudaMemcpy(d_valsILU0, d_val, nz * sizeof(float),
|
||||
cudaMemcpyDeviceToDevice));
|
||||
|
||||
/* generate the ILU(0) factors */
|
||||
checkCudaErrors(cusparseScsrilu02(cusparseHandle, N, nz, descr, d_valsILU0,
|
||||
d_row, d_col, infoILU,
|
||||
CUSPARSE_SOLVE_POLICY_USE_LEVEL, buffer));
|
||||
|
||||
/* perform triangular solve analysis */
|
||||
checkCudaErrors(
|
||||
cusparseScsrsv2_analysis(cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE,
|
||||
N, nz, descrL, d_valsILU0, d_row, d_col, infoL,
|
||||
CUSPARSE_SOLVE_POLICY_USE_LEVEL, buffer));
|
||||
checkCudaErrors(
|
||||
cusparseScsrsv2_analysis(cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE,
|
||||
N, nz, descrU, d_valsILU0, d_row, d_col, infoU,
|
||||
CUSPARSE_SOLVE_POLICY_USE_LEVEL, buffer));
|
||||
|
||||
/* reset the initial guess of the solution to zero */
|
||||
for (int i = 0; i < N; i++) {
|
||||
x[i] = 0.0;
|
||||
}
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(d_r, rhs, N * sizeof(float), cudaMemcpyHostToDevice));
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(d_x, x, N * sizeof(float), cudaMemcpyHostToDevice));
|
||||
|
||||
k = 0;
|
||||
checkCudaErrors(cublasSdot(cublasHandle, N, d_r, 1, d_r, 1, &r1));
|
||||
|
||||
while (r1 > tol * tol && k <= max_iter) {
|
||||
// preconditioner application: d_zm1 = U^-1 L^-1 d_r
|
||||
checkCudaErrors(cusparseScsrsv2_solve(
|
||||
cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE, N, nz, &floatone,
|
||||
descrL, d_valsILU0, d_row, d_col, infoL, d_r, d_y,
|
||||
CUSPARSE_SOLVE_POLICY_USE_LEVEL, buffer));
|
||||
checkCudaErrors(cusparseScsrsv2_solve(
|
||||
cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE, N, nz, &floatone,
|
||||
descrU, d_valsILU0, d_row, d_col, infoU, d_y, d_zm1,
|
||||
CUSPARSE_SOLVE_POLICY_USE_LEVEL, buffer));
|
||||
|
||||
k++;
|
||||
|
||||
if (k == 1) {
|
||||
checkCudaErrors(cublasScopy(cublasHandle, N, d_zm1, 1, d_p, 1));
|
||||
} else {
|
||||
checkCudaErrors(
|
||||
cublasSdot(cublasHandle, N, d_r, 1, d_zm1, 1, &numerator));
|
||||
checkCudaErrors(
|
||||
cublasSdot(cublasHandle, N, d_rm2, 1, d_zm2, 1, &denominator));
|
||||
beta = numerator / denominator;
|
||||
checkCudaErrors(cublasSscal(cublasHandle, N, &beta, d_p, 1));
|
||||
checkCudaErrors(
|
||||
cublasSaxpy(cublasHandle, N, &floatone, d_zm1, 1, d_p, 1));
|
||||
}
|
||||
|
||||
checkCudaErrors(cusparseSpMV(
|
||||
cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE, &floatone, matA, vecp,
|
||||
&floatzero, vecomega, CUDA_R_32F, CUSPARSE_SPMV_ALG_DEFAULT, buffer));
|
||||
checkCudaErrors(cublasSdot(cublasHandle, N, d_r, 1, d_zm1, 1, &numerator));
|
||||
checkCudaErrors(
|
||||
cublasSdot(cublasHandle, N, d_p, 1, d_omega, 1, &denominator));
|
||||
alpha = numerator / denominator;
|
||||
checkCudaErrors(cublasSaxpy(cublasHandle, N, &alpha, d_p, 1, d_x, 1));
|
||||
checkCudaErrors(cublasScopy(cublasHandle, N, d_r, 1, d_rm2, 1));
|
||||
checkCudaErrors(cublasScopy(cublasHandle, N, d_zm1, 1, d_zm2, 1));
|
||||
nalpha = -alpha;
|
||||
checkCudaErrors(cublasSaxpy(cublasHandle, N, &nalpha, d_omega, 1, d_r, 1));
|
||||
checkCudaErrors(cublasSdot(cublasHandle, N, d_r, 1, d_r, 1, &r1));
|
||||
}
|
||||
|
||||
printf(" iteration = %3d, residual = %e \n", k, sqrt(r1));
|
||||
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(x, d_x, N * sizeof(float), cudaMemcpyDeviceToHost));
|
||||
|
||||
/* check result */
|
||||
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[j] * x[J[j]];
|
||||
}
|
||||
|
||||
diff = fabs(rsum - rhs[i]);
|
||||
|
||||
if (diff > err) {
|
||||
err = diff;
|
||||
}
|
||||
}
|
||||
|
||||
printf(" Convergence Test: %s \n", (k <= max_iter) ? "OK" : "FAIL");
|
||||
nErrors += (k > max_iter) ? 1 : 0;
|
||||
qaerr2 = err;
|
||||
|
||||
/* Destroy descriptors */
|
||||
checkCudaErrors(cusparseDestroyCsrsv2Info(infoU));
|
||||
checkCudaErrors(cusparseDestroyCsrsv2Info(infoL));
|
||||
checkCudaErrors(cusparseDestroyCsrilu02Info(infoILU));
|
||||
checkCudaErrors(cusparseDestroyMatDescr(descrL));
|
||||
checkCudaErrors(cusparseDestroyMatDescr(descrU));
|
||||
checkCudaErrors(cusparseDestroyMatDescr(descr));
|
||||
checkCudaErrors(cusparseDestroySpMat(matA));
|
||||
checkCudaErrors(cusparseDestroyDnVec(vecp));
|
||||
checkCudaErrors(cusparseDestroyDnVec(vecomega));
|
||||
|
||||
/* Destroy contexts */
|
||||
checkCudaErrors(cusparseDestroy(cusparseHandle));
|
||||
checkCudaErrors(cublasDestroy(cublasHandle));
|
||||
|
||||
/* Free device memory */
|
||||
free(I);
|
||||
free(J);
|
||||
free(val);
|
||||
free(x);
|
||||
free(rhs);
|
||||
checkCudaErrors(cudaFree(buffer));
|
||||
checkCudaErrors(cudaFree(d_col));
|
||||
checkCudaErrors(cudaFree(d_row));
|
||||
checkCudaErrors(cudaFree(d_val));
|
||||
checkCudaErrors(cudaFree(d_x));
|
||||
checkCudaErrors(cudaFree(d_y));
|
||||
checkCudaErrors(cudaFree(d_r));
|
||||
checkCudaErrors(cudaFree(d_p));
|
||||
checkCudaErrors(cudaFree(d_omega));
|
||||
checkCudaErrors(cudaFree(d_valsILU0));
|
||||
checkCudaErrors(cudaFree(d_zm1));
|
||||
checkCudaErrors(cudaFree(d_zm2));
|
||||
checkCudaErrors(cudaFree(d_rm2));
|
||||
|
||||
printf("\n");
|
||||
printf("Test Summary:\n");
|
||||
printf(" Counted total of %d errors\n", nErrors);
|
||||
printf(" qaerr1 = %f qaerr2 = %f\n\n", fabs(qaerr1), fabs(qaerr2));
|
||||
exit((nErrors == 0 && fabs(qaerr1) < 1e-5 && fabs(qaerr2) < 1e-5
|
||||
? EXIT_SUCCESS
|
||||
: EXIT_FAILURE));
|
||||
}
|
||||
Reference in New Issue
Block a user