mirror of
https://github.com/NVIDIA/cuda-samples.git
synced 2026-10-23 21:28:17 +08:00
add and update samples for CUDA 11.5
This commit is contained in:
Rutwik Choughule
18
Samples/transpose/.vscode/c_cpp_properties.json
vendored
Normal file
18
Samples/transpose/.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/transpose/.vscode/extensions.json
vendored
Normal file
7
Samples/transpose/.vscode/extensions.json
vendored
Normal file
@@ -0,0 +1,7 @@
|
||||
{
|
||||
"recommendations": [
|
||||
"nvidia.nsight-vscode-edition",
|
||||
"ms-vscode.cpptools",
|
||||
"ms-vscode.makefile-tools"
|
||||
]
|
||||
}
|
||||
10
Samples/transpose/.vscode/launch.json
vendored
Normal file
10
Samples/transpose/.vscode/launch.json
vendored
Normal file
@@ -0,0 +1,10 @@
|
||||
{
|
||||
"configurations": [
|
||||
{
|
||||
"name": "CUDA C++: Launch",
|
||||
"type": "cuda-gdb",
|
||||
"request": "launch",
|
||||
"program": "${workspaceFolder}/transpose"
|
||||
}
|
||||
]
|
||||
}
|
||||
15
Samples/transpose/.vscode/tasks.json
vendored
Normal file
15
Samples/transpose/.vscode/tasks.json
vendored
Normal file
@@ -0,0 +1,15 @@
|
||||
{
|
||||
"version": "2.0.0",
|
||||
"tasks": [
|
||||
{
|
||||
"label": "sample",
|
||||
"type": "shell",
|
||||
"command": "make dbg=1",
|
||||
"problemMatcher": ["$nvcc"],
|
||||
"group": {
|
||||
"kind": "build",
|
||||
"isDefault": true
|
||||
}
|
||||
}
|
||||
]
|
||||
}
|
||||
362
Samples/transpose/Makefile
Normal file
362
Samples/transpose/Makefile
Normal file
@@ -0,0 +1,362 @@
|
||||
################################################################################
|
||||
# Copyright (c) 2019, 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 += --qpp-config 5.4.0,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
|
||||
|
||||
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 $(GCCVERSION)` \>= 51000)
|
||||
|
||||
ifeq ($(IS_MIN_VERSION), 1)
|
||||
$(info >>> GCC Version is greater or equal to 5.1.0 <<<)
|
||||
else
|
||||
$(info >>> Waiving build. Minimum GCC version required is 5.1.0<<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
endif
|
||||
|
||||
# Gencode arguments
|
||||
ifeq ($(TARGET_ARCH),$(filter $(TARGET_ARCH),armv7l aarch64))
|
||||
SMS ?= 35 37 50 52 60 61 70 72 75 80 86
|
||||
else
|
||||
SMS ?= 35 37 50 52 60 61 70 75 80 86
|
||||
endif
|
||||
|
||||
ifeq ($(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 += --threads 0 --std=c++11
|
||||
|
||||
ifeq ($(SAMPLE_ENABLED),0)
|
||||
EXEC ?= @echo "[@]"
|
||||
endif
|
||||
|
||||
################################################################################
|
||||
|
||||
# Target rules
|
||||
all: build
|
||||
|
||||
build: transpose
|
||||
|
||||
check.deps:
|
||||
ifeq ($(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
|
||||
|
||||
transpose.o:transpose.cu
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
transpose: transpose.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
|
||||
$(EXEC) ./transpose
|
||||
|
||||
clean:
|
||||
rm -f transpose transpose.o
|
||||
rm -rf ../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)/transpose
|
||||
|
||||
clobber: clean
|
||||
69
Samples/transpose/NsightEclipse.xml
Normal file
69
Samples/transpose/NsightEclipse.xml
Normal file
@@ -0,0 +1,69 @@
|
||||
<?xml version="1.0" encoding="UTF-8"?>
|
||||
<!DOCTYPE entry SYSTEM "SamplesInfo.dtd">
|
||||
<entry>
|
||||
<name>transpose</name>
|
||||
<description><![CDATA[This sample demonstrates Matrix Transpose. Different performance are shown to achieve high performance.]]></description>
|
||||
<devicecompilation>whole</devicecompilation>
|
||||
<includepaths>
|
||||
<path>./</path>
|
||||
<path>../</path>
|
||||
<path>../../Common</path>
|
||||
</includepaths>
|
||||
<keyconcepts>
|
||||
<concept level="advanced">Performance Strategies</concept>
|
||||
<concept level="advanced">Linear Algebra</concept>
|
||||
</keyconcepts>
|
||||
<keywords>
|
||||
<keyword>GPGPU</keyword>
|
||||
<keyword>matrix</keyword>
|
||||
<keyword>transpose</keyword>
|
||||
</keywords>
|
||||
<libraries>
|
||||
</libraries>
|
||||
<librarypaths>
|
||||
</librarypaths>
|
||||
<nsight_eclipse>true</nsight_eclipse>
|
||||
<primary_file>transpose.cu</primary_file>
|
||||
<scopes>
|
||||
<scope>1:CUDA Advanced Topics</scope>
|
||||
<scope>1:Performance Strategies</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>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>
|
||||
<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>ppc64le</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
</supported_envs>
|
||||
<supported_sm_architectures>
|
||||
<include>all</include>
|
||||
</supported_sm_architectures>
|
||||
<title>Matrix Transpose</title>
|
||||
<type>exe</type>
|
||||
<whitepaper>doc\MatrixTranspose.pdf</whitepaper>
|
||||
</entry>
|
||||
67
Samples/transpose/README.md
Normal file
67
Samples/transpose/README.md
Normal file
@@ -0,0 +1,67 @@
|
||||
# transpose - Matrix Transpose
|
||||
|
||||
## Description
|
||||
|
||||
This sample demonstrates Matrix Transpose. Different performance are shown to achieve high performance.
|
||||
|
||||
## Key Concepts
|
||||
|
||||
Performance Strategies, Linear Algebra
|
||||
|
||||
## 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 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)
|
||||
|
||||
## Supported OSes
|
||||
|
||||
Linux, Windows
|
||||
|
||||
## Supported CPU Architecture
|
||||
|
||||
x86_64, ppc64le, armv7l
|
||||
|
||||
## CUDA APIs involved
|
||||
|
||||
## Prerequisites
|
||||
|
||||
Download and install the [CUDA Toolkit 11.5](https://developer.nvidia.com/cuda-downloads) for your corresponding platform.
|
||||
|
||||
## 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)
|
||||
|
||||
BIN
Samples/transpose/doc/MatrixTranspose.docx
Normal file
BIN
Samples/transpose/doc/MatrixTranspose.docx
Normal file
Binary file not shown.
BIN
Samples/transpose/doc/MatrixTranspose.pdf
Normal file
BIN
Samples/transpose/doc/MatrixTranspose.pdf
Normal file
Binary file not shown.
613
Samples/transpose/transpose.cu
Normal file
613
Samples/transpose/transpose.cu
Normal file
@@ -0,0 +1,613 @@
|
||||
/* Copyright (c) 2021, 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.
|
||||
*/
|
||||
|
||||
// -----------------------------------------------------------------------------
|
||||
// Transpose
|
||||
//
|
||||
// This file contains both device and host code for transposing a floating-point
|
||||
// matrix. It performs several transpose kernels, which incrementally improve
|
||||
// performance through coalescing, removing shared memory bank conflicts, and
|
||||
// eliminating partition camping. Several of the kernels perform a copy, used
|
||||
// to represent the best case performance that a transpose can achieve.
|
||||
//
|
||||
// Please see the whitepaper in the docs folder of the transpose project for a
|
||||
// detailed description of this performance study.
|
||||
// -----------------------------------------------------------------------------
|
||||
|
||||
#include <cooperative_groups.h>
|
||||
|
||||
namespace cg = cooperative_groups;
|
||||
// Utilities and system includes
|
||||
#include <helper_string.h> // helper for string parsing
|
||||
#include <helper_image.h> // helper for image and data comparison
|
||||
#include <helper_cuda.h> // helper for cuda error checking functions
|
||||
|
||||
const char *sSDKsample = "Transpose";
|
||||
|
||||
// Each block transposes/copies a tile of TILE_DIM x TILE_DIM elements
|
||||
// using TILE_DIM x BLOCK_ROWS threads, so that each thread transposes
|
||||
// TILE_DIM/BLOCK_ROWS elements. TILE_DIM must be an integral multiple of
|
||||
// BLOCK_ROWS
|
||||
|
||||
#define TILE_DIM 16
|
||||
#define BLOCK_ROWS 16
|
||||
|
||||
// This sample assumes that MATRIX_SIZE_X = MATRIX_SIZE_Y
|
||||
int MATRIX_SIZE_X = 1024;
|
||||
int MATRIX_SIZE_Y = 1024;
|
||||
int MUL_FACTOR = TILE_DIM;
|
||||
|
||||
#define FLOOR(a, b) (a - (a % b))
|
||||
|
||||
// Compute the tile size necessary to illustrate performance cases for SM20+
|
||||
// hardware
|
||||
int MAX_TILES = (FLOOR(MATRIX_SIZE_X, 512) * FLOOR(MATRIX_SIZE_Y, 512)) /
|
||||
(TILE_DIM * TILE_DIM);
|
||||
|
||||
// Number of repetitions used for timing. Two sets of repetitions are
|
||||
// performed: 1) over kernel launches and 2) inside the kernel over just the
|
||||
// loads and stores
|
||||
|
||||
#define NUM_REPS 100
|
||||
|
||||
// -------------------------------------------------------
|
||||
// Copies
|
||||
// width and height must be integral multiples of TILE_DIM
|
||||
// -------------------------------------------------------
|
||||
|
||||
__global__ void copy(float *odata, float *idata, int width, int height) {
|
||||
int xIndex = blockIdx.x * TILE_DIM + threadIdx.x;
|
||||
int yIndex = blockIdx.y * TILE_DIM + threadIdx.y;
|
||||
|
||||
int index = xIndex + width * yIndex;
|
||||
|
||||
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS) {
|
||||
odata[index + i * width] = idata[index + i * width];
|
||||
}
|
||||
}
|
||||
|
||||
__global__ void copySharedMem(float *odata, float *idata, int width,
|
||||
int height) {
|
||||
// Handle to thread block group
|
||||
cg::thread_block cta = cg::this_thread_block();
|
||||
__shared__ float tile[TILE_DIM][TILE_DIM];
|
||||
|
||||
int xIndex = blockIdx.x * TILE_DIM + threadIdx.x;
|
||||
int yIndex = blockIdx.y * TILE_DIM + threadIdx.y;
|
||||
|
||||
int index = xIndex + width * yIndex;
|
||||
|
||||
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS) {
|
||||
if (xIndex < width && yIndex < height) {
|
||||
tile[threadIdx.y][threadIdx.x] = idata[index];
|
||||
}
|
||||
}
|
||||
|
||||
cg::sync(cta);
|
||||
|
||||
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS) {
|
||||
if (xIndex < height && yIndex < width) {
|
||||
odata[index] = tile[threadIdx.y][threadIdx.x];
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// -------------------------------------------------------
|
||||
// Transposes
|
||||
// width and height must be integral multiples of TILE_DIM
|
||||
// -------------------------------------------------------
|
||||
|
||||
__global__ void transposeNaive(float *odata, float *idata, int width,
|
||||
int height) {
|
||||
int xIndex = blockIdx.x * TILE_DIM + threadIdx.x;
|
||||
int yIndex = blockIdx.y * TILE_DIM + threadIdx.y;
|
||||
|
||||
int index_in = xIndex + width * yIndex;
|
||||
int index_out = yIndex + height * xIndex;
|
||||
|
||||
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS) {
|
||||
odata[index_out + i] = idata[index_in + i * width];
|
||||
}
|
||||
}
|
||||
|
||||
// coalesced transpose (with bank conflicts)
|
||||
|
||||
__global__ void transposeCoalesced(float *odata, float *idata, int width,
|
||||
int height) {
|
||||
// Handle to thread block group
|
||||
cg::thread_block cta = cg::this_thread_block();
|
||||
__shared__ float tile[TILE_DIM][TILE_DIM];
|
||||
|
||||
int xIndex = blockIdx.x * TILE_DIM + threadIdx.x;
|
||||
int yIndex = blockIdx.y * TILE_DIM + threadIdx.y;
|
||||
int index_in = xIndex + (yIndex)*width;
|
||||
|
||||
xIndex = blockIdx.y * TILE_DIM + threadIdx.x;
|
||||
yIndex = blockIdx.x * TILE_DIM + threadIdx.y;
|
||||
int index_out = xIndex + (yIndex)*height;
|
||||
|
||||
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS) {
|
||||
tile[threadIdx.y + i][threadIdx.x] = idata[index_in + i * width];
|
||||
}
|
||||
|
||||
cg::sync(cta);
|
||||
|
||||
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS) {
|
||||
odata[index_out + i * height] = tile[threadIdx.x][threadIdx.y + i];
|
||||
}
|
||||
}
|
||||
|
||||
// Coalesced transpose with no bank conflicts
|
||||
|
||||
__global__ void transposeNoBankConflicts(float *odata, float *idata, int width,
|
||||
int height) {
|
||||
// Handle to thread block group
|
||||
cg::thread_block cta = cg::this_thread_block();
|
||||
__shared__ float tile[TILE_DIM][TILE_DIM + 1];
|
||||
|
||||
int xIndex = blockIdx.x * TILE_DIM + threadIdx.x;
|
||||
int yIndex = blockIdx.y * TILE_DIM + threadIdx.y;
|
||||
int index_in = xIndex + (yIndex)*width;
|
||||
|
||||
xIndex = blockIdx.y * TILE_DIM + threadIdx.x;
|
||||
yIndex = blockIdx.x * TILE_DIM + threadIdx.y;
|
||||
int index_out = xIndex + (yIndex)*height;
|
||||
|
||||
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS) {
|
||||
tile[threadIdx.y + i][threadIdx.x] = idata[index_in + i * width];
|
||||
}
|
||||
|
||||
cg::sync(cta);
|
||||
|
||||
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS) {
|
||||
odata[index_out + i * height] = tile[threadIdx.x][threadIdx.y + i];
|
||||
}
|
||||
}
|
||||
|
||||
// Transpose that effectively reorders execution of thread blocks along
|
||||
// diagonals of the matrix (also coalesced and has no bank conflicts)
|
||||
//
|
||||
// Here blockIdx.x is interpreted as the distance along a diagonal and
|
||||
// blockIdx.y as corresponding to different diagonals
|
||||
//
|
||||
// blockIdx_x and blockIdx_y expressions map the diagonal coordinates to the
|
||||
// more commonly used cartesian coordinates so that the only changes to the code
|
||||
// from the coalesced version are the calculation of the blockIdx_x and
|
||||
// blockIdx_y and replacement of blockIdx.x and bloclIdx.y with the subscripted
|
||||
// versions in the remaining code
|
||||
|
||||
__global__ void transposeDiagonal(float *odata, float *idata, int width,
|
||||
int height) {
|
||||
// Handle to thread block group
|
||||
cg::thread_block cta = cg::this_thread_block();
|
||||
__shared__ float tile[TILE_DIM][TILE_DIM + 1];
|
||||
|
||||
int blockIdx_x, blockIdx_y;
|
||||
|
||||
// do diagonal reordering
|
||||
if (width == height) {
|
||||
blockIdx_y = blockIdx.x;
|
||||
blockIdx_x = (blockIdx.x + blockIdx.y) % gridDim.x;
|
||||
} else {
|
||||
int bid = blockIdx.x + gridDim.x * blockIdx.y;
|
||||
blockIdx_y = bid % gridDim.y;
|
||||
blockIdx_x = ((bid / gridDim.y) + blockIdx_y) % gridDim.x;
|
||||
}
|
||||
|
||||
// from here on the code is same as previous kernel except blockIdx_x replaces
|
||||
// blockIdx.x and similarly for y
|
||||
|
||||
int xIndex = blockIdx_x * TILE_DIM + threadIdx.x;
|
||||
int yIndex = blockIdx_y * TILE_DIM + threadIdx.y;
|
||||
int index_in = xIndex + (yIndex)*width;
|
||||
|
||||
xIndex = blockIdx_y * TILE_DIM + threadIdx.x;
|
||||
yIndex = blockIdx_x * TILE_DIM + threadIdx.y;
|
||||
int index_out = xIndex + (yIndex)*height;
|
||||
|
||||
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS) {
|
||||
tile[threadIdx.y + i][threadIdx.x] = idata[index_in + i * width];
|
||||
}
|
||||
|
||||
cg::sync(cta);
|
||||
|
||||
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS) {
|
||||
odata[index_out + i * height] = tile[threadIdx.x][threadIdx.y + i];
|
||||
}
|
||||
}
|
||||
|
||||
// --------------------------------------------------------------------
|
||||
// Partial transposes
|
||||
// NB: the coarse- and fine-grained routines only perform part of a
|
||||
// transpose and will fail the test against the reference solution
|
||||
//
|
||||
// They are used to assess performance characteristics of different
|
||||
// components of a full transpose
|
||||
// --------------------------------------------------------------------
|
||||
|
||||
__global__ void transposeFineGrained(float *odata, float *idata, int width,
|
||||
int height) {
|
||||
// Handle to thread block group
|
||||
cg::thread_block cta = cg::this_thread_block();
|
||||
__shared__ float block[TILE_DIM][TILE_DIM + 1];
|
||||
|
||||
int xIndex = blockIdx.x * TILE_DIM + threadIdx.x;
|
||||
int yIndex = blockIdx.y * TILE_DIM + threadIdx.y;
|
||||
int index = xIndex + (yIndex)*width;
|
||||
|
||||
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS) {
|
||||
block[threadIdx.y + i][threadIdx.x] = idata[index + i * width];
|
||||
}
|
||||
|
||||
cg::sync(cta);
|
||||
|
||||
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS) {
|
||||
odata[index + i * height] = block[threadIdx.x][threadIdx.y + i];
|
||||
}
|
||||
}
|
||||
|
||||
__global__ void transposeCoarseGrained(float *odata, float *idata, int width,
|
||||
int height) {
|
||||
// Handle to thread block group
|
||||
cg::thread_block cta = cg::this_thread_block();
|
||||
__shared__ float block[TILE_DIM][TILE_DIM + 1];
|
||||
|
||||
int xIndex = blockIdx.x * TILE_DIM + threadIdx.x;
|
||||
int yIndex = blockIdx.y * TILE_DIM + threadIdx.y;
|
||||
int index_in = xIndex + (yIndex)*width;
|
||||
|
||||
xIndex = blockIdx.y * TILE_DIM + threadIdx.x;
|
||||
yIndex = blockIdx.x * TILE_DIM + threadIdx.y;
|
||||
int index_out = xIndex + (yIndex)*height;
|
||||
|
||||
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS) {
|
||||
block[threadIdx.y + i][threadIdx.x] = idata[index_in + i * width];
|
||||
}
|
||||
|
||||
cg::sync(cta);
|
||||
|
||||
for (int i = 0; i < TILE_DIM; i += BLOCK_ROWS) {
|
||||
odata[index_out + i * height] = block[threadIdx.y + i][threadIdx.x];
|
||||
}
|
||||
}
|
||||
|
||||
// ---------------------
|
||||
// host utility routines
|
||||
// ---------------------
|
||||
|
||||
void computeTransposeGold(float *gold, float *idata, const int size_x,
|
||||
const int size_y) {
|
||||
for (int y = 0; y < size_y; ++y) {
|
||||
for (int x = 0; x < size_x; ++x) {
|
||||
gold[(x * size_y) + y] = idata[(y * size_x) + x];
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
void getParams(int argc, char **argv, cudaDeviceProp &deviceProp, int &size_x,
|
||||
int &size_y, int max_tile_dim) {
|
||||
// set matrix size (if (x,y) dim of matrix is not square, then this will have
|
||||
// to be modified
|
||||
if (checkCmdLineFlag(argc, (const char **)argv, "dimX")) {
|
||||
size_x = getCmdLineArgumentInt(argc, (const char **)argv, "dimX");
|
||||
|
||||
if (size_x > max_tile_dim) {
|
||||
printf("> MatrixSize X = %d is greater than the recommended size = %d\n",
|
||||
size_x, max_tile_dim);
|
||||
} else {
|
||||
printf("> MatrixSize X = %d\n", size_x);
|
||||
}
|
||||
} else {
|
||||
size_x = max_tile_dim;
|
||||
size_x = FLOOR(size_x, 512);
|
||||
}
|
||||
|
||||
if (checkCmdLineFlag(argc, (const char **)argv, "dimY")) {
|
||||
size_y = getCmdLineArgumentInt(argc, (const char **)argv, "dimY");
|
||||
|
||||
if (size_y > max_tile_dim) {
|
||||
printf("> MatrixSize Y = %d is greater than the recommended size = %d\n",
|
||||
size_y, max_tile_dim);
|
||||
} else {
|
||||
printf("> MatrixSize Y = %d\n", size_y);
|
||||
}
|
||||
} else {
|
||||
size_y = max_tile_dim;
|
||||
size_y = FLOOR(size_y, 512);
|
||||
}
|
||||
}
|
||||
|
||||
void showHelp() {
|
||||
printf("\n%s : Command line options\n", sSDKsample);
|
||||
printf("\t-device=n (where n=0,1,2.... for the GPU device)\n\n");
|
||||
printf("> The default matrix size can be overridden with these parameters\n");
|
||||
printf("\t-dimX=row_dim_size (matrix row dimensions)\n");
|
||||
printf("\t-dimY=col_dim_size (matrix column dimensions)\n");
|
||||
}
|
||||
|
||||
// ----
|
||||
// main
|
||||
// ----
|
||||
|
||||
int main(int argc, char **argv) {
|
||||
// Start logs
|
||||
printf("%s Starting...\n\n", sSDKsample);
|
||||
|
||||
if (checkCmdLineFlag(argc, (const char **)argv, "help")) {
|
||||
showHelp();
|
||||
return 0;
|
||||
}
|
||||
|
||||
int devID = findCudaDevice(argc, (const char **)argv);
|
||||
cudaDeviceProp deviceProp;
|
||||
|
||||
// get number of SMs on this GPU
|
||||
checkCudaErrors(cudaGetDevice(&devID));
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, devID));
|
||||
|
||||
// compute the scaling factor (for GPUs with fewer MPs)
|
||||
float scale_factor, total_tiles;
|
||||
scale_factor =
|
||||
max((192.0f / (_ConvertSMVer2Cores(deviceProp.major, deviceProp.minor) *
|
||||
(float)deviceProp.multiProcessorCount)),
|
||||
1.0f);
|
||||
|
||||
printf("> Device %d: \"%s\"\n", devID, deviceProp.name);
|
||||
printf("> SM Capability %d.%d detected:\n", deviceProp.major,
|
||||
deviceProp.minor);
|
||||
|
||||
// Calculate number of tiles we will run for the Matrix Transpose performance
|
||||
// tests
|
||||
int size_x, size_y, max_matrix_dim, matrix_size_test;
|
||||
|
||||
matrix_size_test = 512; // we round down max_matrix_dim for this perf test
|
||||
total_tiles = (float)MAX_TILES / scale_factor;
|
||||
|
||||
max_matrix_dim =
|
||||
FLOOR((int)(floor(sqrt(total_tiles)) * TILE_DIM), matrix_size_test);
|
||||
|
||||
// This is the minimum size allowed
|
||||
if (max_matrix_dim == 0) {
|
||||
max_matrix_dim = matrix_size_test;
|
||||
}
|
||||
|
||||
printf("> [%s] has %d MP(s) x %d (Cores/MP) = %d (Cores)\n", deviceProp.name,
|
||||
deviceProp.multiProcessorCount,
|
||||
_ConvertSMVer2Cores(deviceProp.major, deviceProp.minor),
|
||||
_ConvertSMVer2Cores(deviceProp.major, deviceProp.minor) *
|
||||
deviceProp.multiProcessorCount);
|
||||
|
||||
printf("> Compute performance scaling factor = %4.2f\n", scale_factor);
|
||||
|
||||
// Extract parameters if there are any, command line -dimx and -dimy can
|
||||
// override any of these settings
|
||||
getParams(argc, argv, deviceProp, size_x, size_y, max_matrix_dim);
|
||||
|
||||
if (size_x != size_y) {
|
||||
printf(
|
||||
"\n[%s] does not support non-square matrices (row_dim_size(%d) != "
|
||||
"col_dim_size(%d))\nExiting...\n\n",
|
||||
sSDKsample, size_x, size_y);
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
|
||||
if (size_x % TILE_DIM != 0 || size_y % TILE_DIM != 0) {
|
||||
printf(
|
||||
"[%s] Matrix size must be integral multiple of tile "
|
||||
"size\nExiting...\n\n",
|
||||
sSDKsample);
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
|
||||
// kernel pointer and descriptor
|
||||
void (*kernel)(float *, float *, int, int);
|
||||
const char *kernelName;
|
||||
|
||||
// execution configuration parameters
|
||||
dim3 grid(size_x / TILE_DIM, size_y / TILE_DIM),
|
||||
threads(TILE_DIM, BLOCK_ROWS);
|
||||
|
||||
if (grid.x < 1 || grid.y < 1) {
|
||||
printf("[%s] grid size computation incorrect in test \nExiting...\n\n",
|
||||
sSDKsample);
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
|
||||
// CUDA events
|
||||
cudaEvent_t start, stop;
|
||||
|
||||
// size of memory required to store the matrix
|
||||
size_t mem_size = static_cast<size_t>(sizeof(float) * size_x * size_y);
|
||||
|
||||
if (2 * mem_size > deviceProp.totalGlobalMem) {
|
||||
printf("Input matrix size is larger than the available device memory!\n");
|
||||
printf("Please choose a smaller size matrix\n");
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
|
||||
// allocate host memory
|
||||
float *h_idata = (float *)malloc(mem_size);
|
||||
float *h_odata = (float *)malloc(mem_size);
|
||||
float *transposeGold = (float *)malloc(mem_size);
|
||||
float *gold;
|
||||
|
||||
// allocate device memory
|
||||
float *d_idata, *d_odata;
|
||||
checkCudaErrors(cudaMalloc((void **)&d_idata, mem_size));
|
||||
checkCudaErrors(cudaMalloc((void **)&d_odata, mem_size));
|
||||
|
||||
// initialize host data
|
||||
for (int i = 0; i < (size_x * size_y); ++i) {
|
||||
h_idata[i] = (float)i;
|
||||
}
|
||||
|
||||
// copy host data to device
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(d_idata, h_idata, mem_size, cudaMemcpyHostToDevice));
|
||||
|
||||
// Compute reference transpose solution
|
||||
computeTransposeGold(transposeGold, h_idata, size_x, size_y);
|
||||
|
||||
// print out common data for all kernels
|
||||
printf(
|
||||
"\nMatrix size: %dx%d (%dx%d tiles), tile size: %dx%d, block size: "
|
||||
"%dx%d\n\n",
|
||||
size_x, size_y, size_x / TILE_DIM, size_y / TILE_DIM, TILE_DIM, TILE_DIM,
|
||||
TILE_DIM, BLOCK_ROWS);
|
||||
|
||||
// initialize events
|
||||
checkCudaErrors(cudaEventCreate(&start));
|
||||
checkCudaErrors(cudaEventCreate(&stop));
|
||||
|
||||
//
|
||||
// loop over different kernels
|
||||
//
|
||||
|
||||
bool success = true;
|
||||
|
||||
for (int k = 0; k < 8; k++) {
|
||||
// set kernel pointer
|
||||
switch (k) {
|
||||
case 0:
|
||||
kernel = ©
|
||||
kernelName = "simple copy ";
|
||||
break;
|
||||
|
||||
case 1:
|
||||
kernel = ©SharedMem;
|
||||
kernelName = "shared memory copy";
|
||||
break;
|
||||
|
||||
case 2:
|
||||
kernel = &transposeNaive;
|
||||
kernelName = "naive ";
|
||||
break;
|
||||
|
||||
case 3:
|
||||
kernel = &transposeCoalesced;
|
||||
kernelName = "coalesced ";
|
||||
break;
|
||||
|
||||
case 4:
|
||||
kernel = &transposeNoBankConflicts;
|
||||
kernelName = "optimized ";
|
||||
break;
|
||||
|
||||
case 5:
|
||||
kernel = &transposeCoarseGrained;
|
||||
kernelName = "coarse-grained ";
|
||||
break;
|
||||
|
||||
case 6:
|
||||
kernel = &transposeFineGrained;
|
||||
kernelName = "fine-grained ";
|
||||
break;
|
||||
|
||||
case 7:
|
||||
kernel = &transposeDiagonal;
|
||||
kernelName = "diagonal ";
|
||||
break;
|
||||
}
|
||||
|
||||
// set reference solution
|
||||
if (kernel == © || kernel == ©SharedMem) {
|
||||
gold = h_idata;
|
||||
} else if (kernel == &transposeCoarseGrained ||
|
||||
kernel == &transposeFineGrained) {
|
||||
gold = h_odata; // fine- and coarse-grained kernels are not full
|
||||
// transposes, so bypass check
|
||||
} else {
|
||||
gold = transposeGold;
|
||||
}
|
||||
|
||||
// Clear error status
|
||||
checkCudaErrors(cudaGetLastError());
|
||||
|
||||
// warmup to avoid timing startup
|
||||
kernel<<<grid, threads>>>(d_odata, d_idata, size_x, size_y);
|
||||
|
||||
// take measurements for loop over kernel launches
|
||||
checkCudaErrors(cudaEventRecord(start, 0));
|
||||
|
||||
for (int i = 0; i < NUM_REPS; i++) {
|
||||
kernel<<<grid, threads>>>(d_odata, d_idata, size_x, size_y);
|
||||
// Ensure no launch failure
|
||||
checkCudaErrors(cudaGetLastError());
|
||||
}
|
||||
|
||||
checkCudaErrors(cudaEventRecord(stop, 0));
|
||||
checkCudaErrors(cudaEventSynchronize(stop));
|
||||
float kernelTime;
|
||||
checkCudaErrors(cudaEventElapsedTime(&kernelTime, start, stop));
|
||||
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(h_odata, d_odata, mem_size, cudaMemcpyDeviceToHost));
|
||||
bool res = compareData(gold, h_odata, size_x * size_y, 0.01f, 0.0f);
|
||||
|
||||
if (res == false) {
|
||||
printf("*** %s kernel FAILED ***\n", kernelName);
|
||||
success = false;
|
||||
}
|
||||
|
||||
// take measurements for loop inside kernel
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(h_odata, d_odata, mem_size, cudaMemcpyDeviceToHost));
|
||||
res = compareData(gold, h_odata, size_x * size_y, 0.01f, 0.0f);
|
||||
|
||||
if (res == false) {
|
||||
printf("*** %s kernel FAILED ***\n", kernelName);
|
||||
success = false;
|
||||
}
|
||||
|
||||
// report effective bandwidths
|
||||
float kernelBandwidth = 2.0f * 1000.0f * mem_size / (1024 * 1024 * 1024) /
|
||||
(kernelTime / NUM_REPS);
|
||||
printf(
|
||||
"transpose %s, Throughput = %.4f GB/s, Time = %.5f ms, Size = %u fp32 "
|
||||
"elements, NumDevsUsed = %u, Workgroup = %u\n",
|
||||
kernelName, kernelBandwidth, kernelTime / NUM_REPS, (size_x * size_y),
|
||||
1, TILE_DIM * BLOCK_ROWS);
|
||||
}
|
||||
|
||||
// cleanup
|
||||
free(h_idata);
|
||||
free(h_odata);
|
||||
free(transposeGold);
|
||||
cudaFree(d_idata);
|
||||
cudaFree(d_odata);
|
||||
|
||||
checkCudaErrors(cudaEventDestroy(start));
|
||||
checkCudaErrors(cudaEventDestroy(stop));
|
||||
|
||||
if (!success) {
|
||||
printf("Test failed!\n");
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
|
||||
printf("Test passed\n");
|
||||
exit(EXIT_SUCCESS);
|
||||
}
|
||||
20
Samples/transpose/transpose_vs2017.sln
Normal file
20
Samples/transpose/transpose_vs2017.sln
Normal file
@@ -0,0 +1,20 @@
|
||||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2017
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "transpose", "transpose_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
|
||||
113
Samples/transpose/transpose_vs2017.vcxproj
Normal file
113
Samples/transpose/transpose_vs2017.vcxproj
Normal file
@@ -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>transpose_vs2017</RootNamespace>
|
||||
<ProjectName>transpose</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 11.5.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;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>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)/transpose.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;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" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<AdditionalOptions>--threads 0</AdditionalOptions>
|
||||
</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="transpose.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.5.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/transpose/transpose_vs2019.sln
Normal file
20
Samples/transpose/transpose_vs2019.sln
Normal file
@@ -0,0 +1,20 @@
|
||||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2019
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "transpose", "transpose_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
|
||||
109
Samples/transpose/transpose_vs2019.vcxproj
Normal file
109
Samples/transpose/transpose_vs2019.vcxproj
Normal file
@@ -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>transpose_vs2019</RootNamespace>
|
||||
<ProjectName>transpose</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="$(CUDAPropsPath)\CUDA 11.5.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(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>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/transpose.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;compute_80,sm_80;compute_86,sm_86;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<AdditionalOptions>--threads 0</AdditionalOptions>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(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="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="transpose.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.5.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
Reference in New Issue
Block a user