mirror of
https://github.com/NVIDIA/cuda-samples.git
synced 2026-10-23 13:18:16 +08:00
add and update samples for CUDA 11.5
This commit is contained in:
Rutwik Choughule
18
Samples/simpleStreams/.vscode/c_cpp_properties.json
vendored
Normal file
18
Samples/simpleStreams/.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/simpleStreams/.vscode/extensions.json
vendored
Normal file
7
Samples/simpleStreams/.vscode/extensions.json
vendored
Normal file
@@ -0,0 +1,7 @@
|
||||
{
|
||||
"recommendations": [
|
||||
"nvidia.nsight-vscode-edition",
|
||||
"ms-vscode.cpptools",
|
||||
"ms-vscode.makefile-tools"
|
||||
]
|
||||
}
|
||||
10
Samples/simpleStreams/.vscode/launch.json
vendored
Normal file
10
Samples/simpleStreams/.vscode/launch.json
vendored
Normal file
@@ -0,0 +1,10 @@
|
||||
{
|
||||
"configurations": [
|
||||
{
|
||||
"name": "CUDA C++: Launch",
|
||||
"type": "cuda-gdb",
|
||||
"request": "launch",
|
||||
"program": "${workspaceFolder}/simpleStreams"
|
||||
}
|
||||
]
|
||||
}
|
||||
15
Samples/simpleStreams/.vscode/tasks.json
vendored
Normal file
15
Samples/simpleStreams/.vscode/tasks.json
vendored
Normal file
@@ -0,0 +1,15 @@
|
||||
{
|
||||
"version": "2.0.0",
|
||||
"tasks": [
|
||||
{
|
||||
"label": "sample",
|
||||
"type": "shell",
|
||||
"command": "make dbg=1",
|
||||
"problemMatcher": ["$nvcc"],
|
||||
"group": {
|
||||
"kind": "build",
|
||||
"isDefault": true
|
||||
}
|
||||
}
|
||||
]
|
||||
}
|
||||
337
Samples/simpleStreams/Makefile
Normal file
337
Samples/simpleStreams/Makefile
Normal file
@@ -0,0 +1,337 @@
|
||||
################################################################################
|
||||
# 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 :=
|
||||
|
||||
################################################################################
|
||||
|
||||
# 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 ($(SAMPLE_ENABLED),0)
|
||||
EXEC ?= @echo "[@]"
|
||||
endif
|
||||
|
||||
################################################################################
|
||||
|
||||
# Target rules
|
||||
all: build
|
||||
|
||||
build: simpleStreams
|
||||
|
||||
check.deps:
|
||||
ifeq ($(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
|
||||
|
||||
simpleStreams.o:simpleStreams.cu
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
simpleStreams: simpleStreams.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
|
||||
$(EXEC) ./simpleStreams
|
||||
|
||||
clean:
|
||||
rm -f simpleStreams simpleStreams.o
|
||||
rm -rf ../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)/simpleStreams
|
||||
|
||||
clobber: clean
|
||||
72
Samples/simpleStreams/NsightEclipse.xml
Normal file
72
Samples/simpleStreams/NsightEclipse.xml
Normal file
@@ -0,0 +1,72 @@
|
||||
<?xml version="1.0" encoding="UTF-8"?>
|
||||
<!DOCTYPE entry SYSTEM "SamplesInfo.dtd">
|
||||
<entry>
|
||||
<name>simpleStreams</name>
|
||||
<cuda_api_list>
|
||||
<toolkit>cudaEventCreate</toolkit>
|
||||
<toolkit>cudaEventRecord</toolkit>
|
||||
<toolkit>cudaEventQuery</toolkit>
|
||||
<toolkit>cudaEventDestroy</toolkit>
|
||||
<toolkit>cudaEventElapsedTime</toolkit>
|
||||
<toolkit>cudaMemcpyAsync</toolkit>
|
||||
</cuda_api_list>
|
||||
<description><![CDATA[This sample uses CUDA streams to overlap kernel executions with memory copies between the host and a GPU device. This sample uses a new CUDA 4.0 feature that supports pinning of generic host memory. Requires Compute Capability 2.0 or higher.]]></description>
|
||||
<devicecompilation>whole</devicecompilation>
|
||||
<includepaths>
|
||||
<path>./</path>
|
||||
<path>../</path>
|
||||
<path>../../Common</path>
|
||||
</includepaths>
|
||||
<keyconcepts>
|
||||
<concept level="basic">Asynchronous Data Transfers</concept>
|
||||
<concept level="basic">CUDA Streams and Events</concept>
|
||||
</keyconcepts>
|
||||
<keywords>
|
||||
</keywords>
|
||||
<libraries>
|
||||
</libraries>
|
||||
<librarypaths>
|
||||
</librarypaths>
|
||||
<nsight_eclipse>true</nsight_eclipse>
|
||||
<primary_file>simpleStreams.cu</primary_file>
|
||||
<scopes>
|
||||
<scope>1:CUDA Basic Topics</scope>
|
||||
<scope>1:Performance Strategies</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>simpleStreams</title>
|
||||
<type>exe</type>
|
||||
</entry>
|
||||
70
Samples/simpleStreams/README.md
Normal file
70
Samples/simpleStreams/README.md
Normal file
@@ -0,0 +1,70 @@
|
||||
# simpleStreams - simpleStreams
|
||||
|
||||
## Description
|
||||
|
||||
This sample uses CUDA streams to overlap kernel executions with memory copies between the host and a GPU device. This sample uses a new CUDA 4.0 feature that supports pinning of generic host memory. Requires Compute Capability 2.0 or higher.
|
||||
|
||||
## Key Concepts
|
||||
|
||||
Asynchronous Data Transfers, CUDA Streams and Events
|
||||
|
||||
## 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
|
||||
|
||||
### [CUDA Runtime API](http://docs.nvidia.com/cuda/cuda-runtime-api/index.html)
|
||||
cudaEventCreate, cudaEventRecord, cudaEventQuery, cudaEventDestroy, cudaEventElapsedTime, cudaMemcpyAsync
|
||||
|
||||
## 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)
|
||||
|
||||
445
Samples/simpleStreams/simpleStreams.cu
Normal file
445
Samples/simpleStreams/simpleStreams.cu
Normal file
@@ -0,0 +1,445 @@
|
||||
/* 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.
|
||||
*/
|
||||
|
||||
/*
|
||||
* This sample illustrates the usage of CUDA streams for overlapping
|
||||
* kernel execution with device/host memcopies. The kernel is used to
|
||||
* initialize an array to a specific value, after which the array is
|
||||
* copied to the host (CPU) memory. To increase performance, multiple
|
||||
* kernel/memcopy pairs are launched asynchronously, each pair in its
|
||||
* own stream. Devices with Compute Capability 1.1 can overlap a kernel
|
||||
* and a memcopy as long as they are issued in different streams. Kernels
|
||||
* are serialized. Thus, if n pairs are launched, streamed approach
|
||||
* can reduce the memcopy cost to the (1/n)th of a single copy of the entire
|
||||
* data set.
|
||||
*
|
||||
* Additionally, this sample uses CUDA events to measure elapsed time for
|
||||
* CUDA calls. Events are a part of CUDA API and provide a system independent
|
||||
* way to measure execution times on CUDA devices with approximately 0.5
|
||||
* microsecond precision.
|
||||
*
|
||||
* Elapsed times are averaged over nreps repetitions (10 by default).
|
||||
*
|
||||
*/
|
||||
|
||||
const char *sSDKsample = "simpleStreams";
|
||||
|
||||
const char *sEventSyncMethod[] = {"cudaEventDefault", "cudaEventBlockingSync",
|
||||
"cudaEventDisableTiming", NULL};
|
||||
|
||||
const char *sDeviceSyncMethod[] = {
|
||||
"cudaDeviceScheduleAuto", "cudaDeviceScheduleSpin",
|
||||
"cudaDeviceScheduleYield", "INVALID",
|
||||
"cudaDeviceScheduleBlockingSync", NULL};
|
||||
|
||||
// System includes
|
||||
#include <stdio.h>
|
||||
#include <assert.h>
|
||||
|
||||
// CUDA runtime
|
||||
#include <cuda_runtime.h>
|
||||
|
||||
// helper functions and utilities to work with CUDA
|
||||
#include <helper_functions.h>
|
||||
#include <helper_cuda.h>
|
||||
|
||||
#ifndef WIN32
|
||||
#include <sys/mman.h> // for mmap() / munmap()
|
||||
#endif
|
||||
|
||||
// Macro to aligned up to the memory size in question
|
||||
#define MEMORY_ALIGNMENT 4096
|
||||
#define ALIGN_UP(x, size) (((size_t)x + (size - 1)) & (~(size - 1)))
|
||||
|
||||
__global__ void init_array(int *g_data, int *factor, int num_iterations) {
|
||||
int idx = blockIdx.x * blockDim.x + threadIdx.x;
|
||||
|
||||
for (int i = 0; i < num_iterations; i++) {
|
||||
g_data[idx] += *factor; // non-coalesced on purpose, to burn time
|
||||
}
|
||||
}
|
||||
|
||||
bool correct_data(int *a, const int n, const int c) {
|
||||
for (int i = 0; i < n; i++) {
|
||||
if (a[i] != c) {
|
||||
printf("%d: %d %d\n", i, a[i], c);
|
||||
return false;
|
||||
}
|
||||
}
|
||||
|
||||
return true;
|
||||
}
|
||||
|
||||
inline void AllocateHostMemory(bool bPinGenericMemory, int **pp_a,
|
||||
int **ppAligned_a, int nbytes) {
|
||||
#if CUDART_VERSION >= 4000
|
||||
#if !defined(__arm__) && !defined(__aarch64__)
|
||||
if (bPinGenericMemory) {
|
||||
// allocate a generic page-aligned chunk of system memory
|
||||
#ifdef WIN32
|
||||
printf(
|
||||
"> VirtualAlloc() allocating %4.2f Mbytes of (generic page-aligned "
|
||||
"system memory)\n",
|
||||
(float)nbytes / 1048576.0f);
|
||||
*pp_a = (int *)VirtualAlloc(NULL, (nbytes + MEMORY_ALIGNMENT),
|
||||
MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE);
|
||||
#else
|
||||
printf(
|
||||
"> mmap() allocating %4.2f Mbytes (generic page-aligned system "
|
||||
"memory)\n",
|
||||
(float)nbytes / 1048576.0f);
|
||||
*pp_a = (int *)mmap(NULL, (nbytes + MEMORY_ALIGNMENT),
|
||||
PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANON, -1, 0);
|
||||
#endif
|
||||
|
||||
*ppAligned_a = (int *)ALIGN_UP(*pp_a, MEMORY_ALIGNMENT);
|
||||
|
||||
printf(
|
||||
"> cudaHostRegister() registering %4.2f Mbytes of generic allocated "
|
||||
"system memory\n",
|
||||
(float)nbytes / 1048576.0f);
|
||||
// pin allocate memory
|
||||
checkCudaErrors(
|
||||
cudaHostRegister(*ppAligned_a, nbytes, cudaHostRegisterMapped));
|
||||
} else
|
||||
#endif
|
||||
#endif
|
||||
{
|
||||
printf("> cudaMallocHost() allocating %4.2f Mbytes of system memory\n",
|
||||
(float)nbytes / 1048576.0f);
|
||||
// allocate host memory (pinned is required for achieve asynchronicity)
|
||||
checkCudaErrors(cudaMallocHost((void **)pp_a, nbytes));
|
||||
*ppAligned_a = *pp_a;
|
||||
}
|
||||
}
|
||||
|
||||
inline void FreeHostMemory(bool bPinGenericMemory, int **pp_a,
|
||||
int **ppAligned_a, int nbytes) {
|
||||
#if CUDART_VERSION >= 4000
|
||||
#if !defined(__arm__) && !defined(__aarch64__)
|
||||
// CUDA 4.0 support pinning of generic host memory
|
||||
if (bPinGenericMemory) {
|
||||
// unpin and delete host memory
|
||||
checkCudaErrors(cudaHostUnregister(*ppAligned_a));
|
||||
#ifdef WIN32
|
||||
VirtualFree(*pp_a, 0, MEM_RELEASE);
|
||||
#else
|
||||
munmap(*pp_a, nbytes);
|
||||
#endif
|
||||
} else
|
||||
#endif
|
||||
#endif
|
||||
{
|
||||
cudaFreeHost(*pp_a);
|
||||
}
|
||||
}
|
||||
|
||||
static const char *sSyncMethod[] = {
|
||||
"0 (Automatic Blocking)",
|
||||
"1 (Spin Blocking)",
|
||||
"2 (Yield Blocking)",
|
||||
"3 (Undefined Blocking Method)",
|
||||
"4 (Blocking Sync Event) = low CPU utilization",
|
||||
NULL};
|
||||
|
||||
void printHelp() {
|
||||
printf("Usage: %s [options below]\n", sSDKsample);
|
||||
printf("\t--sync_method=n for CPU/GPU synchronization\n");
|
||||
printf("\t n=%s\n", sSyncMethod[0]);
|
||||
printf("\t n=%s\n", sSyncMethod[1]);
|
||||
printf("\t n=%s\n", sSyncMethod[2]);
|
||||
printf("\t <Default> n=%s\n", sSyncMethod[4]);
|
||||
printf(
|
||||
"\t--use_generic_memory (default) use generic page-aligned for system "
|
||||
"memory\n");
|
||||
printf(
|
||||
"\t--use_cuda_malloc_host (optional) use cudaMallocHost to allocate "
|
||||
"system memory\n");
|
||||
}
|
||||
|
||||
#if defined(__APPLE__) || defined(MACOSX)
|
||||
#define DEFAULT_PINNED_GENERIC_MEMORY false
|
||||
#else
|
||||
#define DEFAULT_PINNED_GENERIC_MEMORY true
|
||||
#endif
|
||||
|
||||
int main(int argc, char **argv) {
|
||||
int cuda_device = 0;
|
||||
int nstreams = 4; // number of streams for CUDA calls
|
||||
int nreps = 10; // number of times each experiment is repeated
|
||||
int n = 16 * 1024 * 1024; // number of ints in the data set
|
||||
int nbytes = n * sizeof(int); // number of data bytes
|
||||
dim3 threads, blocks; // kernel launch configuration
|
||||
float elapsed_time, time_memcpy, time_kernel; // timing variables
|
||||
float scale_factor = 1.0f;
|
||||
|
||||
// allocate generic memory and pin it laster instead of using cudaHostAlloc()
|
||||
|
||||
bool bPinGenericMemory =
|
||||
DEFAULT_PINNED_GENERIC_MEMORY; // we want this to be the default behavior
|
||||
int device_sync_method =
|
||||
cudaDeviceBlockingSync; // by default we use BlockingSync
|
||||
|
||||
int niterations; // number of iterations for the loop inside the kernel
|
||||
|
||||
printf("[ %s ]\n\n", sSDKsample);
|
||||
|
||||
if (checkCmdLineFlag(argc, (const char **)argv, "help")) {
|
||||
printHelp();
|
||||
return EXIT_SUCCESS;
|
||||
}
|
||||
|
||||
if ((device_sync_method = getCmdLineArgumentInt(argc, (const char **)argv,
|
||||
"sync_method")) >= 0) {
|
||||
if (device_sync_method == 0 || device_sync_method == 1 ||
|
||||
device_sync_method == 2 || device_sync_method == 4) {
|
||||
printf("Device synchronization method set to = %s\n",
|
||||
sSyncMethod[device_sync_method]);
|
||||
printf("Setting reps to 100 to demonstrate steady state\n");
|
||||
nreps = 100;
|
||||
} else {
|
||||
printf("Invalid command line option sync_method=\"%d\"\n",
|
||||
device_sync_method);
|
||||
return EXIT_FAILURE;
|
||||
}
|
||||
} else {
|
||||
printHelp();
|
||||
return EXIT_SUCCESS;
|
||||
}
|
||||
|
||||
if (checkCmdLineFlag(argc, (const char **)argv, "use_generic_memory")) {
|
||||
#if defined(__APPLE__) || defined(MACOSX)
|
||||
bPinGenericMemory = false; // Generic Pinning of System Paged memory not
|
||||
// currently supported on Mac OSX
|
||||
#else
|
||||
bPinGenericMemory = true;
|
||||
#endif
|
||||
}
|
||||
|
||||
if (checkCmdLineFlag(argc, (const char **)argv, "use_cuda_malloc_host")) {
|
||||
bPinGenericMemory = false;
|
||||
}
|
||||
|
||||
printf("\n> ");
|
||||
cuda_device = findCudaDevice(argc, (const char **)argv);
|
||||
|
||||
// check the compute capability of the device
|
||||
int num_devices = 0;
|
||||
checkCudaErrors(cudaGetDeviceCount(&num_devices));
|
||||
|
||||
if (0 == num_devices) {
|
||||
printf(
|
||||
"your system does not have a CUDA capable device, waiving test...\n");
|
||||
return EXIT_WAIVED;
|
||||
}
|
||||
|
||||
// check if the command-line chosen device ID is within range, exit if not
|
||||
if (cuda_device >= num_devices) {
|
||||
printf(
|
||||
"cuda_device=%d is invalid, must choose device ID between 0 and %d\n",
|
||||
cuda_device, num_devices - 1);
|
||||
return EXIT_FAILURE;
|
||||
}
|
||||
|
||||
checkCudaErrors(cudaSetDevice(cuda_device));
|
||||
|
||||
// Checking for compute capabilities
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, cuda_device));
|
||||
|
||||
niterations = 5;
|
||||
|
||||
// Check if GPU can map host memory (Generic Method), if not then we override
|
||||
// bPinGenericMemory to be false
|
||||
if (bPinGenericMemory) {
|
||||
printf("Device: <%s> canMapHostMemory: %s\n", deviceProp.name,
|
||||
deviceProp.canMapHostMemory ? "Yes" : "No");
|
||||
|
||||
if (deviceProp.canMapHostMemory == 0) {
|
||||
printf(
|
||||
"Using cudaMallocHost, CUDA device does not support mapping of "
|
||||
"generic host memory\n");
|
||||
bPinGenericMemory = false;
|
||||
}
|
||||
}
|
||||
|
||||
// Anything that is less than 32 Cores will have scaled down workload
|
||||
scale_factor =
|
||||
max((32.0f / (_ConvertSMVer2Cores(deviceProp.major, deviceProp.minor) *
|
||||
(float)deviceProp.multiProcessorCount)),
|
||||
1.0f);
|
||||
n = (int)rint((float)n / scale_factor);
|
||||
|
||||
printf("> CUDA Capable: SM %d.%d hardware\n", deviceProp.major,
|
||||
deviceProp.minor);
|
||||
printf("> %d Multiprocessor(s) x %d (Cores/Multiprocessor) = %d (Cores)\n",
|
||||
deviceProp.multiProcessorCount,
|
||||
_ConvertSMVer2Cores(deviceProp.major, deviceProp.minor),
|
||||
_ConvertSMVer2Cores(deviceProp.major, deviceProp.minor) *
|
||||
deviceProp.multiProcessorCount);
|
||||
|
||||
printf("> scale_factor = %1.4f\n", 1.0f / scale_factor);
|
||||
printf("> array_size = %d\n\n", n);
|
||||
|
||||
// enable use of blocking sync, to reduce CPU usage
|
||||
printf("> Using CPU/GPU Device Synchronization method (%s)\n",
|
||||
sDeviceSyncMethod[device_sync_method]);
|
||||
checkCudaErrors(cudaSetDeviceFlags(
|
||||
device_sync_method | (bPinGenericMemory ? cudaDeviceMapHost : 0)));
|
||||
|
||||
// allocate host memory
|
||||
int c = 5; // value to which the array will be initialized
|
||||
int *h_a = 0; // pointer to the array data in host memory
|
||||
int *hAligned_a = 0; // pointer to the array data in host memory (aligned to
|
||||
// MEMORY_ALIGNMENT)
|
||||
|
||||
// Allocate Host memory (could be using cudaMallocHost or VirtualAlloc/mmap if
|
||||
// using the new CUDA 4.0 features
|
||||
AllocateHostMemory(bPinGenericMemory, &h_a, &hAligned_a, nbytes);
|
||||
|
||||
// allocate device memory
|
||||
int *d_a = 0,
|
||||
*d_c = 0; // pointers to data and init value in the device memory
|
||||
checkCudaErrors(cudaMalloc((void **)&d_a, nbytes));
|
||||
checkCudaErrors(cudaMemset(d_a, 0x0, nbytes));
|
||||
checkCudaErrors(cudaMalloc((void **)&d_c, sizeof(int)));
|
||||
checkCudaErrors(cudaMemcpy(d_c, &c, sizeof(int), cudaMemcpyHostToDevice));
|
||||
|
||||
printf("\nStarting Test\n");
|
||||
|
||||
// allocate and initialize an array of stream handles
|
||||
cudaStream_t *streams =
|
||||
(cudaStream_t *)malloc(nstreams * sizeof(cudaStream_t));
|
||||
|
||||
for (int i = 0; i < nstreams; i++) {
|
||||
checkCudaErrors(cudaStreamCreate(&(streams[i])));
|
||||
}
|
||||
|
||||
// create CUDA event handles
|
||||
// use blocking sync
|
||||
cudaEvent_t start_event, stop_event;
|
||||
int eventflags =
|
||||
((device_sync_method == cudaDeviceBlockingSync) ? cudaEventBlockingSync
|
||||
: cudaEventDefault);
|
||||
|
||||
checkCudaErrors(cudaEventCreateWithFlags(&start_event, eventflags));
|
||||
checkCudaErrors(cudaEventCreateWithFlags(&stop_event, eventflags));
|
||||
|
||||
// time memcopy from device
|
||||
checkCudaErrors(cudaEventRecord(start_event, 0)); // record in stream-0, to
|
||||
// ensure that all previous
|
||||
// CUDA calls have
|
||||
// completed
|
||||
checkCudaErrors(cudaMemcpyAsync(hAligned_a, d_a, nbytes,
|
||||
cudaMemcpyDeviceToHost, streams[0]));
|
||||
checkCudaErrors(cudaEventRecord(stop_event, 0));
|
||||
checkCudaErrors(cudaEventSynchronize(
|
||||
stop_event)); // block until the event is actually recorded
|
||||
checkCudaErrors(cudaEventElapsedTime(&time_memcpy, start_event, stop_event));
|
||||
printf("memcopy:\t%.2f\n", time_memcpy);
|
||||
|
||||
// time kernel
|
||||
threads = dim3(512, 1);
|
||||
blocks = dim3(n / threads.x, 1);
|
||||
checkCudaErrors(cudaEventRecord(start_event, 0));
|
||||
init_array<<<blocks, threads, 0, streams[0]>>>(d_a, d_c, niterations);
|
||||
checkCudaErrors(cudaEventRecord(stop_event, 0));
|
||||
checkCudaErrors(cudaEventSynchronize(stop_event));
|
||||
checkCudaErrors(cudaEventElapsedTime(&time_kernel, start_event, stop_event));
|
||||
printf("kernel:\t\t%.2f\n", time_kernel);
|
||||
|
||||
//////////////////////////////////////////////////////////////////////
|
||||
// time non-streamed execution for reference
|
||||
threads = dim3(512, 1);
|
||||
blocks = dim3(n / threads.x, 1);
|
||||
checkCudaErrors(cudaEventRecord(start_event, 0));
|
||||
|
||||
for (int k = 0; k < nreps; k++) {
|
||||
init_array<<<blocks, threads>>>(d_a, d_c, niterations);
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(hAligned_a, d_a, nbytes, cudaMemcpyDeviceToHost));
|
||||
}
|
||||
|
||||
checkCudaErrors(cudaEventRecord(stop_event, 0));
|
||||
checkCudaErrors(cudaEventSynchronize(stop_event));
|
||||
checkCudaErrors(cudaEventElapsedTime(&elapsed_time, start_event, stop_event));
|
||||
printf("non-streamed:\t%.2f\n", elapsed_time / nreps);
|
||||
|
||||
//////////////////////////////////////////////////////////////////////
|
||||
// time execution with nstreams streams
|
||||
threads = dim3(512, 1);
|
||||
blocks = dim3(n / (nstreams * threads.x), 1);
|
||||
memset(hAligned_a, 255,
|
||||
nbytes); // set host memory bits to all 1s, for testing correctness
|
||||
checkCudaErrors(cudaMemset(
|
||||
d_a, 0, nbytes)); // set device memory to all 0s, for testing correctness
|
||||
checkCudaErrors(cudaEventRecord(start_event, 0));
|
||||
|
||||
for (int k = 0; k < nreps; k++) {
|
||||
// asynchronously launch nstreams kernels, each operating on its own portion
|
||||
// of data
|
||||
for (int i = 0; i < nstreams; i++) {
|
||||
init_array<<<blocks, threads, 0, streams[i]>>>(d_a + i * n / nstreams,
|
||||
d_c, niterations);
|
||||
}
|
||||
|
||||
// asynchronously launch nstreams memcopies. Note that memcopy in stream x
|
||||
// will only
|
||||
// commence executing when all previous CUDA calls in stream x have
|
||||
// completed
|
||||
for (int i = 0; i < nstreams; i++) {
|
||||
checkCudaErrors(cudaMemcpyAsync(hAligned_a + i * n / nstreams,
|
||||
d_a + i * n / nstreams, nbytes / nstreams,
|
||||
cudaMemcpyDeviceToHost, streams[i]));
|
||||
}
|
||||
}
|
||||
|
||||
checkCudaErrors(cudaEventRecord(stop_event, 0));
|
||||
checkCudaErrors(cudaEventSynchronize(stop_event));
|
||||
checkCudaErrors(cudaEventElapsedTime(&elapsed_time, start_event, stop_event));
|
||||
printf("%d streams:\t%.2f\n", nstreams, elapsed_time / nreps);
|
||||
|
||||
// check whether the output is correct
|
||||
printf("-------------------------------\n");
|
||||
bool bResults = correct_data(hAligned_a, n, c * nreps * niterations);
|
||||
|
||||
// release resources
|
||||
for (int i = 0; i < nstreams; i++) {
|
||||
checkCudaErrors(cudaStreamDestroy(streams[i]));
|
||||
}
|
||||
|
||||
checkCudaErrors(cudaEventDestroy(start_event));
|
||||
checkCudaErrors(cudaEventDestroy(stop_event));
|
||||
|
||||
// Free cudaMallocHost or Generic Host allocated memory (from CUDA 4.0)
|
||||
FreeHostMemory(bPinGenericMemory, &h_a, &hAligned_a, nbytes);
|
||||
|
||||
checkCudaErrors(cudaFree(d_a));
|
||||
checkCudaErrors(cudaFree(d_c));
|
||||
|
||||
return bResults ? EXIT_SUCCESS : EXIT_FAILURE;
|
||||
}
|
||||
20
Samples/simpleStreams/simpleStreams_vs2017.sln
Normal file
20
Samples/simpleStreams/simpleStreams_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}") = "simpleStreams", "simpleStreams_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/simpleStreams/simpleStreams_vs2017.vcxproj
Normal file
113
Samples/simpleStreams/simpleStreams_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>simpleStreams_vs2017</RootNamespace>
|
||||
<ProjectName>simpleStreams</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)/simpleStreams.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="simpleStreams.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.5.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/simpleStreams/simpleStreams_vs2019.sln
Normal file
20
Samples/simpleStreams/simpleStreams_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}") = "simpleStreams", "simpleStreams_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/simpleStreams/simpleStreams_vs2019.vcxproj
Normal file
109
Samples/simpleStreams/simpleStreams_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>simpleStreams_vs2019</RootNamespace>
|
||||
<ProjectName>simpleStreams</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>$(OutDir)/simpleStreams.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;compute_80,sm_80;compute_86,sm_86;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<AdditionalOptions>--threads 0</AdditionalOptions>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(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="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="simpleStreams.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.5.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
Reference in New Issue
Block a user