mirror of
https://github.com/NVIDIA/cuda-samples.git
synced 2026-10-11 23:38:25 +08:00
Add and Update samples for CUDA 10.0
This commit is contained in:
Mahesh Doijade
318
Samples/systemWideAtomics/Makefile
Normal file
318
Samples/systemWideAtomics/Makefile
Normal file
@@ -0,0 +1,318 @@
|
||||
################################################################################
|
||||
# Copyright (c) 2018, 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 $(TARGET_ARCH),x86_64 aarch64 ppc64le armv7l))
|
||||
ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
||||
ifneq (,$(filter $(TARGET_ARCH),x86_64 aarch64 ppc64le))
|
||||
TARGET_SIZE := 64
|
||||
else ifneq (,$(filter $(TARGET_ARCH),armv7l))
|
||||
TARGET_SIZE := 32
|
||||
endif
|
||||
else
|
||||
TARGET_SIZE := $(shell getconf LONG_BIT)
|
||||
endif
|
||||
else
|
||||
$(error ERROR - unsupported value $(TARGET_ARCH) for TARGET_ARCH!)
|
||||
endif
|
||||
ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
||||
ifeq (,$(filter $(HOST_ARCH)-$(TARGET_ARCH),aarch64-armv7l x86_64-armv7l x86_64-aarch64 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 ?= $(QNX_HOST)/usr/bin/aarch64-unknown-nto-qnx7.0.0-g++
|
||||
else ifeq ($(TARGET_OS), android)
|
||||
HOST_COMPILER ?= aarch64-linux-android-clang++
|
||||
endif
|
||||
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)/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
|
||||
CCFLAGS += -isystem=$(TARGET_FS)/usr/include/aarch64-linux-gnu
|
||||
endif
|
||||
endif
|
||||
endif
|
||||
|
||||
ifeq ($(TARGET_OS),qnx)
|
||||
CCFLAGS += -DWIN_INTERFACE_CUSTOM
|
||||
LDFLAGS += -lsocket
|
||||
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),armv7l-android)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/armv7-linux-androideabi/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-android)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/aarch64-linux-androideabi/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-qnx)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/ARMv7-linux-QNX/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-qnx)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/aarch64-qnx/
|
||||
else ifeq ($(TARGET_ARCH),ppc64le)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/ppc64le-linux/
|
||||
endif
|
||||
|
||||
# Debug build flags
|
||||
ifeq ($(dbg),1)
|
||||
NVCCFLAGS += -g -G
|
||||
BUILD_TYPE := debug
|
||||
else
|
||||
BUILD_TYPE := release
|
||||
endif
|
||||
|
||||
ALL_CCFLAGS :=
|
||||
ALL_CCFLAGS += $(NVCCFLAGS)
|
||||
ALL_CCFLAGS += $(EXTRA_NVCCFLAGS)
|
||||
ALL_CCFLAGS += $(addprefix -Xcompiler ,$(CCFLAGS))
|
||||
ALL_CCFLAGS += $(addprefix -Xcompiler ,$(EXTRA_CCFLAGS))
|
||||
|
||||
SAMPLE_ENABLED := 1
|
||||
|
||||
# This sample is not supported on Mac OSX
|
||||
ifeq ($(TARGET_OS),darwin)
|
||||
$(info >>> WARNING - systemWideAtomics is not supported on Mac OSX - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
# This sample is not supported on ARMv7
|
||||
ifeq ($(TARGET_ARCH),armv7l)
|
||||
$(info >>> WARNING - systemWideAtomics is not supported on ARMv7 - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
# This sample is not supported on aarch64
|
||||
ifeq ($(TARGET_ARCH),aarch64)
|
||||
$(info >>> WARNING - systemWideAtomics is not supported on aarch64 - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
ALL_LDFLAGS :=
|
||||
ALL_LDFLAGS += $(ALL_CCFLAGS)
|
||||
ALL_LDFLAGS += $(addprefix -Xlinker ,$(LDFLAGS))
|
||||
ALL_LDFLAGS += $(addprefix -Xlinker ,$(EXTRA_LDFLAGS))
|
||||
|
||||
# Common includes and paths for CUDA
|
||||
INCLUDES := -I../../Common
|
||||
LIBRARIES :=
|
||||
|
||||
################################################################################
|
||||
|
||||
# Gencode arguments
|
||||
SMS ?= 60 61 70 75
|
||||
|
||||
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
|
||||
|
||||
ifeq ($(SAMPLE_ENABLED),0)
|
||||
EXEC ?= @echo "[@]"
|
||||
endif
|
||||
|
||||
################################################################################
|
||||
|
||||
# Target rules
|
||||
all: build
|
||||
|
||||
build: systemWideAtomics
|
||||
|
||||
check.deps:
|
||||
ifeq ($(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
|
||||
|
||||
systemWideAtomics.o:systemWideAtomics.cu
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
systemWideAtomics: systemWideAtomics.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
|
||||
$(EXEC) ./systemWideAtomics
|
||||
|
||||
clean:
|
||||
rm -f systemWideAtomics systemWideAtomics.o
|
||||
rm -rf ../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)/systemWideAtomics
|
||||
|
||||
clobber: clean
|
||||
58
Samples/systemWideAtomics/NsightEclipse.xml
Normal file
58
Samples/systemWideAtomics/NsightEclipse.xml
Normal file
@@ -0,0 +1,58 @@
|
||||
<?xml version="1.0" encoding="UTF-8"?>
|
||||
<!DOCTYPE entry SYSTEM "SamplesInfo.dtd">
|
||||
<entry>
|
||||
<name>systemWideAtomics</name>
|
||||
<cuda_api_list>
|
||||
<toolkit>cudaMalloc</toolkit>
|
||||
<toolkit>cudaFree</toolkit>
|
||||
<toolkit>cudaMemcpy</toolkit>
|
||||
<toolkit>cudaFreeHost</toolkit>
|
||||
</cuda_api_list>
|
||||
<description><![CDATA[A simple demonstration of system wide atomic instructions.]]></description>
|
||||
<devicecompilation>whole</devicecompilation>
|
||||
<includepaths>
|
||||
<path>./</path>
|
||||
<path>../</path>
|
||||
<path>../../common/inc</path>
|
||||
</includepaths>
|
||||
<keyconcepts>
|
||||
<concept level="basic">Atomic Intrinsics</concept>
|
||||
<concept level="basic">Unified Memory</concept>
|
||||
</keyconcepts>
|
||||
<keywords>
|
||||
<keyword>CUDA</keyword>
|
||||
<keyword>GPGPU</keyword>
|
||||
<keyword>system-wide atomic</keyword>
|
||||
</keywords>
|
||||
<libraries>
|
||||
</libraries>
|
||||
<librarypaths>
|
||||
</librarypaths>
|
||||
<nsight_eclipse>true</nsight_eclipse>
|
||||
<primary_file>systemWideAtomics.cu</primary_file>
|
||||
<required_dependencies>
|
||||
<dependency>UVM</dependency>
|
||||
</required_dependencies>
|
||||
<scopes>
|
||||
<scope>1:CUDA Basic Topics</scope>
|
||||
</scopes>
|
||||
<sm-arch>sm60</sm-arch>
|
||||
<sm-arch>sm61</sm-arch>
|
||||
<sm-arch>sm70</sm-arch>
|
||||
<sm-arch>sm75</sm-arch>
|
||||
<supported_envs>
|
||||
<env>
|
||||
<arch>x86_64</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
<env>
|
||||
<arch>ppc64le</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
</supported_envs>
|
||||
<supported_sm_architectures>
|
||||
<from>6.0</from>
|
||||
</supported_sm_architectures>
|
||||
<title>System wide Atomics</title>
|
||||
<type>exe</type>
|
||||
</entry>
|
||||
64
Samples/systemWideAtomics/README.md
Normal file
64
Samples/systemWideAtomics/README.md
Normal file
@@ -0,0 +1,64 @@
|
||||
# systemWideAtomics - System wide Atomics
|
||||
|
||||
## Description
|
||||
|
||||
A simple demonstration of system wide atomic instructions.
|
||||
|
||||
## Key Concepts
|
||||
|
||||
Atomic Intrinsics, Unified Memory
|
||||
|
||||
## Supported SM Architectures
|
||||
|
||||
[SM 6.0 ](https://developer.nvidia.com/cuda-gpus) [SM 6.1 ](https://developer.nvidia.com/cuda-gpus) [SM 7.0 ](https://developer.nvidia.com/cuda-gpus) [SM 7.5 ](https://developer.nvidia.com/cuda-gpus)
|
||||
|
||||
## Supported OSes
|
||||
|
||||
Linux
|
||||
|
||||
## Supported CPU Architecture
|
||||
|
||||
x86_64, ppc64le
|
||||
|
||||
## CUDA APIs involved
|
||||
|
||||
### [CUDA Runtime API](http://docs.nvidia.com/cuda/cuda-runtime-api/index.html)
|
||||
cudaMalloc, cudaFree, cudaMemcpy, cudaFreeHost
|
||||
|
||||
## Dependencies needed to build/run
|
||||
[UVM](../../README.md#uvm)
|
||||
|
||||
## Prerequisites
|
||||
|
||||
Download and install the [CUDA Toolkit 10.0](https://developer.nvidia.com/cuda-downloads) for your corresponding platform.
|
||||
Make sure the dependencies mentioned in [Dependencies]() section above are installed.
|
||||
|
||||
## Build and Run
|
||||
|
||||
### 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.
|
||||
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/>
|
||||
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)
|
||||
|
||||
342
Samples/systemWideAtomics/systemWideAtomics.cu
Normal file
342
Samples/systemWideAtomics/systemWideAtomics.cu
Normal file
@@ -0,0 +1,342 @@
|
||||
/* Copyright (c) 2018, 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.
|
||||
*/
|
||||
|
||||
/* A program demonstrating trivial use of system-wide atomics on migratable
|
||||
* memory.
|
||||
*/
|
||||
|
||||
#include <cuda_runtime.h>
|
||||
#include <helper_cuda.h>
|
||||
#include <math.h>
|
||||
#include <stdint.h>
|
||||
#include <cstdio>
|
||||
#include <ctime>
|
||||
|
||||
#define min(a, b) (a) < (b) ? (a) : (b)
|
||||
#define max(a, b) (a) > (b) ? (a) : (b)
|
||||
|
||||
#define LOOP_NUM 50
|
||||
|
||||
__global__ void atomicKernel(int *atom_arr) {
|
||||
unsigned int tid = blockDim.x * blockIdx.x + threadIdx.x;
|
||||
|
||||
for (int i = 0; i < LOOP_NUM; i++) {
|
||||
// Atomic addition
|
||||
atomicAdd_system(&atom_arr[0], 10);
|
||||
|
||||
// Atomic exchange
|
||||
atomicExch_system(&atom_arr[1], tid);
|
||||
|
||||
// Atomic maximum
|
||||
atomicMax_system(&atom_arr[2], tid);
|
||||
|
||||
// Atomic minimum
|
||||
atomicMin_system(&atom_arr[3], tid);
|
||||
|
||||
// Atomic increment (modulo 17+1)
|
||||
atomicInc_system((unsigned int *)&atom_arr[4], 17);
|
||||
|
||||
// Atomic decrement
|
||||
atomicDec_system((unsigned int *)&atom_arr[5], 137);
|
||||
|
||||
// Atomic compare-and-swap
|
||||
atomicCAS_system(&atom_arr[6], tid - 1, tid);
|
||||
|
||||
// Bitwise atomic instructions
|
||||
|
||||
// Atomic AND
|
||||
atomicAnd_system(&atom_arr[7], 2 * tid + 7);
|
||||
|
||||
// Atomic OR
|
||||
atomicOr_system(&atom_arr[8], 1 << tid);
|
||||
|
||||
// Atomic XOR
|
||||
atomicXor_system(&atom_arr[9], tid);
|
||||
}
|
||||
}
|
||||
|
||||
void atomicKernel_CPU(int *atom_arr, int no_of_threads) {
|
||||
for (int i = no_of_threads; i < 2 * no_of_threads; i++) {
|
||||
for (int j = 0; j < LOOP_NUM; j++) {
|
||||
// Atomic addition
|
||||
__sync_fetch_and_add(&atom_arr[0], 10);
|
||||
|
||||
// Atomic exchange
|
||||
__sync_lock_test_and_set(&atom_arr[1], i);
|
||||
|
||||
// Atomic maximum
|
||||
int old, expected;
|
||||
do {
|
||||
expected = atom_arr[2];
|
||||
old = __sync_val_compare_and_swap(&atom_arr[2], expected,
|
||||
max(expected, i));
|
||||
} while (old != expected);
|
||||
|
||||
// Atomic minimum
|
||||
do {
|
||||
expected = atom_arr[3];
|
||||
old = __sync_val_compare_and_swap(&atom_arr[3], expected,
|
||||
min(expected, i));
|
||||
} while (old != expected);
|
||||
|
||||
// Atomic increment (modulo 17+1)
|
||||
int limit = 17;
|
||||
do {
|
||||
expected = atom_arr[4];
|
||||
old = __sync_val_compare_and_swap(
|
||||
&atom_arr[4], expected, (expected >= limit) ? 0 : expected + 1);
|
||||
} while (old != expected);
|
||||
|
||||
// Atomic decrement
|
||||
limit = 137;
|
||||
do {
|
||||
expected = atom_arr[5];
|
||||
old = __sync_val_compare_and_swap(
|
||||
&atom_arr[5], expected,
|
||||
((expected == 0) || (expected > limit)) ? limit : expected - 1);
|
||||
} while (old != expected);
|
||||
|
||||
// Atomic compare-and-swap
|
||||
__sync_val_compare_and_swap(&atom_arr[6], i - 1, i);
|
||||
|
||||
// Bitwise atomic instructions
|
||||
|
||||
// Atomic AND
|
||||
__sync_fetch_and_and(&atom_arr[7], 2 * i + 7);
|
||||
|
||||
// Atomic OR
|
||||
__sync_fetch_and_or(&atom_arr[8], 1 << i);
|
||||
|
||||
// Atomic XOR
|
||||
// 11th element should be 0xff
|
||||
__sync_fetch_and_xor(&atom_arr[9], i);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
//! Compute reference data set
|
||||
//! Each element is multiplied with the number of threads / array length
|
||||
//! @param reference reference data, computed but preallocated
|
||||
//! @param idata input data as provided to device
|
||||
//! @param len number of elements in reference / idata
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
int verify(int *testData, const int len) {
|
||||
int val = 0;
|
||||
|
||||
for (int i = 0; i < len * LOOP_NUM; ++i) {
|
||||
val += 10;
|
||||
}
|
||||
|
||||
if (val != testData[0]) {
|
||||
printf("atomicAdd failed val = %d testData = %d\n", val, testData[0]);
|
||||
return false;
|
||||
}
|
||||
|
||||
val = 0;
|
||||
|
||||
bool found = false;
|
||||
|
||||
for (int i = 0; i < len; ++i) {
|
||||
// second element should be a member of [0, len)
|
||||
if (i == testData[1]) {
|
||||
found = true;
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
if (!found) {
|
||||
printf("atomicExch failed\n");
|
||||
return false;
|
||||
}
|
||||
|
||||
val = -(1 << 8);
|
||||
|
||||
for (int i = 0; i < len; ++i) {
|
||||
// third element should be len-1
|
||||
val = max(val, i);
|
||||
}
|
||||
|
||||
if (val != testData[2]) {
|
||||
printf("atomicMax failed\n");
|
||||
return false;
|
||||
}
|
||||
|
||||
val = 1 << 8;
|
||||
|
||||
for (int i = 0; i < len; ++i) {
|
||||
val = min(val, i);
|
||||
}
|
||||
|
||||
if (val != testData[3]) {
|
||||
printf("atomicMin failed\n");
|
||||
return false;
|
||||
}
|
||||
|
||||
int limit = 17;
|
||||
val = 0;
|
||||
|
||||
for (int i = 0; i < len * LOOP_NUM; ++i) {
|
||||
val = (val >= limit) ? 0 : val + 1;
|
||||
}
|
||||
|
||||
if (val != testData[4]) {
|
||||
printf("atomicInc failed\n");
|
||||
return false;
|
||||
}
|
||||
|
||||
limit = 137;
|
||||
val = 0;
|
||||
|
||||
for (int i = 0; i < len * LOOP_NUM; ++i) {
|
||||
val = ((val == 0) || (val > limit)) ? limit : val - 1;
|
||||
}
|
||||
|
||||
if (val != testData[5]) {
|
||||
printf("atomicDec failed\n");
|
||||
return false;
|
||||
}
|
||||
|
||||
found = false;
|
||||
|
||||
for (int i = 0; i < len; ++i) {
|
||||
// seventh element should be a member of [0, len)
|
||||
if (i == testData[6]) {
|
||||
found = true;
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
if (!found) {
|
||||
printf("atomicCAS failed\n");
|
||||
return false;
|
||||
}
|
||||
|
||||
val = 0xff;
|
||||
|
||||
for (int i = 0; i < len; ++i) {
|
||||
// 8th element should be 1
|
||||
val &= (2 * i + 7);
|
||||
}
|
||||
|
||||
if (val != testData[7]) {
|
||||
printf("atomicAnd failed\n");
|
||||
return false;
|
||||
}
|
||||
|
||||
val = 0;
|
||||
|
||||
for (int i = 0; i < len; ++i) {
|
||||
// 9th element should be 0xff
|
||||
val |= (1 << i);
|
||||
}
|
||||
|
||||
if (val != testData[8]) {
|
||||
printf("atomicOr failed\n");
|
||||
return false;
|
||||
}
|
||||
|
||||
val = 0xff;
|
||||
|
||||
for (int i = 0; i < len; ++i) {
|
||||
// 11th element should be 0xff
|
||||
val ^= i;
|
||||
}
|
||||
|
||||
if (val != testData[9]) {
|
||||
printf("atomicXor failed\n");
|
||||
return false;
|
||||
}
|
||||
|
||||
return true;
|
||||
}
|
||||
|
||||
int main(int argc, char **argv) {
|
||||
// set device
|
||||
cudaDeviceProp device_prop;
|
||||
int dev_id = findCudaDevice(argc, (const char **)argv);
|
||||
checkCudaErrors(cudaGetDeviceProperties(&device_prop, dev_id));
|
||||
|
||||
if (!device_prop.managedMemory) {
|
||||
// This samples requires being run on a device that supports Unified Memory
|
||||
fprintf(stderr, "Unified Memory not supported on this device\n");
|
||||
exit(EXIT_WAIVED);
|
||||
}
|
||||
|
||||
if (device_prop.computeMode == cudaComputeModeProhibited) {
|
||||
// This sample requires being run with a default or process exclusive mode
|
||||
fprintf(stderr,
|
||||
"This sample requires a device in either default or process "
|
||||
"exclusive mode\n");
|
||||
exit(EXIT_WAIVED);
|
||||
}
|
||||
|
||||
if (device_prop.major < 6) {
|
||||
printf(
|
||||
"%s: requires a minimum CUDA compute 6.0 capability, waiving "
|
||||
"testing.\n",
|
||||
argv[0]);
|
||||
exit(EXIT_WAIVED);
|
||||
}
|
||||
|
||||
unsigned int numThreads = 256;
|
||||
unsigned int numBlocks = 64;
|
||||
unsigned int numData = 10;
|
||||
|
||||
int *atom_arr;
|
||||
|
||||
if (device_prop.pageableMemoryAccess) {
|
||||
printf("CAN access pageable memory\n");
|
||||
atom_arr = (int *)malloc(sizeof(int) * numData);
|
||||
} else {
|
||||
printf("CANNOT access pageable memory\n");
|
||||
checkCudaErrors(cudaMallocManaged(&atom_arr, sizeof(int) * numData));
|
||||
}
|
||||
|
||||
for (unsigned int i = 0; i < numData; i++) atom_arr[i] = 0;
|
||||
|
||||
// To make the AND and XOR tests generate something other than 0...
|
||||
atom_arr[7] = atom_arr[9] = 0xff;
|
||||
|
||||
atomicKernel<<<numBlocks, numThreads>>>(atom_arr);
|
||||
atomicKernel_CPU(atom_arr, numBlocks * numThreads);
|
||||
|
||||
checkCudaErrors(cudaDeviceSynchronize());
|
||||
|
||||
// Compute & verify reference solution
|
||||
int testResult = verify(atom_arr, 2 * numThreads * numBlocks);
|
||||
|
||||
if (device_prop.pageableMemoryAccess) {
|
||||
free(atom_arr);
|
||||
} else {
|
||||
cudaFree(atom_arr);
|
||||
}
|
||||
|
||||
printf("systemWideAtomics completed, returned %s \n",
|
||||
testResult ? "OK" : "ERROR!");
|
||||
exit(testResult ? EXIT_SUCCESS : EXIT_FAILURE);
|
||||
}
|
||||
Reference in New Issue
Block a user