mirror of
https://github.com/NVIDIA/cuda-samples.git
synced 2026-10-23 21:28:17 +08:00
add and update samples for CUDA 11.6
This commit is contained in:
Rutwik Choughule
18
Samples/3_CUDA_Features/memMapIPCDrv/.vscode/c_cpp_properties.json
vendored
Normal file
18
Samples/3_CUDA_Features/memMapIPCDrv/.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/3_CUDA_Features/memMapIPCDrv/.vscode/extensions.json
vendored
Normal file
7
Samples/3_CUDA_Features/memMapIPCDrv/.vscode/extensions.json
vendored
Normal file
@@ -0,0 +1,7 @@
|
||||
{
|
||||
"recommendations": [
|
||||
"nvidia.nsight-vscode-edition",
|
||||
"ms-vscode.cpptools",
|
||||
"ms-vscode.makefile-tools"
|
||||
]
|
||||
}
|
||||
10
Samples/3_CUDA_Features/memMapIPCDrv/.vscode/launch.json
vendored
Normal file
10
Samples/3_CUDA_Features/memMapIPCDrv/.vscode/launch.json
vendored
Normal file
@@ -0,0 +1,10 @@
|
||||
{
|
||||
"configurations": [
|
||||
{
|
||||
"name": "CUDA C++: Launch",
|
||||
"type": "cuda-gdb",
|
||||
"request": "launch",
|
||||
"program": "${workspaceFolder}/memMapIPCDrv"
|
||||
}
|
||||
]
|
||||
}
|
||||
15
Samples/3_CUDA_Features/memMapIPCDrv/.vscode/tasks.json
vendored
Normal file
15
Samples/3_CUDA_Features/memMapIPCDrv/.vscode/tasks.json
vendored
Normal file
@@ -0,0 +1,15 @@
|
||||
{
|
||||
"version": "2.0.0",
|
||||
"tasks": [
|
||||
{
|
||||
"label": "sample",
|
||||
"type": "shell",
|
||||
"command": "make dbg=1",
|
||||
"problemMatcher": ["$nvcc"],
|
||||
"group": {
|
||||
"kind": "build",
|
||||
"isDefault": true
|
||||
}
|
||||
}
|
||||
]
|
||||
}
|
||||
438
Samples/3_CUDA_Features/memMapIPCDrv/Makefile
Normal file
438
Samples/3_CUDA_Features/memMapIPCDrv/Makefile
Normal file
@@ -0,0 +1,438 @@
|
||||
################################################################################
|
||||
# Copyright (c) 2022, NVIDIA CORPORATION. All rights reserved.
|
||||
#
|
||||
# Redistribution and use in source and binary forms, with or without
|
||||
# modification, are permitted provided that the following conditions
|
||||
# are met:
|
||||
# * Redistributions of source code must retain the above copyright
|
||||
# notice, this list of conditions and the following disclaimer.
|
||||
# * Redistributions in binary form must reproduce the above copyright
|
||||
# notice, this list of conditions and the following disclaimer in the
|
||||
# documentation and/or other materials provided with the distribution.
|
||||
# * Neither the name of NVIDIA CORPORATION nor the names of its
|
||||
# contributors may be used to endorse or promote products derived
|
||||
# from this software without specific prior written permission.
|
||||
#
|
||||
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
||||
# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||||
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
# PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
||||
# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
||||
# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||||
# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||||
# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
||||
# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
#
|
||||
################################################################################
|
||||
#
|
||||
# Makefile project only supported on Mac OS X and Linux Platforms)
|
||||
#
|
||||
################################################################################
|
||||
|
||||
# Location of the CUDA Toolkit
|
||||
CUDA_PATH ?= /usr/local/cuda
|
||||
|
||||
##############################
|
||||
# start deprecated interface #
|
||||
##############################
|
||||
ifeq ($(x86_64),1)
|
||||
$(info WARNING - x86_64 variable has been deprecated)
|
||||
$(info WARNING - please use TARGET_ARCH=x86_64 instead)
|
||||
TARGET_ARCH ?= x86_64
|
||||
endif
|
||||
ifeq ($(ARMv7),1)
|
||||
$(info WARNING - ARMv7 variable has been deprecated)
|
||||
$(info WARNING - please use TARGET_ARCH=armv7l instead)
|
||||
TARGET_ARCH ?= armv7l
|
||||
endif
|
||||
ifeq ($(aarch64),1)
|
||||
$(info WARNING - aarch64 variable has been deprecated)
|
||||
$(info WARNING - please use TARGET_ARCH=aarch64 instead)
|
||||
TARGET_ARCH ?= aarch64
|
||||
endif
|
||||
ifeq ($(ppc64le),1)
|
||||
$(info WARNING - ppc64le variable has been deprecated)
|
||||
$(info WARNING - please use TARGET_ARCH=ppc64le instead)
|
||||
TARGET_ARCH ?= ppc64le
|
||||
endif
|
||||
ifneq ($(GCC),)
|
||||
$(info WARNING - GCC variable has been deprecated)
|
||||
$(info WARNING - please use HOST_COMPILER=$(GCC) instead)
|
||||
HOST_COMPILER ?= $(GCC)
|
||||
endif
|
||||
ifneq ($(abi),)
|
||||
$(error ERROR - abi variable has been removed)
|
||||
endif
|
||||
############################
|
||||
# end deprecated interface #
|
||||
############################
|
||||
|
||||
# architecture
|
||||
HOST_ARCH := $(shell uname -m)
|
||||
TARGET_ARCH ?= $(HOST_ARCH)
|
||||
ifneq (,$(filter $(TARGET_ARCH),x86_64 aarch64 sbsa ppc64le armv7l))
|
||||
ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
||||
ifneq (,$(filter $(TARGET_ARCH),x86_64 aarch64 sbsa ppc64le))
|
||||
TARGET_SIZE := 64
|
||||
else ifneq (,$(filter $(TARGET_ARCH),armv7l))
|
||||
TARGET_SIZE := 32
|
||||
endif
|
||||
else
|
||||
TARGET_SIZE := $(shell getconf LONG_BIT)
|
||||
endif
|
||||
else
|
||||
$(error ERROR - unsupported value $(TARGET_ARCH) for TARGET_ARCH!)
|
||||
endif
|
||||
|
||||
# sbsa and aarch64 systems look similar. Need to differentiate them at host level for now.
|
||||
ifeq ($(HOST_ARCH),aarch64)
|
||||
ifeq ($(CUDA_PATH)/targets/sbsa-linux,$(shell ls -1d $(CUDA_PATH)/targets/sbsa-linux 2>/dev/null))
|
||||
HOST_ARCH := sbsa
|
||||
TARGET_ARCH := sbsa
|
||||
endif
|
||||
endif
|
||||
|
||||
ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
||||
ifeq (,$(filter $(HOST_ARCH)-$(TARGET_ARCH),aarch64-armv7l x86_64-armv7l x86_64-aarch64 x86_64-sbsa x86_64-ppc64le))
|
||||
$(error ERROR - cross compiling from $(HOST_ARCH) to $(TARGET_ARCH) is not supported!)
|
||||
endif
|
||||
endif
|
||||
|
||||
# When on native aarch64 system with userspace of 32-bit, change TARGET_ARCH to armv7l
|
||||
ifeq ($(HOST_ARCH)-$(TARGET_ARCH)-$(TARGET_SIZE),aarch64-aarch64-32)
|
||||
TARGET_ARCH = armv7l
|
||||
endif
|
||||
|
||||
# operating system
|
||||
HOST_OS := $(shell uname -s 2>/dev/null | tr "[:upper:]" "[:lower:]")
|
||||
TARGET_OS ?= $(HOST_OS)
|
||||
ifeq (,$(filter $(TARGET_OS),linux darwin qnx android))
|
||||
$(error ERROR - unsupported value $(TARGET_OS) for TARGET_OS!)
|
||||
endif
|
||||
|
||||
# host compiler
|
||||
ifeq ($(TARGET_OS),darwin)
|
||||
ifeq ($(shell expr `xcodebuild -version | grep -i xcode | awk '{print $$2}' | cut -d'.' -f1` \>= 5),1)
|
||||
HOST_COMPILER ?= clang++
|
||||
endif
|
||||
else ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
||||
ifeq ($(HOST_ARCH)-$(TARGET_ARCH),x86_64-armv7l)
|
||||
ifeq ($(TARGET_OS),linux)
|
||||
HOST_COMPILER ?= arm-linux-gnueabihf-g++
|
||||
else ifeq ($(TARGET_OS),qnx)
|
||||
ifeq ($(QNX_HOST),)
|
||||
$(error ERROR - QNX_HOST must be passed to the QNX host toolchain)
|
||||
endif
|
||||
ifeq ($(QNX_TARGET),)
|
||||
$(error ERROR - QNX_TARGET must be passed to the QNX target toolchain)
|
||||
endif
|
||||
export QNX_HOST
|
||||
export QNX_TARGET
|
||||
HOST_COMPILER ?= $(QNX_HOST)/usr/bin/arm-unknown-nto-qnx6.6.0eabi-g++
|
||||
else ifeq ($(TARGET_OS),android)
|
||||
HOST_COMPILER ?= arm-linux-androideabi-g++
|
||||
endif
|
||||
else ifeq ($(TARGET_ARCH),aarch64)
|
||||
ifeq ($(TARGET_OS), linux)
|
||||
HOST_COMPILER ?= aarch64-linux-gnu-g++
|
||||
else ifeq ($(TARGET_OS),qnx)
|
||||
ifeq ($(QNX_HOST),)
|
||||
$(error ERROR - QNX_HOST must be passed to the QNX host toolchain)
|
||||
endif
|
||||
ifeq ($(QNX_TARGET),)
|
||||
$(error ERROR - QNX_TARGET must be passed to the QNX target toolchain)
|
||||
endif
|
||||
export QNX_HOST
|
||||
export QNX_TARGET
|
||||
HOST_COMPILER ?= $(QNX_HOST)/usr/bin/q++
|
||||
else ifeq ($(TARGET_OS), android)
|
||||
HOST_COMPILER ?= aarch64-linux-android-clang++
|
||||
endif
|
||||
else ifeq ($(TARGET_ARCH),sbsa)
|
||||
HOST_COMPILER ?= aarch64-linux-gnu-g++
|
||||
else ifeq ($(TARGET_ARCH),ppc64le)
|
||||
HOST_COMPILER ?= powerpc64le-linux-gnu-g++
|
||||
endif
|
||||
endif
|
||||
HOST_COMPILER ?= g++
|
||||
NVCC := $(CUDA_PATH)/bin/nvcc -ccbin $(HOST_COMPILER)
|
||||
|
||||
# internal flags
|
||||
NVCCFLAGS := -m${TARGET_SIZE}
|
||||
CCFLAGS :=
|
||||
LDFLAGS :=
|
||||
|
||||
# build flags
|
||||
ifeq ($(TARGET_OS),darwin)
|
||||
LDFLAGS += -rpath $(CUDA_PATH)/lib
|
||||
CCFLAGS += -arch $(HOST_ARCH)
|
||||
else ifeq ($(HOST_ARCH)-$(TARGET_ARCH)-$(TARGET_OS),x86_64-armv7l-linux)
|
||||
LDFLAGS += --dynamic-linker=/lib/ld-linux-armhf.so.3
|
||||
CCFLAGS += -mfloat-abi=hard
|
||||
else ifeq ($(TARGET_OS),android)
|
||||
LDFLAGS += -pie
|
||||
CCFLAGS += -fpie -fpic -fexceptions
|
||||
endif
|
||||
|
||||
ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-linux)
|
||||
ifneq ($(TARGET_FS),)
|
||||
GCCVERSIONLTEQ46 := $(shell expr `$(HOST_COMPILER) -dumpversion` \<= 4.6)
|
||||
ifeq ($(GCCVERSIONLTEQ46),1)
|
||||
CCFLAGS += --sysroot=$(TARGET_FS)
|
||||
endif
|
||||
LDFLAGS += --sysroot=$(TARGET_FS)
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/lib
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib/arm-linux-gnueabihf
|
||||
endif
|
||||
endif
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-linux)
|
||||
ifneq ($(TARGET_FS),)
|
||||
GCCVERSIONLTEQ46 := $(shell expr `$(HOST_COMPILER) -dumpversion` \<= 4.6)
|
||||
ifeq ($(GCCVERSIONLTEQ46),1)
|
||||
CCFLAGS += --sysroot=$(TARGET_FS)
|
||||
endif
|
||||
LDFLAGS += --sysroot=$(TARGET_FS)
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/lib -L$(TARGET_FS)/lib
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/lib/aarch64-linux-gnu -L$(TARGET_FS)/lib/aarch64-linux-gnu
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib -L$(TARGET_FS)/usr/lib
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib/aarch64-linux-gnu -L$(TARGET_FS)/usr/lib/aarch64-linux-gnu
|
||||
LDFLAGS += --unresolved-symbols=ignore-in-shared-libs
|
||||
CCFLAGS += -isystem=$(TARGET_FS)/usr/include -I$(TARGET_FS)/usr/include -I$(TARGET_FS)/usr/include/libdrm
|
||||
CCFLAGS += -isystem=$(TARGET_FS)/usr/include/aarch64-linux-gnu -I$(TARGET_FS)/usr/include/aarch64-linux-gnu
|
||||
endif
|
||||
endif
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-qnx)
|
||||
NVCCFLAGS += -D_QNX_SOURCE
|
||||
NVCCFLAGS += --qpp-config 8.3.0,gcc_ntoaarch64le
|
||||
CCFLAGS += -DWIN_INTERFACE_CUSTOM -I/usr/include/aarch64-qnx-gnu
|
||||
LDFLAGS += -lsocket
|
||||
LDFLAGS += -L/usr/lib/aarch64-qnx-gnu
|
||||
CCFLAGS += "-Wl\,-rpath-link\,/usr/lib/aarch64-qnx-gnu"
|
||||
ifdef TARGET_OVERRIDE
|
||||
LDFLAGS += -lslog2
|
||||
endif
|
||||
|
||||
ifneq ($(TARGET_FS),)
|
||||
LDFLAGS += -L$(TARGET_FS)/usr/lib
|
||||
CCFLAGS += "-Wl\,-rpath-link\,$(TARGET_FS)/usr/lib"
|
||||
LDFLAGS += -L$(TARGET_FS)/usr/libnvidia
|
||||
CCFLAGS += "-Wl\,-rpath-link\,$(TARGET_FS)/usr/libnvidia"
|
||||
CCFLAGS += -I$(TARGET_FS)/../include
|
||||
endif
|
||||
endif
|
||||
endif
|
||||
|
||||
ifdef TARGET_OVERRIDE # cuda toolkit targets override
|
||||
NVCCFLAGS += -target-dir $(TARGET_OVERRIDE)
|
||||
endif
|
||||
|
||||
# Install directory of different arch
|
||||
CUDA_INSTALL_TARGET_DIR :=
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-linux)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/armv7-linux-gnueabihf/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-linux)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/aarch64-linux/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),sbsa-linux)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/sbsa-linux/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-android)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/armv7-linux-androideabi/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-android)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/aarch64-linux-androideabi/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-qnx)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/ARMv7-linux-QNX/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-qnx)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/aarch64-qnx/
|
||||
else ifeq ($(TARGET_ARCH),ppc64le)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/ppc64le-linux/
|
||||
endif
|
||||
|
||||
# Debug build flags
|
||||
ifeq ($(dbg),1)
|
||||
NVCCFLAGS += -g -G
|
||||
BUILD_TYPE := debug
|
||||
else
|
||||
BUILD_TYPE := release
|
||||
endif
|
||||
|
||||
ALL_CCFLAGS :=
|
||||
ALL_CCFLAGS += $(NVCCFLAGS)
|
||||
ALL_CCFLAGS += $(EXTRA_NVCCFLAGS)
|
||||
ALL_CCFLAGS += $(addprefix -Xcompiler ,$(CCFLAGS))
|
||||
ALL_CCFLAGS += $(addprefix -Xcompiler ,$(EXTRA_CCFLAGS))
|
||||
|
||||
UBUNTU = $(shell lsb_release -i -s 2>/dev/null | grep -i ubuntu)
|
||||
|
||||
SAMPLE_ENABLED := 1
|
||||
|
||||
# This sample is not supported on Mac OSX
|
||||
ifeq ($(TARGET_OS),darwin)
|
||||
$(info >>> WARNING - memMapIPCDrv is not supported on Mac OSX - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
# This sample is not supported on ARMv7
|
||||
ifeq ($(TARGET_ARCH),armv7l)
|
||||
$(info >>> WARNING - memMapIPCDrv is not supported on ARMv7 - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
# This sample is not supported on aarch64
|
||||
ifeq ($(TARGET_ARCH),aarch64)
|
||||
$(info >>> WARNING - memMapIPCDrv is not supported on aarch64 - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
# This sample is not supported on sbsa
|
||||
ifeq ($(TARGET_ARCH),sbsa)
|
||||
$(info >>> WARNING - memMapIPCDrv is not supported on sbsa - 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 :=
|
||||
|
||||
################################################################################
|
||||
|
||||
PTX_FILE := memMapIpc_kernel${TARGET_SIZE}.ptx
|
||||
|
||||
# Gencode arguments
|
||||
SMS ?=
|
||||
|
||||
ifeq ($(GENCODE_FLAGS),)
|
||||
# Generate SASS code for each SM architecture listed in $(SMS)
|
||||
$(foreach sm,$(SMS),$(eval GENCODE_FLAGS += -gencode arch=compute_$(sm),code=sm_$(sm)))
|
||||
|
||||
ifeq ($(SMS),)
|
||||
ifeq ($(TARGET_ARCH),$(filter $(TARGET_ARCH),armv7l aarch64 sbsa))
|
||||
# Generate PTX code from SM 53
|
||||
GENCODE_FLAGS += -gencode arch=compute_53,code=compute_53
|
||||
else
|
||||
# Generate PTX code from SM 35
|
||||
GENCODE_FLAGS += -gencode arch=compute_35,code=compute_35
|
||||
endif
|
||||
endif
|
||||
|
||||
# Generate PTX code from the highest SM architecture in $(SMS) to guarantee forward-compatibility
|
||||
HIGHEST_SM := $(lastword $(sort $(SMS)))
|
||||
ifneq ($(HIGHEST_SM),)
|
||||
GENCODE_FLAGS += -gencode arch=compute_$(HIGHEST_SM),code=compute_$(HIGHEST_SM)
|
||||
endif
|
||||
endif
|
||||
|
||||
ifeq ($(TARGET_OS),darwin)
|
||||
ALL_LDFLAGS += -Xcompiler -F/Library/Frameworks -Xlinker -framework -Xlinker CUDA
|
||||
else
|
||||
ifeq ($(TARGET_ARCH),x86_64)
|
||||
CUDA_SEARCH_PATH ?= $(CUDA_PATH)/lib64/stubs
|
||||
CUDA_SEARCH_PATH += $(CUDA_PATH)/lib/stubs
|
||||
CUDA_SEARCH_PATH += $(CUDA_PATH)/targets/x86_64-linux/lib/stubs
|
||||
endif
|
||||
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-linux)
|
||||
CUDA_SEARCH_PATH ?= $(CUDA_PATH)/targets/armv7-linux-gnueabihf/lib/stubs
|
||||
endif
|
||||
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-linux)
|
||||
CUDA_SEARCH_PATH ?= $(CUDA_PATH)/targets/aarch64-linux/lib/stubs
|
||||
endif
|
||||
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),sbsa-linux)
|
||||
CUDA_SEARCH_PATH ?= $(CUDA_PATH)/targets/sbsa-linux/lib/stubs
|
||||
endif
|
||||
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-android)
|
||||
CUDA_SEARCH_PATH ?= $(CUDA_PATH)/targets/armv7-linux-androideabi/lib/stubs
|
||||
endif
|
||||
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-android)
|
||||
CUDA_SEARCH_PATH ?= $(CUDA_PATH)/targets/aarch64-linux-androideabi/lib/stubs
|
||||
endif
|
||||
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-qnx)
|
||||
CUDA_SEARCH_PATH ?= $(CUDA_PATH)/targets/ARMv7-linux-QNX/lib/stubs
|
||||
endif
|
||||
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-qnx)
|
||||
CUDA_SEARCH_PATH ?= $(CUDA_PATH)/targets/aarch64-qnx/lib/stubs
|
||||
ifdef TARGET_OVERRIDE
|
||||
CUDA_SEARCH_PATH := $(CUDA_PATH)/targets/$(TARGET_OVERRIDE)/lib/stubs
|
||||
endif
|
||||
endif
|
||||
|
||||
ifeq ($(TARGET_ARCH),ppc64le)
|
||||
CUDA_SEARCH_PATH ?= $(CUDA_PATH)/targets/ppc64le-linux/lib/stubs
|
||||
endif
|
||||
|
||||
ifeq ($(HOST_ARCH),ppc64le)
|
||||
CUDA_SEARCH_PATH += $(CUDA_PATH)/lib64/stubs
|
||||
endif
|
||||
|
||||
CUDALIB ?= $(shell find -L $(CUDA_SEARCH_PATH) -maxdepth 1 -name libcuda.so 2> /dev/null)
|
||||
ifeq ("$(CUDALIB)","")
|
||||
$(info >>> WARNING - libcuda.so not found, CUDA Driver is not installed. Please re-install the driver. <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
else
|
||||
CUDALIB := $(shell echo $(CUDALIB) | sed "s/ .*//" | sed "s/\/libcuda.so//" )
|
||||
LIBRARIES += -L$(CUDALIB) -lcuda
|
||||
endif
|
||||
endif
|
||||
|
||||
ALL_CCFLAGS += --threads 0 --std=c++11
|
||||
|
||||
ifeq ($(SAMPLE_ENABLED),0)
|
||||
EXEC ?= @echo "[@]"
|
||||
endif
|
||||
|
||||
################################################################################
|
||||
|
||||
# Target rules
|
||||
all: build
|
||||
|
||||
build: memMapIPCDrv $(PTX_FILE)
|
||||
|
||||
check.deps:
|
||||
ifeq ($(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
|
||||
|
||||
$(PTX_FILE): memMapIpc_kernel.cu
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -ptx $<
|
||||
$(EXEC) mkdir -p data
|
||||
$(EXEC) cp -f $@ ./data
|
||||
$(EXEC) mkdir -p ../../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)
|
||||
$(EXEC) cp -f $@ ../../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)
|
||||
|
||||
helper_multiprocess.o:../../../Common/helper_multiprocess.cpp
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
memMapIpc.o:memMapIpc.cpp
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
memMapIPCDrv: helper_multiprocess.o memMapIpc.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
|
||||
$(EXEC) ./memMapIPCDrv
|
||||
|
||||
testrun: build
|
||||
|
||||
clean:
|
||||
rm -f memMapIPCDrv helper_multiprocess.o memMapIpc.o data/$(PTX_FILE) $(PTX_FILE)
|
||||
rm -rf ../../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)/memMapIPCDrv
|
||||
rm -rf ../../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)/$(PTX_FILE)
|
||||
|
||||
clobber: clean
|
||||
74
Samples/3_CUDA_Features/memMapIPCDrv/README.md
Normal file
74
Samples/3_CUDA_Features/memMapIPCDrv/README.md
Normal file
@@ -0,0 +1,74 @@
|
||||
# memMapIPCDrv - Memmap IPC Driver API
|
||||
|
||||
## Description
|
||||
|
||||
This CUDA Driver API sample is a very basic sample that demonstrates Inter Process Communication using cuMemMap APIs with one process per GPU for computation. Requires Compute Capability 3.0 or higher and a Linux Operating System, or a Windows Operating System
|
||||
|
||||
## Key Concepts
|
||||
|
||||
CUDA Driver API, cuMemMap IPC, MMAP
|
||||
|
||||
## Supported SM Architectures
|
||||
|
||||
[SM 3.5 ](https://developer.nvidia.com/cuda-gpus) [SM 3.7 ](https://developer.nvidia.com/cuda-gpus) [SM 5.0 ](https://developer.nvidia.com/cuda-gpus) [SM 5.2 ](https://developer.nvidia.com/cuda-gpus) [SM 5.3 ](https://developer.nvidia.com/cuda-gpus) [SM 6.0 ](https://developer.nvidia.com/cuda-gpus) [SM 6.1 ](https://developer.nvidia.com/cuda-gpus) [SM 7.0 ](https://developer.nvidia.com/cuda-gpus) [SM 7.2 ](https://developer.nvidia.com/cuda-gpus) [SM 7.5 ](https://developer.nvidia.com/cuda-gpus) [SM 8.0 ](https://developer.nvidia.com/cuda-gpus) [SM 8.6 ](https://developer.nvidia.com/cuda-gpus) [SM 8.7 ](https://developer.nvidia.com/cuda-gpus)
|
||||
|
||||
## Supported OSes
|
||||
|
||||
Linux, Windows
|
||||
|
||||
## Supported CPU Architecture
|
||||
|
||||
x86_64, ppc64le
|
||||
|
||||
## CUDA APIs involved
|
||||
|
||||
### [CUDA Driver API](http://docs.nvidia.com/cuda/cuda-driver-api/index.html)
|
||||
cuCtxSetCurrent, cuMemSetAccess, cuMemcpyDtoHAsync, cuStreamDestroy, cuInit, cuMemAddressReserve, cuCtxDestroy, cuModuleGetFunction, cuModuleLoad, cuStreamCreate, cuCtxCreate, cuMemExportToShareableHandle, cuMemAddressFree, cuMemGetAllocationGranularity, cuModuleLoadDataEx, cuDeviceGet, cuMemUnmap, cuDeviceGetAttribute, cuMemRelease, cuCtxEnablePeerAccess, cuMemMap, cuMemImportFromShareableHandle, cuMemCreate, cuStreamSynchronize, cuDeviceCanAccessPeer, cuDeviceGetCount, cuLaunchKernel, cuOccupancyMaxActiveBlocksPerMultiprocessor
|
||||
|
||||
## Dependencies needed to build/run
|
||||
[IPC](../../README.md#ipc)
|
||||
|
||||
## Prerequisites
|
||||
|
||||
Download and install the [CUDA Toolkit 11.6](https://developer.nvidia.com/cuda-downloads) for your corresponding platform.
|
||||
Make sure the dependencies mentioned in [Dependencies]() section above are installed.
|
||||
|
||||
## Build and Run
|
||||
|
||||
### Windows
|
||||
The Windows samples are built using the Visual Studio IDE. Solution files (.sln) are provided for each supported version of Visual Studio, using the format:
|
||||
```
|
||||
*_vs<version>.sln - for Visual Studio <version>
|
||||
```
|
||||
Each individual sample has its own set of solution files in its directory:
|
||||
|
||||
To build/examine all the samples at once, the complete solution files should be used. To build/examine a single sample, the individual sample solution files should be used.
|
||||
> **Note:** Some samples require that the Microsoft DirectX SDK (June 2010 or newer) be installed and that the VC++ directory paths are properly set up (**Tools > Options...**). Check DirectX Dependencies section for details."
|
||||
|
||||
### Linux
|
||||
The Linux samples are built using makefiles. To use the makefiles, change the current directory to the sample directory you wish to build, and run make:
|
||||
```
|
||||
$ cd <sample_dir>
|
||||
$ make
|
||||
```
|
||||
The samples makefiles can take advantage of certain options:
|
||||
* **TARGET_ARCH=<arch>** - cross-compile targeting a specific architecture. Allowed architectures are x86_64, ppc64le.
|
||||
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)
|
||||
|
||||
20
Samples/3_CUDA_Features/memMapIPCDrv/memMapIPCDrv_vs2017.sln
Normal file
20
Samples/3_CUDA_Features/memMapIPCDrv/memMapIPCDrv_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}") = "memMapIPCDrv", "memMapIPCDrv_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
|
||||
117
Samples/3_CUDA_Features/memMapIPCDrv/memMapIPCDrv_vs2017.vcxproj
Normal file
117
Samples/3_CUDA_Features/memMapIPCDrv/memMapIPCDrv_vs2017.vcxproj
Normal file
@@ -0,0 +1,117 @@
|
||||
<?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>memMapIPCDrv_vs2017</RootNamespace>
|
||||
<ProjectName>memMapIPCDrv</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(WindowsTargetPlatformVersion)'==''">
|
||||
<LatestTargetPlatformVersion>$([Microsoft.Build.Utilities.ToolLocationHelper]::GetLatestSDKTargetPlatformVersion('Windows', '10.0'))</LatestTargetPlatformVersion>
|
||||
<WindowsTargetPlatformVersion Condition="'$(WindowsTargetPlatformVersion)' == ''">$(LatestTargetPlatformVersion)</WindowsTargetPlatformVersion>
|
||||
<TargetPlatformVersion>$(WindowsTargetPlatformVersion)</TargetPlatformVersion>
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v141</PlatformToolset>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<UseDebugLibraries>true</UseDebugLibraries>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Release'">
|
||||
<WholeProgramOptimization>true</WholeProgramOptimization>
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
|
||||
<ImportGroup Label="ExtensionSettings">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cuda.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)/memMapIPCDrv.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_35,compute_35;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" --threads 0 </AdditionalOptions>
|
||||
<Include>./;../../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<ClCompile Include="memMapIpc.cpp" />
|
||||
<CudaCompile Include="memMapIpc_kernel.cu">
|
||||
<CompileOut Condition="'$(Platform)'=='x64'">data/%(Filename)64.ptx</CompileOut>
|
||||
<NvccCompilation>ptx</NvccCompilation>
|
||||
</CudaCompile>
|
||||
<ClCompile Include="../../../Common/helper_multiprocess.cpp" />
|
||||
<ClInclude Include="../../../Common/helper_multiprocess.h" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/3_CUDA_Features/memMapIPCDrv/memMapIPCDrv_vs2019.sln
Normal file
20
Samples/3_CUDA_Features/memMapIPCDrv/memMapIPCDrv_vs2019.sln
Normal file
@@ -0,0 +1,20 @@
|
||||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2019
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "memMapIPCDrv", "memMapIPCDrv_vs2019.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
113
Samples/3_CUDA_Features/memMapIPCDrv/memMapIPCDrv_vs2019.vcxproj
Normal file
113
Samples/3_CUDA_Features/memMapIPCDrv/memMapIPCDrv_vs2019.vcxproj
Normal file
@@ -0,0 +1,113 @@
|
||||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>memMapIPCDrv_vs2019</RootNamespace>
|
||||
<ProjectName>memMapIPCDrv</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v142</PlatformToolset>
|
||||
<WindowsTargetPlatformVersion>10.0</WindowsTargetPlatformVersion>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<UseDebugLibraries>true</UseDebugLibraries>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Release'">
|
||||
<WholeProgramOptimization>true</WholeProgramOptimization>
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
|
||||
<ImportGroup Label="ExtensionSettings">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cuda.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)/memMapIPCDrv.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_35,compute_35;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" --threads 0 </AdditionalOptions>
|
||||
<Include>./;../../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<ClCompile Include="memMapIpc.cpp" />
|
||||
<CudaCompile Include="memMapIpc_kernel.cu">
|
||||
<CompileOut Condition="'$(Platform)'=='x64'">data/%(Filename)64.ptx</CompileOut>
|
||||
<NvccCompilation>ptx</NvccCompilation>
|
||||
</CudaCompile>
|
||||
<ClCompile Include="../../../Common/helper_multiprocess.cpp" />
|
||||
<ClInclude Include="../../../Common/helper_multiprocess.h" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/3_CUDA_Features/memMapIPCDrv/memMapIPCDrv_vs2022.sln
Normal file
20
Samples/3_CUDA_Features/memMapIPCDrv/memMapIPCDrv_vs2022.sln
Normal file
@@ -0,0 +1,20 @@
|
||||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2022
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "memMapIPCDrv", "memMapIPCDrv_vs2022.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
113
Samples/3_CUDA_Features/memMapIPCDrv/memMapIPCDrv_vs2022.vcxproj
Normal file
113
Samples/3_CUDA_Features/memMapIPCDrv/memMapIPCDrv_vs2022.vcxproj
Normal file
@@ -0,0 +1,113 @@
|
||||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>memMapIPCDrv_vs2022</RootNamespace>
|
||||
<ProjectName>memMapIPCDrv</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v143</PlatformToolset>
|
||||
<WindowsTargetPlatformVersion>10.0</WindowsTargetPlatformVersion>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<UseDebugLibraries>true</UseDebugLibraries>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Release'">
|
||||
<WholeProgramOptimization>true</WholeProgramOptimization>
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
|
||||
<ImportGroup Label="ExtensionSettings">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cuda.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)/memMapIPCDrv.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_35,compute_35;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" --threads 0 </AdditionalOptions>
|
||||
<Include>./;../../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<ClCompile Include="memMapIpc.cpp" />
|
||||
<CudaCompile Include="memMapIpc_kernel.cu">
|
||||
<CompileOut Condition="'$(Platform)'=='x64'">data/%(Filename)64.ptx</CompileOut>
|
||||
<NvccCompilation>ptx</NvccCompilation>
|
||||
</CudaCompile>
|
||||
<ClCompile Include="../../../Common/helper_multiprocess.cpp" />
|
||||
<ClInclude Include="../../../Common/helper_multiprocess.h" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.6.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
646
Samples/3_CUDA_Features/memMapIPCDrv/memMapIpc.cpp
Normal file
646
Samples/3_CUDA_Features/memMapIPCDrv/memMapIpc.cpp
Normal file
@@ -0,0 +1,646 @@
|
||||
/* Copyright (c) 2022, NVIDIA CORPORATION. All rights reserved.
|
||||
*
|
||||
* Redistribution and use in source and binary forms, with or without
|
||||
* modification, are permitted provided that the following conditions
|
||||
* are met:
|
||||
* * Redistributions of source code must retain the above copyright
|
||||
* notice, this list of conditions and the following disclaimer.
|
||||
* * Redistributions in binary form must reproduce the above copyright
|
||||
* notice, this list of conditions and the following disclaimer in the
|
||||
* documentation and/or other materials provided with the distribution.
|
||||
* * Neither the name of NVIDIA CORPORATION nor the names of its
|
||||
* contributors may be used to endorse or promote products derived
|
||||
* from this software without specific prior written permission.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
||||
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||||
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
||||
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
||||
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||||
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||||
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
||||
* OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
*/
|
||||
|
||||
/*
|
||||
* This sample demonstrates Inter Process Communication
|
||||
* using cuMemMap APIs and with one process per GPU for computation.
|
||||
*/
|
||||
|
||||
#include <stdio.h>
|
||||
#include <string.h>
|
||||
#include <cstring>
|
||||
#include <iostream>
|
||||
#include "cuda.h"
|
||||
|
||||
#include "helper_multiprocess.h"
|
||||
|
||||
// includes, project
|
||||
#include <helper_functions.h>
|
||||
#include "helper_cuda_drvapi.h"
|
||||
|
||||
// includes, CUDA
|
||||
#include <builtin_types.h>
|
||||
|
||||
using namespace std;
|
||||
|
||||
// For direct NVLINK and PCI-E peers, at max 8 simultaneous peers are allowed
|
||||
// For NVSWITCH connected peers like DGX-2, simultaneous peers are not limited
|
||||
// in the same way.
|
||||
#define MAX_DEVICES (32)
|
||||
|
||||
#define PROCESSES_PER_DEVICE 1
|
||||
#define DATA_BUF_SIZE 4ULL * 1024ULL * 1024ULL
|
||||
|
||||
static const char ipcName[] = "memmap_ipc_pipe";
|
||||
static const char shmName[] = "memmap_ipc_shm";
|
||||
|
||||
typedef struct shmStruct_st {
|
||||
size_t nprocesses;
|
||||
int barrier;
|
||||
int sense;
|
||||
} shmStruct;
|
||||
|
||||
bool findModulePath(const char *, string &, char **, string &);
|
||||
|
||||
// define input ptx file for different platforms
|
||||
#if defined(_WIN64) || defined(__LP64__)
|
||||
#define PTX_FILE "memMapIpc_kernel64.ptx"
|
||||
#else
|
||||
#define PTX_FILE "memMapIpc_kernel32.ptx"
|
||||
#endif
|
||||
|
||||
// `ipcHandleTypeFlag` specifies the platform specific handle type this sample
|
||||
// uses for importing and exporting memory allocation. On Linux this sample
|
||||
// specifies the type as CU_MEM_HANDLE_TYPE_POSIX_FILE_DESCRIPTOR meaning that
|
||||
// file descriptors will be used. On Windows this sample specifies the type as
|
||||
// CU_MEM_HANDLE_TYPE_WIN32 meaning that NT HANDLEs will be used. The
|
||||
// ipcHandleTypeFlag variable is a convenience variable and is passed by value
|
||||
// to individual requests.
|
||||
#if defined(__linux__)
|
||||
CUmemAllocationHandleType ipcHandleTypeFlag =
|
||||
CU_MEM_HANDLE_TYPE_POSIX_FILE_DESCRIPTOR;
|
||||
#else
|
||||
CUmemAllocationHandleType ipcHandleTypeFlag = CU_MEM_HANDLE_TYPE_WIN32;
|
||||
#endif
|
||||
|
||||
#if defined(__linux__)
|
||||
#define cpu_atomic_add32(a, x) __sync_add_and_fetch(a, x)
|
||||
#elif defined(WIN32) || defined(_WIN32) || defined(WIN64) || defined(_WIN64)
|
||||
#define cpu_atomic_add32(a, x) InterlockedAdd((volatile LONG *)a, x)
|
||||
#else
|
||||
#error Unsupported system
|
||||
#endif
|
||||
|
||||
CUmodule cuModule;
|
||||
CUfunction _memMapIpc_kernel;
|
||||
|
||||
static void barrierWait(volatile int *barrier, volatile int *sense,
|
||||
unsigned int n) {
|
||||
int count;
|
||||
|
||||
// Check-in
|
||||
count = cpu_atomic_add32(barrier, 1);
|
||||
if (count == n) { // Last one in
|
||||
*sense = 1;
|
||||
}
|
||||
while (!*sense)
|
||||
;
|
||||
|
||||
// Check-out
|
||||
count = cpu_atomic_add32(barrier, -1);
|
||||
if (count == 0) { // Last one out
|
||||
*sense = 0;
|
||||
}
|
||||
while (*sense)
|
||||
;
|
||||
}
|
||||
|
||||
// Windows-specific LPSECURITYATTRIBUTES
|
||||
void getDefaultSecurityDescriptor(CUmemAllocationProp *prop) {
|
||||
#if defined(__linux__)
|
||||
return;
|
||||
#elif defined(WIN32) || defined(_WIN32) || defined(WIN64) || defined(_WIN64)
|
||||
static const char sddl[] = "D:P(OA;;GARCSDWDWOCCDCLCSWLODTWPRPCRFA;;;WD)";
|
||||
static OBJECT_ATTRIBUTES objAttributes;
|
||||
static bool objAttributesConfigured = false;
|
||||
|
||||
if (!objAttributesConfigured) {
|
||||
PSECURITY_DESCRIPTOR secDesc;
|
||||
BOOL result = ConvertStringSecurityDescriptorToSecurityDescriptorA(
|
||||
sddl, SDDL_REVISION_1, &secDesc, NULL);
|
||||
if (result == 0) {
|
||||
printf("IPC failure: getDefaultSecurityDescriptor Failed! (%d)\n",
|
||||
GetLastError());
|
||||
}
|
||||
|
||||
InitializeObjectAttributes(&objAttributes, NULL, 0, NULL, secDesc);
|
||||
|
||||
objAttributesConfigured = true;
|
||||
}
|
||||
|
||||
prop->win32HandleMetaData = &objAttributes;
|
||||
return;
|
||||
#endif
|
||||
}
|
||||
|
||||
static void memMapAllocateAndExportMemory(
|
||||
unsigned char backingDevice, size_t allocSize,
|
||||
std::vector<CUmemGenericAllocationHandle> &allocationHandles,
|
||||
std::vector<ShareableHandle> &shareableHandles) {
|
||||
// This property structure describes the physical location where the memory
|
||||
// will be allocated via cuMemCreate along with additional properties.
|
||||
CUmemAllocationProp prop = {};
|
||||
|
||||
// The allocations will be device pinned memory backed on backingDevice and
|
||||
// exportable with the specified handle type.
|
||||
prop.type = CU_MEM_ALLOCATION_TYPE_PINNED;
|
||||
prop.location.type = CU_MEM_LOCATION_TYPE_DEVICE;
|
||||
|
||||
// Back all allocations on backingDevice.
|
||||
prop.location.id = (int)backingDevice;
|
||||
|
||||
// Passing a requestedHandleTypes indicates intention to export this
|
||||
// allocation to a platform-specific handle. This sample requests a file
|
||||
// descriptor on Linux and NT Handle on Windows.
|
||||
prop.requestedHandleTypes = ipcHandleTypeFlag;
|
||||
|
||||
// Get the minimum granularity supported for allocation with cuMemCreate()
|
||||
size_t granularity = 0;
|
||||
checkCudaErrors(cuMemGetAllocationGranularity(
|
||||
&granularity, &prop, CU_MEM_ALLOC_GRANULARITY_MINIMUM));
|
||||
if (allocSize % granularity) {
|
||||
printf(
|
||||
"Allocation size is not a multiple of minimum supported granularity "
|
||||
"for this device. Exiting...\n");
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
|
||||
// Windows-specific LPSECURITYATTRIBUTES is required when
|
||||
// CU_MEM_HANDLE_TYPE_WIN32 is used. The security attribute defines the scope
|
||||
// of which exported allocations may be tranferred to other processes. For all
|
||||
// other handle types, pass NULL.
|
||||
getDefaultSecurityDescriptor(&prop);
|
||||
|
||||
for (int i = 0; i < allocationHandles.size(); i++) {
|
||||
// Create the allocation as a pinned allocation on device specified in
|
||||
// prop.location.id
|
||||
checkCudaErrors(cuMemCreate(&allocationHandles[i], allocSize, &prop, 0));
|
||||
|
||||
// Export the allocation to a platform-specific handle. The type of handle
|
||||
// requested here must match the requestedHandleTypes field in the prop
|
||||
// structure passed to cuMemCreate.
|
||||
checkCudaErrors(cuMemExportToShareableHandle((void *)&shareableHandles[i],
|
||||
allocationHandles[i],
|
||||
ipcHandleTypeFlag, 0));
|
||||
}
|
||||
}
|
||||
|
||||
static void memMapImportAndMapMemory(
|
||||
CUdeviceptr d_ptr, size_t mapSize,
|
||||
std::vector<ShareableHandle> &shareableHandles, int mapDevice) {
|
||||
std::vector<CUmemGenericAllocationHandle> allocationHandles;
|
||||
allocationHandles.resize(shareableHandles.size());
|
||||
|
||||
// The accessDescriptor will describe the mapping requirement for the
|
||||
// mapDevice passed as argument
|
||||
CUmemAccessDesc accessDescriptor;
|
||||
|
||||
// Specify location for mapping the imported allocations.
|
||||
accessDescriptor.location.type = CU_MEM_LOCATION_TYPE_DEVICE;
|
||||
accessDescriptor.location.id = mapDevice;
|
||||
|
||||
// Specify both read and write accesses.
|
||||
accessDescriptor.flags = CU_MEM_ACCESS_FLAGS_PROT_READWRITE;
|
||||
|
||||
for (int i = 0; i < shareableHandles.size(); i++) {
|
||||
// Import the memory allocation back into a CUDA handle from the platform
|
||||
// specific handle.
|
||||
checkCudaErrors(cuMemImportFromShareableHandle(
|
||||
&allocationHandles[i], (void *)(uintptr_t)shareableHandles[i],
|
||||
ipcHandleTypeFlag));
|
||||
|
||||
// Assign the chunk to the appropriate VA range and release the handle.
|
||||
// After mapping the memory, it can be referenced by virtual address.
|
||||
checkCudaErrors(
|
||||
cuMemMap(d_ptr + (i * mapSize), mapSize, 0, allocationHandles[i], 0));
|
||||
|
||||
// Since we do not need to make any other mappings of this memory or export
|
||||
// it, we no longer need and can release the allocationHandle. The
|
||||
// allocation will be kept live until it is unmapped.
|
||||
checkCudaErrors(cuMemRelease(allocationHandles[i]));
|
||||
}
|
||||
|
||||
// Retain peer access and map all chunks to mapDevice
|
||||
checkCudaErrors(cuMemSetAccess(d_ptr, shareableHandles.size() * mapSize,
|
||||
&accessDescriptor, 1));
|
||||
}
|
||||
|
||||
static void memMapUnmapAndFreeMemory(CUdeviceptr dptr, size_t size) {
|
||||
CUresult status = CUDA_SUCCESS;
|
||||
|
||||
// Unmap the mapped virtual memory region
|
||||
// Since the handles to the mapped backing stores have already been released
|
||||
// by cuMemRelease, and these are the only/last mappings referencing them,
|
||||
// The backing stores will be freed.
|
||||
// Since the memory has been unmapped after this call, accessing the specified
|
||||
// va range will result in a fault (unitll it is remapped).
|
||||
checkCudaErrors(cuMemUnmap(dptr, size));
|
||||
|
||||
// Free the virtual address region. This allows the virtual address region
|
||||
// to be reused by future cuMemAddressReserve calls. This also allows the
|
||||
// virtual address region to be used by other allocation made through
|
||||
// opperating system calls like malloc & mmap.
|
||||
checkCudaErrors(cuMemAddressFree(dptr, size));
|
||||
}
|
||||
|
||||
static void memMapGetDeviceFunction(char **argv) {
|
||||
// first search for the module path before we load the results
|
||||
string module_path, ptx_source;
|
||||
if (!findModulePath(PTX_FILE, module_path, argv, ptx_source)) {
|
||||
if (!findModulePath("memMapIpc_kernel.cubin", module_path, argv,
|
||||
ptx_source)) {
|
||||
printf(
|
||||
"> findModulePath could not find <simpleMemMapIpc> ptx or cubin\n");
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
} else {
|
||||
printf("> initCUDA loading module: <%s>\n", module_path.c_str());
|
||||
}
|
||||
|
||||
// Create module from binary file (PTX or CUBIN)
|
||||
if (module_path.rfind("ptx") != string::npos) {
|
||||
// in this branch we use compilation with parameters
|
||||
const unsigned int jitNumOptions = 3;
|
||||
CUjit_option *jitOptions = new CUjit_option[jitNumOptions];
|
||||
void **jitOptVals = new void *[jitNumOptions];
|
||||
// set up size of compilation log buffer
|
||||
jitOptions[0] = CU_JIT_INFO_LOG_BUFFER_SIZE_BYTES;
|
||||
int jitLogBufferSize = 1024;
|
||||
jitOptVals[0] = (void *)(size_t)jitLogBufferSize;
|
||||
// set up pointer to the compilation log buffer
|
||||
jitOptions[1] = CU_JIT_INFO_LOG_BUFFER;
|
||||
char *jitLogBuffer = new char[jitLogBufferSize];
|
||||
jitOptVals[1] = jitLogBuffer;
|
||||
// set up pointer to set the Maximum # of registers for a particular kernel
|
||||
jitOptions[2] = CU_JIT_MAX_REGISTERS;
|
||||
int jitRegCount = 32;
|
||||
jitOptVals[2] = (void *)(size_t)jitRegCount;
|
||||
checkCudaErrors(cuModuleLoadDataEx(&cuModule, ptx_source.c_str(),
|
||||
jitNumOptions, jitOptions,
|
||||
(void **)jitOptVals));
|
||||
printf("> PTX JIT log:\n%s\n", jitLogBuffer);
|
||||
} else {
|
||||
checkCudaErrors(cuModuleLoad(&cuModule, module_path.c_str()));
|
||||
}
|
||||
|
||||
// Get function handle from module
|
||||
checkCudaErrors(
|
||||
cuModuleGetFunction(&_memMapIpc_kernel, cuModule, "memMapIpc_kernel"));
|
||||
}
|
||||
|
||||
static void childProcess(int devId, int id, char **argv) {
|
||||
volatile shmStruct *shm = NULL;
|
||||
sharedMemoryInfo info;
|
||||
ipcHandle *ipcChildHandle = NULL;
|
||||
int blocks = 0;
|
||||
int threads = 128;
|
||||
|
||||
checkIpcErrors(ipcOpenSocket(ipcChildHandle));
|
||||
|
||||
if (sharedMemoryOpen(shmName, sizeof(shmStruct), &info) != 0) {
|
||||
printf("Failed to create shared memory slab\n");
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
shm = (volatile shmStruct *)info.addr;
|
||||
int procCount = (int)shm->nprocesses;
|
||||
|
||||
barrierWait(&shm->barrier, &shm->sense, (unsigned int)(procCount + 1));
|
||||
|
||||
// Receive all allocation handles shared by Parent.
|
||||
std::vector<ShareableHandle> shHandle(procCount);
|
||||
checkIpcErrors(ipcRecvShareableHandles(ipcChildHandle, shHandle));
|
||||
|
||||
CUcontext ctx;
|
||||
CUdevice device;
|
||||
CUstream stream;
|
||||
int multiProcessorCount;
|
||||
|
||||
checkCudaErrors(cuDeviceGet(&device, devId));
|
||||
checkCudaErrors(cuCtxCreate(&ctx, 0, device));
|
||||
checkCudaErrors(cuStreamCreate(&stream, CU_STREAM_NON_BLOCKING));
|
||||
|
||||
// Obtain kernel function for the sample
|
||||
memMapGetDeviceFunction(argv);
|
||||
|
||||
checkCudaErrors(cuOccupancyMaxActiveBlocksPerMultiprocessor(
|
||||
&blocks, _memMapIpc_kernel, threads, 0));
|
||||
checkCudaErrors(cuDeviceGetAttribute(
|
||||
&multiProcessorCount, CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT, device));
|
||||
blocks *= multiProcessorCount;
|
||||
|
||||
CUdeviceptr d_ptr = 0ULL;
|
||||
|
||||
// Reserve the required contiguous VA space for the allocations
|
||||
checkCudaErrors(cuMemAddressReserve(&d_ptr, procCount * DATA_BUF_SIZE,
|
||||
DATA_BUF_SIZE, 0, 0));
|
||||
|
||||
// Import the memory allocations shared by the parent with us and map them in
|
||||
// our address space.
|
||||
memMapImportAndMapMemory(d_ptr, DATA_BUF_SIZE, shHandle, devId);
|
||||
|
||||
// Since we have imported allocations shared by the parent with us, we can
|
||||
// close all the ShareableHandles.
|
||||
for (int i = 0; i < procCount; i++) {
|
||||
checkIpcErrors(ipcCloseShareableHandle(shHandle[i]));
|
||||
}
|
||||
checkIpcErrors(ipcCloseSocket(ipcChildHandle));
|
||||
|
||||
for (int i = 0; i < procCount; i++) {
|
||||
size_t bufferId = (i + id) % procCount;
|
||||
|
||||
// Build arguments to be passed to cuda kernel.
|
||||
CUdeviceptr ptr = d_ptr + (bufferId * DATA_BUF_SIZE);
|
||||
int size = DATA_BUF_SIZE;
|
||||
char val = (char)id;
|
||||
|
||||
void *args[] = {&ptr, &size, &val};
|
||||
|
||||
// Push a simple kernel on th buffer.
|
||||
checkCudaErrors(cuLaunchKernel(_memMapIpc_kernel, blocks, 1, 1, threads, 1,
|
||||
1, 0, stream, args, 0));
|
||||
checkCudaErrors(cuStreamSynchronize(stream));
|
||||
|
||||
// Wait for all my sibling processes to push this stage of their work
|
||||
// before proceeding to the next. This makes the data in the buffer
|
||||
// deterministic.
|
||||
barrierWait(&shm->barrier, &shm->sense, (unsigned int)procCount);
|
||||
if (id == 0) {
|
||||
printf("Step %lld done\n", (unsigned long long)i);
|
||||
}
|
||||
}
|
||||
|
||||
printf("Process %d: verifying...\n", id);
|
||||
|
||||
// Copy the data onto host and verify value if it matches expected value or
|
||||
// not.
|
||||
std::vector<char> verification_buffer(DATA_BUF_SIZE);
|
||||
checkCudaErrors(cuMemcpyDtoHAsync(&verification_buffer[0],
|
||||
d_ptr + (id * DATA_BUF_SIZE), DATA_BUF_SIZE,
|
||||
stream));
|
||||
checkCudaErrors(cuStreamSynchronize(stream));
|
||||
|
||||
// The contents should have the id of the sibling just after me
|
||||
char compareId = (char)((id + 1) % procCount);
|
||||
for (unsigned long long j = 0; j < DATA_BUF_SIZE; j++) {
|
||||
if (verification_buffer[j] != compareId) {
|
||||
printf("Process %d: Verification mismatch at %lld: %d != %d\n", id, j,
|
||||
(int)verification_buffer[j], (int)compareId);
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
// Clean up!
|
||||
checkCudaErrors(cuStreamDestroy(stream));
|
||||
checkCudaErrors(cuCtxDestroy(ctx));
|
||||
|
||||
// Unmap the allocations from our address space. Unmapping will also free the
|
||||
// handle as we have already released the imported handle with the call to
|
||||
// cuMemRelease. Finally, free up the Virtual Address space we reserved with
|
||||
// cuMemAddressReserve.
|
||||
memMapUnmapAndFreeMemory(d_ptr, procCount * DATA_BUF_SIZE);
|
||||
|
||||
exit(EXIT_SUCCESS);
|
||||
}
|
||||
|
||||
static void parentProcess(char *app) {
|
||||
int devCount, i, nprocesses = 0;
|
||||
volatile shmStruct *shm = NULL;
|
||||
sharedMemoryInfo info;
|
||||
std::vector<Process> processes;
|
||||
|
||||
checkCudaErrors(cuDeviceGetCount(&devCount));
|
||||
std::vector<CUdevice> devices(devCount);
|
||||
|
||||
if (sharedMemoryCreate(shmName, sizeof(*shm), &info) != 0) {
|
||||
printf("Failed to create shared memory slab\n");
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
|
||||
shm = (volatile shmStruct *)info.addr;
|
||||
memset((void *)shm, 0, sizeof(*shm));
|
||||
|
||||
for (i = 0; i < devCount; i++) {
|
||||
checkCudaErrors(cuDeviceGet(&devices[i], i));
|
||||
}
|
||||
|
||||
std::vector<CUcontext> ctxs;
|
||||
std::vector<unsigned char> selectedDevices;
|
||||
|
||||
// Pick all the devices that can access each other's memory for this test
|
||||
// Keep in mind that CUDA has minimal support for fork() without a
|
||||
// corresponding exec() in the child process, but in this case our
|
||||
// spawnProcess will always exec, so no need to worry.
|
||||
for (i = 0; i < devCount; i++) {
|
||||
bool allPeers = true;
|
||||
int deviceComputeMode;
|
||||
int deviceSupportsIpcHandle;
|
||||
int attributeVal = 0;
|
||||
|
||||
checkCudaErrors(cuDeviceGet(&devices[i], i));
|
||||
checkCudaErrors(cuDeviceGetAttribute(
|
||||
&deviceComputeMode, CU_DEVICE_ATTRIBUTE_COMPUTE_MODE, devices[i]));
|
||||
checkCudaErrors(cuDeviceGetAttribute(
|
||||
&attributeVal, CU_DEVICE_ATTRIBUTE_VIRTUAL_ADDRESS_MANAGEMENT_SUPPORTED,
|
||||
devices[i]));
|
||||
#if defined(__linux__)
|
||||
checkCudaErrors(cuDeviceGetAttribute(
|
||||
&deviceSupportsIpcHandle,
|
||||
CU_DEVICE_ATTRIBUTE_HANDLE_TYPE_POSIX_FILE_DESCRIPTOR_SUPPORTED,
|
||||
devices[i]));
|
||||
#else
|
||||
checkCudaErrors(cuDeviceGetAttribute(
|
||||
&deviceSupportsIpcHandle,
|
||||
CU_DEVICE_ATTRIBUTE_HANDLE_TYPE_WIN32_HANDLE_SUPPORTED, devices[i]));
|
||||
#endif
|
||||
// Check that the selected device supports virtual address management
|
||||
if (attributeVal == 0) {
|
||||
printf("Device %d doesn't support VIRTUAL ADDRESS MANAGEMENT.\n",
|
||||
devices[i]);
|
||||
continue;
|
||||
}
|
||||
|
||||
// This sample requires two processes accessing each device, so we need
|
||||
// to ensure exclusive or prohibited mode is not set
|
||||
if (deviceComputeMode != CU_COMPUTEMODE_DEFAULT) {
|
||||
printf("Device %d is in an unsupported compute mode for this sample\n",
|
||||
i);
|
||||
continue;
|
||||
}
|
||||
|
||||
if (!deviceSupportsIpcHandle) {
|
||||
printf(
|
||||
"Device %d does not support requested handle type for IPC, "
|
||||
"skipping...\n",
|
||||
i);
|
||||
continue;
|
||||
}
|
||||
|
||||
for (int j = 0; j < nprocesses; j++) {
|
||||
int canAccessPeerIJ, canAccessPeerJI;
|
||||
checkCudaErrors(
|
||||
cuDeviceCanAccessPeer(&canAccessPeerJI, devices[j], devices[i]));
|
||||
checkCudaErrors(
|
||||
cuDeviceCanAccessPeer(&canAccessPeerIJ, devices[i], devices[j]));
|
||||
if (!canAccessPeerIJ || !canAccessPeerJI) {
|
||||
allPeers = false;
|
||||
break;
|
||||
}
|
||||
}
|
||||
if (allPeers) {
|
||||
CUcontext ctx;
|
||||
checkCudaErrors(cuCtxCreate(&ctx, 0, devices[i]));
|
||||
ctxs.push_back(ctx);
|
||||
|
||||
// Enable peers here. This isn't necessary for IPC, but it will
|
||||
// setup the peers for the device. For systems that only allow 8
|
||||
// peers per GPU at a time, this acts to remove devices from CanAccessPeer
|
||||
for (int j = 0; j < nprocesses; j++) {
|
||||
checkCudaErrors(cuCtxSetCurrent(ctxs[i]));
|
||||
checkCudaErrors(cuCtxEnablePeerAccess(ctxs[j], 0));
|
||||
checkCudaErrors(cuCtxSetCurrent(ctxs[j]));
|
||||
checkCudaErrors(cuCtxEnablePeerAccess(ctxs[i], 0));
|
||||
}
|
||||
selectedDevices.push_back(i);
|
||||
nprocesses++;
|
||||
if (nprocesses >= MAX_DEVICES) {
|
||||
break;
|
||||
}
|
||||
} else {
|
||||
printf(
|
||||
"Device %d is not peer capable with some other selected peers, "
|
||||
"skipping\n",
|
||||
i);
|
||||
}
|
||||
}
|
||||
|
||||
for (int i = 0; i < ctxs.size(); ++i) {
|
||||
checkCudaErrors(cuCtxDestroy(ctxs[i]));
|
||||
};
|
||||
|
||||
if (nprocesses == 0) {
|
||||
printf("No CUDA devices support IPC\n");
|
||||
exit(EXIT_WAIVED);
|
||||
}
|
||||
shm->nprocesses = nprocesses;
|
||||
|
||||
unsigned char firstSelectedDevice = selectedDevices[0];
|
||||
|
||||
std::vector<ShareableHandle> shHandles(nprocesses);
|
||||
std::vector<CUmemGenericAllocationHandle> allocationHandles(nprocesses);
|
||||
|
||||
// Allocate `nprocesses` number of memory chunks and obtain a shareable handle
|
||||
// for each allocation. Share all memory allocations with all children.
|
||||
memMapAllocateAndExportMemory(firstSelectedDevice, DATA_BUF_SIZE,
|
||||
allocationHandles, shHandles);
|
||||
|
||||
// Launch the child processes!
|
||||
for (i = 0; i < nprocesses; i++) {
|
||||
char devIdx[10];
|
||||
char procIdx[10];
|
||||
char *const args[] = {app, devIdx, procIdx, NULL};
|
||||
Process process;
|
||||
|
||||
SPRINTF(devIdx, "%d", selectedDevices[i]);
|
||||
SPRINTF(procIdx, "%d", i);
|
||||
|
||||
if (spawnProcess(&process, app, args)) {
|
||||
printf("Failed to create process\n");
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
|
||||
processes.push_back(process);
|
||||
}
|
||||
|
||||
barrierWait(&shm->barrier, &shm->sense, (unsigned int)(nprocesses + 1));
|
||||
|
||||
ipcHandle *ipcParentHandle = NULL;
|
||||
checkIpcErrors(ipcCreateSocket(ipcParentHandle, ipcName, processes));
|
||||
checkIpcErrors(
|
||||
ipcSendShareableHandles(ipcParentHandle, shHandles, processes));
|
||||
|
||||
// Close the shareable handles as they are not needed anymore.
|
||||
for (int i = 0; i < nprocesses; i++) {
|
||||
checkIpcErrors(ipcCloseShareableHandle(shHandles[i]));
|
||||
}
|
||||
|
||||
// And wait for them to finish
|
||||
for (i = 0; i < processes.size(); i++) {
|
||||
if (waitProcess(&processes[i]) != EXIT_SUCCESS) {
|
||||
printf("Process %d failed!\n", i);
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
}
|
||||
|
||||
for (i = 0; i < nprocesses; i++) {
|
||||
checkCudaErrors(cuMemRelease(allocationHandles[i]));
|
||||
}
|
||||
|
||||
checkIpcErrors(ipcCloseSocket(ipcParentHandle));
|
||||
sharedMemoryClose(&info);
|
||||
}
|
||||
|
||||
// Host code
|
||||
int main(int argc, char **argv) {
|
||||
#if defined(__arm__) || defined(__aarch64__)
|
||||
printf("Not supported on ARM\n");
|
||||
return EXIT_WAIVED;
|
||||
#else
|
||||
// Initialize
|
||||
checkCudaErrors(cuInit(0));
|
||||
|
||||
if (argc == 1) {
|
||||
parentProcess(argv[0]);
|
||||
} else {
|
||||
childProcess(atoi(argv[1]), atoi(argv[2]), argv);
|
||||
}
|
||||
return EXIT_SUCCESS;
|
||||
#endif
|
||||
}
|
||||
|
||||
bool inline findModulePath(const char *module_file, string &module_path,
|
||||
char **argv, string &ptx_source) {
|
||||
char *actual_path = sdkFindFilePath(module_file, argv[0]);
|
||||
|
||||
if (actual_path) {
|
||||
module_path = actual_path;
|
||||
} else {
|
||||
printf("> findModulePath file not found: <%s> \n", module_file);
|
||||
return false;
|
||||
}
|
||||
|
||||
if (module_path.empty()) {
|
||||
printf("> findModulePath could not find file: <%s> \n", module_file);
|
||||
return false;
|
||||
} else {
|
||||
printf("> findModulePath found file at <%s>\n", module_path.c_str());
|
||||
|
||||
if (module_path.rfind(".ptx") != string::npos) {
|
||||
FILE *fp = fopen(module_path.c_str(), "rb");
|
||||
fseek(fp, 0, SEEK_END);
|
||||
int file_size = ftell(fp);
|
||||
char *buf = new char[file_size + 1];
|
||||
fseek(fp, 0, SEEK_SET);
|
||||
fread(buf, sizeof(char), file_size, fp);
|
||||
fclose(fp);
|
||||
buf[file_size] = '\0';
|
||||
ptx_source = buf;
|
||||
delete[] buf;
|
||||
}
|
||||
|
||||
return true;
|
||||
}
|
||||
}
|
||||
37
Samples/3_CUDA_Features/memMapIPCDrv/memMapIpc_kernel.cu
Normal file
37
Samples/3_CUDA_Features/memMapIPCDrv/memMapIpc_kernel.cu
Normal file
@@ -0,0 +1,37 @@
|
||||
/* Copyright (c) 2022, NVIDIA CORPORATION. All rights reserved.
|
||||
*
|
||||
* Redistribution and use in source and binary forms, with or without
|
||||
* modification, are permitted provided that the following conditions
|
||||
* are met:
|
||||
* * Redistributions of source code must retain the above copyright
|
||||
* notice, this list of conditions and the following disclaimer.
|
||||
* * Redistributions in binary form must reproduce the above copyright
|
||||
* notice, this list of conditions and the following disclaimer in the
|
||||
* documentation and/or other materials provided with the distribution.
|
||||
* * Neither the name of NVIDIA CORPORATION nor the names of its
|
||||
* contributors may be used to endorse or promote products derived
|
||||
* from this software without specific prior written permission.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
||||
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||||
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
||||
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
||||
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||||
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||||
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
||||
* OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
*/
|
||||
|
||||
|
||||
// Device code
|
||||
extern "C" __global__ void memMapIpc_kernel(char *ptr, int sz, char val)
|
||||
{
|
||||
// Dummy kernel
|
||||
int idx = blockIdx.x * blockDim.x + threadIdx.x;
|
||||
for (; idx < sz; idx += (gridDim.x * blockDim.x)) {
|
||||
ptr[idx] = val;
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user