mirror of
https://github.com/NVIDIA/cuda-samples.git
synced 2024-10-06 10:29:14 +08:00
Update samples list to include additional samples.
This commit is contained in:
Andy Dick
parent
8bb8c5fac0
commit
d08d485c67
18
README.md
18
README.md
|
|
@ -9,6 +9,7 @@ This section describes the release notes for the CUDA Samples on GitHub only.
|
|||
### CUDA 9.2
|
||||
|
||||
This is the first release of CUDA Samples on GitHub:
|
||||
* Added `vectorAdd_nvrtc`. Demonstrates runtime compilation library using NVRTC of a simple vectorAdd kernel.
|
||||
* Added `warpAggregatedAtomicsCG`. Demonstrates warp aggregated atomics using Cooperative Groups.
|
||||
* Added `deviceQuery`. Enumerates the properties of the CUDA devices present in the system.
|
||||
* Added `matrixMul`. Demonstrates a matrix multiplication using shared memory through tiled approach.
|
||||
|
|
@ -16,6 +17,10 @@ This is the first release of CUDA Samples on GitHub:
|
|||
* Added `cudaTensorCoreGemm`. Demonstrates a GEMM computation using the Warp Matrix Multiply and Accumulate (WMMA) API introduced in CUDA 9, as well as the new Tensor Cores introduced in the Volta chip family.
|
||||
* Added `simpleVoteIntrinsics` which uses *_sync equivalent of the vote intrinsics _any, _all added since CUDA 9.0.
|
||||
* Added `shfl_scan` which uses *_sync equivalent of the shfl intrinsics added since CUDA 9.0.
|
||||
* Added `conjugateGradientMultiBlockCG`. Demonstrates a conjugate gradient solver on GPU using Multi Block Cooperative Groups.
|
||||
* Added `conjugateGradientMultiDeviceCG`. Demonstrates a conjugate gradient solver on multiple GPUs using Multi Device Cooperative Groups, also uses unified memory prefetching and usage hints APIs.
|
||||
* Added `simpleCUBLAS`. Demonstrates how perform GEMM operations using CUBLAS library.
|
||||
* Added `simpleCUFFT`. Demonstrates how perform FFT operations using CUFFT library.
|
||||
|
||||
## Getting Started
|
||||
|
||||
|
|
@ -103,19 +108,22 @@ The samples makefiles can take advantage of certain options:
|
|||
### Samples by OS
|
||||
|
||||
#### Linux
|
||||
**[warpAggregatedAtomicsCG](./Samples/warpAggregatedAtomicsCG)** | **[matrixMulDrv](./Samples/matrixMulDrv)** | **[vectorAdd_nvrtc](./Samples/vectorAdd_nvrtc)** | **[simpleCUFFT](./Samples/simpleCUFFT)** |
|
||||
**[cudaTensorCoreGemm](./Samples/cudaTensorCoreGemm)** | **[conjugateGradientMultiBlockCG](./Samples/conjugateGradientMultiBlockCG)** | **[warpAggregatedAtomicsCG](./Samples/warpAggregatedAtomicsCG)** | **[matrixMulDrv](./Samples/matrixMulDrv)** |
|
||||
---|---|---|---|
|
||||
**[simpleCUBLAS](./Samples/simpleCUBLAS)** | **[matrixMul](./Samples/matrixMul)** | **[deviceQuery](./Samples/deviceQuery)** |
|
||||
**[vectorAdd_nvrtc](./Samples/vectorAdd_nvrtc)** | **[simpleCUFFT](./Samples/simpleCUFFT)** | **[simpleCUBLAS](./Samples/simpleCUBLAS)** | **[conjugateGradientMultiDeviceCG](./Samples/conjugateGradientMultiDeviceCG)** |
|
||||
**[matrixMul](./Samples/matrixMul)** | **[shfl_scan](./Samples/shfl_scan)** | **[simpleVoteIntrinsics](./Samples/simpleVoteIntrinsics)** | **[deviceQuery](./Samples/deviceQuery)** |
|
||||
|
||||
#### Windows
|
||||
**[warpAggregatedAtomicsCG](./Samples/warpAggregatedAtomicsCG)** | **[matrixMulDrv](./Samples/matrixMulDrv)** | **[vectorAdd_nvrtc](./Samples/vectorAdd_nvrtc)** | **[simpleCUFFT](./Samples/simpleCUFFT)** |
|
||||
**[cudaTensorCoreGemm](./Samples/cudaTensorCoreGemm)** | **[conjugateGradientMultiBlockCG](./Samples/conjugateGradientMultiBlockCG)** | **[warpAggregatedAtomicsCG](./Samples/warpAggregatedAtomicsCG)** | **[matrixMulDrv](./Samples/matrixMulDrv)** |
|
||||
---|---|---|---|
|
||||
**[simpleCUBLAS](./Samples/simpleCUBLAS)** | **[matrixMul](./Samples/matrixMul)** | **[deviceQuery](./Samples/deviceQuery)** |
|
||||
**[vectorAdd_nvrtc](./Samples/vectorAdd_nvrtc)** | **[simpleCUFFT](./Samples/simpleCUFFT)** | **[simpleCUBLAS](./Samples/simpleCUBLAS)** | **[conjugateGradientMultiDeviceCG](./Samples/conjugateGradientMultiDeviceCG)** |
|
||||
**[matrixMul](./Samples/matrixMul)** | **[shfl_scan](./Samples/shfl_scan)** | **[simpleVoteIntrinsics](./Samples/simpleVoteIntrinsics)** | **[deviceQuery](./Samples/deviceQuery)** |
|
||||
|
||||
#### Mac OSX
|
||||
**[warpAggregatedAtomicsCG](./Samples/warpAggregatedAtomicsCG)** | **[matrixMulDrv](./Samples/matrixMulDrv)** | **[vectorAdd_nvrtc](./Samples/vectorAdd_nvrtc)** | **[simpleCUFFT](./Samples/simpleCUFFT)** |
|
||||
---|---|---|---|
|
||||
**[simpleCUBLAS](./Samples/simpleCUBLAS)** | **[matrixMul](./Samples/matrixMul)** | **[deviceQuery](./Samples/deviceQuery)** |
|
||||
**[simpleCUBLAS](./Samples/simpleCUBLAS)** | **[matrixMul](./Samples/matrixMul)** | **[shfl_scan](./Samples/shfl_scan)** | **[simpleVoteIntrinsics](./Samples/simpleVoteIntrinsics)** |
|
||||
**[deviceQuery](./Samples/deviceQuery)** |
|
||||
|
||||
## Dependencies
|
||||
|
||||
|
|
|
|||
324
Samples/conjugateGradientMultiBlockCG/Makefile
Normal file
324
Samples/conjugateGradientMultiBlockCG/Makefile
Normal file
|
|
@ -0,0 +1,324 @@
|
|||
################################################################################
|
||||
#
|
||||
# Copyright 1993-2015 NVIDIA Corporation. All rights reserved.
|
||||
#
|
||||
# NOTICE TO USER:
|
||||
#
|
||||
# This source code is subject to NVIDIA ownership rights under U.S. and
|
||||
# international Copyright laws.
|
||||
#
|
||||
# NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOURCE
|
||||
# CODE FOR ANY PURPOSE. IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR
|
||||
# IMPLIED WARRANTY OF ANY KIND. NVIDIA DISCLAIMS ALL WARRANTIES WITH
|
||||
# REGARD TO THIS SOURCE CODE, INCLUDING ALL IMPLIED WARRANTIES OF
|
||||
# MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
|
||||
# IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL,
|
||||
# OR CONSEQUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS
|
||||
# OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
|
||||
# OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE
|
||||
# OR PERFORMANCE OF THIS SOURCE CODE.
|
||||
#
|
||||
# U.S. Government End Users. This source code is a "commercial item" as
|
||||
# that term is defined at 48 C.F.R. 2.101 (OCT 1995), consisting of
|
||||
# "commercial computer software" and "commercial computer software
|
||||
# documentation" as such terms are used in 48 C.F.R. 12.212 (SEPT 1995)
|
||||
# and is provided to the U.S. Government only as a commercial end item.
|
||||
# Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through
|
||||
# 227.7202-4 (JUNE 1995), all U.S. Government End Users acquire the
|
||||
# source code with only those rights set forth herein.
|
||||
#
|
||||
################################################################################
|
||||
#
|
||||
# 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-g++
|
||||
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 - conjugateGradientMultiBlockCG 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 - conjugateGradientMultiBlockCG is not supported on ARMv7 - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
# This sample is not supported on aarch64
|
||||
ifeq ($(TARGET_ARCH),aarch64)
|
||||
$(info >>> WARNING - conjugateGradientMultiBlockCG 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
|
||||
|
||||
ifeq ($(SMS),)
|
||||
$(info >>> WARNING - no SM architectures have been specified - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
ifeq ($(GENCODE_FLAGS),)
|
||||
# Generate SASS code for each SM architecture listed in $(SMS)
|
||||
$(foreach sm,$(SMS),$(eval GENCODE_FLAGS += -gencode arch=compute_$(sm),code=sm_$(sm)))
|
||||
|
||||
# Generate PTX code from the highest SM architecture in $(SMS) to guarantee forward-compatibility
|
||||
HIGHEST_SM := $(lastword $(sort $(SMS)))
|
||||
ifneq ($(HIGHEST_SM),)
|
||||
GENCODE_FLAGS += -gencode arch=compute_$(HIGHEST_SM),code=compute_$(HIGHEST_SM)
|
||||
endif
|
||||
endif
|
||||
|
||||
ALL_CCFLAGS += -dc
|
||||
|
||||
LIBRARIES += -lcudadevrt
|
||||
|
||||
ifeq ($(SAMPLE_ENABLED),0)
|
||||
EXEC ?= @echo "[@]"
|
||||
endif
|
||||
|
||||
################################################################################
|
||||
|
||||
# Target rules
|
||||
all: build
|
||||
|
||||
build: conjugateGradientMultiBlockCG
|
||||
|
||||
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
|
||||
|
||||
conjugateGradientMultiBlockCG.o:conjugateGradientMultiBlockCG.cu
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
conjugateGradientMultiBlockCG: conjugateGradientMultiBlockCG.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) ./conjugateGradientMultiBlockCG
|
||||
|
||||
clean:
|
||||
rm -f conjugateGradientMultiBlockCG conjugateGradientMultiBlockCG.o
|
||||
rm -rf ../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)/conjugateGradientMultiBlockCG
|
||||
|
||||
clobber: clean
|
||||
63
Samples/conjugateGradientMultiBlockCG/NsightEclipse.xml
Normal file
63
Samples/conjugateGradientMultiBlockCG/NsightEclipse.xml
Normal file
|
|
@ -0,0 +1,63 @@
|
|||
<?xml version="1.0" encoding="UTF-8"?>
|
||||
<!DOCTYPE entry SYSTEM "SamplesInfo.dtd">
|
||||
<entry>
|
||||
<name>conjugateGradientMultiBlockCG</name>
|
||||
<cflags>
|
||||
<flag>-dc</flag>
|
||||
</cflags>
|
||||
<description><![CDATA[This sample implements a conjugate gradient solver on GPU using Multi Block Cooperative Groups, also uses Unified Memory.]]></description>
|
||||
<devicecompilation>whole</devicecompilation>
|
||||
<includepaths>
|
||||
<path>./</path>
|
||||
<path>../</path>
|
||||
<path>../../common/inc</path>
|
||||
</includepaths>
|
||||
<keyconcepts>
|
||||
<concept level="basic">Unified Memory</concept>
|
||||
<concept level="advanced">Linear Algebra</concept>
|
||||
<concept level="basic">Cooperative Groups</concept>
|
||||
<concept level="advanced">MultiBlock Cooperative Groups</concept>
|
||||
</keyconcepts>
|
||||
<keywords>
|
||||
<keyword>CUDA</keyword>
|
||||
<keyword>Sparse Matrix</keyword>
|
||||
<keyword>Unified Memory</keyword>
|
||||
<keyword>GPGPU</keyword>
|
||||
</keywords>
|
||||
<libraries>
|
||||
<library>cudadevrt</library>
|
||||
</libraries>
|
||||
<librarypaths>
|
||||
</librarypaths>
|
||||
<nsight_eclipse>true</nsight_eclipse>
|
||||
<primary_file>conjugateGradientMultiBlockCG.cu</primary_file>
|
||||
<required_dependencies>
|
||||
<dependency>UVM</dependency>
|
||||
<dependency>MBCG</dependency>
|
||||
</required_dependencies>
|
||||
<scopes>
|
||||
<scope>1:CUDA Advanced Topics</scope>
|
||||
<scope>3:Linear Algebra</scope>
|
||||
</scopes>
|
||||
<sm-arch>sm60</sm-arch>
|
||||
<sm-arch>sm61</sm-arch>
|
||||
<sm-arch>sm70</sm-arch>
|
||||
<supported_envs>
|
||||
<env>
|
||||
<arch>x86_64</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
<env>
|
||||
<arch>ppc64le</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
<env>
|
||||
<platform>windows7</platform>
|
||||
</env>
|
||||
</supported_envs>
|
||||
<supported_sm_architectures>
|
||||
<from>6.0</from>
|
||||
</supported_sm_architectures>
|
||||
<title>conjugateGradient using MultiBlock Cooperative Groups</title>
|
||||
<type>exe</type>
|
||||
</entry>
|
||||
71
Samples/conjugateGradientMultiBlockCG/README.md
Normal file
71
Samples/conjugateGradientMultiBlockCG/README.md
Normal file
|
|
@ -0,0 +1,71 @@
|
|||
# conjugateGradientMultiBlockCG - conjugateGradient using MultiBlock Cooperative Groups
|
||||
|
||||
## Description
|
||||
|
||||
This sample implements a conjugate gradient solver on GPU using Multi Block Cooperative Groups, also uses Unified Memory.
|
||||
|
||||
## Key Concepts
|
||||
|
||||
Unified Memory, Linear Algebra, Cooperative Groups, MultiBlock Cooperative Groups
|
||||
|
||||
## 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)
|
||||
|
||||
## Supported OSes
|
||||
|
||||
Linux, Windows
|
||||
|
||||
## Supported CPU Architecture
|
||||
|
||||
x86_64, ppc64le
|
||||
|
||||
## CUDA APIs involved
|
||||
|
||||
## Dependencies needed to build/run
|
||||
[UVM](../../README.md#uvm), [MBCG](../../README.md#mbcg)
|
||||
|
||||
## Prerequisites
|
||||
|
||||
Download and install the [CUDA Toolkit 9.2](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)
|
||||
|
||||
|
|
@ -0,0 +1,500 @@
|
|||
/* 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.
|
||||
*/
|
||||
|
||||
/*
|
||||
* This sample implements a conjugate gradient solver on GPU using
|
||||
* Multi Block Cooperative Groups, also uses Unified Memory.
|
||||
*
|
||||
*/
|
||||
|
||||
// includes, system
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
|
||||
#include <cuda_runtime.h>
|
||||
|
||||
// Utilities and system includes
|
||||
#include <helper_cuda.h> // helper function CUDA error checking and initialization
|
||||
#include <helper_functions.h> // helper for shared functions common to CUDA Samples
|
||||
|
||||
#include <cooperative_groups.h>
|
||||
|
||||
namespace cg = cooperative_groups;
|
||||
|
||||
const char *sSDKname = "conjugateGradientMultiBlockCG";
|
||||
|
||||
#define ENABLE_CPU_DEBUG_CODE 0
|
||||
#define THREADS_PER_BLOCK 512
|
||||
|
||||
/* genTridiag: generate a random tridiagonal symmetric matrix */
|
||||
void genTridiag(int *I, int *J, float *val, int N, int nz) {
|
||||
I[0] = 0, J[0] = 0, J[1] = 1;
|
||||
val[0] = static_cast<float>(rand()) / RAND_MAX + 10.0f;
|
||||
val[1] = static_cast<float>(rand()) / RAND_MAX;
|
||||
int start;
|
||||
|
||||
for (int i = 1; i < N; i++) {
|
||||
if (i > 1) {
|
||||
I[i] = I[i - 1] + 3;
|
||||
} else {
|
||||
I[1] = 2;
|
||||
}
|
||||
|
||||
start = (i - 1) * 3 + 2;
|
||||
J[start] = i - 1;
|
||||
J[start + 1] = i;
|
||||
|
||||
if (i < N - 1) {
|
||||
J[start + 2] = i + 1;
|
||||
}
|
||||
|
||||
val[start] = val[start - 1];
|
||||
val[start + 1] = static_cast<float>(rand()) / RAND_MAX + 10.0f;
|
||||
|
||||
if (i < N - 1) {
|
||||
val[start + 2] = static_cast<float>(rand()) / RAND_MAX;
|
||||
}
|
||||
}
|
||||
|
||||
I[N] = nz;
|
||||
}
|
||||
|
||||
// I - contains location of the given non-zero element in the row of the matrix
|
||||
// J - contains location of the given non-zero element in the column of the
|
||||
// matrix val - contains values of the given non-zero elements of the matrix
|
||||
// inputVecX - input vector to be multiplied
|
||||
// outputVecY - resultant vector
|
||||
void cpuSpMV(int *I, int *J, float *val, int nnz, int num_rows, float alpha,
|
||||
float *inputVecX, float *outputVecY) {
|
||||
for (int i = 0; i < num_rows; i++) {
|
||||
int num_elems_this_row = I[i + 1] - I[i];
|
||||
|
||||
float output = 0.0;
|
||||
for (int j = 0; j < num_elems_this_row; j++) {
|
||||
output += alpha * val[I[i] + j] * inputVecX[J[I[i] + j]];
|
||||
}
|
||||
outputVecY[i] = output;
|
||||
}
|
||||
|
||||
return;
|
||||
}
|
||||
|
||||
double dotProduct(float *vecA, float *vecB, int size) {
|
||||
double result = 0.0;
|
||||
|
||||
for (int i = 0; i < size; i++) {
|
||||
result = result + (vecA[i] * vecB[i]);
|
||||
}
|
||||
|
||||
return result;
|
||||
}
|
||||
|
||||
void scaleVector(float *vec, float alpha, int size) {
|
||||
for (int i = 0; i < size; i++) {
|
||||
vec[i] = alpha * vec[i];
|
||||
}
|
||||
}
|
||||
|
||||
void saxpy(float *x, float *y, float a, int size) {
|
||||
for (int i = 0; i < size; i++) {
|
||||
y[i] = a * x[i] + y[i];
|
||||
}
|
||||
}
|
||||
|
||||
void cpuConjugateGrad(int *I, int *J, float *val, float *x, float *Ax, float *p,
|
||||
float *r, int nnz, int N, float tol) {
|
||||
int max_iter = 10000;
|
||||
|
||||
float alpha = 1.0;
|
||||
float alpham1 = -1.0;
|
||||
float r0 = 0.0, b, a, na;
|
||||
|
||||
cpuSpMV(I, J, val, nnz, N, alpha, x, Ax);
|
||||
saxpy(Ax, r, alpham1, N);
|
||||
|
||||
float r1 = dotProduct(r, r, N);
|
||||
|
||||
int k = 1;
|
||||
|
||||
while (r1 > tol * tol && k <= max_iter) {
|
||||
if (k > 1) {
|
||||
b = r1 / r0;
|
||||
scaleVector(p, b, N);
|
||||
|
||||
saxpy(r, p, alpha, N);
|
||||
} else {
|
||||
for (int i = 0; i < N; i++) p[i] = r[i];
|
||||
}
|
||||
|
||||
cpuSpMV(I, J, val, nnz, N, alpha, p, Ax);
|
||||
|
||||
float dot = dotProduct(p, Ax, N);
|
||||
a = r1 / dot;
|
||||
|
||||
saxpy(p, x, a, N);
|
||||
na = -a;
|
||||
saxpy(Ax, r, na, N);
|
||||
|
||||
r0 = r1;
|
||||
r1 = dotProduct(r, r, N);
|
||||
|
||||
printf("\nCPU code iteration = %3d, residual = %e\n", k, sqrt(r1));
|
||||
k++;
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void gpuSpMV(int *I, int *J, float *val, int nnz, int num_rows,
|
||||
float alpha, float *inputVecX, float *outputVecY,
|
||||
cg::thread_block &cta, const cg::grid_group &grid) {
|
||||
for (int i = grid.thread_rank(); i < num_rows; i += grid.size()) {
|
||||
int row_elem = I[i];
|
||||
int next_row_elem = I[i + 1];
|
||||
int num_elems_this_row = next_row_elem - row_elem;
|
||||
|
||||
float output = 0.0;
|
||||
for (int j = 0; j < num_elems_this_row; j++) {
|
||||
// I or J or val arrays - can be put in shared memory
|
||||
// as the access is random and reused in next calls of gpuSpMV function.
|
||||
output += alpha * val[row_elem + j] * inputVecX[J[row_elem + j]];
|
||||
}
|
||||
|
||||
outputVecY[i] = output;
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void gpuSaxpy(float *x, float *y, float a, int size,
|
||||
const cg::grid_group &grid) {
|
||||
for (int i = grid.thread_rank(); i < size; i += grid.size()) {
|
||||
y[i] = a * x[i] + y[i];
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void gpuDotProduct(float *vecA, float *vecB, double *result,
|
||||
int size, const cg::thread_block &cta,
|
||||
const cg::grid_group &grid) {
|
||||
__shared__ double tmp[THREADS_PER_BLOCK];
|
||||
|
||||
double temp_sum = 0.0;
|
||||
for (int i = grid.thread_rank(); i < size; i += grid.size()) {
|
||||
temp_sum += static_cast<double>(vecA[i] * vecB[i]);
|
||||
}
|
||||
tmp[cta.thread_rank()] = temp_sum;
|
||||
|
||||
cg::sync(cta);
|
||||
|
||||
cg::thread_block_tile<32> tile32 = cg::tiled_partition<32>(cta);
|
||||
|
||||
double beta = temp_sum;
|
||||
double temp;
|
||||
|
||||
for (int i = tile32.size() / 2; i > 0; i >>= 1) {
|
||||
if (tile32.thread_rank() < i) {
|
||||
temp = tmp[cta.thread_rank() + i];
|
||||
beta += temp;
|
||||
tmp[cta.thread_rank()] = beta;
|
||||
}
|
||||
cg::sync(tile32);
|
||||
}
|
||||
cg::sync(cta);
|
||||
|
||||
if (cta.thread_rank() == 0) {
|
||||
beta = 0.0;
|
||||
for (int i = 0; i < cta.size(); i += tile32.size()) {
|
||||
beta += tmp[i];
|
||||
}
|
||||
atomicAdd(result, beta);
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void gpuCopyVector(float *srcA, float *destB, int size,
|
||||
const cg::grid_group &grid) {
|
||||
for (int i = grid.thread_rank(); i < size; i += grid.size()) {
|
||||
destB[i] = srcA[i];
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void gpuScaleVector(float *vec, float alpha, int size,
|
||||
const cg::grid_group &grid) {
|
||||
for (int i = grid.thread_rank(); i < size; i += grid.size()) {
|
||||
vec[i] = alpha * vec[i];
|
||||
}
|
||||
}
|
||||
|
||||
extern "C" __global__ void gpuConjugateGradient(int *I, int *J, float *val,
|
||||
float *x, float *Ax, float *p,
|
||||
float *r, double *dot_result,
|
||||
int nnz, int N, float tol) {
|
||||
cg::thread_block cta = cg::this_thread_block();
|
||||
cg::grid_group grid = cg::this_grid();
|
||||
|
||||
int max_iter = 10000;
|
||||
|
||||
float alpha = 1.0;
|
||||
float alpham1 = -1.0;
|
||||
float r0 = 0.0, r1, b, a, na;
|
||||
|
||||
gpuSpMV(I, J, val, nnz, N, alpha, x, Ax, cta, grid);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
gpuSaxpy(Ax, r, alpham1, N, grid);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
gpuDotProduct(r, r, dot_result, N, cta, grid);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
r1 = *dot_result;
|
||||
|
||||
int k = 1;
|
||||
while (r1 > tol * tol && k <= max_iter) {
|
||||
if (k > 1) {
|
||||
b = r1 / r0;
|
||||
gpuScaleVector(p, b, N, grid);
|
||||
|
||||
cg::sync(grid);
|
||||
gpuSaxpy(r, p, alpha, N, grid);
|
||||
} else {
|
||||
gpuCopyVector(r, p, N, grid);
|
||||
}
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
gpuSpMV(I, J, val, nnz, N, alpha, p, Ax, cta, grid);
|
||||
|
||||
if (threadIdx.x == 0 && blockIdx.x == 0) *dot_result = 0.0;
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
gpuDotProduct(p, Ax, dot_result, N, cta, grid);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
a = r1 / *dot_result;
|
||||
|
||||
gpuSaxpy(p, x, a, N, grid);
|
||||
na = -a;
|
||||
gpuSaxpy(Ax, r, na, N, grid);
|
||||
|
||||
r0 = r1;
|
||||
|
||||
if (threadIdx.x == 0 && blockIdx.x == 0) *dot_result = 0.0;
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
gpuDotProduct(r, r, dot_result, N, cta, grid);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
r1 = *dot_result;
|
||||
k++;
|
||||
}
|
||||
}
|
||||
|
||||
bool areAlmostEqual(float a, float b, float maxRelDiff) {
|
||||
float diff = fabsf(a - b);
|
||||
float abs_a = fabsf(a);
|
||||
float abs_b = fabsf(b);
|
||||
float largest = abs_a > abs_b ? abs_a : abs_b;
|
||||
|
||||
if (diff <= largest * maxRelDiff) {
|
||||
return true;
|
||||
} else {
|
||||
printf("maxRelDiff = %.8e\n", maxRelDiff);
|
||||
printf(
|
||||
"diff %.8e > largest * maxRelDiff %.8e therefore %.8e and %.8e are not "
|
||||
"same\n",
|
||||
diff, largest * maxRelDiff, a, b);
|
||||
return false;
|
||||
}
|
||||
}
|
||||
|
||||
int main(int argc, char **argv) {
|
||||
int N = 0, nz = 0, *I = NULL, *J = NULL;
|
||||
float *val = NULL;
|
||||
const float tol = 1e-5f;
|
||||
float *x;
|
||||
float *rhs;
|
||||
float r1;
|
||||
float *r, *p, *Ax;
|
||||
cudaEvent_t start, stop;
|
||||
|
||||
printf("Starting [%s]...\n", sSDKname);
|
||||
|
||||
// This will pick the best possible CUDA capable device
|
||||
cudaDeviceProp deviceProp;
|
||||
int devID = findCudaDevice(argc, (const char **)argv);
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, devID));
|
||||
|
||||
if (!deviceProp.managedMemory) {
|
||||
// This sample requires being run on a device that supports Unified Memory
|
||||
fprintf(stderr, "Unified Memory not supported on this device\n");
|
||||
exit(EXIT_WAIVED);
|
||||
}
|
||||
|
||||
// This sample requires being run on a device that supports Cooperative Kernel
|
||||
// Launch
|
||||
if (!deviceProp.cooperativeLaunch) {
|
||||
printf(
|
||||
"\nSelected GPU (%d) does not support Cooperative Kernel Launch, "
|
||||
"Waiving the run\n",
|
||||
devID);
|
||||
exit(EXIT_WAIVED);
|
||||
}
|
||||
|
||||
// Statistics about the GPU device
|
||||
printf(
|
||||
"> GPU device has %d Multi-Processors, SM %d.%d compute capabilities\n\n",
|
||||
deviceProp.multiProcessorCount, deviceProp.major, deviceProp.minor);
|
||||
|
||||
/* Generate a random tridiagonal symmetric matrix in CSR format */
|
||||
N = 1048576;
|
||||
nz = (N - 2) * 3 + 4;
|
||||
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&I), sizeof(int) * (N + 1));
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&J), sizeof(int) * nz);
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&val), sizeof(float) * nz);
|
||||
|
||||
genTridiag(I, J, val, N, nz);
|
||||
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&x), sizeof(float) * N);
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&rhs), sizeof(float) * N);
|
||||
|
||||
double *dot_result;
|
||||
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&dot_result), sizeof(double));
|
||||
|
||||
*dot_result = 0.0;
|
||||
|
||||
// temp memory for CG
|
||||
checkCudaErrors(
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&r), N * sizeof(float)));
|
||||
checkCudaErrors(
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&p), N * sizeof(float)));
|
||||
checkCudaErrors(
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&Ax), N * sizeof(float)));
|
||||
|
||||
cudaDeviceSynchronize();
|
||||
|
||||
checkCudaErrors(cudaEventCreate(&start));
|
||||
checkCudaErrors(cudaEventCreate(&stop));
|
||||
|
||||
#if ENABLE_CPU_DEBUG_CODE
|
||||
float *Ax_cpu = reinterpret_cast<float *>(malloc(sizeof(float) * N));
|
||||
float *r_cpu = reinterpret_cast<float *>(malloc(sizeof(float) * N));
|
||||
float *p_cpu = reinterpret_cast<float *>(malloc(sizeof(float) * N));
|
||||
float *x_cpu = reinterpret_cast<float *>(malloc(sizeof(float) * N));
|
||||
|
||||
for (int i = 0; i < N; i++) {
|
||||
r_cpu[i] = 1.0;
|
||||
Ax_cpu[i] = x_cpu[i] = 0.0;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
for (int i = 0; i < N; i++) {
|
||||
r[i] = rhs[i] = 1.0;
|
||||
x[i] = 0.0;
|
||||
}
|
||||
|
||||
void *kernelArgs[] = {
|
||||
(void *)&I, (void *)&J, (void *)&val, (void *)&x,
|
||||
(void *)&Ax, (void *)&p, (void *)&r, (void *)&dot_result,
|
||||
(void *)&nz, (void *)&N, (void *)&tol,
|
||||
};
|
||||
|
||||
int sMemSize = sizeof(double) * THREADS_PER_BLOCK;
|
||||
int numBlocksPerSm = 0;
|
||||
int numThreads = THREADS_PER_BLOCK;
|
||||
|
||||
checkCudaErrors(cudaOccupancyMaxActiveBlocksPerMultiprocessor(
|
||||
&numBlocksPerSm, gpuConjugateGradient, numThreads, sMemSize));
|
||||
|
||||
int numSms = deviceProp.multiProcessorCount;
|
||||
dim3 dimGrid(numSms * numBlocksPerSm, 1, 1),
|
||||
dimBlock(THREADS_PER_BLOCK, 1, 1);
|
||||
checkCudaErrors(cudaEventRecord(start, 0));
|
||||
checkCudaErrors(cudaLaunchCooperativeKernel((void *)gpuConjugateGradient,
|
||||
dimGrid, dimBlock, kernelArgs,
|
||||
sMemSize, NULL));
|
||||
checkCudaErrors(cudaEventRecord(stop, 0));
|
||||
checkCudaErrors(cudaDeviceSynchronize());
|
||||
|
||||
float time;
|
||||
checkCudaErrors(cudaEventElapsedTime(&time, start, stop));
|
||||
|
||||
r1 = *dot_result;
|
||||
|
||||
printf("GPU Final, residual = %e, kernel execution time = %f ms\n", sqrt(r1),
|
||||
time);
|
||||
|
||||
#if ENABLE_CPU_DEBUG_CODE
|
||||
cpuConjugateGrad(I, J, val, x_cpu, Ax_cpu, p_cpu, r_cpu, nz, N, tol);
|
||||
#endif
|
||||
|
||||
float rsum, diff, err = 0.0;
|
||||
|
||||
for (int i = 0; i < N; i++) {
|
||||
rsum = 0.0;
|
||||
|
||||
for (int j = I[i]; j < I[i + 1]; j++) {
|
||||
rsum += val[j] * x[J[j]];
|
||||
}
|
||||
|
||||
diff = fabs(rsum - rhs[i]);
|
||||
|
||||
if (diff > err) {
|
||||
err = diff;
|
||||
}
|
||||
}
|
||||
|
||||
checkCudaErrors(cudaFree(I));
|
||||
checkCudaErrors(cudaFree(J));
|
||||
checkCudaErrors(cudaFree(val));
|
||||
checkCudaErrors(cudaFree(x));
|
||||
checkCudaErrors(cudaFree(rhs));
|
||||
checkCudaErrors(cudaFree(r));
|
||||
checkCudaErrors(cudaFree(p));
|
||||
checkCudaErrors(cudaFree(Ax));
|
||||
checkCudaErrors(cudaFree(dot_result));
|
||||
checkCudaErrors(cudaEventDestroy(start));
|
||||
checkCudaErrors(cudaEventDestroy(stop));
|
||||
|
||||
#if ENABLE_CPU_DEBUG_CODE
|
||||
free(Ax_cpu);
|
||||
free(r_cpu);
|
||||
free(p_cpu);
|
||||
free(x_cpu);
|
||||
#endif
|
||||
|
||||
printf("Test Summary: Error amount = %f \n", err);
|
||||
fprintf(stdout, "&&&& conjugateGradientMultiBlockCG %s\n",
|
||||
(sqrt(r1) < tol) ? "PASSED" : "FAILED");
|
||||
exit((sqrt(r1) < tol) ? EXIT_SUCCESS : EXIT_FAILURE);
|
||||
}
|
||||
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 11.00
|
||||
# Visual Studio 2010
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientMultiBlockCG", "conjugateGradientMultiBlockCG_vs2010.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
|
|
@ -0,0 +1,107 @@
|
|||
<?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>conjugateGradientMultiBlockCG_vs2010</RootNamespace>
|
||||
<ProjectName>conjugateGradientMultiBlockCG</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
</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 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientMultiBlockCG.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="conjugateGradientMultiBlockCG.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2012
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientMultiBlockCG", "conjugateGradientMultiBlockCG_vs2012.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
|
|
@ -0,0 +1,108 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>conjugateGradientMultiBlockCG_vs2012</RootNamespace>
|
||||
<ProjectName>conjugateGradientMultiBlockCG</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v110</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 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientMultiBlockCG.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="conjugateGradientMultiBlockCG.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 13.00
|
||||
# Visual Studio 2013
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientMultiBlockCG", "conjugateGradientMultiBlockCG_vs2013.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
|
|
@ -0,0 +1,108 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>conjugateGradientMultiBlockCG_vs2013</RootNamespace>
|
||||
<ProjectName>conjugateGradientMultiBlockCG</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v120</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 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientMultiBlockCG.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="conjugateGradientMultiBlockCG.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 14.00
|
||||
# Visual Studio 2015
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientMultiBlockCG", "conjugateGradientMultiBlockCG_vs2015.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
|
|
@ -0,0 +1,108 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>conjugateGradientMultiBlockCG_vs2015</RootNamespace>
|
||||
<ProjectName>conjugateGradientMultiBlockCG</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v140</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 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientMultiBlockCG.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="conjugateGradientMultiBlockCG.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2017
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientMultiBlockCG", "conjugateGradientMultiBlockCG_vs2017.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
|
|
@ -0,0 +1,109 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>conjugateGradientMultiBlockCG_vs2017</RootNamespace>
|
||||
<ProjectName>conjugateGradientMultiBlockCG</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v141</PlatformToolset>
|
||||
<WindowsTargetPlatformVersion>10.0.15063.0</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 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientMultiBlockCG.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="conjugateGradientMultiBlockCG.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
324
Samples/conjugateGradientMultiDeviceCG/Makefile
Normal file
324
Samples/conjugateGradientMultiDeviceCG/Makefile
Normal file
|
|
@ -0,0 +1,324 @@
|
|||
################################################################################
|
||||
#
|
||||
# Copyright 1993-2015 NVIDIA Corporation. All rights reserved.
|
||||
#
|
||||
# NOTICE TO USER:
|
||||
#
|
||||
# This source code is subject to NVIDIA ownership rights under U.S. and
|
||||
# international Copyright laws.
|
||||
#
|
||||
# NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOURCE
|
||||
# CODE FOR ANY PURPOSE. IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR
|
||||
# IMPLIED WARRANTY OF ANY KIND. NVIDIA DISCLAIMS ALL WARRANTIES WITH
|
||||
# REGARD TO THIS SOURCE CODE, INCLUDING ALL IMPLIED WARRANTIES OF
|
||||
# MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
|
||||
# IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL,
|
||||
# OR CONSEQUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS
|
||||
# OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
|
||||
# OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE
|
||||
# OR PERFORMANCE OF THIS SOURCE CODE.
|
||||
#
|
||||
# U.S. Government End Users. This source code is a "commercial item" as
|
||||
# that term is defined at 48 C.F.R. 2.101 (OCT 1995), consisting of
|
||||
# "commercial computer software" and "commercial computer software
|
||||
# documentation" as such terms are used in 48 C.F.R. 12.212 (SEPT 1995)
|
||||
# and is provided to the U.S. Government only as a commercial end item.
|
||||
# Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through
|
||||
# 227.7202-4 (JUNE 1995), all U.S. Government End Users acquire the
|
||||
# source code with only those rights set forth herein.
|
||||
#
|
||||
################################################################################
|
||||
#
|
||||
# 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-g++
|
||||
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)
|
||||
$(info >>> WARNING - conjugateGradientMultiDeviceCG is not supported on Mac OSX - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
# This sample is not supported on ARMv7
|
||||
ifeq ($(TARGET_ARCH),armv7l)
|
||||
$(info >>> WARNING - conjugateGradientMultiDeviceCG is not supported on ARMv7 - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
# This sample is not supported on aarch64
|
||||
ifeq ($(TARGET_ARCH),aarch64)
|
||||
$(info >>> WARNING - conjugateGradientMultiDeviceCG 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
|
||||
|
||||
ifeq ($(SMS),)
|
||||
$(info >>> WARNING - no SM architectures have been specified - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
ifeq ($(GENCODE_FLAGS),)
|
||||
# Generate SASS code for each SM architecture listed in $(SMS)
|
||||
$(foreach sm,$(SMS),$(eval GENCODE_FLAGS += -gencode arch=compute_$(sm),code=sm_$(sm)))
|
||||
|
||||
# Generate PTX code from the highest SM architecture in $(SMS) to guarantee forward-compatibility
|
||||
HIGHEST_SM := $(lastword $(sort $(SMS)))
|
||||
ifneq ($(HIGHEST_SM),)
|
||||
GENCODE_FLAGS += -gencode arch=compute_$(HIGHEST_SM),code=compute_$(HIGHEST_SM)
|
||||
endif
|
||||
endif
|
||||
|
||||
ALL_CCFLAGS += -dc
|
||||
|
||||
LIBRARIES += -lcudadevrt
|
||||
|
||||
ifeq ($(SAMPLE_ENABLED),0)
|
||||
EXEC ?= @echo "[@]"
|
||||
endif
|
||||
|
||||
################################################################################
|
||||
|
||||
# Target rules
|
||||
all: build
|
||||
|
||||
build: conjugateGradientMultiDeviceCG
|
||||
|
||||
check.deps:
|
||||
ifeq ($(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
|
||||
|
||||
conjugateGradientMultiDeviceCG.o:conjugateGradientMultiDeviceCG.cu
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
conjugateGradientMultiDeviceCG: conjugateGradientMultiDeviceCG.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
|
||||
$(EXEC) ./conjugateGradientMultiDeviceCG
|
||||
|
||||
clean:
|
||||
rm -f conjugateGradientMultiDeviceCG conjugateGradientMultiDeviceCG.o
|
||||
rm -rf ../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)/conjugateGradientMultiDeviceCG
|
||||
|
||||
clobber: clean
|
||||
70
Samples/conjugateGradientMultiDeviceCG/NsightEclipse.xml
Normal file
70
Samples/conjugateGradientMultiDeviceCG/NsightEclipse.xml
Normal file
|
|
@ -0,0 +1,70 @@
|
|||
<?xml version="1.0" encoding="UTF-8"?>
|
||||
<!DOCTYPE entry SYSTEM "SamplesInfo.dtd">
|
||||
<entry>
|
||||
<name>conjugateGradientMultiDeviceCG</name>
|
||||
<cflags>
|
||||
<flag>-dc</flag>
|
||||
</cflags>
|
||||
<cuda_api_list>
|
||||
<toolkit>cudaMemAdvise</toolkit>
|
||||
<toolkit>cudaMemPrefetchAsync</toolkit>
|
||||
<toolkit>cudaLaunchCooperativeKernelMultiDevice</toolkit>
|
||||
<toolkit>cudaStreamSynchronize</toolkit>
|
||||
<toolkit>cudaOccupancyMaxActiveBlocksPerMultiprocessor</toolkit>
|
||||
</cuda_api_list>
|
||||
<description><![CDATA[This sample implements a conjugate gradient solver on multiple GPUs using Multi Device Cooperative Groups, also uses Unified Memory optimized using prefetching and usage hints.]]></description>
|
||||
<devicecompilation>whole</devicecompilation>
|
||||
<includepaths>
|
||||
<path>./</path>
|
||||
<path>../</path>
|
||||
<path>../../common/inc</path>
|
||||
</includepaths>
|
||||
<keyconcepts>
|
||||
<concept level="basic">Unified Memory</concept>
|
||||
<concept level="advanced">Linear Algebra</concept>
|
||||
<concept level="basic">Cooperative Groups</concept>
|
||||
<concept level="advanced">MultiDevice Cooperative Groups</concept>
|
||||
</keyconcepts>
|
||||
<keywords>
|
||||
<keyword>CUDA</keyword>
|
||||
<keyword>Sparse Matrix</keyword>
|
||||
<keyword>Unified Memory</keyword>
|
||||
<keyword>Multi-GPU</keyword>
|
||||
</keywords>
|
||||
<libraries>
|
||||
<library>cudadevrt</library>
|
||||
</libraries>
|
||||
<librarypaths>
|
||||
</librarypaths>
|
||||
<nsight_eclipse>true</nsight_eclipse>
|
||||
<primary_file>conjugateGradientMultiBlockCG.cu</primary_file>
|
||||
<required_dependencies>
|
||||
<dependency>UVM</dependency>
|
||||
<dependency>MDCG</dependency>
|
||||
</required_dependencies>
|
||||
<scopes>
|
||||
<scope>1:CUDA Advanced Topics</scope>
|
||||
<scope>3:Linear Algebra</scope>
|
||||
</scopes>
|
||||
<sm-arch>sm60</sm-arch>
|
||||
<sm-arch>sm61</sm-arch>
|
||||
<sm-arch>sm70</sm-arch>
|
||||
<supported_envs>
|
||||
<env>
|
||||
<arch>x86_64</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
<env>
|
||||
<arch>ppc64le</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
<env>
|
||||
<platform>windows</platform>
|
||||
</env>
|
||||
</supported_envs>
|
||||
<supported_sm_architectures>
|
||||
<from>6.0</from>
|
||||
</supported_sm_architectures>
|
||||
<title>conjugateGradient using MultiDevice Cooperative Groups</title>
|
||||
<type>exe</type>
|
||||
</entry>
|
||||
74
Samples/conjugateGradientMultiDeviceCG/README.md
Normal file
74
Samples/conjugateGradientMultiDeviceCG/README.md
Normal file
|
|
@ -0,0 +1,74 @@
|
|||
# conjugateGradientMultiDeviceCG - conjugateGradient using MultiDevice Cooperative Groups
|
||||
|
||||
## Description
|
||||
|
||||
This sample implements a conjugate gradient solver on multiple GPUs using Multi Device Cooperative Groups, also uses Unified Memory optimized using prefetching and usage hints.
|
||||
|
||||
## Key Concepts
|
||||
|
||||
Unified Memory, Linear Algebra, Cooperative Groups, MultiDevice Cooperative Groups
|
||||
|
||||
## 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)
|
||||
|
||||
## Supported OSes
|
||||
|
||||
Linux, Windows
|
||||
|
||||
## Supported CPU Architecture
|
||||
|
||||
x86_64, ppc64le
|
||||
|
||||
## CUDA APIs involved
|
||||
|
||||
### [CUDA Runtime API](http://docs.nvidia.com/cuda/cuda-runtime-api/index.html)
|
||||
cudaMemAdvise, cudaMemPrefetchAsync, cudaLaunchCooperativeKernelMultiDevice, cudaStreamSynchronize, cudaOccupancyMaxActiveBlocksPerMultiprocessor
|
||||
|
||||
## Dependencies needed to build/run
|
||||
[UVM](../../README.md#uvm), [MDCG](../../README.md#mdcg)
|
||||
|
||||
## Prerequisites
|
||||
|
||||
Download and install the [CUDA Toolkit 9.2](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)
|
||||
|
||||
|
|
@ -0,0 +1,677 @@
|
|||
/* 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.
|
||||
*/
|
||||
|
||||
/*
|
||||
* This sample implements a conjugate gradient solver on multiple GPU using
|
||||
* Multi Device Cooperative Groups, also uses Unified Memory optimized using
|
||||
* prefetching and usage hints.
|
||||
*
|
||||
*/
|
||||
|
||||
// includes, system
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
#include <iostream>
|
||||
#include <set>
|
||||
|
||||
#include <cuda_runtime.h>
|
||||
|
||||
// Utilities and system includes
|
||||
#include <helper_cuda.h> // helper function CUDA error checking and initialization
|
||||
#include <helper_functions.h> // helper for shared functions common to CUDA Samples
|
||||
|
||||
#include <cooperative_groups.h>
|
||||
|
||||
namespace cg = cooperative_groups;
|
||||
|
||||
const char *sSDKname = "conjugateGradientMultiDeviceCG";
|
||||
|
||||
#define ENABLE_CPU_DEBUG_CODE 0
|
||||
#define THREADS_PER_BLOCK 512
|
||||
|
||||
__device__ double grid_dot_result = 0.0;
|
||||
|
||||
/* genTridiag: generate a random tridiagonal symmetric matrix */
|
||||
void genTridiag(int *I, int *J, float *val, int N, int nz) {
|
||||
I[0] = 0, J[0] = 0, J[1] = 1;
|
||||
val[0] = static_cast<float>(rand()) / RAND_MAX + 10.0f;
|
||||
val[1] = static_cast<float>(rand()) / RAND_MAX;
|
||||
int start;
|
||||
|
||||
for (int i = 1; i < N; i++) {
|
||||
if (i > 1) {
|
||||
I[i] = I[i - 1] + 3;
|
||||
} else {
|
||||
I[1] = 2;
|
||||
}
|
||||
|
||||
start = (i - 1) * 3 + 2;
|
||||
J[start] = i - 1;
|
||||
J[start + 1] = i;
|
||||
|
||||
if (i < N - 1) {
|
||||
J[start + 2] = i + 1;
|
||||
}
|
||||
|
||||
val[start] = val[start - 1];
|
||||
val[start + 1] = static_cast<float>(rand()) / RAND_MAX + 10.0f;
|
||||
|
||||
if (i < N - 1) {
|
||||
val[start + 2] = static_cast<float>(rand()) / RAND_MAX;
|
||||
}
|
||||
}
|
||||
|
||||
I[N] = nz;
|
||||
}
|
||||
|
||||
// I - contains location of the given non-zero element in the row of the matrix
|
||||
// J - contains location of the given non-zero element in the column of the
|
||||
// matrix val - contains values of the given non-zero elements of the matrix
|
||||
// inputVecX - input vector to be multiplied
|
||||
// outputVecY - resultant vector
|
||||
void cpuSpMV(int *I, int *J, float *val, int nnz, int num_rows, float alpha,
|
||||
float *inputVecX, float *outputVecY) {
|
||||
for (int i = 0; i < num_rows; i++) {
|
||||
int num_elems_this_row = I[i + 1] - I[i];
|
||||
|
||||
float output = 0.0;
|
||||
for (int j = 0; j < num_elems_this_row; j++) {
|
||||
output += alpha * val[I[i] + j] * inputVecX[J[I[i] + j]];
|
||||
}
|
||||
outputVecY[i] = output;
|
||||
}
|
||||
|
||||
return;
|
||||
}
|
||||
|
||||
double dotProduct(float *vecA, float *vecB, int size) {
|
||||
double result = 0.0;
|
||||
|
||||
for (int i = 0; i < size; i++) {
|
||||
result = result + (vecA[i] * vecB[i]);
|
||||
}
|
||||
|
||||
return result;
|
||||
}
|
||||
|
||||
void scaleVector(float *vec, float alpha, int size) {
|
||||
for (int i = 0; i < size; i++) {
|
||||
vec[i] = alpha * vec[i];
|
||||
}
|
||||
}
|
||||
|
||||
void saxpy(float *x, float *y, float a, int size) {
|
||||
for (int i = 0; i < size; i++) {
|
||||
y[i] = a * x[i] + y[i];
|
||||
}
|
||||
}
|
||||
|
||||
void cpuConjugateGrad(int *I, int *J, float *val, float *x, float *Ax, float *p,
|
||||
float *r, int nnz, int N, float tol) {
|
||||
int max_iter = 10000;
|
||||
|
||||
float alpha = 1.0;
|
||||
float alpham1 = -1.0;
|
||||
float r0 = 0.0, b, a, na;
|
||||
|
||||
cpuSpMV(I, J, val, nnz, N, alpha, x, Ax);
|
||||
saxpy(Ax, r, alpham1, N);
|
||||
|
||||
float r1 = dotProduct(r, r, N);
|
||||
|
||||
int k = 1;
|
||||
|
||||
while (r1 > tol * tol && k <= max_iter) {
|
||||
if (k > 1) {
|
||||
b = r1 / r0;
|
||||
scaleVector(p, b, N);
|
||||
|
||||
saxpy(r, p, alpha, N);
|
||||
} else {
|
||||
for (int i = 0; i < N; i++) p[i] = r[i];
|
||||
}
|
||||
|
||||
cpuSpMV(I, J, val, nnz, N, alpha, p, Ax);
|
||||
|
||||
float dot = dotProduct(p, Ax, N);
|
||||
a = r1 / dot;
|
||||
|
||||
saxpy(p, x, a, N);
|
||||
na = -a;
|
||||
saxpy(Ax, r, na, N);
|
||||
|
||||
r0 = r1;
|
||||
r1 = dotProduct(r, r, N);
|
||||
|
||||
printf("\nCPU code iteration = %3d, residual = %e\n", k, sqrt(r1));
|
||||
k++;
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void gpuSpMV(int *I, int *J, float *val, int nnz, int num_rows,
|
||||
float alpha, float *inputVecX, float *outputVecY,
|
||||
cg::thread_block &cta,
|
||||
const cg::multi_grid_group &multi_grid) {
|
||||
for (int i = multi_grid.thread_rank(); i < num_rows; i += multi_grid.size()) {
|
||||
int row_elem = I[i];
|
||||
int next_row_elem = I[i + 1];
|
||||
int num_elems_this_row = next_row_elem - row_elem;
|
||||
|
||||
float output = 0.0;
|
||||
for (int j = 0; j < num_elems_this_row; j++) {
|
||||
output += alpha * val[row_elem + j] * inputVecX[J[row_elem + j]];
|
||||
}
|
||||
|
||||
outputVecY[i] = output;
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void gpuSaxpy(float *x, float *y, float a, int size,
|
||||
const cg::multi_grid_group &multi_grid) {
|
||||
for (int i = multi_grid.thread_rank(); i < size; i += multi_grid.size()) {
|
||||
y[i] = a * x[i] + y[i];
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void gpuDotProduct(float *vecA, float *vecB, int size,
|
||||
const cg::thread_block &cta,
|
||||
const cg::multi_grid_group &multi_grid) {
|
||||
__shared__ double tmp[THREADS_PER_BLOCK];
|
||||
|
||||
double temp_sum = 0.0;
|
||||
for (int i = multi_grid.thread_rank(); i < size; i += multi_grid.size()) {
|
||||
temp_sum += static_cast<double>(vecA[i] * vecB[i]);
|
||||
}
|
||||
tmp[cta.thread_rank()] = temp_sum;
|
||||
|
||||
cg::sync(cta);
|
||||
|
||||
cg::thread_block_tile<32> tile32 = cg::tiled_partition<32>(cta);
|
||||
|
||||
double beta = temp_sum;
|
||||
double temp;
|
||||
|
||||
for (int i = tile32.size() / 2; i > 0; i >>= 1) {
|
||||
if (tile32.thread_rank() < i) {
|
||||
temp = tmp[cta.thread_rank() + i];
|
||||
beta += temp;
|
||||
tmp[cta.thread_rank()] = beta;
|
||||
}
|
||||
cg::sync(tile32);
|
||||
}
|
||||
cg::sync(cta);
|
||||
|
||||
if (cta.thread_rank() == 0) {
|
||||
beta = 0.0;
|
||||
for (int i = 0; i < cta.size(); i += tile32.size()) {
|
||||
beta += tmp[i];
|
||||
}
|
||||
atomicAdd(&grid_dot_result, beta);
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void gpuCopyVector(float *srcA, float *destB, int size,
|
||||
const cg::multi_grid_group &multi_grid) {
|
||||
for (int i = multi_grid.thread_rank(); i < size; i += multi_grid.size()) {
|
||||
destB[i] = srcA[i];
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void gpuScaleVector(float *vec, float alpha, int size,
|
||||
const cg::multi_grid_group &multi_grid) {
|
||||
for (int i = multi_grid.thread_rank(); i < size; i += multi_grid.size()) {
|
||||
vec[i] = alpha * vec[i];
|
||||
}
|
||||
}
|
||||
|
||||
__device__ void setDotResultToZero(double *dot_result) {
|
||||
unsigned long long int *address_as_ull = (unsigned long long int *)dot_result;
|
||||
unsigned long long int old = *address_as_ull, assumed;
|
||||
|
||||
do {
|
||||
assumed = old;
|
||||
old = atomicCAS_system(address_as_ull, assumed, 0);
|
||||
|
||||
} while (assumed != old);
|
||||
}
|
||||
|
||||
extern "C" __global__ void multiGpuConjugateGradient(
|
||||
int *I, int *J, float *val, float *x, float *Ax, float *p, float *r,
|
||||
double *dot_result, int nnz, int N, float tol) {
|
||||
cg::thread_block cta = cg::this_thread_block();
|
||||
cg::grid_group grid = cg::this_grid();
|
||||
cg::multi_grid_group multi_grid = cg::this_multi_grid();
|
||||
|
||||
const int max_iter = 10000;
|
||||
|
||||
float alpha = 1.0;
|
||||
float alpham1 = -1.0;
|
||||
float r0 = 0.0, r1, b, a, na;
|
||||
|
||||
for (int i = multi_grid.thread_rank(); i < N; i += multi_grid.size()) {
|
||||
r[i] = 1.0;
|
||||
x[i] = 0.0;
|
||||
}
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
gpuSpMV(I, J, val, nnz, N, alpha, x, Ax, cta, multi_grid);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
gpuSaxpy(Ax, r, alpham1, N, multi_grid);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
gpuDotProduct(r, r, N, cta, multi_grid);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
if (grid.thread_rank() == 0) {
|
||||
atomicAdd_system(dot_result, grid_dot_result);
|
||||
grid_dot_result = 0.0;
|
||||
}
|
||||
cg::sync(multi_grid);
|
||||
|
||||
r1 = *dot_result;
|
||||
|
||||
int k = 1;
|
||||
while (r1 > tol * tol && k <= max_iter) {
|
||||
if (k > 1) {
|
||||
b = r1 / r0;
|
||||
|
||||
gpuScaleVector(p, b, N, multi_grid);
|
||||
cg::sync(grid);
|
||||
gpuSaxpy(r, p, alpha, N, multi_grid);
|
||||
} else {
|
||||
gpuCopyVector(r, p, N, multi_grid);
|
||||
}
|
||||
|
||||
cg::sync(multi_grid);
|
||||
|
||||
gpuSpMV(I, J, val, nnz, N, alpha, p, Ax, cta, multi_grid);
|
||||
|
||||
if (multi_grid.thread_rank() == 0) {
|
||||
setDotResultToZero(dot_result);
|
||||
}
|
||||
cg::sync(multi_grid);
|
||||
|
||||
gpuDotProduct(p, Ax, N, cta, multi_grid);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
if (grid.thread_rank() == 0) {
|
||||
atomicAdd_system(dot_result, grid_dot_result);
|
||||
grid_dot_result = 0.0;
|
||||
}
|
||||
cg::sync(multi_grid);
|
||||
|
||||
a = r1 / *dot_result;
|
||||
|
||||
gpuSaxpy(p, x, a, N, multi_grid);
|
||||
|
||||
na = -a;
|
||||
|
||||
gpuSaxpy(Ax, r, na, N, multi_grid);
|
||||
|
||||
r0 = r1;
|
||||
|
||||
cg::sync(multi_grid);
|
||||
if (multi_grid.thread_rank() == 0) {
|
||||
setDotResultToZero(dot_result);
|
||||
}
|
||||
|
||||
cg::sync(multi_grid);
|
||||
|
||||
gpuDotProduct(r, r, N, cta, multi_grid);
|
||||
|
||||
cg::sync(grid);
|
||||
|
||||
if (grid.thread_rank() == 0) {
|
||||
atomicAdd_system(dot_result, grid_dot_result);
|
||||
grid_dot_result = 0.0;
|
||||
}
|
||||
cg::sync(multi_grid);
|
||||
|
||||
r1 = *dot_result;
|
||||
k++;
|
||||
}
|
||||
}
|
||||
|
||||
void getIdenticalGPUs(int num_of_gpus, std::set<int> &identicalGPUs) {
|
||||
int *major_minor =
|
||||
reinterpret_cast<int *>(malloc(sizeof(int) * num_of_gpus * 2));
|
||||
int foundIdenticalGPUs = 0;
|
||||
|
||||
for (int i = 0; i < num_of_gpus; i++) {
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, i));
|
||||
major_minor[i * 2] = deviceProp.major;
|
||||
major_minor[i * 2 + 1] = deviceProp.minor;
|
||||
printf("GPU Device %d: \"%s\" with compute capability %d.%d\n", i,
|
||||
deviceProp.name, deviceProp.major, deviceProp.minor);
|
||||
}
|
||||
|
||||
int maxMajorMinor[2] = {0, 0};
|
||||
|
||||
for (int i = 0; i < num_of_gpus; i++) {
|
||||
for (int j = i + 1; j < num_of_gpus; j++) {
|
||||
if ((major_minor[i * 2] == major_minor[j * 2]) &&
|
||||
(major_minor[i * 2 + 1] == major_minor[j * 2 + 1])) {
|
||||
identicalGPUs.insert(i);
|
||||
identicalGPUs.insert(j);
|
||||
foundIdenticalGPUs = 1;
|
||||
if (maxMajorMinor[0] < major_minor[i * 2] &&
|
||||
maxMajorMinor[1] < major_minor[i * 2 + 1]) {
|
||||
maxMajorMinor[0] = major_minor[i * 2];
|
||||
maxMajorMinor[1] = major_minor[i * 2 + 1];
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
free(major_minor);
|
||||
if (!foundIdenticalGPUs) {
|
||||
printf(
|
||||
"No Two or more GPUs with same architecture found\nWaiving the "
|
||||
"sample\n");
|
||||
exit(EXIT_WAIVED);
|
||||
}
|
||||
|
||||
std::set<int>::iterator it = identicalGPUs.begin();
|
||||
|
||||
// Iterate over all the identical GPUs found
|
||||
while (it != identicalGPUs.end()) {
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, *it));
|
||||
// Remove all the GPUs which are less than the best arch available
|
||||
if (deviceProp.major != maxMajorMinor[0] &&
|
||||
deviceProp.minor != maxMajorMinor[1]) {
|
||||
identicalGPUs.erase(it);
|
||||
}
|
||||
if (!deviceProp.cooperativeMultiDeviceLaunch) {
|
||||
identicalGPUs.erase(it);
|
||||
}
|
||||
it++;
|
||||
}
|
||||
|
||||
return;
|
||||
}
|
||||
|
||||
int main(int argc, char **argv) {
|
||||
int N = 0, nz = 0, *I = NULL, *J = NULL;
|
||||
float *val = NULL;
|
||||
const float tol = 1e-5f;
|
||||
float *x;
|
||||
float rhs = 1.0;
|
||||
float r1;
|
||||
float *r, *p, *Ax;
|
||||
|
||||
printf("Starting [%s]...\n", sSDKname);
|
||||
|
||||
int num_of_gpus = 0;
|
||||
checkCudaErrors(cudaGetDeviceCount(&num_of_gpus));
|
||||
|
||||
if (num_of_gpus <= 1) {
|
||||
printf("No. of GPU on node %d\n", num_of_gpus);
|
||||
printf("Minimum Two or more GPUs are required to run this sample code\n");
|
||||
exit(EXIT_WAIVED);
|
||||
}
|
||||
|
||||
std::set<int> identicalGPUs;
|
||||
getIdenticalGPUs(num_of_gpus, identicalGPUs);
|
||||
|
||||
if (identicalGPUs.size() <= 1) {
|
||||
printf(
|
||||
"No Two or more GPUs with same architecture capable of "
|
||||
"cooperativeMultiDeviceLaunch found. \nWaiving the sample\n");
|
||||
exit(EXIT_WAIVED);
|
||||
}
|
||||
|
||||
std::set<int>::iterator deviceId = identicalGPUs.begin();
|
||||
|
||||
// We use only 2 GPUs as for input size of N = 10485760*2 two GPUs are enough.
|
||||
while (identicalGPUs.size() > 2) {
|
||||
identicalGPUs.erase(deviceId);
|
||||
deviceId++;
|
||||
}
|
||||
/* Generate a random tridiagonal symmetric matrix in CSR format */
|
||||
N = 10485760 * 2;
|
||||
nz = (N - 2) * 3 + 4;
|
||||
|
||||
checkCudaErrors(
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&I), sizeof(int) * (N + 1)));
|
||||
checkCudaErrors(
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&J), sizeof(int) * nz));
|
||||
checkCudaErrors(
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&val), sizeof(float) * nz));
|
||||
|
||||
float *val_cpu = reinterpret_cast<float *>(malloc(sizeof(float) * nz));
|
||||
|
||||
genTridiag(I, J, val_cpu, N, nz);
|
||||
|
||||
memcpy(val, val_cpu, sizeof(float) * nz);
|
||||
checkCudaErrors(
|
||||
cudaMemAdvise(I, sizeof(int) * (N + 1), cudaMemAdviseSetReadMostly, 0));
|
||||
checkCudaErrors(
|
||||
cudaMemAdvise(J, sizeof(int) * nz, cudaMemAdviseSetReadMostly, 0));
|
||||
checkCudaErrors(
|
||||
cudaMemAdvise(val, sizeof(float) * nz, cudaMemAdviseSetReadMostly, 0));
|
||||
|
||||
checkCudaErrors(
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&x), sizeof(float) * N));
|
||||
|
||||
double *dot_result;
|
||||
checkCudaErrors(cudaMallocManaged(reinterpret_cast<void **>(&dot_result),
|
||||
sizeof(double)));
|
||||
|
||||
checkCudaErrors(cudaMemset(dot_result, 0.0, sizeof(double)));
|
||||
|
||||
// temp memory for ConjugateGradient
|
||||
checkCudaErrors(
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&r), N * sizeof(float)));
|
||||
checkCudaErrors(
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&p), N * sizeof(float)));
|
||||
checkCudaErrors(
|
||||
cudaMallocManaged(reinterpret_cast<void **>(&Ax), N * sizeof(float)));
|
||||
|
||||
std::cout << "\nRunning on GPUs = " << identicalGPUs.size() << std::endl;
|
||||
cudaStream_t *nStreams = reinterpret_cast<cudaStream_t *>(
|
||||
malloc(sizeof(cudaStream_t) * identicalGPUs.size()));
|
||||
|
||||
void *kernelArgs[] = {
|
||||
(void *)&I, (void *)&J, (void *)&val, (void *)&x,
|
||||
(void *)&Ax, (void *)&p, (void *)&r, (void *)&dot_result,
|
||||
(void *)&nz, (void *)&N, (void *)&tol,
|
||||
};
|
||||
|
||||
int sMemSize = sizeof(double) * THREADS_PER_BLOCK;
|
||||
int numBlocksPerSm = 0;
|
||||
int numThreads = THREADS_PER_BLOCK;
|
||||
|
||||
deviceId = identicalGPUs.begin();
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(cudaSetDevice(*deviceId));
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, *deviceId));
|
||||
|
||||
checkCudaErrors(cudaOccupancyMaxActiveBlocksPerMultiprocessor(
|
||||
&numBlocksPerSm, multiGpuConjugateGradient, numThreads, sMemSize));
|
||||
|
||||
int numSms = deviceProp.multiProcessorCount;
|
||||
dim3 dimGrid(numSms * numBlocksPerSm, 1, 1),
|
||||
dimBlock(THREADS_PER_BLOCK, 1, 1);
|
||||
|
||||
int device_count = 0;
|
||||
|
||||
int totalThreadsPerGPU = numSms * numBlocksPerSm * THREADS_PER_BLOCK;
|
||||
|
||||
while (deviceId != identicalGPUs.end()) {
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(cudaSetDevice(*deviceId));
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, *deviceId));
|
||||
checkCudaErrors(cudaStreamCreate(&nStreams[device_count]));
|
||||
|
||||
if (deviceProp.concurrentManagedAccess) {
|
||||
int perGPUIter = N / (totalThreadsPerGPU * identicalGPUs.size());
|
||||
int offset_Ax = device_count * totalThreadsPerGPU;
|
||||
int offset_r = device_count * totalThreadsPerGPU;
|
||||
int offset_p = device_count * totalThreadsPerGPU;
|
||||
int offset_x = device_count * totalThreadsPerGPU;
|
||||
|
||||
checkCudaErrors(cudaMemPrefetchAsync(I, sizeof(int) * N, *deviceId,
|
||||
nStreams[device_count]));
|
||||
checkCudaErrors(cudaMemPrefetchAsync(val, sizeof(float) * nz, *deviceId,
|
||||
nStreams[device_count]));
|
||||
checkCudaErrors(cudaMemPrefetchAsync(J, sizeof(float) * nz, *deviceId,
|
||||
nStreams[device_count]));
|
||||
|
||||
if (offset_Ax <= N) {
|
||||
for (int i = 0; i < perGPUIter; i++) {
|
||||
cudaMemAdvise(Ax + offset_Ax, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetPreferredLocation, *deviceId);
|
||||
cudaMemAdvise(r + offset_r, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetPreferredLocation, *deviceId);
|
||||
cudaMemAdvise(x + offset_x, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetPreferredLocation, *deviceId);
|
||||
cudaMemAdvise(p + offset_p, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetPreferredLocation, *deviceId);
|
||||
|
||||
cudaMemAdvise(Ax + offset_Ax, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetAccessedBy, *deviceId);
|
||||
cudaMemAdvise(r + offset_r, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetAccessedBy, *deviceId);
|
||||
cudaMemAdvise(p + offset_p, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetAccessedBy, *deviceId);
|
||||
cudaMemAdvise(x + offset_x, sizeof(float) * totalThreadsPerGPU,
|
||||
cudaMemAdviseSetAccessedBy, *deviceId);
|
||||
|
||||
offset_Ax += totalThreadsPerGPU * identicalGPUs.size();
|
||||
offset_r += totalThreadsPerGPU * identicalGPUs.size();
|
||||
offset_p += totalThreadsPerGPU * identicalGPUs.size();
|
||||
offset_x += totalThreadsPerGPU * identicalGPUs.size();
|
||||
|
||||
if (offset_Ax >= N) {
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
device_count++;
|
||||
deviceId++;
|
||||
}
|
||||
|
||||
#if ENABLE_CPU_DEBUG_CODE
|
||||
float *Ax_cpu = reinterpret_cast<float *>(malloc(sizeof(float) * N));
|
||||
float *r_cpu = reinterpret_cast<float *>(malloc(sizeof(float) * N));
|
||||
float *p_cpu = reinterpret_cast<float *>(malloc(sizeof(float) * N));
|
||||
float *x_cpu = reinterpret_cast<float *>(malloc(sizeof(float) * N));
|
||||
|
||||
for (int i = 0; i < N; i++) {
|
||||
r_cpu[i] = 1.0;
|
||||
Ax_cpu[i] = x_cpu[i] = 0.0;
|
||||
}
|
||||
#endif
|
||||
|
||||
printf("Total threads per GPU = %d numBlocksPerSm = %d\n",
|
||||
numSms * numBlocksPerSm * THREADS_PER_BLOCK, numBlocksPerSm);
|
||||
cudaLaunchParams *launchParamsList = reinterpret_cast<cudaLaunchParams *>(
|
||||
malloc(sizeof(cudaLaunchParams) * identicalGPUs.size()));
|
||||
for (int i = 0; i < identicalGPUs.size(); i++) {
|
||||
launchParamsList[i].func =
|
||||
reinterpret_cast<void *>(multiGpuConjugateGradient);
|
||||
launchParamsList[i].gridDim = dimGrid;
|
||||
launchParamsList[i].blockDim = dimBlock;
|
||||
launchParamsList[i].sharedMem = sMemSize;
|
||||
launchParamsList[i].stream = nStreams[i];
|
||||
launchParamsList[i].args = kernelArgs;
|
||||
}
|
||||
|
||||
printf("Launching kernel\n");
|
||||
checkCudaErrors(cudaLaunchCooperativeKernelMultiDevice(
|
||||
launchParamsList, identicalGPUs.size(),
|
||||
cudaCooperativeLaunchMultiDeviceNoPreSync |
|
||||
cudaCooperativeLaunchMultiDeviceNoPostSync));
|
||||
|
||||
checkCudaErrors(cudaMemPrefetchAsync(x, sizeof(float) * N, cudaCpuDeviceId));
|
||||
checkCudaErrors(
|
||||
cudaMemPrefetchAsync(dot_result, sizeof(double), cudaCpuDeviceId));
|
||||
|
||||
deviceId = identicalGPUs.begin();
|
||||
device_count = 0;
|
||||
while (deviceId != identicalGPUs.end()) {
|
||||
checkCudaErrors(cudaSetDevice(*deviceId));
|
||||
checkCudaErrors(cudaStreamSynchronize(nStreams[device_count++]));
|
||||
deviceId++;
|
||||
}
|
||||
|
||||
r1 = *dot_result;
|
||||
|
||||
printf("GPU Final, residual = %e \n ", sqrt(r1));
|
||||
|
||||
#if ENABLE_CPU_DEBUG_CODE
|
||||
cpuConjugateGrad(I, J, val, x_cpu, Ax_cpu, p_cpu, r_cpu, nz, N, tol);
|
||||
#endif
|
||||
|
||||
float rsum, diff, err = 0.0;
|
||||
|
||||
for (int i = 0; i < N; i++) {
|
||||
rsum = 0.0;
|
||||
|
||||
for (int j = I[i]; j < I[i + 1]; j++) {
|
||||
rsum += val_cpu[j] * x[J[j]];
|
||||
}
|
||||
|
||||
diff = fabs(rsum - rhs);
|
||||
|
||||
if (diff > err) {
|
||||
err = diff;
|
||||
}
|
||||
}
|
||||
|
||||
checkCudaErrors(cudaFree(I));
|
||||
checkCudaErrors(cudaFree(J));
|
||||
checkCudaErrors(cudaFree(val));
|
||||
checkCudaErrors(cudaFree(x));
|
||||
checkCudaErrors(cudaFree(r));
|
||||
checkCudaErrors(cudaFree(p));
|
||||
checkCudaErrors(cudaFree(Ax));
|
||||
checkCudaErrors(cudaFree(dot_result));
|
||||
free(val_cpu);
|
||||
|
||||
#if ENABLE_CPU_DEBUG_CODE
|
||||
free(Ax_cpu);
|
||||
free(r_cpu);
|
||||
free(p_cpu);
|
||||
free(x_cpu);
|
||||
#endif
|
||||
|
||||
printf("Test Summary: Error amount = %f \n", err);
|
||||
fprintf(stdout, "&&&& conjugateGradientMultiDeviceCG %s\n",
|
||||
(sqrt(r1) < tol) ? "PASSED" : "FAILED");
|
||||
exit((sqrt(r1) < tol) ? EXIT_SUCCESS : EXIT_FAILURE);
|
||||
}
|
||||
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 11.00
|
||||
# Visual Studio 2010
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientMultiDeviceCG", "conjugateGradientMultiDeviceCG_vs2010.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
|
|
@ -0,0 +1,107 @@
|
|||
<?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>conjugateGradientMultiDeviceCG_vs2010</RootNamespace>
|
||||
<ProjectName>conjugateGradientMultiDeviceCG</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(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="$(CUDAPropsPath)\CUDA 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientMultiDeviceCG.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="conjugateGradientMultiDeviceCG.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2012
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientMultiDeviceCG", "conjugateGradientMultiDeviceCG_vs2012.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
|
|
@ -0,0 +1,108 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>conjugateGradientMultiDeviceCG_vs2012</RootNamespace>
|
||||
<ProjectName>conjugateGradientMultiDeviceCG</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v110</PlatformToolset>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(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="$(CUDAPropsPath)\CUDA 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientMultiDeviceCG.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="conjugateGradientMultiDeviceCG.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 13.00
|
||||
# Visual Studio 2013
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientMultiDeviceCG", "conjugateGradientMultiDeviceCG_vs2013.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
|
|
@ -0,0 +1,108 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>conjugateGradientMultiDeviceCG_vs2013</RootNamespace>
|
||||
<ProjectName>conjugateGradientMultiDeviceCG</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v120</PlatformToolset>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(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="$(CUDAPropsPath)\CUDA 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientMultiDeviceCG.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="conjugateGradientMultiDeviceCG.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 14.00
|
||||
# Visual Studio 2015
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientMultiDeviceCG", "conjugateGradientMultiDeviceCG_vs2015.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
|
|
@ -0,0 +1,108 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>conjugateGradientMultiDeviceCG_vs2015</RootNamespace>
|
||||
<ProjectName>conjugateGradientMultiDeviceCG</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v140</PlatformToolset>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(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="$(CUDAPropsPath)\CUDA 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientMultiDeviceCG.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="conjugateGradientMultiDeviceCG.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2017
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "conjugateGradientMultiDeviceCG", "conjugateGradientMultiDeviceCG_vs2017.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
|
|
@ -0,0 +1,109 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>conjugateGradientMultiDeviceCG_vs2017</RootNamespace>
|
||||
<ProjectName>conjugateGradientMultiDeviceCG</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v141</PlatformToolset>
|
||||
<WindowsTargetPlatformVersion>10.0.15063.0</WindowsTargetPlatformVersion>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(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="$(CUDAPropsPath)\CUDA 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudadevrt.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/conjugateGradientMultiDeviceCG.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="conjugateGradientMultiDeviceCG.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
322
Samples/cudaTensorCoreGemm/Makefile
Normal file
322
Samples/cudaTensorCoreGemm/Makefile
Normal file
|
|
@ -0,0 +1,322 @@
|
|||
################################################################################
|
||||
#
|
||||
# Copyright 1993-2015 NVIDIA Corporation. All rights reserved.
|
||||
#
|
||||
# NOTICE TO USER:
|
||||
#
|
||||
# This source code is subject to NVIDIA ownership rights under U.S. and
|
||||
# international Copyright laws.
|
||||
#
|
||||
# NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOURCE
|
||||
# CODE FOR ANY PURPOSE. IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR
|
||||
# IMPLIED WARRANTY OF ANY KIND. NVIDIA DISCLAIMS ALL WARRANTIES WITH
|
||||
# REGARD TO THIS SOURCE CODE, INCLUDING ALL IMPLIED WARRANTIES OF
|
||||
# MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
|
||||
# IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL,
|
||||
# OR CONSEQUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS
|
||||
# OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
|
||||
# OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE
|
||||
# OR PERFORMANCE OF THIS SOURCE CODE.
|
||||
#
|
||||
# U.S. Government End Users. This source code is a "commercial item" as
|
||||
# that term is defined at 48 C.F.R. 2.101 (OCT 1995), consisting of
|
||||
# "commercial computer software" and "commercial computer software
|
||||
# documentation" as such terms are used in 48 C.F.R. 12.212 (SEPT 1995)
|
||||
# and is provided to the U.S. Government only as a commercial end item.
|
||||
# Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through
|
||||
# 227.7202-4 (JUNE 1995), all U.S. Government End Users acquire the
|
||||
# source code with only those rights set forth herein.
|
||||
#
|
||||
################################################################################
|
||||
#
|
||||
# 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-g++
|
||||
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 - cudaTensorCoreGemm 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 - cudaTensorCoreGemm is not supported on ARMv7 - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
# This sample is not supported on aarch64
|
||||
ifeq ($(TARGET_ARCH),aarch64)
|
||||
$(info >>> WARNING - cudaTensorCoreGemm 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 ?= 70
|
||||
|
||||
ifeq ($(SMS),)
|
||||
$(info >>> WARNING - no SM architectures have been specified - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
ifeq ($(GENCODE_FLAGS),)
|
||||
# Generate SASS code for each SM architecture listed in $(SMS)
|
||||
$(foreach sm,$(SMS),$(eval GENCODE_FLAGS += -gencode arch=compute_$(sm),code=sm_$(sm)))
|
||||
|
||||
# Generate PTX code from the highest SM architecture in $(SMS) to guarantee forward-compatibility
|
||||
HIGHEST_SM := $(lastword $(sort $(SMS)))
|
||||
ifneq ($(HIGHEST_SM),)
|
||||
GENCODE_FLAGS += -gencode arch=compute_$(HIGHEST_SM),code=compute_$(HIGHEST_SM)
|
||||
endif
|
||||
endif
|
||||
|
||||
ALL_CCFLAGS += -maxrregcount=255
|
||||
|
||||
ifeq ($(SAMPLE_ENABLED),0)
|
||||
EXEC ?= @echo "[@]"
|
||||
endif
|
||||
|
||||
################################################################################
|
||||
|
||||
# Target rules
|
||||
all: build
|
||||
|
||||
build: cudaTensorCoreGemm
|
||||
|
||||
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
|
||||
|
||||
cudaTensorCoreGemm.o:cudaTensorCoreGemm.cu
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
cudaTensorCoreGemm: cudaTensorCoreGemm.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) ./cudaTensorCoreGemm
|
||||
|
||||
clean:
|
||||
rm -f cudaTensorCoreGemm cudaTensorCoreGemm.o
|
||||
rm -rf ../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)/cudaTensorCoreGemm
|
||||
|
||||
clobber: clean
|
||||
64
Samples/cudaTensorCoreGemm/NsightEclipse.xml
Normal file
64
Samples/cudaTensorCoreGemm/NsightEclipse.xml
Normal file
|
|
@ -0,0 +1,64 @@
|
|||
<?xml version="1.0" encoding="UTF-8"?>
|
||||
<!DOCTYPE entry SYSTEM "SamplesInfo.dtd">
|
||||
<entry>
|
||||
<name>cudaTensorCoreGemm</name>
|
||||
<cflags>
|
||||
<flag>-maxrregcount=255</flag>
|
||||
</cflags>
|
||||
<cuda_api_list>
|
||||
<toolkit>cudaMallocManaged</toolkit>
|
||||
<toolkit>cudaDeviceSynchronize</toolkit>
|
||||
<toolkit>cudaFuncSetAttribute</toolkit>
|
||||
<toolkit>cudaEventCreate</toolkit>
|
||||
<toolkit>cudaEventRecord</toolkit>
|
||||
<toolkit>cudaEventSynchronize</toolkit>
|
||||
<toolkit>cudaEventElapsedTime</toolkit>
|
||||
<toolkit>cudaFree</toolkit>
|
||||
</cuda_api_list>
|
||||
<description><![CDATA[CUDA sample demonstrating a GEMM computation using the Warp Matrix Multiply and Accumulate (WMMA) API introduced in CUDA 9.
|
||||
|
||||
This sample demonstrates the use of the new CUDA WMMA API employing the Tensor Cores introduced in the Volta chip family for faster matrix operations.
|
||||
|
||||
In addition to that, it demonstrates the use of the new CUDA function attribute cudaFuncAttributeMaxDynamicSharedMemorySize that allows the application to reserve an extended amount of shared memory than it is available by default.]]></description>
|
||||
<devicecompilation>whole</devicecompilation>
|
||||
<includepaths>
|
||||
<path>./</path>
|
||||
<path>../</path>
|
||||
<path>../../common/inc</path>
|
||||
</includepaths>
|
||||
<keyconcepts>
|
||||
<concept level="basic">Matrix Multiply</concept>
|
||||
<concept level="advanced">WMMA</concept>
|
||||
<concept level="advanced">Tensor Cores</concept>
|
||||
</keyconcepts>
|
||||
<keywords>
|
||||
</keywords>
|
||||
<libraries>
|
||||
</libraries>
|
||||
<librarypaths>
|
||||
</librarypaths>
|
||||
<nsight_eclipse>true</nsight_eclipse>
|
||||
<primary_file>cudaTensorCoreGemm.cu</primary_file>
|
||||
<scopes>
|
||||
<scope>1:CUDA Basic Topics</scope>
|
||||
</scopes>
|
||||
<sm-arch>sm70</sm-arch>
|
||||
<supported_envs>
|
||||
<env>
|
||||
<arch>x86_64</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
<env>
|
||||
<platform>windows7</platform>
|
||||
</env>
|
||||
<env>
|
||||
<arch>ppc64le</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
</supported_envs>
|
||||
<supported_sm_architectures>
|
||||
<from>7.0</from>
|
||||
</supported_sm_architectures>
|
||||
<title>CUDA Tensor Core GEMM</title>
|
||||
<type>exe</type>
|
||||
</entry>
|
||||
74
Samples/cudaTensorCoreGemm/README.md
Normal file
74
Samples/cudaTensorCoreGemm/README.md
Normal file
|
|
@ -0,0 +1,74 @@
|
|||
# cudaTensorCoreGemm - CUDA Tensor Core GEMM
|
||||
|
||||
## Description
|
||||
|
||||
CUDA sample demonstrating a GEMM computation using the Warp Matrix Multiply and Accumulate (WMMA) API introduced in CUDA 9.
|
||||
|
||||
This sample demonstrates the use of the new CUDA WMMA API employing the Tensor Cores introduced in the Volta chip family for faster matrix operations.
|
||||
|
||||
In addition to that, it demonstrates the use of the new CUDA function attribute cudaFuncAttributeMaxDynamicSharedMemorySize that allows the application to reserve an extended amount of shared memory than it is available by default.
|
||||
|
||||
## Key Concepts
|
||||
|
||||
Matrix Multiply, WMMA, Tensor Cores
|
||||
|
||||
## Supported SM Architectures
|
||||
|
||||
[SM 7.0 ](https://developer.nvidia.com/cuda-gpus)
|
||||
|
||||
## Supported OSes
|
||||
|
||||
Linux, Windows
|
||||
|
||||
## Supported CPU Architecture
|
||||
|
||||
x86_64, ppc64le
|
||||
|
||||
## CUDA APIs involved
|
||||
|
||||
### [CUDA Runtime API](http://docs.nvidia.com/cuda/cuda-runtime-api/index.html)
|
||||
cudaMallocManaged, cudaDeviceSynchronize, cudaFuncSetAttribute, cudaEventCreate, cudaEventRecord, cudaEventSynchronize, cudaEventElapsedTime, cudaFree
|
||||
|
||||
## Prerequisites
|
||||
|
||||
Download and install the [CUDA Toolkit 9.2](https://developer.nvidia.com/cuda-downloads) for your corresponding platform.
|
||||
|
||||
## Build and Run
|
||||
|
||||
### Windows
|
||||
The Windows samples are built using the Visual Studio IDE. Solution files (.sln) are provided for each supported version of Visual Studio, using the format:
|
||||
```
|
||||
*_vs<version>.sln - for Visual Studio <version>
|
||||
```
|
||||
Each individual sample has its own set of solution files in its directory:
|
||||
|
||||
To build/examine all the samples at once, the complete solution files should be used. To build/examine a single sample, the individual sample solution files should be used.
|
||||
> **Note:** Some samples require that the Microsoft DirectX SDK (June 2010 or newer) be installed and that the VC++ directory paths are properly set up (**Tools > Options...**). Check DirectX Dependencies section for details."
|
||||
|
||||
### Linux
|
||||
The Linux samples are built using makefiles. To use the makefiles, change the current directory to the sample directory you wish to build, and run make:
|
||||
```
|
||||
$ cd <sample_dir>
|
||||
$ make
|
||||
```
|
||||
The samples makefiles can take advantage of certain options:
|
||||
* **TARGET_ARCH=<arch>** - cross-compile targeting a specific architecture. Allowed architectures are x86_64, ppc64le.
|
||||
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)
|
||||
|
||||
495
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm.cu
Normal file
495
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm.cu
Normal file
|
|
@ -0,0 +1,495 @@
|
|||
/* 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.
|
||||
*/
|
||||
|
||||
// CUDA sample demonstrating a GEMM computation using the Warp Matrix Multiply
|
||||
// and Accumulate API introduced in CUDA 9.
|
||||
|
||||
// In this program, the compute_gemm kernel computes the result of a matrix
|
||||
// multiplication and addition: D = alpha * A * B + beta * C. The dimensions of
|
||||
// both C and D matrices are M_GLOBAL x N_GLOBAL. The A matrix is M_GLOBAL x
|
||||
// K_GLOBAL (row-major), the B matrix is K_GLOBAL x N_GLOBAL (column-major). In
|
||||
// that kernel, each CTA computes one 128 x 128 tile of the resulting matrix per
|
||||
// iteration. When the tile is computed, the CTA stores it to the global memory
|
||||
// and begins a new iteration, selecting a new 128 x 128 tile to compute.
|
||||
// Each CTA consists of eight warps. For the 128 x 128 tile, each warp computes
|
||||
// eight 16 x 16 subtiles, organized in a 2 x 4 two-dimensional array. Warps
|
||||
// compute the 16 x 16 subtiles using nvcuda::wmma::mma_sync operations by
|
||||
// moving through the K_GLOBAL dimension of the A and B matrices and
|
||||
// accumulating the intermediate result in the local thread state.
|
||||
|
||||
// There are a number of simple optimizations used in the algorithm:
|
||||
// - The CTA copies the 128 x 128 tile of the C matrix from the global memory to
|
||||
// shared memory. After that is done, each warp loads the C matrix fragments
|
||||
// from shared memory, thus avoiding a random global memory access.
|
||||
// - On each internal iteration, the CTA copies a portion of the A and B
|
||||
// matrices from
|
||||
// global memory to shared memory. After that, all warps in the CTA reuse the
|
||||
// A and B data from shared memory, thus reducing the number of data copies
|
||||
// from global memory.
|
||||
// - The portions of the A and B matrices are stored in shared memory with an
|
||||
// additional
|
||||
// padding (skew) to reduce the number of shared memory access bank conflicts.
|
||||
// (See a detailed explanation near the SKEW_HALF macro definition.)
|
||||
// - When the CTA finishes computing the tiles of the resulting matrix, each
|
||||
// warp stores
|
||||
// its subtiles to shared memory. The CTA then copies the shared memory
|
||||
// contents to global memory, again avoiding redundant random global memory
|
||||
// accesses.
|
||||
// - Note that the CTA tile size is chosen to maximize the GPU register
|
||||
// utilization,
|
||||
// but carefully enough to avoid local memory use.
|
||||
|
||||
#include <assert.h>
|
||||
#include <cuda.h>
|
||||
#include <mma.h>
|
||||
#include <stdio.h>
|
||||
|
||||
// helper functions and utilities to work with CUDA
|
||||
#include <helper_cuda.h>
|
||||
#include <helper_functions.h>
|
||||
|
||||
// GPU configuration.
|
||||
|
||||
#define WARP_SIZE 32
|
||||
|
||||
// MMA matrix tile dimensions.
|
||||
|
||||
#define M 16
|
||||
#define N 16
|
||||
#define K 16
|
||||
|
||||
// GEMM configuration.
|
||||
|
||||
#define M_TILES 256
|
||||
#define N_TILES 256
|
||||
#define K_TILES 256
|
||||
|
||||
#define M_GLOBAL (M * M_TILES)
|
||||
#define N_GLOBAL (N * N_TILES)
|
||||
#define K_GLOBAL (K * K_TILES)
|
||||
|
||||
#define C_LAYOUT wmma::mem_row_major
|
||||
|
||||
// Implementation constants.
|
||||
|
||||
#define WARPS_PER_BLOCK 8
|
||||
#define THREADS_PER_BLOCK (WARP_SIZE * WARPS_PER_BLOCK)
|
||||
|
||||
#define CHUNK_K 8
|
||||
|
||||
#define BLOCK_ROW_WARPS 2
|
||||
#define BLOCK_COL_WARPS 4
|
||||
|
||||
#define WARP_ROW_TILES 4
|
||||
#define WARP_COL_TILES 2
|
||||
|
||||
#define BLOCK_ROW_TILES (WARP_ROW_TILES * BLOCK_ROW_WARPS)
|
||||
#define BLOCK_COL_TILES (WARP_COL_TILES * BLOCK_COL_WARPS)
|
||||
|
||||
#define GLOBAL_MEM_STRIDE N_GLOBAL
|
||||
|
||||
#define SHMEM_STRIDE (N * BLOCK_ROW_TILES)
|
||||
#define SHMEM_OFFSET (N * WARP_ROW_TILES)
|
||||
|
||||
// The macro below is used to shift rows of the A matrix and columns of the B
|
||||
// matrix in shared memory to minimize possible bank conflicts. Before
|
||||
// performing the nvcuda::wmma::mma_sync operation, the warp must load the
|
||||
// matrix data using the nvcuda::wmma::load_matrix_sync operation. Although the
|
||||
// memory access pattern is not specified for that function, each lane in the
|
||||
// warp can read one or multiple matrix elements from different matrix rows or
|
||||
// columns. For shared memory, such access can result in bank conflicts if
|
||||
// different rows / columns of the matrix map to the same bank. By shifting each
|
||||
// row and column by a few bytes, we make sure that they map to different banks,
|
||||
// thus reducing the number of possible bank conflicts. The number of 8 two-byte
|
||||
// "half" elements is chosen as the minimum possible shift because we must keep
|
||||
// each row and column 128-bit aligned, as required by
|
||||
// nvcuda::wmma::load_matrix_sync.
|
||||
#define SKEW_HALF 8
|
||||
|
||||
#define checkKernelErrors(expr) \
|
||||
do { \
|
||||
expr; \
|
||||
\
|
||||
cudaError_t __err = cudaGetLastError(); \
|
||||
if (__err != cudaSuccess) { \
|
||||
printf("Line %d: '%s' failed: %s\n", __LINE__, #expr, \
|
||||
cudaGetErrorString(__err)); \
|
||||
abort(); \
|
||||
} \
|
||||
} while (0)
|
||||
|
||||
using namespace nvcuda;
|
||||
|
||||
__host__ void init_host_matrices(float *a, float *b, float *c) {
|
||||
for (int i = 0; i < M_GLOBAL; i++) {
|
||||
for (int j = 0; j < K_GLOBAL; j++) {
|
||||
a[i * K_GLOBAL + j] = static_cast<float>(rand() % 3);
|
||||
}
|
||||
}
|
||||
|
||||
for (int i = 0; i < N_GLOBAL; i++) {
|
||||
for (int j = 0; j < K_GLOBAL; j++) {
|
||||
b[i * K_GLOBAL + j] = static_cast<float>(rand() % 3);
|
||||
}
|
||||
}
|
||||
|
||||
for (int t = 0; t < M_GLOBAL * N_GLOBAL; t++) {
|
||||
c[t] = static_cast<float>(rand() % 3);
|
||||
}
|
||||
}
|
||||
|
||||
__global__ void init_device_matrices(const float *A_h, const float *B_h,
|
||||
const float *C_h, half *A, half *B,
|
||||
float *C, float *D) {
|
||||
for (int i = blockDim.x * blockIdx.x + threadIdx.x; i < M_GLOBAL * K_GLOBAL;
|
||||
i += gridDim.x * blockDim.x)
|
||||
A[i] = __float2half(A_h[i]);
|
||||
|
||||
for (int i = blockDim.x * blockIdx.x + threadIdx.x; i < N_GLOBAL * K_GLOBAL;
|
||||
i += gridDim.x * blockDim.x)
|
||||
B[i] = __float2half(B_h[i]);
|
||||
|
||||
for (int i = blockDim.x * blockIdx.x + threadIdx.x; i < M_GLOBAL * N_GLOBAL;
|
||||
i += gridDim.x * blockDim.x)
|
||||
C[i] = C_h[i];
|
||||
|
||||
for (int i = blockDim.x * blockIdx.x + threadIdx.x; i < M_GLOBAL * N_GLOBAL;
|
||||
i += gridDim.x * blockDim.x)
|
||||
D[i] = 0;
|
||||
}
|
||||
|
||||
__global__ void compute_gemm(const half *A, const half *B, const float *C,
|
||||
float *D, float alpha, float beta) {
|
||||
extern __shared__ half shmem[][CHUNK_K * K + SKEW_HALF];
|
||||
|
||||
// Warp and lane identification.
|
||||
const unsigned int warpId = threadIdx.x / WARP_SIZE;
|
||||
const unsigned int laneId = threadIdx.x % WARP_SIZE;
|
||||
|
||||
// Offset in shared memory from which the B matrix is stored.
|
||||
const size_t shmem_idx_b_off = BLOCK_COL_TILES * M;
|
||||
|
||||
// This pointer is used to access the C and D matrix tiles this warp computes.
|
||||
float *shmem_warp_tile_ptr = reinterpret_cast<float *>(
|
||||
&shmem[0][0] + (warpId / 2) * SHMEM_STRIDE * K * 2 +
|
||||
(warpId % 2) * SHMEM_OFFSET);
|
||||
|
||||
// This pointer is used to stream the C and D matrices block-wide tile to and
|
||||
// from shared memory.
|
||||
float *shmem_warp_stream_ptr =
|
||||
reinterpret_cast<float *>(&shmem[0][0] + warpId * SHMEM_STRIDE * K);
|
||||
|
||||
// Adjust the beta scaler, as it'll be multiplied by alpha at the end of
|
||||
// each tile computation. Technically this is not generally correct (may
|
||||
// result in a loss of precision). Zero still needs to be specially handled
|
||||
// though.
|
||||
beta /= alpha;
|
||||
|
||||
// Each CTA slides along the 128 x 128 tiles from the top left corner of the
|
||||
// matrix to the right and down, and selects the next tile to compute. Once
|
||||
// there's no such tile, all warps in this CTA exit.
|
||||
for (unsigned int block_pos = blockIdx.x;; block_pos += gridDim.x) {
|
||||
const unsigned int block_tile_i =
|
||||
((block_pos * BLOCK_ROW_TILES) / N_TILES) * (BLOCK_COL_TILES);
|
||||
const unsigned int block_tile_j = (block_pos * BLOCK_COL_TILES) % N_TILES;
|
||||
|
||||
// Stop when there are no more D matrix tiles to compute in this CTA.
|
||||
if (block_tile_i >= M_TILES) {
|
||||
break;
|
||||
}
|
||||
|
||||
// This warp's pointer to the C matrix data to copy memory from to shared
|
||||
// memory.
|
||||
const size_t gmem_idx =
|
||||
(block_tile_i + warpId) * M * GLOBAL_MEM_STRIDE + block_tile_j * N;
|
||||
const float *src_gmem_warp_stream_ptr = &C[gmem_idx];
|
||||
|
||||
// Stream multiple C tiles to shared memory.
|
||||
#pragma unroll
|
||||
for (int i = 0; i < K; i++) {
|
||||
typedef int4 copy_t;
|
||||
|
||||
*((copy_t *)(shmem_warp_stream_ptr + SHMEM_STRIDE * i) + laneId) =
|
||||
*((copy_t *)(src_gmem_warp_stream_ptr + GLOBAL_MEM_STRIDE * i) +
|
||||
laneId);
|
||||
}
|
||||
|
||||
__syncthreads();
|
||||
|
||||
// These fragments will accumulate the result of A and B matrix fragment
|
||||
// multiplications along the K_GLOBAL dimension.
|
||||
wmma::fragment<wmma::accumulator, M, N, K, float> c[WARP_COL_TILES]
|
||||
[WARP_ROW_TILES];
|
||||
|
||||
// Load the C matrix tiles into fragments from shared memory.
|
||||
#pragma unroll
|
||||
for (int i = 0; i < WARP_COL_TILES; i++) {
|
||||
#pragma unroll
|
||||
for (int j = 0; j < WARP_ROW_TILES; j++) {
|
||||
const float *tile_ptr =
|
||||
shmem_warp_tile_ptr + i * SHMEM_STRIDE * K + j * N;
|
||||
|
||||
wmma::load_matrix_sync(c[i][j], tile_ptr, SHMEM_STRIDE, C_LAYOUT);
|
||||
}
|
||||
}
|
||||
|
||||
__syncthreads();
|
||||
|
||||
// Scale the C matrix.
|
||||
#pragma unroll
|
||||
for (int i = 0; i < WARP_COL_TILES; i++) {
|
||||
#pragma unroll
|
||||
for (int j = 0; j < WARP_ROW_TILES; j++) {
|
||||
#pragma unroll
|
||||
for (int t = 0; t < c[i][j].num_elements; t++) {
|
||||
c[i][j].x[t] *= beta;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Select what warp copies what matrix to shared memory.
|
||||
// Warps 0-3 copy the A matrix, warps 4-7 copy the B matrix.
|
||||
const half *warp_ptr = (warpId < 4) ? (&A[block_tile_i * M * K_GLOBAL] +
|
||||
M * K_GLOBAL * (warpId % 4) * 2)
|
||||
: (&B[block_tile_j * N * K_GLOBAL] +
|
||||
N * K_GLOBAL * (warpId % 4) * 2);
|
||||
|
||||
// Go through the global K dimension by a fixed step at a time.
|
||||
#pragma unroll
|
||||
for (int tile_k = 0; tile_k < K_TILES; tile_k += CHUNK_K) {
|
||||
// Copy slices of the A and B matrices to shared memory.
|
||||
// The first half of the warps in the CTA copy the A matrix, the rest copy
|
||||
// the B matrix.
|
||||
size_t shmem_idx =
|
||||
warpId < (WARPS_PER_BLOCK / 2)
|
||||
? (M * (warpId % (WARPS_PER_BLOCK / 2)) * 2)
|
||||
: (N * (warpId % (WARPS_PER_BLOCK / 2)) * 2 + shmem_idx_b_off);
|
||||
|
||||
// First half of the warp copies the first row / column of the matrix,
|
||||
// the second half of the warp copies the next.
|
||||
int4 *lane_ptr = (int4 *)(warp_ptr + tile_k * K +
|
||||
(laneId / (WARP_SIZE / 2)) * K_GLOBAL) +
|
||||
(laneId % (WARP_SIZE / 2));
|
||||
|
||||
// Shift the second half of the warp to the next row / column in the
|
||||
// shared memory.
|
||||
shmem_idx += laneId / (WARP_SIZE / 2);
|
||||
|
||||
#pragma unroll
|
||||
for (int i = 0; i < (WARP_SIZE / 2); i++) {
|
||||
// Copy 16 bytes at once in each lane.
|
||||
*((int4 *)&shmem[shmem_idx][0] + (laneId % (WARP_SIZE / 2))) =
|
||||
*lane_ptr;
|
||||
|
||||
// Advance the global memory pointer and the shared memory index.
|
||||
lane_ptr = reinterpret_cast<int4 *>(
|
||||
reinterpret_cast<half *>(lane_ptr + K_GLOBAL * 2));
|
||||
shmem_idx += 2;
|
||||
}
|
||||
|
||||
__syncthreads();
|
||||
|
||||
// Compute a grid of C matrix tiles in each warp.
|
||||
#pragma unroll
|
||||
for (int k_step = 0; k_step < CHUNK_K; k_step++) {
|
||||
wmma::fragment<wmma::matrix_a, M, N, K, half, wmma::row_major>
|
||||
a[WARP_COL_TILES];
|
||||
wmma::fragment<wmma::matrix_b, M, N, K, half, wmma::col_major>
|
||||
b[WARP_ROW_TILES];
|
||||
|
||||
#pragma unroll
|
||||
for (int i = 0; i < WARP_COL_TILES; i++) {
|
||||
size_t shmem_idx_a = (warpId / 2) * M * 2 + (i * M);
|
||||
const half *tile_ptr = &shmem[shmem_idx_a][k_step * K];
|
||||
|
||||
wmma::load_matrix_sync(a[i], tile_ptr, K * CHUNK_K + SKEW_HALF);
|
||||
|
||||
#pragma unroll
|
||||
for (int j = 0; j < WARP_ROW_TILES; j++) {
|
||||
if (i == 0) {
|
||||
// Load the B matrix fragment once, because it is going to be
|
||||
// reused against the other A matrix fragments.
|
||||
size_t shmem_idx_b = shmem_idx_b_off +
|
||||
(WARP_ROW_TILES * N) * (warpId % 2) +
|
||||
(j * N);
|
||||
const half *tile_ptr = &shmem[shmem_idx_b][k_step * K];
|
||||
|
||||
wmma::load_matrix_sync(b[j], tile_ptr, K * CHUNK_K + SKEW_HALF);
|
||||
}
|
||||
|
||||
wmma::mma_sync(c[i][j], a[i], b[j], c[i][j]);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
__syncthreads();
|
||||
}
|
||||
|
||||
// Store the D fragments to shared memory.
|
||||
#pragma unroll
|
||||
for (int i = 0; i < WARP_COL_TILES; i++) {
|
||||
#pragma unroll
|
||||
for (int j = 0; j < WARP_ROW_TILES; j++) {
|
||||
#pragma unroll
|
||||
// Uniform, point-wise transformations of ALL fragment elements by ALL
|
||||
// threads in the warp are well-defined even though element indices
|
||||
// within fragment storage are not defined.
|
||||
for (int t = 0; t < c[i][j].num_elements; t++) c[i][j].x[t] *= alpha;
|
||||
|
||||
float *tile_ptr = shmem_warp_tile_ptr + i * SHMEM_STRIDE * K + j * N;
|
||||
|
||||
wmma::store_matrix_sync(tile_ptr, c[i][j], SHMEM_STRIDE, C_LAYOUT);
|
||||
}
|
||||
}
|
||||
|
||||
__syncthreads();
|
||||
|
||||
// Now that shared memory contains all the D tiles, stream them to global
|
||||
// memory.
|
||||
float *dst_gmem_warp_stream_ptr = &D[gmem_idx];
|
||||
|
||||
#pragma unroll
|
||||
for (int i = 0; i < K; i++) {
|
||||
*(reinterpret_cast<int4 *>(dst_gmem_warp_stream_ptr +
|
||||
GLOBAL_MEM_STRIDE * i) +
|
||||
laneId) =
|
||||
*(reinterpret_cast<int4 *>(shmem_warp_stream_ptr + SHMEM_STRIDE * i) +
|
||||
laneId);
|
||||
}
|
||||
|
||||
__syncthreads();
|
||||
}
|
||||
}
|
||||
|
||||
int main(int argc, char **argv) {
|
||||
printf("Initializing...\n");
|
||||
|
||||
int dev = findCudaDevice(argc, (const char **)argv);
|
||||
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, dev));
|
||||
|
||||
// Tensor cores require a GPU of Volta (SM7X) architecture or higher.
|
||||
if (deviceProp.major < 7) {
|
||||
printf(
|
||||
"cudaTensorCoreGemm requires requires SM 7.0 or higher to use Tensor "
|
||||
"Cores. Exiting...\n");
|
||||
exit(EXIT_WAIVED);
|
||||
}
|
||||
|
||||
printf("M: %d (%d x %d)\n", M_GLOBAL, M, M_TILES);
|
||||
printf("N: %d (%d x %d)\n", N_GLOBAL, N, N_TILES);
|
||||
printf("K: %d (%d x %d)\n", K_GLOBAL, K, K_TILES);
|
||||
|
||||
float *A_h = NULL;
|
||||
float *B_h = NULL;
|
||||
float *C_h = NULL;
|
||||
|
||||
checkCudaErrors(cudaMallocManaged(reinterpret_cast<void **>(&A_h),
|
||||
sizeof(float) * M_GLOBAL * K_GLOBAL));
|
||||
checkCudaErrors(cudaMallocManaged(reinterpret_cast<void **>(&B_h),
|
||||
sizeof(float) * K_GLOBAL * N_GLOBAL));
|
||||
checkCudaErrors(cudaMallocManaged(reinterpret_cast<void **>(&C_h),
|
||||
sizeof(float) * M_GLOBAL * N_GLOBAL));
|
||||
|
||||
half *A = NULL;
|
||||
half *B = NULL;
|
||||
float *C = NULL;
|
||||
float *D = NULL;
|
||||
|
||||
checkCudaErrors(cudaMalloc(reinterpret_cast<void **>(&A),
|
||||
sizeof(half) * M_GLOBAL * K_GLOBAL));
|
||||
checkCudaErrors(cudaMalloc(reinterpret_cast<void **>(&B),
|
||||
sizeof(half) * N_GLOBAL * K_GLOBAL));
|
||||
checkCudaErrors(cudaMalloc(reinterpret_cast<void **>(&C),
|
||||
sizeof(float) * M_GLOBAL * N_GLOBAL));
|
||||
checkCudaErrors(cudaMalloc(reinterpret_cast<void **>(&D),
|
||||
sizeof(float) * M_GLOBAL * N_GLOBAL));
|
||||
|
||||
assert(((unsigned long long)A) % 128 == 0);
|
||||
assert(((unsigned long long)B) % 128 == 0);
|
||||
assert(((unsigned long long)C) % 128 == 0);
|
||||
assert(((unsigned long long)D) % 128 == 0);
|
||||
|
||||
init_host_matrices(A_h, B_h, C_h);
|
||||
|
||||
printf("Preparing data for GPU...\n");
|
||||
|
||||
checkKernelErrors(
|
||||
(init_device_matrices<<<deviceProp.multiProcessorCount,
|
||||
THREADS_PER_BLOCK>>>(A_h, B_h, C_h, A, B, C, D)));
|
||||
|
||||
checkCudaErrors(cudaDeviceSynchronize());
|
||||
|
||||
enum {
|
||||
SHMEM_SZ =
|
||||
sizeof(half) * (BLOCK_COL_TILES * M) * (CHUNK_K * K + SKEW_HALF) * 2
|
||||
};
|
||||
|
||||
printf("Required shared memory size: %lu Kb\n", SHMEM_SZ / 1024UL);
|
||||
|
||||
checkCudaErrors(cudaFuncSetAttribute(
|
||||
compute_gemm, cudaFuncAttributeMaxDynamicSharedMemorySize, SHMEM_SZ));
|
||||
|
||||
printf("Computing...\n");
|
||||
|
||||
cudaEvent_t start, stop;
|
||||
|
||||
checkCudaErrors(cudaEventCreate(&start));
|
||||
checkCudaErrors(cudaEventCreate(&stop));
|
||||
checkCudaErrors(cudaEventRecord(start));
|
||||
|
||||
const float alpha = 1.1f;
|
||||
const float beta = 1.2f;
|
||||
|
||||
checkKernelErrors(
|
||||
(compute_gemm<<<deviceProp.multiProcessorCount, THREADS_PER_BLOCK,
|
||||
SHMEM_SZ>>>(A, B, C, D, alpha, beta)));
|
||||
|
||||
checkCudaErrors(cudaEventRecord(stop));
|
||||
checkCudaErrors(cudaEventSynchronize(stop));
|
||||
|
||||
float milliseconds = 0;
|
||||
|
||||
checkCudaErrors(cudaEventElapsedTime(&milliseconds, start, stop));
|
||||
|
||||
printf("Time: %f ms\n", milliseconds);
|
||||
printf("TFLOPS: %.2f\n", static_cast<double>((static_cast<double>(M_GLOBAL) *
|
||||
N_GLOBAL * K_GLOBAL * 2) /
|
||||
(milliseconds / 1000.)) /
|
||||
1e12);
|
||||
|
||||
checkCudaErrors(cudaFree(reinterpret_cast<void *>(A_h)));
|
||||
checkCudaErrors(cudaFree(reinterpret_cast<void *>(B_h)));
|
||||
checkCudaErrors(cudaFree(reinterpret_cast<void *>(C_h)));
|
||||
checkCudaErrors(cudaFree(reinterpret_cast<void *>(A)));
|
||||
checkCudaErrors(cudaFree(reinterpret_cast<void *>(B)));
|
||||
checkCudaErrors(cudaFree(reinterpret_cast<void *>(C)));
|
||||
checkCudaErrors(cudaFree(reinterpret_cast<void *>(D)));
|
||||
|
||||
return 0;
|
||||
}
|
||||
20
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2010.sln
Normal file
20
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2010.sln
Normal file
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 11.00
|
||||
# Visual Studio 2010
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "cudaTensorCoreGemm", "cudaTensorCoreGemm_vs2010.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
|
||||
106
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2010.vcxproj
Normal file
106
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2010.vcxproj
Normal file
|
|
@ -0,0 +1,106 @@
|
|||
<?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>cudaTensorCoreGemm_vs2010</RootNamespace>
|
||||
<ProjectName>cudaTensorCoreGemm</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
</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 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/cudaTensorCoreGemm.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</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>
|
||||
<CudaCompile Include="cudaTensorCoreGemm.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2012.sln
Normal file
20
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2012.sln
Normal file
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2012
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "cudaTensorCoreGemm", "cudaTensorCoreGemm_vs2012.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
|
||||
107
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2012.vcxproj
Normal file
107
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2012.vcxproj
Normal file
|
|
@ -0,0 +1,107 @@
|
|||
<?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>cudaTensorCoreGemm_vs2012</RootNamespace>
|
||||
<ProjectName>cudaTensorCoreGemm</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v110</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 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/cudaTensorCoreGemm.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</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>
|
||||
<CudaCompile Include="cudaTensorCoreGemm.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2013.sln
Normal file
20
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2013.sln
Normal file
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 13.00
|
||||
# Visual Studio 2013
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "cudaTensorCoreGemm", "cudaTensorCoreGemm_vs2013.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
|
||||
107
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2013.vcxproj
Normal file
107
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2013.vcxproj
Normal file
|
|
@ -0,0 +1,107 @@
|
|||
<?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>cudaTensorCoreGemm_vs2013</RootNamespace>
|
||||
<ProjectName>cudaTensorCoreGemm</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v120</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 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/cudaTensorCoreGemm.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</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>
|
||||
<CudaCompile Include="cudaTensorCoreGemm.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2015.sln
Normal file
20
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2015.sln
Normal file
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 14.00
|
||||
# Visual Studio 2015
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "cudaTensorCoreGemm", "cudaTensorCoreGemm_vs2015.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
|
||||
107
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2015.vcxproj
Normal file
107
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2015.vcxproj
Normal file
|
|
@ -0,0 +1,107 @@
|
|||
<?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>cudaTensorCoreGemm_vs2015</RootNamespace>
|
||||
<ProjectName>cudaTensorCoreGemm</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v140</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 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/cudaTensorCoreGemm.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</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>
|
||||
<CudaCompile Include="cudaTensorCoreGemm.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2017.sln
Normal file
20
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_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}") = "cudaTensorCoreGemm", "cudaTensorCoreGemm_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
|
||||
108
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2017.vcxproj
Normal file
108
Samples/cudaTensorCoreGemm/cudaTensorCoreGemm_vs2017.vcxproj
Normal file
|
|
@ -0,0 +1,108 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>cudaTensorCoreGemm_vs2017</RootNamespace>
|
||||
<ProjectName>cudaTensorCoreGemm</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v141</PlatformToolset>
|
||||
<WindowsTargetPlatformVersion>10.0.15063.0</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 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/cudaTensorCoreGemm.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</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>
|
||||
<CudaCompile Include="cudaTensorCoreGemm.cu" />
|
||||
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
|
|
@ -180,6 +180,21 @@ ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
|||
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)
|
||||
|
|
|
|||
|
|
@ -180,6 +180,21 @@ ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
|||
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)
|
||||
|
|
|
|||
|
|
@ -180,6 +180,21 @@ ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
|||
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)
|
||||
|
|
|
|||
304
Samples/shfl_scan/Makefile
Normal file
304
Samples/shfl_scan/Makefile
Normal file
|
|
@ -0,0 +1,304 @@
|
|||
################################################################################
|
||||
#
|
||||
# Copyright 1993-2015 NVIDIA Corporation. All rights reserved.
|
||||
#
|
||||
# NOTICE TO USER:
|
||||
#
|
||||
# This source code is subject to NVIDIA ownership rights under U.S. and
|
||||
# international Copyright laws.
|
||||
#
|
||||
# NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOURCE
|
||||
# CODE FOR ANY PURPOSE. IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR
|
||||
# IMPLIED WARRANTY OF ANY KIND. NVIDIA DISCLAIMS ALL WARRANTIES WITH
|
||||
# REGARD TO THIS SOURCE CODE, INCLUDING ALL IMPLIED WARRANTIES OF
|
||||
# MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
|
||||
# IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL,
|
||||
# OR CONSEQUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS
|
||||
# OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
|
||||
# OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE
|
||||
# OR PERFORMANCE OF THIS SOURCE CODE.
|
||||
#
|
||||
# U.S. Government End Users. This source code is a "commercial item" as
|
||||
# that term is defined at 48 C.F.R. 2.101 (OCT 1995), consisting of
|
||||
# "commercial computer software" and "commercial computer software
|
||||
# documentation" as such terms are used in 48 C.F.R. 12.212 (SEPT 1995)
|
||||
# and is provided to the U.S. Government only as a commercial end item.
|
||||
# Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through
|
||||
# 227.7202-4 (JUNE 1995), all U.S. Government End Users acquire the
|
||||
# source code with only those rights set forth herein.
|
||||
#
|
||||
################################################################################
|
||||
#
|
||||
# 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-g++
|
||||
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
|
||||
|
||||
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 ?= 30 35 37 50 52 60 61 70
|
||||
|
||||
ifeq ($(SMS),)
|
||||
$(info >>> WARNING - no SM architectures have been specified - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
ifeq ($(GENCODE_FLAGS),)
|
||||
# Generate SASS code for each SM architecture listed in $(SMS)
|
||||
$(foreach sm,$(SMS),$(eval GENCODE_FLAGS += -gencode arch=compute_$(sm),code=sm_$(sm)))
|
||||
|
||||
# Generate PTX code from the highest SM architecture in $(SMS) to guarantee forward-compatibility
|
||||
HIGHEST_SM := $(lastword $(sort $(SMS)))
|
||||
ifneq ($(HIGHEST_SM),)
|
||||
GENCODE_FLAGS += -gencode arch=compute_$(HIGHEST_SM),code=compute_$(HIGHEST_SM)
|
||||
endif
|
||||
endif
|
||||
|
||||
ALL_CCFLAGS += -O3
|
||||
|
||||
ifeq ($(SAMPLE_ENABLED),0)
|
||||
EXEC ?= @echo "[@]"
|
||||
endif
|
||||
|
||||
################################################################################
|
||||
|
||||
# Target rules
|
||||
all: build
|
||||
|
||||
build: shfl_scan
|
||||
|
||||
check.deps:
|
||||
ifeq ($(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
|
||||
|
||||
shfl_scan.o:shfl_scan.cu
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
shfl_scan: shfl_scan.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
|
||||
$(EXEC) ./shfl_scan
|
||||
|
||||
clean:
|
||||
rm -f shfl_scan shfl_scan.o
|
||||
rm -rf ../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)/shfl_scan
|
||||
|
||||
clobber: clean
|
||||
73
Samples/shfl_scan/NsightEclipse.xml
Normal file
73
Samples/shfl_scan/NsightEclipse.xml
Normal file
|
|
@ -0,0 +1,73 @@
|
|||
<?xml version="1.0" encoding="UTF-8"?>
|
||||
<!DOCTYPE entry SYSTEM "SamplesInfo.dtd">
|
||||
<entry>
|
||||
<name>shfl_scan</name>
|
||||
<cflags>
|
||||
<flag>-O3</flag>
|
||||
</cflags>
|
||||
<description><![CDATA[This example demonstrates how to use the shuffle intrinsic __shfl_up_sync to perform a scan operation across a thread block. ]]></description>
|
||||
<devicecompilation>whole</devicecompilation>
|
||||
<includepaths>
|
||||
<path>./</path>
|
||||
<path>../</path>
|
||||
<path>../../common/inc</path>
|
||||
</includepaths>
|
||||
<keyconcepts>
|
||||
<concept level="advanced">Data-Parallel Algorithms</concept>
|
||||
<concept level="advanced">Performance Strategies</concept>
|
||||
</keyconcepts>
|
||||
<keywords>
|
||||
<keyword>GPGPU</keyword>
|
||||
<keyword>CUDA</keyword>
|
||||
<keyword>scan</keyword>
|
||||
<keyword>parallel prefix sum</keyword>
|
||||
<keyword>Data-Parallel Algorithms</keyword>
|
||||
</keywords>
|
||||
<libraries>
|
||||
</libraries>
|
||||
<librarypaths>
|
||||
</librarypaths>
|
||||
<nsight_eclipse>true</nsight_eclipse>
|
||||
<primary_file>shfl_scan.cu</primary_file>
|
||||
<scopes>
|
||||
<scope>1:CUDA Advanced Topics</scope>
|
||||
<scope>1:Data-Parallel Algorithms</scope>
|
||||
<scope>1:Performance Strategies</scope>
|
||||
</scopes>
|
||||
<sm-arch>sm30</sm-arch>
|
||||
<sm-arch>sm35</sm-arch>
|
||||
<sm-arch>sm37</sm-arch>
|
||||
<sm-arch>sm50</sm-arch>
|
||||
<sm-arch>sm52</sm-arch>
|
||||
<sm-arch>sm60</sm-arch>
|
||||
<sm-arch>sm61</sm-arch>
|
||||
<sm-arch>sm70</sm-arch>
|
||||
<supported_envs>
|
||||
<env>
|
||||
<arch>x86_64</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
<env>
|
||||
<platform>windows7</platform>
|
||||
</env>
|
||||
<env>
|
||||
<arch>x86_64</arch>
|
||||
<platform>macosx</platform>
|
||||
</env>
|
||||
<env>
|
||||
<arch>arm</arch>
|
||||
</env>
|
||||
<env>
|
||||
<arch>aarch64</arch>
|
||||
</env>
|
||||
<env>
|
||||
<arch>ppc64le</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
</supported_envs>
|
||||
<supported_sm_architectures>
|
||||
<from>3.0</from>
|
||||
</supported_sm_architectures>
|
||||
<title>CUDA Parallel Prefix Sum with Shuffle Intrinsics (SHFL_Scan)</title>
|
||||
<type>exe</type>
|
||||
</entry>
|
||||
91
Samples/shfl_scan/README.md
Normal file
91
Samples/shfl_scan/README.md
Normal file
|
|
@ -0,0 +1,91 @@
|
|||
# shfl_scan - CUDA Parallel Prefix Sum with Shuffle Intrinsics (SHFL_Scan)
|
||||
|
||||
## Description
|
||||
|
||||
This example demonstrates how to use the shuffle intrinsic __shfl_up_sync to perform a scan operation across a thread block.
|
||||
|
||||
## Key Concepts
|
||||
|
||||
Data-Parallel Algorithms, Performance Strategies
|
||||
|
||||
## Supported SM Architectures
|
||||
|
||||
[SM 3.0 ](https://developer.nvidia.com/cuda-gpus) [SM 3.5 ](https://developer.nvidia.com/cuda-gpus) [SM 3.7 ](https://developer.nvidia.com/cuda-gpus) [SM 5.0 ](https://developer.nvidia.com/cuda-gpus) [SM 5.2 ](https://developer.nvidia.com/cuda-gpus) [SM 6.0 ](https://developer.nvidia.com/cuda-gpus) [SM 6.1 ](https://developer.nvidia.com/cuda-gpus) [SM 7.0 ](https://developer.nvidia.com/cuda-gpus)
|
||||
|
||||
## Supported OSes
|
||||
|
||||
Linux, Windows, MacOSX
|
||||
|
||||
## Supported CPU Architecture
|
||||
|
||||
x86_64, ppc64le, armv7l, aarch64
|
||||
|
||||
## CUDA APIs involved
|
||||
|
||||
## Prerequisites
|
||||
|
||||
Download and install the [CUDA Toolkit 9.2](https://developer.nvidia.com/cuda-downloads) for your corresponding platform.
|
||||
|
||||
## Build and Run
|
||||
|
||||
### Windows
|
||||
The Windows samples are built using the Visual Studio IDE. Solution files (.sln) are provided for each supported version of Visual Studio, using the format:
|
||||
```
|
||||
*_vs<version>.sln - for Visual Studio <version>
|
||||
```
|
||||
Each individual sample has its own set of solution files in its directory:
|
||||
|
||||
To build/examine all the samples at once, the complete solution files should be used. To build/examine a single sample, the individual sample solution files should be used.
|
||||
> **Note:** Some samples require that the Microsoft DirectX SDK (June 2010 or newer) be installed and that the VC++ directory paths are properly set up (**Tools > Options...**). Check DirectX Dependencies section for details."
|
||||
|
||||
### Linux
|
||||
The Linux samples are built using makefiles. To use the makefiles, change the current directory to the sample directory you wish to build, and run make:
|
||||
```
|
||||
$ cd <sample_dir>
|
||||
$ make
|
||||
```
|
||||
The samples makefiles can take advantage of certain options:
|
||||
* **TARGET_ARCH=<arch>** - cross-compile targeting a specific architecture. Allowed architectures are x86_64, ppc64le, armv7l, aarch64.
|
||||
By default, TARGET_ARCH is set to HOST_ARCH. On a x86_64 machine, not setting TARGET_ARCH is the equivalent of setting TARGET_ARCH=x86_64.<br/>
|
||||
`$ make TARGET_ARCH=x86_64` <br/> `$ make TARGET_ARCH=ppc64le` <br/> `$ make TARGET_ARCH=armv7l` <br/> `$ make TARGET_ARCH=aarch64` <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++
|
||||
```
|
||||
|
||||
### Mac
|
||||
The Mac samples are built using makefiles. To use the makefiles, change directory into the sample directory you wish to build, and run make:
|
||||
```
|
||||
$ cd <sample_dir>
|
||||
$ make
|
||||
```
|
||||
|
||||
The samples makefiles can take advantage of certain options:
|
||||
|
||||
* **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="A B ..."
|
||||
```
|
||||
|
||||
* **HOST_COMPILER=<host_compiler>** - override the default clang host compiler. See the [Mac Installation Guide](http://docs.nvidia.com/cuda/cuda-installation-guide-mac-os-x/index.html#system-requirements) for a list of supported host compilers.
|
||||
```
|
||||
$ make HOST_COMPILER=clang
|
||||
```
|
||||
|
||||
## References (for more details)
|
||||
|
||||
359
Samples/shfl_scan/shfl_integral_image.cuh
Normal file
359
Samples/shfl_scan/shfl_integral_image.cuh
Normal file
|
|
@ -0,0 +1,359 @@
|
|||
/* 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.
|
||||
*/
|
||||
|
||||
// Utility function to extract unsigned chars from an
|
||||
// unsigned integer
|
||||
|
||||
__device__ uchar4 int_to_uchar4(unsigned int in) {
|
||||
uchar4 bytes;
|
||||
bytes.x = in & 0x000000ff >> 0;
|
||||
bytes.y = in & 0x0000ff00 >> 8;
|
||||
bytes.z = in & 0x00ff0000 >> 16;
|
||||
bytes.w = in & 0xff000000 >> 24;
|
||||
return bytes;
|
||||
}
|
||||
|
||||
// This function demonstrates some uses of the shuffle instruction
|
||||
// in the generation of an integral image (also
|
||||
// called a summed area table)
|
||||
// The approach is two pass, a horizontal (scanline) then a vertical
|
||||
// (column) pass.
|
||||
// This is the horizontal pass kernel.
|
||||
__global__ void shfl_intimage_rows(uint4 *img, uint4 *integral_image) {
|
||||
__shared__ int sums[128];
|
||||
|
||||
int id = threadIdx.x;
|
||||
// pointer to head of current scanline
|
||||
uint4 *scanline = &img[blockIdx.x * 120];
|
||||
uint4 data;
|
||||
data = scanline[id];
|
||||
int result[16];
|
||||
int sum;
|
||||
unsigned int lane_id = id % warpSize;
|
||||
int warp_id = threadIdx.x / warpSize;
|
||||
|
||||
uchar4 a = int_to_uchar4(data.x);
|
||||
uchar4 b = int_to_uchar4(data.y);
|
||||
uchar4 c = int_to_uchar4(data.z);
|
||||
uchar4 d = int_to_uchar4(data.w);
|
||||
|
||||
result[0] = a.x;
|
||||
result[1] = a.x + a.y;
|
||||
result[2] = a.x + a.y + a.z;
|
||||
result[3] = a.x + a.y + a.z + a.w;
|
||||
|
||||
result[4] = b.x;
|
||||
result[5] = b.x + b.y;
|
||||
result[6] = b.x + b.y + b.z;
|
||||
result[7] = b.x + b.y + b.z + b.w;
|
||||
|
||||
result[8] = c.x;
|
||||
result[9] = c.x + c.y;
|
||||
result[10] = c.x + c.y + c.z;
|
||||
result[11] = c.x + c.y + c.z + c.w;
|
||||
|
||||
result[12] = d.x;
|
||||
result[13] = d.x + d.y;
|
||||
result[14] = d.x + d.y + d.z;
|
||||
result[15] = d.x + d.y + d.z + d.w;
|
||||
|
||||
#pragma unroll
|
||||
|
||||
for (int i = 4; i <= 7; i++) result[i] += result[3];
|
||||
|
||||
#pragma unroll
|
||||
|
||||
for (int i = 8; i <= 11; i++) result[i] += result[7];
|
||||
|
||||
#pragma unroll
|
||||
|
||||
for (int i = 12; i <= 15; i++) result[i] += result[11];
|
||||
|
||||
sum = result[15];
|
||||
|
||||
// the prefix sum for each thread's 16 value is computed,
|
||||
// now the final sums (result[15]) need to be shared
|
||||
// with the other threads and add. To do this,
|
||||
// the __shfl_up() instruction is used and a shuffle scan
|
||||
// operation is performed to distribute the sums to the correct
|
||||
// threads
|
||||
#pragma unroll
|
||||
|
||||
for (int i = 1; i < 32; i *= 2) {
|
||||
unsigned int mask = 0xffffffff;
|
||||
int n = __shfl_up_sync(mask, sum, i, 32);
|
||||
|
||||
if (lane_id >= i) {
|
||||
#pragma unroll
|
||||
|
||||
for (int i = 0; i < 16; i++) {
|
||||
result[i] += n;
|
||||
}
|
||||
|
||||
sum += n;
|
||||
}
|
||||
}
|
||||
|
||||
// Now the final sum for the warp must be shared
|
||||
// between warps. This is done by each warp
|
||||
// having a thread store to shared memory, then
|
||||
// having some other warp load the values and
|
||||
// compute a prefix sum, again by using __shfl_up.
|
||||
// The results are uniformly added back to the warps.
|
||||
// last thread in the warp holding sum of the warp
|
||||
// places that in shared
|
||||
if (threadIdx.x % warpSize == warpSize - 1) {
|
||||
sums[warp_id] = result[15];
|
||||
}
|
||||
|
||||
__syncthreads();
|
||||
|
||||
if (warp_id == 0) {
|
||||
int warp_sum = sums[lane_id];
|
||||
#pragma unroll
|
||||
|
||||
for (int i = 1; i <= 32; i *= 2) {
|
||||
unsigned int mask = 0xffffffff;
|
||||
int n = __shfl_up_sync(mask, warp_sum, i, 32);
|
||||
|
||||
if (lane_id >= i) warp_sum += n;
|
||||
}
|
||||
|
||||
sums[lane_id] = warp_sum;
|
||||
}
|
||||
|
||||
__syncthreads();
|
||||
|
||||
int blockSum = 0;
|
||||
|
||||
// fold in unused warp
|
||||
if (warp_id > 0) {
|
||||
blockSum = sums[warp_id - 1];
|
||||
#pragma unroll
|
||||
|
||||
for (int i = 0; i < 16; i++) {
|
||||
result[i] += blockSum;
|
||||
}
|
||||
}
|
||||
|
||||
// assemble result
|
||||
// Each thread has 16 values to write, which are
|
||||
// now integer data (to avoid overflow). Instead of
|
||||
// each thread writing consecutive uint4s, the
|
||||
// approach shown here experiments using
|
||||
// the shuffle command to reformat the data
|
||||
// inside the registers so that each thread holds
|
||||
// consecutive data to be written so larger contiguous
|
||||
// segments can be assembled for writing.
|
||||
/*
|
||||
For example data that needs to be written as
|
||||
|
||||
GMEM[16] <- x0 x1 x2 x3 y0 y1 y2 y3 z0 z1 z2 z3 w0 w1 w2 w3
|
||||
but is stored in registers (r0..r3), in four threads (0..3) as:
|
||||
|
||||
threadId 0 1 2 3
|
||||
r0 x0 y0 z0 w0
|
||||
r1 x1 y1 z1 w1
|
||||
r2 x2 y2 z2 w2
|
||||
r3 x3 y3 z3 w3
|
||||
|
||||
after apply __shfl_xor operations to move data between registers r1..r3:
|
||||
|
||||
threadId 00 01 10 11
|
||||
x0 y0 z0 w0
|
||||
xor(01)->y1 x1 w1 z1
|
||||
xor(10)->z2 w2 x2 y2
|
||||
xor(11)->w3 z3 y3 x3
|
||||
|
||||
and now x0..x3, and z0..z3 can be written out in order by all threads.
|
||||
|
||||
In the current code, each register above is actually representing
|
||||
four integers to be written as uint4's to GMEM.
|
||||
*/
|
||||
|
||||
unsigned int mask = 0xffffffff;
|
||||
uint4 output;
|
||||
result[4] = __shfl_xor_sync(mask, result[4], 1, 32);
|
||||
result[5] = __shfl_xor_sync(mask, result[5], 1, 32);
|
||||
result[6] = __shfl_xor_sync(mask, result[6], 1, 32);
|
||||
result[7] = __shfl_xor_sync(mask, result[7], 1, 32);
|
||||
|
||||
result[8] = __shfl_xor_sync(mask, result[8], 2, 32);
|
||||
result[9] = __shfl_xor_sync(mask, result[9], 2, 32);
|
||||
result[10] = __shfl_xor_sync(mask, result[10], 2, 32);
|
||||
result[11] = __shfl_xor_sync(mask, result[11], 2, 32);
|
||||
|
||||
result[12] = __shfl_xor_sync(mask, result[12], 3, 32);
|
||||
result[13] = __shfl_xor_sync(mask, result[13], 3, 32);
|
||||
result[14] = __shfl_xor_sync(mask, result[14], 3, 32);
|
||||
result[15] = __shfl_xor_sync(mask, result[15], 3, 32);
|
||||
|
||||
if (threadIdx.x % 4 == 0) {
|
||||
output = make_uint4(result[0], result[1], result[2], result[3]);
|
||||
}
|
||||
|
||||
if (threadIdx.x % 4 == 1) {
|
||||
output = make_uint4(result[4], result[5], result[6], result[7]);
|
||||
}
|
||||
|
||||
if (threadIdx.x % 4 == 2) {
|
||||
output = make_uint4(result[8], result[9], result[10], result[11]);
|
||||
}
|
||||
|
||||
if (threadIdx.x % 4 == 3) {
|
||||
output = make_uint4(result[12], result[13], result[14], result[15]);
|
||||
}
|
||||
|
||||
integral_image[blockIdx.x * 480 + threadIdx.x % 4 + (threadIdx.x / 4) * 16] =
|
||||
output;
|
||||
|
||||
if (threadIdx.x % 4 == 2) {
|
||||
output = make_uint4(result[0], result[1], result[2], result[3]);
|
||||
}
|
||||
|
||||
if (threadIdx.x % 4 == 3) {
|
||||
output = make_uint4(result[4], result[5], result[6], result[7]);
|
||||
}
|
||||
|
||||
if (threadIdx.x % 4 == 0) {
|
||||
output = make_uint4(result[8], result[9], result[10], result[11]);
|
||||
}
|
||||
|
||||
if (threadIdx.x % 4 == 1) {
|
||||
output = make_uint4(result[12], result[13], result[14], result[15]);
|
||||
}
|
||||
|
||||
integral_image[blockIdx.x * 480 + (threadIdx.x + 2) % 4 +
|
||||
(threadIdx.x / 4) * 16 + 8] = output;
|
||||
// continuing from the above example,
|
||||
// this use of __shfl_xor() places the y0..y3 and w0..w3 data
|
||||
// in order.
|
||||
#pragma unroll
|
||||
|
||||
for (int i = 0; i < 16; i++) {
|
||||
result[i] = __shfl_xor_sync(mask, result[i], 1, 32);
|
||||
}
|
||||
|
||||
if (threadIdx.x % 4 == 0) {
|
||||
output = make_uint4(result[0], result[1], result[2], result[3]);
|
||||
}
|
||||
|
||||
if (threadIdx.x % 4 == 1) {
|
||||
output = make_uint4(result[4], result[5], result[6], result[7]);
|
||||
}
|
||||
|
||||
if (threadIdx.x % 4 == 2) {
|
||||
output = make_uint4(result[8], result[9], result[10], result[11]);
|
||||
}
|
||||
|
||||
if (threadIdx.x % 4 == 3) {
|
||||
output = make_uint4(result[12], result[13], result[14], result[15]);
|
||||
}
|
||||
|
||||
integral_image[blockIdx.x * 480 + threadIdx.x % 4 + (threadIdx.x / 4) * 16 +
|
||||
4] = output;
|
||||
|
||||
if (threadIdx.x % 4 == 2) {
|
||||
output = make_uint4(result[0], result[1], result[2], result[3]);
|
||||
}
|
||||
|
||||
if (threadIdx.x % 4 == 3) {
|
||||
output = make_uint4(result[4], result[5], result[6], result[7]);
|
||||
}
|
||||
|
||||
if (threadIdx.x % 4 == 0) {
|
||||
output = make_uint4(result[8], result[9], result[10], result[11]);
|
||||
}
|
||||
|
||||
if (threadIdx.x % 4 == 1) {
|
||||
output = make_uint4(result[12], result[13], result[14], result[15]);
|
||||
}
|
||||
|
||||
integral_image[blockIdx.x * 480 + (threadIdx.x + 2) % 4 +
|
||||
(threadIdx.x / 4) * 16 + 12] = output;
|
||||
}
|
||||
|
||||
// This kernel computes columnwise prefix sums. When the data input is
|
||||
// the row sums from above, this completes the integral image.
|
||||
// The approach here is to have each block compute a local set of sums.
|
||||
// First , the data covered by the block is loaded into shared memory,
|
||||
// then instead of performing a sum in shared memory using __syncthreads
|
||||
// between stages, the data is reformatted so that the necessary sums
|
||||
// occur inside warps and the shuffle scan operation is used.
|
||||
// The final set of sums from the block is then propagated, with the block
|
||||
// computing "down" the image and adding the running sum to the local
|
||||
// block sums.
|
||||
__global__ void shfl_vertical_shfl(unsigned int *img, int width, int height) {
|
||||
__shared__ unsigned int sums[32][9];
|
||||
int tidx = blockIdx.x * blockDim.x + threadIdx.x;
|
||||
// int warp_id = threadIdx.x / warpSize ;
|
||||
unsigned int lane_id = tidx % 8;
|
||||
// int rows_per_thread = (height / blockDim. y) ;
|
||||
// int start_row = rows_per_thread * threadIdx.y;
|
||||
unsigned int stepSum = 0;
|
||||
unsigned int mask = 0xffffffff;
|
||||
|
||||
sums[threadIdx.x][threadIdx.y] = 0;
|
||||
__syncthreads();
|
||||
|
||||
for (int step = 0; step < 135; step++) {
|
||||
unsigned int sum = 0;
|
||||
unsigned int *p = img + (threadIdx.y + step * 8) * width + tidx;
|
||||
|
||||
sum = *p;
|
||||
sums[threadIdx.x][threadIdx.y] = sum;
|
||||
__syncthreads();
|
||||
|
||||
// place into SMEM
|
||||
// shfl scan reduce the SMEM, reformating so the column
|
||||
// sums are computed in a warp
|
||||
// then read out properly
|
||||
int partial_sum = 0;
|
||||
int j = threadIdx.x % 8;
|
||||
int k = threadIdx.x / 8 + threadIdx.y * 4;
|
||||
|
||||
partial_sum = sums[k][j];
|
||||
|
||||
for (int i = 1; i <= 8; i *= 2) {
|
||||
int n = __shfl_up_sync(mask, partial_sum, i, 32);
|
||||
|
||||
if (lane_id >= i) partial_sum += n;
|
||||
}
|
||||
|
||||
sums[k][j] = partial_sum;
|
||||
__syncthreads();
|
||||
|
||||
if (threadIdx.y > 0) {
|
||||
sum += sums[threadIdx.x][threadIdx.y - 1];
|
||||
}
|
||||
|
||||
sum += stepSum;
|
||||
stepSum += sums[threadIdx.x][blockDim.y - 1];
|
||||
__syncthreads();
|
||||
*p = sum;
|
||||
}
|
||||
}
|
||||
418
Samples/shfl_scan/shfl_scan.cu
Normal file
418
Samples/shfl_scan/shfl_scan.cu
Normal file
|
|
@ -0,0 +1,418 @@
|
|||
/* 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.
|
||||
*/
|
||||
|
||||
// Shuffle intrinsics CUDA Sample
|
||||
// This sample demonstrates the use of the shuffle intrinsic
|
||||
// First, a simple example of a prefix sum using the shuffle to
|
||||
// perform a scan operation is provided.
|
||||
// Secondly, a more involved example of computing an integral image
|
||||
// using the shuffle intrinsic is provided, where the shuffle
|
||||
// scan operation and shuffle xor operations are used
|
||||
|
||||
#include <stdio.h>
|
||||
|
||||
#include <cuda_runtime.h>
|
||||
|
||||
#include <helper_cuda.h>
|
||||
#include <helper_functions.h>
|
||||
#include "shfl_integral_image.cuh"
|
||||
|
||||
// Scan using shfl - takes log2(n) steps
|
||||
// This function demonstrates basic use of the shuffle intrinsic, __shfl_up,
|
||||
// to perform a scan operation across a block.
|
||||
// First, it performs a scan (prefix sum in this case) inside a warp
|
||||
// Then to continue the scan operation across the block,
|
||||
// each warp's sum is placed into shared memory. A single warp
|
||||
// then performs a shuffle scan on that shared memory. The results
|
||||
// are then uniformly added to each warp's threads.
|
||||
// This pyramid type approach is continued by placing each block's
|
||||
// final sum in global memory and prefix summing that via another kernel call,
|
||||
// then uniformly adding across the input data via the uniform_add<<<>>> kernel.
|
||||
|
||||
__global__ void shfl_scan_test(int *data, int width, int *partial_sums = NULL) {
|
||||
extern __shared__ int sums[];
|
||||
int id = ((blockIdx.x * blockDim.x) + threadIdx.x);
|
||||
int lane_id = id % warpSize;
|
||||
// determine a warp_id within a block
|
||||
int warp_id = threadIdx.x / warpSize;
|
||||
|
||||
// Below is the basic structure of using a shfl instruction
|
||||
// for a scan.
|
||||
// Record "value" as a variable - we accumulate it along the way
|
||||
int value = data[id];
|
||||
|
||||
// Now accumulate in log steps up the chain
|
||||
// compute sums, with another thread's value who is
|
||||
// distance delta away (i). Note
|
||||
// those threads where the thread 'i' away would have
|
||||
// been out of bounds of the warp are unaffected. This
|
||||
// creates the scan sum.
|
||||
|
||||
#pragma unroll
|
||||
for (int i = 1; i <= width; i *= 2) {
|
||||
unsigned int mask = 0xffffffff;
|
||||
int n = __shfl_up_sync(mask, value, i, width);
|
||||
|
||||
if (lane_id >= i) value += n;
|
||||
}
|
||||
|
||||
// value now holds the scan value for the individual thread
|
||||
// next sum the largest values for each warp
|
||||
|
||||
// write the sum of the warp to smem
|
||||
if (threadIdx.x % warpSize == warpSize - 1) {
|
||||
sums[warp_id] = value;
|
||||
}
|
||||
|
||||
__syncthreads();
|
||||
|
||||
//
|
||||
// scan sum the warp sums
|
||||
// the same shfl scan operation, but performed on warp sums
|
||||
//
|
||||
if (warp_id == 0 && lane_id < (blockDim.x / warpSize)) {
|
||||
int warp_sum = sums[lane_id];
|
||||
|
||||
int mask = (1 << (blockDim.x / warpSize)) - 1;
|
||||
for (int i = 1; i <= (blockDim.x / warpSize); i *= 2) {
|
||||
int n = __shfl_up_sync(mask, warp_sum, i, (blockDim.x / warpSize));
|
||||
|
||||
if (lane_id >= i) warp_sum += n;
|
||||
}
|
||||
|
||||
sums[lane_id] = warp_sum;
|
||||
}
|
||||
|
||||
__syncthreads();
|
||||
|
||||
// perform a uniform add across warps in the block
|
||||
// read neighbouring warp's sum and add it to threads value
|
||||
int blockSum = 0;
|
||||
|
||||
if (warp_id > 0) {
|
||||
blockSum = sums[warp_id - 1];
|
||||
}
|
||||
|
||||
value += blockSum;
|
||||
|
||||
// Now write out our result
|
||||
data[id] = value;
|
||||
|
||||
// last thread has sum, write write out the block's sum
|
||||
if (partial_sums != NULL && threadIdx.x == blockDim.x - 1) {
|
||||
partial_sums[blockIdx.x] = value;
|
||||
}
|
||||
}
|
||||
|
||||
// Uniform add: add partial sums array
|
||||
__global__ void uniform_add(int *data, int *partial_sums, int len) {
|
||||
__shared__ int buf;
|
||||
int id = ((blockIdx.x * blockDim.x) + threadIdx.x);
|
||||
|
||||
if (id > len) return;
|
||||
|
||||
if (threadIdx.x == 0) {
|
||||
buf = partial_sums[blockIdx.x];
|
||||
}
|
||||
|
||||
__syncthreads();
|
||||
data[id] += buf;
|
||||
}
|
||||
|
||||
static unsigned int iDivUp(unsigned int dividend, unsigned int divisor) {
|
||||
return ((dividend % divisor) == 0) ? (dividend / divisor)
|
||||
: (dividend / divisor + 1);
|
||||
}
|
||||
|
||||
// This function verifies the shuffle scan result, for the simple
|
||||
// prefix sum case.
|
||||
bool CPUverify(int *h_data, int *h_result, int n_elements) {
|
||||
// cpu verify
|
||||
for (int i = 0; i < n_elements - 1; i++) {
|
||||
h_data[i + 1] = h_data[i] + h_data[i + 1];
|
||||
}
|
||||
|
||||
int diff = 0;
|
||||
|
||||
for (int i = 0; i < n_elements; i++) {
|
||||
diff += h_data[i] - h_result[i];
|
||||
}
|
||||
|
||||
printf("CPU verify result diff (GPUvsCPU) = %d\n", diff);
|
||||
bool bTestResult = false;
|
||||
|
||||
if (diff == 0) bTestResult = true;
|
||||
|
||||
StopWatchInterface *hTimer = NULL;
|
||||
sdkCreateTimer(&hTimer);
|
||||
sdkResetTimer(&hTimer);
|
||||
sdkStartTimer(&hTimer);
|
||||
|
||||
for (int j = 0; j < 100; j++)
|
||||
for (int i = 0; i < n_elements - 1; i++) {
|
||||
h_data[i + 1] = h_data[i] + h_data[i + 1];
|
||||
}
|
||||
|
||||
sdkStopTimer(&hTimer);
|
||||
double cput = sdkGetTimerValue(&hTimer);
|
||||
printf("CPU sum (naive) took %f ms\n", cput / 100);
|
||||
return bTestResult;
|
||||
}
|
||||
|
||||
// this verifies the row scan result for synthetic data of all 1's
|
||||
unsigned int verifyDataRowSums(unsigned int *h_image, int w, int h) {
|
||||
unsigned int diff = 0;
|
||||
|
||||
for (int j = 0; j < h; j++) {
|
||||
for (int i = 0; i < w; i++) {
|
||||
int gold = i + 1;
|
||||
diff +=
|
||||
abs(static_cast<int>(gold) - static_cast<int>(h_image[j * w + i]));
|
||||
}
|
||||
}
|
||||
|
||||
return diff;
|
||||
}
|
||||
|
||||
bool shuffle_simple_test(int argc, char **argv) {
|
||||
int *h_data, *h_partial_sums, *h_result;
|
||||
int *d_data, *d_partial_sums;
|
||||
const int n_elements = 65536;
|
||||
int sz = sizeof(int) * n_elements;
|
||||
int cuda_device = 0;
|
||||
|
||||
printf("Starting shfl_scan\n");
|
||||
|
||||
// use command-line specified CUDA device, otherwise use device with highest
|
||||
// Gflops/s
|
||||
cuda_device = findCudaDevice(argc, (const char **)argv);
|
||||
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(cudaGetDevice(&cuda_device));
|
||||
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, cuda_device));
|
||||
|
||||
printf("> Detected Compute SM %d.%d hardware with %d multi-processors\n",
|
||||
deviceProp.major, deviceProp.minor, deviceProp.multiProcessorCount);
|
||||
|
||||
// __shfl intrinsic needs SM 3.0 or higher
|
||||
if (deviceProp.major < 3) {
|
||||
printf("> __shfl() intrinsic requires device SM 3.0+\n");
|
||||
printf("> Waiving test.\n");
|
||||
exit(EXIT_WAIVED);
|
||||
}
|
||||
|
||||
checkCudaErrors(cudaMallocHost(reinterpret_cast<void **>(&h_data),
|
||||
sizeof(int) * n_elements));
|
||||
checkCudaErrors(cudaMallocHost(reinterpret_cast<void **>(&h_result),
|
||||
sizeof(int) * n_elements));
|
||||
|
||||
// initialize data:
|
||||
printf("Computing Simple Sum test\n");
|
||||
printf("---------------------------------------------------\n");
|
||||
|
||||
printf("Initialize test data [1, 1, 1...]\n");
|
||||
|
||||
for (int i = 0; i < n_elements; i++) {
|
||||
h_data[i] = 1;
|
||||
}
|
||||
|
||||
int blockSize = 256;
|
||||
int gridSize = n_elements / blockSize;
|
||||
int nWarps = blockSize / 32;
|
||||
int shmem_sz = nWarps * sizeof(int);
|
||||
int n_partialSums = n_elements / blockSize;
|
||||
int partial_sz = n_partialSums * sizeof(int);
|
||||
|
||||
printf("Scan summation for %d elements, %d partial sums\n", n_elements,
|
||||
n_elements / blockSize);
|
||||
|
||||
int p_blockSize = min(n_partialSums, blockSize);
|
||||
int p_gridSize = iDivUp(n_partialSums, p_blockSize);
|
||||
printf("Partial summing %d elements with %d blocks of size %d\n",
|
||||
n_partialSums, p_gridSize, p_blockSize);
|
||||
|
||||
// initialize a timer
|
||||
cudaEvent_t start, stop;
|
||||
checkCudaErrors(cudaEventCreate(&start));
|
||||
checkCudaErrors(cudaEventCreate(&stop));
|
||||
float et = 0;
|
||||
float inc = 0;
|
||||
|
||||
checkCudaErrors(cudaMalloc(reinterpret_cast<void **>(&d_data), sz));
|
||||
checkCudaErrors(
|
||||
cudaMalloc(reinterpret_cast<void **>(&d_partial_sums), partial_sz));
|
||||
checkCudaErrors(cudaMemset(d_partial_sums, 0, partial_sz));
|
||||
|
||||
checkCudaErrors(
|
||||
cudaMallocHost(reinterpret_cast<void **>(&h_partial_sums), partial_sz));
|
||||
checkCudaErrors(cudaMemcpy(d_data, h_data, sz, cudaMemcpyHostToDevice));
|
||||
|
||||
checkCudaErrors(cudaEventRecord(start, 0));
|
||||
shfl_scan_test<<<gridSize, blockSize, shmem_sz>>>(d_data, 32, d_partial_sums);
|
||||
shfl_scan_test<<<p_gridSize, p_blockSize, shmem_sz>>>(d_partial_sums, 32);
|
||||
uniform_add<<<gridSize - 1, blockSize>>>(d_data + blockSize, d_partial_sums,
|
||||
n_elements);
|
||||
checkCudaErrors(cudaEventRecord(stop, 0));
|
||||
checkCudaErrors(cudaEventSynchronize(stop));
|
||||
checkCudaErrors(cudaEventElapsedTime(&inc, start, stop));
|
||||
et += inc;
|
||||
|
||||
checkCudaErrors(cudaMemcpy(h_result, d_data, sz, cudaMemcpyDeviceToHost));
|
||||
checkCudaErrors(cudaMemcpy(h_partial_sums, d_partial_sums, partial_sz,
|
||||
cudaMemcpyDeviceToHost));
|
||||
|
||||
printf("Test Sum: %d\n", h_partial_sums[n_partialSums - 1]);
|
||||
printf("Time (ms): %f\n", et);
|
||||
printf("%d elements scanned in %f ms -> %f MegaElements/s\n", n_elements, et,
|
||||
n_elements / (et / 1000.0f) / 1000000.0f);
|
||||
|
||||
bool bTestResult = CPUverify(h_data, h_result, n_elements);
|
||||
|
||||
checkCudaErrors(cudaFreeHost(h_data));
|
||||
checkCudaErrors(cudaFreeHost(h_result));
|
||||
checkCudaErrors(cudaFreeHost(h_partial_sums));
|
||||
checkCudaErrors(cudaFree(d_data));
|
||||
checkCudaErrors(cudaFree(d_partial_sums));
|
||||
|
||||
return bTestResult;
|
||||
}
|
||||
|
||||
// This function tests creation of an integral image using
|
||||
// synthetic data, of size 1920x1080 pixels greyscale.
|
||||
bool shuffle_integral_image_test() {
|
||||
char *d_data;
|
||||
unsigned int *h_image;
|
||||
unsigned int *d_integral_image;
|
||||
int w = 1920;
|
||||
int h = 1080;
|
||||
int n_elements = w * h;
|
||||
int sz = sizeof(unsigned int) * n_elements;
|
||||
|
||||
printf("\nComputing Integral Image Test on size %d x %d synthetic data\n", w,
|
||||
h);
|
||||
printf("---------------------------------------------------\n");
|
||||
checkCudaErrors(cudaMallocHost(reinterpret_cast<void **>(&h_image), sz));
|
||||
// fill test "image" with synthetic 1's data
|
||||
memset(h_image, 0, sz);
|
||||
|
||||
// each thread handles 16 values, use 1 block/row
|
||||
int blockSize = iDivUp(w, 16);
|
||||
// launch 1 block / row
|
||||
int gridSize = h;
|
||||
|
||||
// Create a synthetic image for testing
|
||||
checkCudaErrors(cudaMalloc(reinterpret_cast<void **>(&d_data), sz));
|
||||
checkCudaErrors(cudaMalloc(reinterpret_cast<void **>(&d_integral_image),
|
||||
n_elements * sizeof(int) * 4));
|
||||
checkCudaErrors(cudaMemset(d_data, 1, sz));
|
||||
checkCudaErrors(cudaMemset(d_integral_image, 0, sz));
|
||||
|
||||
cudaEvent_t start, stop;
|
||||
cudaEventCreate(&start);
|
||||
cudaEventCreate(&stop);
|
||||
float et = 0;
|
||||
unsigned int err;
|
||||
|
||||
// Execute scan line prefix sum kernel, and time it
|
||||
cudaEventRecord(start);
|
||||
shfl_intimage_rows<<<gridSize, blockSize>>>(
|
||||
reinterpret_cast<uint4 *>(d_data),
|
||||
reinterpret_cast<uint4 *>(d_integral_image));
|
||||
cudaEventRecord(stop);
|
||||
checkCudaErrors(cudaEventSynchronize(stop));
|
||||
checkCudaErrors(cudaEventElapsedTime(&et, start, stop));
|
||||
printf("Method: Fast Time (GPU Timer): %f ms ", et);
|
||||
|
||||
// verify the scan line results
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(h_image, d_integral_image, sz, cudaMemcpyDeviceToHost));
|
||||
err = verifyDataRowSums(h_image, w, h);
|
||||
printf("Diff = %d\n", err);
|
||||
|
||||
// Execute column prefix sum kernel and time it
|
||||
dim3 blockSz(32, 8);
|
||||
dim3 testGrid(w / blockSz.x, 1);
|
||||
|
||||
cudaEventRecord(start);
|
||||
shfl_vertical_shfl<<<testGrid, blockSz>>>((unsigned int *)d_integral_image, w,
|
||||
h);
|
||||
cudaEventRecord(stop);
|
||||
checkCudaErrors(cudaEventSynchronize(stop));
|
||||
checkCudaErrors(cudaEventElapsedTime(&et, start, stop));
|
||||
printf("Method: Vertical Scan Time (GPU Timer): %f ms ", et);
|
||||
|
||||
// Verify the column results
|
||||
checkCudaErrors(
|
||||
cudaMemcpy(h_image, d_integral_image, sz, cudaMemcpyDeviceToHost));
|
||||
printf("\n");
|
||||
|
||||
int finalSum = h_image[w * h - 1];
|
||||
printf("CheckSum: %d, (expect %dx%d=%d)\n", finalSum, w, h, w * h);
|
||||
|
||||
checkCudaErrors(cudaFree(d_data));
|
||||
checkCudaErrors(cudaFree(d_integral_image));
|
||||
checkCudaErrors(cudaFreeHost(h_image));
|
||||
// verify final sum: if the final value in the corner is the same as the size
|
||||
// of the buffer (all 1's) then the integral image was generated successfully
|
||||
return (finalSum == w * h) ? true : false;
|
||||
}
|
||||
|
||||
int main(int argc, char *argv[]) {
|
||||
// Initialization. The shuffle intrinsic is not available on SM < 3.0
|
||||
// so waive the test if the hardware is not present.
|
||||
int cuda_device = 0;
|
||||
|
||||
printf("Starting shfl_scan\n");
|
||||
|
||||
// use command-line specified CUDA device, otherwise use device with highest
|
||||
// Gflops/s
|
||||
cuda_device = findCudaDevice(argc, (const char **)argv);
|
||||
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(cudaGetDevice(&cuda_device));
|
||||
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, cuda_device));
|
||||
|
||||
printf("> Detected Compute SM %d.%d hardware with %d multi-processors\n",
|
||||
deviceProp.major, deviceProp.minor, deviceProp.multiProcessorCount);
|
||||
|
||||
// __shfl intrinsic needs SM 3.0 or higher
|
||||
if (deviceProp.major < 3) {
|
||||
printf("> __shfl() intrinsic requires device SM 3.0+\n");
|
||||
printf("> Waiving test.\n");
|
||||
exit(EXIT_WAIVED);
|
||||
}
|
||||
|
||||
bool bTestResult = true;
|
||||
bool simpleTest = shuffle_simple_test(argc, argv);
|
||||
bool intTest = shuffle_integral_image_test();
|
||||
|
||||
bTestResult = simpleTest & intTest;
|
||||
|
||||
exit((bTestResult) ? EXIT_SUCCESS : EXIT_FAILURE);
|
||||
}
|
||||
20
Samples/shfl_scan/shfl_scan_vs2010.sln
Normal file
20
Samples/shfl_scan/shfl_scan_vs2010.sln
Normal file
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 11.00
|
||||
# Visual Studio 2010
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "shfl_scan", "shfl_scan_vs2010.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
|
||||
107
Samples/shfl_scan/shfl_scan_vs2010.vcxproj
Normal file
107
Samples/shfl_scan/shfl_scan_vs2010.vcxproj
Normal file
|
|
@ -0,0 +1,107 @@
|
|||
<?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>shfl_scan_vs2010</RootNamespace>
|
||||
<ProjectName>shfl_scan</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(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="$(CUDAPropsPath)\CUDA 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/shfl_scan.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(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="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="shfl_scan.cu" />
|
||||
<ClInclude Include="util.h" />
|
||||
<None Include="shfl_integral_image.cuh" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/shfl_scan/shfl_scan_vs2012.sln
Normal file
20
Samples/shfl_scan/shfl_scan_vs2012.sln
Normal file
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2012
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "shfl_scan", "shfl_scan_vs2012.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
108
Samples/shfl_scan/shfl_scan_vs2012.vcxproj
Normal file
108
Samples/shfl_scan/shfl_scan_vs2012.vcxproj
Normal file
|
|
@ -0,0 +1,108 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>shfl_scan_vs2012</RootNamespace>
|
||||
<ProjectName>shfl_scan</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v110</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 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/shfl_scan.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</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>
|
||||
<CudaCompile Include="shfl_scan.cu" />
|
||||
<ClInclude Include="util.h" />
|
||||
<None Include="shfl_integral_image.cuh" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/shfl_scan/shfl_scan_vs2013.sln
Normal file
20
Samples/shfl_scan/shfl_scan_vs2013.sln
Normal file
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 13.00
|
||||
# Visual Studio 2013
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "shfl_scan", "shfl_scan_vs2013.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
|
||||
108
Samples/shfl_scan/shfl_scan_vs2013.vcxproj
Normal file
108
Samples/shfl_scan/shfl_scan_vs2013.vcxproj
Normal file
|
|
@ -0,0 +1,108 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>shfl_scan_vs2013</RootNamespace>
|
||||
<ProjectName>shfl_scan</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v120</PlatformToolset>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(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="$(CUDAPropsPath)\CUDA 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/shfl_scan.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(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="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="shfl_scan.cu" />
|
||||
<ClInclude Include="util.h" />
|
||||
<None Include="shfl_integral_image.cuh" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/shfl_scan/shfl_scan_vs2015.sln
Normal file
20
Samples/shfl_scan/shfl_scan_vs2015.sln
Normal file
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 14.00
|
||||
# Visual Studio 2015
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "shfl_scan", "shfl_scan_vs2015.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
108
Samples/shfl_scan/shfl_scan_vs2015.vcxproj
Normal file
108
Samples/shfl_scan/shfl_scan_vs2015.vcxproj
Normal file
|
|
@ -0,0 +1,108 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>shfl_scan_vs2015</RootNamespace>
|
||||
<ProjectName>shfl_scan</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v140</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 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/shfl_scan.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</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>
|
||||
<CudaCompile Include="shfl_scan.cu" />
|
||||
<ClInclude Include="util.h" />
|
||||
<None Include="shfl_integral_image.cuh" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/shfl_scan/shfl_scan_vs2017.sln
Normal file
20
Samples/shfl_scan/shfl_scan_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}") = "shfl_scan", "shfl_scan_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
|
||||
109
Samples/shfl_scan/shfl_scan_vs2017.vcxproj
Normal file
109
Samples/shfl_scan/shfl_scan_vs2017.vcxproj
Normal file
|
|
@ -0,0 +1,109 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>shfl_scan_vs2017</RootNamespace>
|
||||
<ProjectName>shfl_scan</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v141</PlatformToolset>
|
||||
<WindowsTargetPlatformVersion>10.0.15063.0</WindowsTargetPlatformVersion>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(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="$(CUDAPropsPath)\CUDA 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/shfl_scan.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(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="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="shfl_scan.cu" />
|
||||
<ClInclude Include="util.h" />
|
||||
<None Include="shfl_integral_image.cuh" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
61
Samples/shfl_scan/util.h
Normal file
61
Samples/shfl_scan/util.h
Normal file
|
|
@ -0,0 +1,61 @@
|
|||
/* 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.
|
||||
*/
|
||||
|
||||
#ifndef SAMPLES_SHFL_SCAN_UTIL_H_
|
||||
#define SAMPLES_SHFL_SCAN_UTIL_H_
|
||||
|
||||
// Macro to catch CUDA errors in kernel launches
|
||||
#define CHECK_LAUNCH_ERROR() \
|
||||
do { \
|
||||
/* Check synchronous errors, i.e. pre-launch */ \
|
||||
cudaError_t err = cudaGetLastError(); \
|
||||
if (cudaSuccess != err) { \
|
||||
fprintf(stderr, "Cuda error in file '%s' in line %i : %s.\n", __FILE__, \
|
||||
__LINE__, cudaGetErrorString(err)); \
|
||||
exit(EXIT_FAILURE); \
|
||||
} \
|
||||
/* Check asynchronous errors, i.e. kernel failed (ULF) */ \
|
||||
err = cudaDeviceSynchronize(); \
|
||||
if (cudaSuccess != err) { \
|
||||
fprintf(stderr, "Cuda error in file '%s' in line %i : %s!\n", __FILE__, \
|
||||
__LINE__, cudaGetErrorString(err)); \
|
||||
exit(EXIT_FAILURE); \
|
||||
} \
|
||||
} while (0)
|
||||
|
||||
// Macro to catch CUDA errors in CUDA runtime calls
|
||||
#define CUDA_CHECK(call) \
|
||||
do { \
|
||||
cudaError_t err = call; \
|
||||
if (cudaSuccess != err) { \
|
||||
fprintf(stderr, "Cuda error in file '%s' in line %i : %s.\n", __FILE__, \
|
||||
__LINE__, cudaGetErrorString(err)); \
|
||||
exit(EXIT_FAILURE); \
|
||||
} \
|
||||
} while (0)
|
||||
|
||||
#endif // SAMPLES_SHFL_SCAN_UTIL_H_
|
||||
|
|
@ -180,6 +180,21 @@ ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
|||
LDFLAGS += -rpath-link=$(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)
|
||||
|
|
|
|||
|
|
@ -2,7 +2,7 @@
|
|||
<!DOCTYPE entry SYSTEM "SamplesInfo.dtd">
|
||||
<entry>
|
||||
<name>simpleCUBLAS</name>
|
||||
<description><![CDATA[Example of using CUBLAS using the new CUBLAS API interface available in CUDA 4.0.]]></description>
|
||||
<description><![CDATA[Example of using CUBLAS API interface to perform GEMM operations.]]></description>
|
||||
<devicecompilation>whole</devicecompilation>
|
||||
<fallback_min_ptx>true</fallback_min_ptx>
|
||||
<includepaths>
|
||||
|
|
|
|||
|
|
@ -2,7 +2,7 @@
|
|||
|
||||
## Description
|
||||
|
||||
Example of using CUBLAS using the new CUBLAS API interface available in CUDA 4.0.
|
||||
Example of using CUBLAS API interface to perform GEMM operations.
|
||||
|
||||
## Key Concepts
|
||||
|
||||
|
|
|
|||
|
|
@ -180,6 +180,21 @@ ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
|||
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)
|
||||
|
|
|
|||
302
Samples/simpleVoteIntrinsics/Makefile
Normal file
302
Samples/simpleVoteIntrinsics/Makefile
Normal file
|
|
@ -0,0 +1,302 @@
|
|||
################################################################################
|
||||
#
|
||||
# Copyright 1993-2015 NVIDIA Corporation. All rights reserved.
|
||||
#
|
||||
# NOTICE TO USER:
|
||||
#
|
||||
# This source code is subject to NVIDIA ownership rights under U.S. and
|
||||
# international Copyright laws.
|
||||
#
|
||||
# NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOURCE
|
||||
# CODE FOR ANY PURPOSE. IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR
|
||||
# IMPLIED WARRANTY OF ANY KIND. NVIDIA DISCLAIMS ALL WARRANTIES WITH
|
||||
# REGARD TO THIS SOURCE CODE, INCLUDING ALL IMPLIED WARRANTIES OF
|
||||
# MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
|
||||
# IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL,
|
||||
# OR CONSEQUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS
|
||||
# OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
|
||||
# OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE
|
||||
# OR PERFORMANCE OF THIS SOURCE CODE.
|
||||
#
|
||||
# U.S. Government End Users. This source code is a "commercial item" as
|
||||
# that term is defined at 48 C.F.R. 2.101 (OCT 1995), consisting of
|
||||
# "commercial computer software" and "commercial computer software
|
||||
# documentation" as such terms are used in 48 C.F.R. 12.212 (SEPT 1995)
|
||||
# and is provided to the U.S. Government only as a commercial end item.
|
||||
# Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through
|
||||
# 227.7202-4 (JUNE 1995), all U.S. Government End Users acquire the
|
||||
# source code with only those rights set forth herein.
|
||||
#
|
||||
################################################################################
|
||||
#
|
||||
# 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-g++
|
||||
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
|
||||
|
||||
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 ?= 30 35 37 50 52 60 61 70
|
||||
|
||||
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 ($(SAMPLE_ENABLED),0)
|
||||
EXEC ?= @echo "[@]"
|
||||
endif
|
||||
|
||||
################################################################################
|
||||
|
||||
# Target rules
|
||||
all: build
|
||||
|
||||
build: simpleVoteIntrinsics
|
||||
|
||||
check.deps:
|
||||
ifeq ($(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
|
||||
|
||||
simpleVoteIntrinsics.o:simpleVoteIntrinsics.cu
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
simpleVoteIntrinsics: simpleVoteIntrinsics.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
|
||||
$(EXEC) ./simpleVoteIntrinsics
|
||||
|
||||
clean:
|
||||
rm -f simpleVoteIntrinsics simpleVoteIntrinsics.o
|
||||
rm -rf ../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)/simpleVoteIntrinsics
|
||||
|
||||
clobber: clean
|
||||
71
Samples/simpleVoteIntrinsics/NsightEclipse.xml
Normal file
71
Samples/simpleVoteIntrinsics/NsightEclipse.xml
Normal file
|
|
@ -0,0 +1,71 @@
|
|||
<?xml version="1.0" encoding="UTF-8"?>
|
||||
<!DOCTYPE entry SYSTEM "SamplesInfo.dtd">
|
||||
<entry>
|
||||
<name>simpleVoteIntrinsics</name>
|
||||
<cuda_api_list>
|
||||
<toolkit>cudaMalloc</toolkit>
|
||||
<toolkit>cudaFree</toolkit>
|
||||
<toolkit>cudaMemcpy</toolkit>
|
||||
<toolkit>cudaFreeHost</toolkit>
|
||||
</cuda_api_list>
|
||||
<description><![CDATA[Simple program which demonstrates how to use the Vote (__any_sync, __all_sync) intrinsic instruction in a CUDA kernel.]]></description>
|
||||
<devicecompilation>whole</devicecompilation>
|
||||
<includepaths>
|
||||
<path>./</path>
|
||||
<path>../</path>
|
||||
<path>../../common/inc</path>
|
||||
</includepaths>
|
||||
<keyconcepts>
|
||||
<concept level="basic">Vote Intrinsics</concept>
|
||||
</keyconcepts>
|
||||
<keywords>
|
||||
<keyword>CUDA</keyword>
|
||||
<keyword>GPGPU</keyword>
|
||||
<keyword>vote</keyword>
|
||||
</keywords>
|
||||
<libraries>
|
||||
</libraries>
|
||||
<librarypaths>
|
||||
</librarypaths>
|
||||
<nsight_eclipse>true</nsight_eclipse>
|
||||
<primary_file>simpleVoteIntrinsics.cu</primary_file>
|
||||
<scopes>
|
||||
<scope>1:CUDA Basic Topics</scope>
|
||||
</scopes>
|
||||
<sm-arch>sm30</sm-arch>
|
||||
<sm-arch>sm35</sm-arch>
|
||||
<sm-arch>sm37</sm-arch>
|
||||
<sm-arch>sm50</sm-arch>
|
||||
<sm-arch>sm52</sm-arch>
|
||||
<sm-arch>sm60</sm-arch>
|
||||
<sm-arch>sm61</sm-arch>
|
||||
<sm-arch>sm70</sm-arch>
|
||||
<supported_envs>
|
||||
<env>
|
||||
<arch>x86_64</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
<env>
|
||||
<platform>windows7</platform>
|
||||
</env>
|
||||
<env>
|
||||
<arch>x86_64</arch>
|
||||
<platform>macosx</platform>
|
||||
</env>
|
||||
<env>
|
||||
<arch>arm</arch>
|
||||
</env>
|
||||
<env>
|
||||
<arch>aarch64</arch>
|
||||
</env>
|
||||
<env>
|
||||
<arch>ppc64le</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
</supported_envs>
|
||||
<supported_sm_architectures>
|
||||
<include>all</include>
|
||||
</supported_sm_architectures>
|
||||
<title>Simple Vote Intrinsics</title>
|
||||
<type>exe</type>
|
||||
</entry>
|
||||
94
Samples/simpleVoteIntrinsics/README.md
Normal file
94
Samples/simpleVoteIntrinsics/README.md
Normal file
|
|
@ -0,0 +1,94 @@
|
|||
# simpleVoteIntrinsics - Simple Vote Intrinsics
|
||||
|
||||
## Description
|
||||
|
||||
Simple program which demonstrates how to use the Vote (__any_sync, __all_sync) intrinsic instruction in a CUDA kernel.
|
||||
|
||||
## Key Concepts
|
||||
|
||||
Vote Intrinsics
|
||||
|
||||
## Supported SM Architectures
|
||||
|
||||
[SM 3.0 ](https://developer.nvidia.com/cuda-gpus) [SM 3.5 ](https://developer.nvidia.com/cuda-gpus) [SM 3.7 ](https://developer.nvidia.com/cuda-gpus) [SM 5.0 ](https://developer.nvidia.com/cuda-gpus) [SM 5.2 ](https://developer.nvidia.com/cuda-gpus) [SM 6.0 ](https://developer.nvidia.com/cuda-gpus) [SM 6.1 ](https://developer.nvidia.com/cuda-gpus) [SM 7.0 ](https://developer.nvidia.com/cuda-gpus)
|
||||
|
||||
## Supported OSes
|
||||
|
||||
Linux, Windows, MacOSX
|
||||
|
||||
## Supported CPU Architecture
|
||||
|
||||
x86_64, ppc64le, armv7l, aarch64
|
||||
|
||||
## CUDA APIs involved
|
||||
|
||||
### [CUDA Runtime API](http://docs.nvidia.com/cuda/cuda-runtime-api/index.html)
|
||||
cudaMalloc, cudaFree, cudaMemcpy, cudaFreeHost
|
||||
|
||||
## Prerequisites
|
||||
|
||||
Download and install the [CUDA Toolkit 9.2](https://developer.nvidia.com/cuda-downloads) for your corresponding platform.
|
||||
|
||||
## Build and Run
|
||||
|
||||
### Windows
|
||||
The Windows samples are built using the Visual Studio IDE. Solution files (.sln) are provided for each supported version of Visual Studio, using the format:
|
||||
```
|
||||
*_vs<version>.sln - for Visual Studio <version>
|
||||
```
|
||||
Each individual sample has its own set of solution files in its directory:
|
||||
|
||||
To build/examine all the samples at once, the complete solution files should be used. To build/examine a single sample, the individual sample solution files should be used.
|
||||
> **Note:** Some samples require that the Microsoft DirectX SDK (June 2010 or newer) be installed and that the VC++ directory paths are properly set up (**Tools > Options...**). Check DirectX Dependencies section for details."
|
||||
|
||||
### Linux
|
||||
The Linux samples are built using makefiles. To use the makefiles, change the current directory to the sample directory you wish to build, and run make:
|
||||
```
|
||||
$ cd <sample_dir>
|
||||
$ make
|
||||
```
|
||||
The samples makefiles can take advantage of certain options:
|
||||
* **TARGET_ARCH=<arch>** - cross-compile targeting a specific architecture. Allowed architectures are x86_64, ppc64le, armv7l, aarch64.
|
||||
By default, TARGET_ARCH is set to HOST_ARCH. On a x86_64 machine, not setting TARGET_ARCH is the equivalent of setting TARGET_ARCH=x86_64.<br/>
|
||||
`$ make TARGET_ARCH=x86_64` <br/> `$ make TARGET_ARCH=ppc64le` <br/> `$ make TARGET_ARCH=armv7l` <br/> `$ make TARGET_ARCH=aarch64` <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++
|
||||
```
|
||||
|
||||
### Mac
|
||||
The Mac samples are built using makefiles. To use the makefiles, change directory into the sample directory you wish to build, and run make:
|
||||
```
|
||||
$ cd <sample_dir>
|
||||
$ make
|
||||
```
|
||||
|
||||
The samples makefiles can take advantage of certain options:
|
||||
|
||||
* **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="A B ..."
|
||||
```
|
||||
|
||||
* **HOST_COMPILER=<host_compiler>** - override the default clang host compiler. See the [Mac Installation Guide](http://docs.nvidia.com/cuda/cuda-installation-guide-mac-os-x/index.html#system-requirements) for a list of supported host compilers.
|
||||
```
|
||||
$ make HOST_COMPILER=clang
|
||||
```
|
||||
|
||||
## References (for more details)
|
||||
|
||||
309
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics.cu
Normal file
309
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics.cu
Normal file
|
|
@ -0,0 +1,309 @@
|
|||
/* 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.
|
||||
*/
|
||||
|
||||
// System includes
|
||||
#include <assert.h>
|
||||
#include <stdio.h>
|
||||
|
||||
// CUDA runtime
|
||||
#include <cuda_runtime.h>
|
||||
|
||||
// helper functions and utilities to work with CUDA
|
||||
#include <helper_cuda.h>
|
||||
#include <helper_functions.h>
|
||||
|
||||
#ifndef MAX
|
||||
#define MAX(a, b) (a > b ? a : b)
|
||||
#endif
|
||||
|
||||
static const char *sSDKsample = "[simpleVoteIntrinsics]\0";
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// Global types and parameters
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
#define VOTE_DATA_GROUP 4
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// CUDA Voting Kernel functions
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
#include "simpleVote_kernel.cuh"
|
||||
|
||||
// Generate the test pattern for Tests 1 and 2
|
||||
void genVoteTestPattern(unsigned int *VOTE_PATTERN, int size) {
|
||||
// For testing VOTE.Any (all of these threads will return 0)
|
||||
for (int i = 0; i < size / 4; i++) {
|
||||
VOTE_PATTERN[i] = 0x00000000;
|
||||
}
|
||||
|
||||
// For testing VOTE.Any (1/2 these threads will return 1)
|
||||
for (int i = 2 * size / 8; i < 4 * size / 8; i++) {
|
||||
VOTE_PATTERN[i] = (i & 0x01) ? i : 0;
|
||||
}
|
||||
|
||||
// For testing VOTE.all (1/2 of these threads will return 0)
|
||||
for (int i = 2 * size / 4; i < 3 * size / 4; i++) {
|
||||
VOTE_PATTERN[i] = (i & 0x01) ? 0 : i;
|
||||
}
|
||||
|
||||
// For testing VOTE.all (all of these threads will return 1)
|
||||
for (int i = 3 * size / 4; i < 4 * size / 4; i++) {
|
||||
VOTE_PATTERN[i] = 0xffffffff;
|
||||
}
|
||||
}
|
||||
|
||||
int checkErrors1(unsigned int *h_result, int start, int end, int warp_size,
|
||||
const char *voteType) {
|
||||
int i, sum = 0;
|
||||
|
||||
for (sum = 0, i = start; i < end; i++) {
|
||||
sum += h_result[i];
|
||||
}
|
||||
|
||||
if (sum > 0) {
|
||||
printf("\t<%s>[%d - %d] = ", voteType, start, end - 1);
|
||||
|
||||
for (i = start; i < end; i++) {
|
||||
printf("%d", h_result[i]);
|
||||
}
|
||||
|
||||
printf("%d values FAILED\n", sum);
|
||||
}
|
||||
|
||||
return (sum > 0);
|
||||
}
|
||||
|
||||
int checkErrors2(unsigned int *h_result, int start, int end, int warp_size,
|
||||
const char *voteType) {
|
||||
int i, sum = 0;
|
||||
|
||||
for (sum = 0, i = start; i < end; i++) {
|
||||
sum += h_result[i];
|
||||
}
|
||||
|
||||
if (sum != warp_size) {
|
||||
printf("\t<%s>[%d - %d] = ", voteType, start, end - 1);
|
||||
|
||||
for (i = start; i < end; i++) {
|
||||
printf("%d", h_result[i]);
|
||||
}
|
||||
|
||||
printf(" - FAILED\n");
|
||||
}
|
||||
|
||||
return (sum != warp_size);
|
||||
}
|
||||
|
||||
// Verification code for Kernel #1
|
||||
int checkResultsVoteAnyKernel1(unsigned int *h_result, int size,
|
||||
int warp_size) {
|
||||
int error_count = 0;
|
||||
|
||||
error_count += checkErrors1(h_result, 0, VOTE_DATA_GROUP * warp_size / 4,
|
||||
warp_size, "Vote.Any");
|
||||
error_count +=
|
||||
checkErrors2(h_result, VOTE_DATA_GROUP * warp_size / 4,
|
||||
2 * VOTE_DATA_GROUP * warp_size / 4, warp_size, "Vote.Any");
|
||||
error_count +=
|
||||
checkErrors2(h_result, 2 * VOTE_DATA_GROUP * warp_size / 4,
|
||||
3 * VOTE_DATA_GROUP * warp_size / 4, warp_size, "Vote.Any");
|
||||
error_count +=
|
||||
checkErrors2(h_result, 3 * VOTE_DATA_GROUP * warp_size / 4,
|
||||
4 * VOTE_DATA_GROUP * warp_size / 4, warp_size, "Vote.Any");
|
||||
|
||||
printf((error_count == 0) ? "\tOK\n" : "\tERROR\n");
|
||||
return error_count;
|
||||
}
|
||||
|
||||
// Verification code for Kernel #2
|
||||
int checkResultsVoteAllKernel2(unsigned int *h_result, int size,
|
||||
int warp_size) {
|
||||
int error_count = 0;
|
||||
|
||||
error_count += checkErrors1(h_result, 0, VOTE_DATA_GROUP * warp_size / 4,
|
||||
warp_size, "Vote.All");
|
||||
error_count +=
|
||||
checkErrors1(h_result, VOTE_DATA_GROUP * warp_size / 4,
|
||||
2 * VOTE_DATA_GROUP * warp_size / 4, warp_size, "Vote.All");
|
||||
error_count +=
|
||||
checkErrors1(h_result, 2 * VOTE_DATA_GROUP * warp_size / 4,
|
||||
3 * VOTE_DATA_GROUP * warp_size / 4, warp_size, "Vote.All");
|
||||
error_count +=
|
||||
checkErrors2(h_result, 3 * VOTE_DATA_GROUP * warp_size / 4,
|
||||
4 * VOTE_DATA_GROUP * warp_size / 4, warp_size, "Vote.All");
|
||||
|
||||
printf((error_count == 0) ? "\tOK\n" : "\tERROR\n");
|
||||
return error_count;
|
||||
}
|
||||
|
||||
// Verification code for Kernel #3
|
||||
int checkResultsVoteAnyKernel3(bool *hinfo, int size) {
|
||||
int i, error_count = 0;
|
||||
|
||||
for (i = 0; i < size * 3; i++) {
|
||||
switch (i % 3) {
|
||||
case 0:
|
||||
|
||||
// First warp should be all zeros.
|
||||
if (hinfo[i] != (i >= size * 1)) {
|
||||
error_count++;
|
||||
}
|
||||
|
||||
break;
|
||||
|
||||
case 1:
|
||||
|
||||
// First warp and half of second should be all zeros.
|
||||
if (hinfo[i] != (i >= size * 3 / 2)) {
|
||||
error_count++;
|
||||
}
|
||||
|
||||
break;
|
||||
|
||||
case 2:
|
||||
|
||||
// First two warps should be all zeros.
|
||||
if (hinfo[i] != (i >= size * 2)) {
|
||||
error_count++;
|
||||
}
|
||||
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
printf((error_count == 0) ? "\tOK\n" : "\tERROR\n");
|
||||
return error_count;
|
||||
}
|
||||
|
||||
int main(int argc, char **argv) {
|
||||
unsigned int *h_input, *h_result;
|
||||
unsigned int *d_input, *d_result;
|
||||
|
||||
bool *dinfo = NULL, *hinfo = NULL;
|
||||
int error_count[3] = {0, 0, 0};
|
||||
|
||||
cudaDeviceProp deviceProp;
|
||||
int devID, warp_size = 32;
|
||||
|
||||
printf("%s\n", sSDKsample);
|
||||
|
||||
// This will pick the best possible CUDA capable device
|
||||
devID = findCudaDevice(argc, (const char **)argv);
|
||||
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, devID));
|
||||
|
||||
// Statistics about the GPU device
|
||||
printf(
|
||||
"> GPU device has %d Multi-Processors, SM %d.%d compute capabilities\n\n",
|
||||
deviceProp.multiProcessorCount, deviceProp.major, deviceProp.minor);
|
||||
|
||||
h_input = (unsigned int *)malloc(VOTE_DATA_GROUP * warp_size *
|
||||
sizeof(unsigned int));
|
||||
h_result = (unsigned int *)malloc(VOTE_DATA_GROUP * warp_size *
|
||||
sizeof(unsigned int));
|
||||
checkCudaErrors(
|
||||
cudaMalloc(reinterpret_cast<void **>(&d_input),
|
||||
VOTE_DATA_GROUP * warp_size * sizeof(unsigned int)));
|
||||
checkCudaErrors(
|
||||
cudaMalloc(reinterpret_cast<void **>(&d_result),
|
||||
VOTE_DATA_GROUP * warp_size * sizeof(unsigned int)));
|
||||
genVoteTestPattern(h_input, VOTE_DATA_GROUP * warp_size);
|
||||
checkCudaErrors(cudaMemcpy(d_input, h_input,
|
||||
VOTE_DATA_GROUP * warp_size * sizeof(unsigned int),
|
||||
cudaMemcpyHostToDevice));
|
||||
|
||||
// Start of Vote Any Test Kernel #1
|
||||
printf("[VOTE Kernel Test 1/3]\n");
|
||||
printf("\tRunning <<Vote.Any>> kernel1 ...\n");
|
||||
{
|
||||
checkCudaErrors(cudaDeviceSynchronize());
|
||||
dim3 gridBlock(1, 1);
|
||||
dim3 threadBlock(VOTE_DATA_GROUP * warp_size, 1);
|
||||
VoteAnyKernel1<<<gridBlock, threadBlock>>>(d_input, d_result,
|
||||
VOTE_DATA_GROUP * warp_size);
|
||||
getLastCudaError("VoteAnyKernel() execution failed\n");
|
||||
checkCudaErrors(cudaDeviceSynchronize());
|
||||
}
|
||||
checkCudaErrors(cudaMemcpy(h_result, d_result,
|
||||
VOTE_DATA_GROUP * warp_size * sizeof(unsigned int),
|
||||
cudaMemcpyDeviceToHost));
|
||||
error_count[0] += checkResultsVoteAnyKernel1(
|
||||
h_result, VOTE_DATA_GROUP * warp_size, warp_size);
|
||||
|
||||
// Start of Vote All Test Kernel #2
|
||||
printf("\n[VOTE Kernel Test 2/3]\n");
|
||||
printf("\tRunning <<Vote.All>> kernel2 ...\n");
|
||||
{
|
||||
checkCudaErrors(cudaDeviceSynchronize());
|
||||
dim3 gridBlock(1, 1);
|
||||
dim3 threadBlock(VOTE_DATA_GROUP * warp_size, 1);
|
||||
VoteAllKernel2<<<gridBlock, threadBlock>>>(d_input, d_result,
|
||||
VOTE_DATA_GROUP * warp_size);
|
||||
getLastCudaError("VoteAllKernel() execution failed\n");
|
||||
checkCudaErrors(cudaDeviceSynchronize());
|
||||
}
|
||||
checkCudaErrors(cudaMemcpy(h_result, d_result,
|
||||
VOTE_DATA_GROUP * warp_size * sizeof(unsigned int),
|
||||
cudaMemcpyDeviceToHost));
|
||||
error_count[1] += checkResultsVoteAllKernel2(
|
||||
h_result, VOTE_DATA_GROUP * warp_size, warp_size);
|
||||
|
||||
// Second Vote Kernel Test #3 (both Any/All)
|
||||
hinfo = reinterpret_cast<bool *>(calloc(warp_size * 3 * 3, sizeof(bool)));
|
||||
cudaMalloc(reinterpret_cast<void **>(&dinfo),
|
||||
warp_size * 3 * 3 * sizeof(bool));
|
||||
cudaMemcpy(dinfo, hinfo, warp_size * 3 * 3 * sizeof(bool),
|
||||
cudaMemcpyHostToDevice);
|
||||
|
||||
printf("\n[VOTE Kernel Test 3/3]\n");
|
||||
printf("\tRunning <<Vote.Any>> kernel3 ...\n");
|
||||
{
|
||||
checkCudaErrors(cudaDeviceSynchronize());
|
||||
VoteAnyKernel3<<<1, warp_size * 3>>>(dinfo, warp_size);
|
||||
checkCudaErrors(cudaDeviceSynchronize());
|
||||
}
|
||||
|
||||
cudaMemcpy(hinfo, dinfo, warp_size * 3 * 3 * sizeof(bool),
|
||||
cudaMemcpyDeviceToHost);
|
||||
|
||||
error_count[2] = checkResultsVoteAnyKernel3(hinfo, warp_size * 3);
|
||||
|
||||
// Now free these resources for Test #1,2
|
||||
checkCudaErrors(cudaFree(d_input));
|
||||
checkCudaErrors(cudaFree(d_result));
|
||||
free(h_input);
|
||||
free(h_result);
|
||||
|
||||
// Free resources from Test #3
|
||||
free(hinfo);
|
||||
cudaFree(dinfo);
|
||||
|
||||
printf("\tShutting down...\n");
|
||||
|
||||
return (error_count[0] == 0 && error_count[1] == 0 && error_count[2] == 0)
|
||||
? EXIT_SUCCESS
|
||||
: EXIT_FAILURE;
|
||||
}
|
||||
20
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2010.sln
Normal file
20
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2010.sln
Normal file
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 11.00
|
||||
# Visual Studio 2010
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "simpleVoteIntrinsics", "simpleVoteIntrinsics_vs2010.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
|
||||
106
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2010.vcxproj
Normal file
106
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2010.vcxproj
Normal file
|
|
@ -0,0 +1,106 @@
|
|||
<?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>simpleVoteIntrinsics_vs2010</RootNamespace>
|
||||
<ProjectName>simpleVoteIntrinsics</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(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="$(CUDAPropsPath)\CUDA 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/simpleVoteIntrinsics.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(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="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="simpleVoteIntrinsics.cu" />
|
||||
<None Include="simpleVote_kernel.cuh" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2012.sln
Normal file
20
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2012.sln
Normal file
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2012
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "simpleVoteIntrinsics", "simpleVoteIntrinsics_vs2012.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
107
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2012.vcxproj
Normal file
107
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2012.vcxproj
Normal file
|
|
@ -0,0 +1,107 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>simpleVoteIntrinsics_vs2012</RootNamespace>
|
||||
<ProjectName>simpleVoteIntrinsics</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v110</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 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/simpleVoteIntrinsics.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</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>
|
||||
<CudaCompile Include="simpleVoteIntrinsics.cu" />
|
||||
<None Include="simpleVote_kernel.cuh" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2013.sln
Normal file
20
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2013.sln
Normal file
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 13.00
|
||||
# Visual Studio 2013
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "simpleVoteIntrinsics", "simpleVoteIntrinsics_vs2013.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
|
||||
107
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2013.vcxproj
Normal file
107
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2013.vcxproj
Normal file
|
|
@ -0,0 +1,107 @@
|
|||
<?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>simpleVoteIntrinsics_vs2013</RootNamespace>
|
||||
<ProjectName>simpleVoteIntrinsics</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v120</PlatformToolset>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(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="$(CUDAPropsPath)\CUDA 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/simpleVoteIntrinsics.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(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="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="simpleVoteIntrinsics.cu" />
|
||||
<None Include="simpleVote_kernel.cuh" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2015.sln
Normal file
20
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2015.sln
Normal file
|
|
@ -0,0 +1,20 @@
|
|||
|
||||
Microsoft Visual Studio Solution File, Format Version 14.00
|
||||
# Visual Studio 2015
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "simpleVoteIntrinsics", "simpleVoteIntrinsics_vs2015.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
107
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2015.vcxproj
Normal file
107
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2015.vcxproj
Normal file
|
|
@ -0,0 +1,107 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>simpleVoteIntrinsics_vs2015</RootNamespace>
|
||||
<ProjectName>simpleVoteIntrinsics</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v140</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 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/simpleVoteIntrinsics.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</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>
|
||||
<CudaCompile Include="simpleVoteIntrinsics.cu" />
|
||||
<None Include="simpleVote_kernel.cuh" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2017.sln
Normal file
20
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_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}") = "simpleVoteIntrinsics", "simpleVoteIntrinsics_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
|
||||
108
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2017.vcxproj
Normal file
108
Samples/simpleVoteIntrinsics/simpleVoteIntrinsics_vs2017.vcxproj
Normal file
|
|
@ -0,0 +1,108 @@
|
|||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>simpleVoteIntrinsics_vs2017</RootNamespace>
|
||||
<ProjectName>simpleVoteIntrinsics</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v141</PlatformToolset>
|
||||
<WindowsTargetPlatformVersion>10.0.15063.0</WindowsTargetPlatformVersion>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(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="$(CUDAPropsPath)\CUDA 9.2.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/simpleVoteIntrinsics.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_30,sm_30;compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(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="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<CudaCompile Include="simpleVoteIntrinsics.cu" />
|
||||
<None Include="simpleVote_kernel.cuh" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 9.2.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
80
Samples/simpleVoteIntrinsics/simpleVote_kernel.cuh
Normal file
80
Samples/simpleVoteIntrinsics/simpleVote_kernel.cuh
Normal file
|
|
@ -0,0 +1,80 @@
|
|||
/* 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.
|
||||
*/
|
||||
|
||||
#ifndef SIMPLEVOTE_KERNEL_CU
|
||||
#define SIMPLEVOTE_KERNEL_CU
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// Vote Any/All intrinsic kernel function tests are supported only by CUDA
|
||||
// capable devices that are CUDA hardware that has SM1.2 or later
|
||||
// Vote Functions (refer to section 4.4.5 in the CUDA Programming Guide)
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
|
||||
// Kernel #1 tests the across-the-warp vote(any) intrinsic.
|
||||
// If ANY one of the threads (within the warp) of the predicated condition
|
||||
// returns a non-zero value, then all threads within this warp will return a
|
||||
// non-zero value
|
||||
__global__ void VoteAnyKernel1(unsigned int *input, unsigned int *result,
|
||||
int size) {
|
||||
int tx = threadIdx.x;
|
||||
|
||||
int mask = 0xffffffff;
|
||||
result[tx] = __any_sync(mask, input[tx]);
|
||||
}
|
||||
|
||||
// Kernel #2 tests the across-the-warp vote(all) intrinsic.
|
||||
// If ALL of the threads (within the warp) of the predicated condition returns
|
||||
// a non-zero value, then all threads within this warp will return a non-zero
|
||||
// value
|
||||
__global__ void VoteAllKernel2(unsigned int *input, unsigned int *result,
|
||||
int size) {
|
||||
int tx = threadIdx.x;
|
||||
|
||||
int mask = 0xffffffff;
|
||||
result[tx] = __all_sync(mask, input[tx]);
|
||||
}
|
||||
|
||||
// Kernel #3 is a directed test for the across-the-warp vote(all) intrinsic.
|
||||
// This kernel will test for conditions across warps, and within half warps
|
||||
__global__ void VoteAnyKernel3(bool *info, int warp_size) {
|
||||
int tx = threadIdx.x;
|
||||
unsigned int mask = 0xffffffff;
|
||||
bool *offs = info + (tx * 3);
|
||||
|
||||
// The following should hold true for the second and third warp
|
||||
*offs = __any_sync(mask, (tx >= (warp_size * 3) / 2));
|
||||
// The following should hold true for the "upper half" of the second warp,
|
||||
// and all of the third warp
|
||||
*(offs + 1) = (tx >= (warp_size * 3) / 2 ? true : false);
|
||||
|
||||
// The following should hold true for the third warp only
|
||||
if (__all_sync(mask, (tx >= (warp_size * 3) / 2))) {
|
||||
*(offs + 2) = true;
|
||||
}
|
||||
}
|
||||
|
||||
#endif
|
||||
|
|
@ -180,6 +180,21 @@ ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
|||
LDFLAGS += -rpath-link=$(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)
|
||||
|
|
|
|||
|
|
@ -180,6 +180,21 @@ ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
|||
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)
|
||||
|
|
|
|||
Loading…
Reference in New Issue
Block a user