mirror of
https://github.com/NVIDIA/cuda-samples.git
synced 2026-10-23 21:28:17 +08:00
add and update samples for CUDA 11.5
This commit is contained in:
Rutwik Choughule
18
Samples/MonteCarloMultiGPU/.vscode/c_cpp_properties.json
vendored
Normal file
18
Samples/MonteCarloMultiGPU/.vscode/c_cpp_properties.json
vendored
Normal file
@@ -0,0 +1,18 @@
|
||||
{
|
||||
"configurations": [
|
||||
{
|
||||
"name": "Linux",
|
||||
"includePath": [
|
||||
"${workspaceFolder}/**",
|
||||
"${workspaceFolder}/../../Common"
|
||||
],
|
||||
"defines": [],
|
||||
"compilerPath": "/usr/local/cuda/bin/nvcc",
|
||||
"cStandard": "gnu17",
|
||||
"cppStandard": "gnu++14",
|
||||
"intelliSenseMode": "linux-gcc-x64",
|
||||
"configurationProvider": "ms-vscode.makefile-tools"
|
||||
}
|
||||
],
|
||||
"version": 4
|
||||
}
|
||||
7
Samples/MonteCarloMultiGPU/.vscode/extensions.json
vendored
Normal file
7
Samples/MonteCarloMultiGPU/.vscode/extensions.json
vendored
Normal file
@@ -0,0 +1,7 @@
|
||||
{
|
||||
"recommendations": [
|
||||
"nvidia.nsight-vscode-edition",
|
||||
"ms-vscode.cpptools",
|
||||
"ms-vscode.makefile-tools"
|
||||
]
|
||||
}
|
||||
10
Samples/MonteCarloMultiGPU/.vscode/launch.json
vendored
Normal file
10
Samples/MonteCarloMultiGPU/.vscode/launch.json
vendored
Normal file
@@ -0,0 +1,10 @@
|
||||
{
|
||||
"configurations": [
|
||||
{
|
||||
"name": "CUDA C++: Launch",
|
||||
"type": "cuda-gdb",
|
||||
"request": "launch",
|
||||
"program": "${workspaceFolder}/MonteCarloMultiGPU"
|
||||
}
|
||||
]
|
||||
}
|
||||
15
Samples/MonteCarloMultiGPU/.vscode/tasks.json
vendored
Normal file
15
Samples/MonteCarloMultiGPU/.vscode/tasks.json
vendored
Normal file
@@ -0,0 +1,15 @@
|
||||
{
|
||||
"version": "2.0.0",
|
||||
"tasks": [
|
||||
{
|
||||
"label": "sample",
|
||||
"type": "shell",
|
||||
"command": "make dbg=1",
|
||||
"problemMatcher": ["$nvcc"],
|
||||
"group": {
|
||||
"kind": "build",
|
||||
"isDefault": true
|
||||
}
|
||||
}
|
||||
]
|
||||
}
|
||||
373
Samples/MonteCarloMultiGPU/Makefile
Normal file
373
Samples/MonteCarloMultiGPU/Makefile
Normal file
@@ -0,0 +1,373 @@
|
||||
################################################################################
|
||||
# Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
|
||||
#
|
||||
# Redistribution and use in source and binary forms, with or without
|
||||
# modification, are permitted provided that the following conditions
|
||||
# are met:
|
||||
# * Redistributions of source code must retain the above copyright
|
||||
# notice, this list of conditions and the following disclaimer.
|
||||
# * Redistributions in binary form must reproduce the above copyright
|
||||
# notice, this list of conditions and the following disclaimer in the
|
||||
# documentation and/or other materials provided with the distribution.
|
||||
# * Neither the name of NVIDIA CORPORATION nor the names of its
|
||||
# contributors may be used to endorse or promote products derived
|
||||
# from this software without specific prior written permission.
|
||||
#
|
||||
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
||||
# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||||
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
# PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
||||
# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
||||
# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||||
# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||||
# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
||||
# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
#
|
||||
################################################################################
|
||||
#
|
||||
# Makefile project only supported on Mac OS X and Linux Platforms)
|
||||
#
|
||||
################################################################################
|
||||
|
||||
# Location of the CUDA Toolkit
|
||||
CUDA_PATH ?= /usr/local/cuda
|
||||
|
||||
##############################
|
||||
# start deprecated interface #
|
||||
##############################
|
||||
ifeq ($(x86_64),1)
|
||||
$(info WARNING - x86_64 variable has been deprecated)
|
||||
$(info WARNING - please use TARGET_ARCH=x86_64 instead)
|
||||
TARGET_ARCH ?= x86_64
|
||||
endif
|
||||
ifeq ($(ARMv7),1)
|
||||
$(info WARNING - ARMv7 variable has been deprecated)
|
||||
$(info WARNING - please use TARGET_ARCH=armv7l instead)
|
||||
TARGET_ARCH ?= armv7l
|
||||
endif
|
||||
ifeq ($(aarch64),1)
|
||||
$(info WARNING - aarch64 variable has been deprecated)
|
||||
$(info WARNING - please use TARGET_ARCH=aarch64 instead)
|
||||
TARGET_ARCH ?= aarch64
|
||||
endif
|
||||
ifeq ($(ppc64le),1)
|
||||
$(info WARNING - ppc64le variable has been deprecated)
|
||||
$(info WARNING - please use TARGET_ARCH=ppc64le instead)
|
||||
TARGET_ARCH ?= ppc64le
|
||||
endif
|
||||
ifneq ($(GCC),)
|
||||
$(info WARNING - GCC variable has been deprecated)
|
||||
$(info WARNING - please use HOST_COMPILER=$(GCC) instead)
|
||||
HOST_COMPILER ?= $(GCC)
|
||||
endif
|
||||
ifneq ($(abi),)
|
||||
$(error ERROR - abi variable has been removed)
|
||||
endif
|
||||
############################
|
||||
# end deprecated interface #
|
||||
############################
|
||||
|
||||
# architecture
|
||||
HOST_ARCH := $(shell uname -m)
|
||||
TARGET_ARCH ?= $(HOST_ARCH)
|
||||
ifneq (,$(filter $(TARGET_ARCH),x86_64 aarch64 sbsa ppc64le armv7l))
|
||||
ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
||||
ifneq (,$(filter $(TARGET_ARCH),x86_64 aarch64 sbsa ppc64le))
|
||||
TARGET_SIZE := 64
|
||||
else ifneq (,$(filter $(TARGET_ARCH),armv7l))
|
||||
TARGET_SIZE := 32
|
||||
endif
|
||||
else
|
||||
TARGET_SIZE := $(shell getconf LONG_BIT)
|
||||
endif
|
||||
else
|
||||
$(error ERROR - unsupported value $(TARGET_ARCH) for TARGET_ARCH!)
|
||||
endif
|
||||
|
||||
# sbsa and aarch64 systems look similar. Need to differentiate them at host level for now.
|
||||
ifeq ($(HOST_ARCH),aarch64)
|
||||
ifeq ($(CUDA_PATH)/targets/sbsa-linux,$(shell ls -1d $(CUDA_PATH)/targets/sbsa-linux 2>/dev/null))
|
||||
HOST_ARCH := sbsa
|
||||
TARGET_ARCH := sbsa
|
||||
endif
|
||||
endif
|
||||
|
||||
ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
||||
ifeq (,$(filter $(HOST_ARCH)-$(TARGET_ARCH),aarch64-armv7l x86_64-armv7l x86_64-aarch64 x86_64-sbsa x86_64-ppc64le))
|
||||
$(error ERROR - cross compiling from $(HOST_ARCH) to $(TARGET_ARCH) is not supported!)
|
||||
endif
|
||||
endif
|
||||
|
||||
# When on native aarch64 system with userspace of 32-bit, change TARGET_ARCH to armv7l
|
||||
ifeq ($(HOST_ARCH)-$(TARGET_ARCH)-$(TARGET_SIZE),aarch64-aarch64-32)
|
||||
TARGET_ARCH = armv7l
|
||||
endif
|
||||
|
||||
# operating system
|
||||
HOST_OS := $(shell uname -s 2>/dev/null | tr "[:upper:]" "[:lower:]")
|
||||
TARGET_OS ?= $(HOST_OS)
|
||||
ifeq (,$(filter $(TARGET_OS),linux darwin qnx android))
|
||||
$(error ERROR - unsupported value $(TARGET_OS) for TARGET_OS!)
|
||||
endif
|
||||
|
||||
# host compiler
|
||||
ifeq ($(TARGET_OS),darwin)
|
||||
ifeq ($(shell expr `xcodebuild -version | grep -i xcode | awk '{print $$2}' | cut -d'.' -f1` \>= 5),1)
|
||||
HOST_COMPILER ?= clang++
|
||||
endif
|
||||
else ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
||||
ifeq ($(HOST_ARCH)-$(TARGET_ARCH),x86_64-armv7l)
|
||||
ifeq ($(TARGET_OS),linux)
|
||||
HOST_COMPILER ?= arm-linux-gnueabihf-g++
|
||||
else ifeq ($(TARGET_OS),qnx)
|
||||
ifeq ($(QNX_HOST),)
|
||||
$(error ERROR - QNX_HOST must be passed to the QNX host toolchain)
|
||||
endif
|
||||
ifeq ($(QNX_TARGET),)
|
||||
$(error ERROR - QNX_TARGET must be passed to the QNX target toolchain)
|
||||
endif
|
||||
export QNX_HOST
|
||||
export QNX_TARGET
|
||||
HOST_COMPILER ?= $(QNX_HOST)/usr/bin/arm-unknown-nto-qnx6.6.0eabi-g++
|
||||
else ifeq ($(TARGET_OS),android)
|
||||
HOST_COMPILER ?= arm-linux-androideabi-g++
|
||||
endif
|
||||
else ifeq ($(TARGET_ARCH),aarch64)
|
||||
ifeq ($(TARGET_OS), linux)
|
||||
HOST_COMPILER ?= aarch64-linux-gnu-g++
|
||||
else ifeq ($(TARGET_OS),qnx)
|
||||
ifeq ($(QNX_HOST),)
|
||||
$(error ERROR - QNX_HOST must be passed to the QNX host toolchain)
|
||||
endif
|
||||
ifeq ($(QNX_TARGET),)
|
||||
$(error ERROR - QNX_TARGET must be passed to the QNX target toolchain)
|
||||
endif
|
||||
export QNX_HOST
|
||||
export QNX_TARGET
|
||||
HOST_COMPILER ?= $(QNX_HOST)/usr/bin/q++
|
||||
else ifeq ($(TARGET_OS), android)
|
||||
HOST_COMPILER ?= aarch64-linux-android-clang++
|
||||
endif
|
||||
else ifeq ($(TARGET_ARCH),sbsa)
|
||||
HOST_COMPILER ?= aarch64-linux-gnu-g++
|
||||
else ifeq ($(TARGET_ARCH),ppc64le)
|
||||
HOST_COMPILER ?= powerpc64le-linux-gnu-g++
|
||||
endif
|
||||
endif
|
||||
HOST_COMPILER ?= g++
|
||||
NVCC := $(CUDA_PATH)/bin/nvcc -ccbin $(HOST_COMPILER)
|
||||
|
||||
# internal flags
|
||||
NVCCFLAGS := -m${TARGET_SIZE}
|
||||
CCFLAGS :=
|
||||
LDFLAGS :=
|
||||
|
||||
# build flags
|
||||
ifeq ($(TARGET_OS),darwin)
|
||||
LDFLAGS += -rpath $(CUDA_PATH)/lib
|
||||
CCFLAGS += -arch $(HOST_ARCH)
|
||||
else ifeq ($(HOST_ARCH)-$(TARGET_ARCH)-$(TARGET_OS),x86_64-armv7l-linux)
|
||||
LDFLAGS += --dynamic-linker=/lib/ld-linux-armhf.so.3
|
||||
CCFLAGS += -mfloat-abi=hard
|
||||
else ifeq ($(TARGET_OS),android)
|
||||
LDFLAGS += -pie
|
||||
CCFLAGS += -fpie -fpic -fexceptions
|
||||
endif
|
||||
|
||||
ifneq ($(TARGET_ARCH),$(HOST_ARCH))
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-linux)
|
||||
ifneq ($(TARGET_FS),)
|
||||
GCCVERSIONLTEQ46 := $(shell expr `$(HOST_COMPILER) -dumpversion` \<= 4.6)
|
||||
ifeq ($(GCCVERSIONLTEQ46),1)
|
||||
CCFLAGS += --sysroot=$(TARGET_FS)
|
||||
endif
|
||||
LDFLAGS += --sysroot=$(TARGET_FS)
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/lib
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib/arm-linux-gnueabihf
|
||||
endif
|
||||
endif
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-linux)
|
||||
ifneq ($(TARGET_FS),)
|
||||
GCCVERSIONLTEQ46 := $(shell expr `$(HOST_COMPILER) -dumpversion` \<= 4.6)
|
||||
ifeq ($(GCCVERSIONLTEQ46),1)
|
||||
CCFLAGS += --sysroot=$(TARGET_FS)
|
||||
endif
|
||||
LDFLAGS += --sysroot=$(TARGET_FS)
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/lib -L$(TARGET_FS)/lib
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/lib/aarch64-linux-gnu -L$(TARGET_FS)/lib/aarch64-linux-gnu
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib -L$(TARGET_FS)/usr/lib
|
||||
LDFLAGS += -rpath-link=$(TARGET_FS)/usr/lib/aarch64-linux-gnu -L$(TARGET_FS)/usr/lib/aarch64-linux-gnu
|
||||
LDFLAGS += --unresolved-symbols=ignore-in-shared-libs
|
||||
CCFLAGS += -isystem=$(TARGET_FS)/usr/include -I$(TARGET_FS)/usr/include -I$(TARGET_FS)/usr/include/libdrm
|
||||
CCFLAGS += -isystem=$(TARGET_FS)/usr/include/aarch64-linux-gnu -I$(TARGET_FS)/usr/include/aarch64-linux-gnu
|
||||
endif
|
||||
endif
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-qnx)
|
||||
NVCCFLAGS += --qpp-config 5.4.0,gcc_ntoaarch64le
|
||||
CCFLAGS += -DWIN_INTERFACE_CUSTOM -I/usr/include/aarch64-qnx-gnu
|
||||
LDFLAGS += -lsocket
|
||||
LDFLAGS += -L/usr/lib/aarch64-qnx-gnu
|
||||
CCFLAGS += "-Wl\,-rpath-link\,/usr/lib/aarch64-qnx-gnu"
|
||||
ifdef TARGET_OVERRIDE
|
||||
LDFLAGS += -lslog2
|
||||
endif
|
||||
|
||||
ifneq ($(TARGET_FS),)
|
||||
LDFLAGS += -L$(TARGET_FS)/usr/lib
|
||||
CCFLAGS += "-Wl\,-rpath-link\,$(TARGET_FS)/usr/lib"
|
||||
LDFLAGS += -L$(TARGET_FS)/usr/libnvidia
|
||||
CCFLAGS += "-Wl\,-rpath-link\,$(TARGET_FS)/usr/libnvidia"
|
||||
CCFLAGS += -I$(TARGET_FS)/../include
|
||||
endif
|
||||
endif
|
||||
endif
|
||||
|
||||
ifdef TARGET_OVERRIDE # cuda toolkit targets override
|
||||
NVCCFLAGS += -target-dir $(TARGET_OVERRIDE)
|
||||
endif
|
||||
|
||||
# Install directory of different arch
|
||||
CUDA_INSTALL_TARGET_DIR :=
|
||||
ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-linux)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/armv7-linux-gnueabihf/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-linux)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/aarch64-linux/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),sbsa-linux)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/sbsa-linux/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-android)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/armv7-linux-androideabi/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-android)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/aarch64-linux-androideabi/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),armv7l-qnx)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/ARMv7-linux-QNX/
|
||||
else ifeq ($(TARGET_ARCH)-$(TARGET_OS),aarch64-qnx)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/aarch64-qnx/
|
||||
else ifeq ($(TARGET_ARCH),ppc64le)
|
||||
CUDA_INSTALL_TARGET_DIR = targets/ppc64le-linux/
|
||||
endif
|
||||
|
||||
# Debug build flags
|
||||
ifeq ($(dbg),1)
|
||||
NVCCFLAGS += -g -G
|
||||
BUILD_TYPE := debug
|
||||
else
|
||||
BUILD_TYPE := release
|
||||
endif
|
||||
|
||||
ALL_CCFLAGS :=
|
||||
ALL_CCFLAGS += $(NVCCFLAGS)
|
||||
ALL_CCFLAGS += $(EXTRA_NVCCFLAGS)
|
||||
ALL_CCFLAGS += $(addprefix -Xcompiler ,$(CCFLAGS))
|
||||
ALL_CCFLAGS += $(addprefix -Xcompiler ,$(EXTRA_CCFLAGS))
|
||||
|
||||
SAMPLE_ENABLED := 1
|
||||
|
||||
ALL_LDFLAGS :=
|
||||
ALL_LDFLAGS += $(ALL_CCFLAGS)
|
||||
ALL_LDFLAGS += $(addprefix -Xlinker ,$(LDFLAGS))
|
||||
ALL_LDFLAGS += $(addprefix -Xlinker ,$(EXTRA_LDFLAGS))
|
||||
|
||||
# Common includes and paths for CUDA
|
||||
INCLUDES := -I../../Common
|
||||
LIBRARIES :=
|
||||
|
||||
################################################################################
|
||||
|
||||
#Detect if installed version of GCC supports required C++11
|
||||
ifeq ($(TARGET_OS),linux)
|
||||
empty :=
|
||||
space := $(empty) $(empty)
|
||||
GCCVERSIONSTRING := $(shell expr `$(HOST_COMPILER) -dumpversion`)
|
||||
#Create version number without "."
|
||||
GCCVERSION := $(shell expr `echo $(GCCVERSIONSTRING)` | cut -f1 -d.)
|
||||
GCCVERSION += $(shell expr `echo $(GCCVERSIONSTRING)` | cut -f2 -d.)
|
||||
GCCVERSION += $(shell expr `echo $(GCCVERSIONSTRING)` | cut -f3 -d.)
|
||||
# Make sure the version number has at least 3 decimals
|
||||
GCCVERSION += 00
|
||||
# Remove spaces from the version number
|
||||
GCCVERSION := $(subst $(space),$(empty),$(GCCVERSION))
|
||||
#$(warning $(GCCVERSION))
|
||||
|
||||
IS_MIN_VERSION := $(shell expr `echo $(GCCVERSION)` \>= 51000)
|
||||
|
||||
ifeq ($(IS_MIN_VERSION), 1)
|
||||
$(info >>> GCC Version is greater or equal to 5.1.0 <<<)
|
||||
else
|
||||
$(info >>> Waiving build. Minimum GCC version required is 5.1.0<<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
endif
|
||||
|
||||
# Gencode arguments
|
||||
ifeq ($(TARGET_ARCH),$(filter $(TARGET_ARCH),armv7l aarch64))
|
||||
SMS ?= 35 37 50 52 60 61 70 72 75 80 86
|
||||
else
|
||||
SMS ?= 35 37 50 52 60 61 70 75 80 86
|
||||
endif
|
||||
|
||||
ifeq ($(SMS),)
|
||||
$(info >>> WARNING - no SM architectures have been specified - waiving sample <<<)
|
||||
SAMPLE_ENABLED := 0
|
||||
endif
|
||||
|
||||
ifeq ($(GENCODE_FLAGS),)
|
||||
# Generate SASS code for each SM architecture listed in $(SMS)
|
||||
$(foreach sm,$(SMS),$(eval GENCODE_FLAGS += -gencode arch=compute_$(sm),code=sm_$(sm)))
|
||||
|
||||
# Generate PTX code from the highest SM architecture in $(SMS) to guarantee forward-compatibility
|
||||
HIGHEST_SM := $(lastword $(sort $(SMS)))
|
||||
ifneq ($(HIGHEST_SM),)
|
||||
GENCODE_FLAGS += -gencode arch=compute_$(HIGHEST_SM),code=compute_$(HIGHEST_SM)
|
||||
endif
|
||||
endif
|
||||
|
||||
ALL_CCFLAGS += --threads 0 --std=c++11
|
||||
|
||||
LIBRARIES += -lcurand
|
||||
|
||||
ifeq ($(SAMPLE_ENABLED),0)
|
||||
EXEC ?= @echo "[@]"
|
||||
endif
|
||||
|
||||
################################################################################
|
||||
|
||||
# Target rules
|
||||
all: build
|
||||
|
||||
build: MonteCarloMultiGPU
|
||||
|
||||
check.deps:
|
||||
ifeq ($(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
|
||||
|
||||
MonteCarloMultiGPU.o:MonteCarloMultiGPU.cpp
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
MonteCarlo_gold.o:MonteCarlo_gold.cpp
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
MonteCarlo_kernel.o:MonteCarlo_kernel.cu
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
multithreading.o:multithreading.cpp
|
||||
$(EXEC) $(NVCC) $(INCLUDES) $(ALL_CCFLAGS) $(GENCODE_FLAGS) -o $@ -c $<
|
||||
|
||||
MonteCarloMultiGPU: MonteCarloMultiGPU.o MonteCarlo_gold.o MonteCarlo_kernel.o multithreading.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
|
||||
$(EXEC) ./MonteCarloMultiGPU
|
||||
|
||||
clean:
|
||||
rm -f MonteCarloMultiGPU MonteCarloMultiGPU.o MonteCarlo_gold.o MonteCarlo_kernel.o multithreading.o
|
||||
rm -rf ../../bin/$(TARGET_ARCH)/$(TARGET_OS)/$(BUILD_TYPE)/MonteCarloMultiGPU
|
||||
|
||||
clobber: clean
|
||||
487
Samples/MonteCarloMultiGPU/MonteCarloMultiGPU.cpp
Normal file
487
Samples/MonteCarloMultiGPU/MonteCarloMultiGPU.cpp
Normal file
@@ -0,0 +1,487 @@
|
||||
/* Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved.
|
||||
*
|
||||
* Redistribution and use in source and binary forms, with or without
|
||||
* modification, are permitted provided that the following conditions
|
||||
* are met:
|
||||
* * Redistributions of source code must retain the above copyright
|
||||
* notice, this list of conditions and the following disclaimer.
|
||||
* * Redistributions in binary form must reproduce the above copyright
|
||||
* notice, this list of conditions and the following disclaimer in the
|
||||
* documentation and/or other materials provided with the distribution.
|
||||
* * Neither the name of NVIDIA CORPORATION nor the names of its
|
||||
* contributors may be used to endorse or promote products derived
|
||||
* from this software without specific prior written permission.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
||||
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||||
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
||||
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
||||
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||||
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||||
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
||||
* OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
*/
|
||||
|
||||
/*
|
||||
* This sample evaluates fair call price for a
|
||||
* given set of European options using Monte Carlo approach.
|
||||
* See supplied whitepaper for more explanations.
|
||||
*/
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <stdio.h>
|
||||
#include <string.h>
|
||||
#include <math.h>
|
||||
#include <cuda_runtime.h>
|
||||
|
||||
// includes, project
|
||||
#include <helper_functions.h> // Helper functions (utilities, parsing, timing)
|
||||
#include <helper_cuda.h> // helper functions (cuda error checking and initialization)
|
||||
#include <multithreading.h>
|
||||
|
||||
#include "MonteCarlo_common.h"
|
||||
|
||||
int *pArgc = NULL;
|
||||
char **pArgv = NULL;
|
||||
|
||||
#ifdef WIN32
|
||||
#define strcasecmp _strcmpi
|
||||
#endif
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// Common functions
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
float randFloat(float low, float high) {
|
||||
float t = (float)rand() / (float)RAND_MAX;
|
||||
return (1.0f - t) * low + t * high;
|
||||
}
|
||||
|
||||
/// Utility function to tweak problem size for small GPUs
|
||||
int adjustProblemSize(int GPU_N, int default_nOptions) {
|
||||
int nOptions = default_nOptions;
|
||||
|
||||
// select problem size
|
||||
for (int i = 0; i < GPU_N; i++) {
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, i));
|
||||
int cudaCores = _ConvertSMVer2Cores(deviceProp.major, deviceProp.minor) *
|
||||
deviceProp.multiProcessorCount;
|
||||
|
||||
if (cudaCores <= 32) {
|
||||
nOptions = (nOptions < cudaCores / 2 ? nOptions : cudaCores / 2);
|
||||
}
|
||||
}
|
||||
|
||||
return nOptions;
|
||||
}
|
||||
|
||||
int adjustGridSize(int GPUIndex, int defaultGridSize) {
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, GPUIndex));
|
||||
int maxGridSize = deviceProp.multiProcessorCount * 40;
|
||||
return ((defaultGridSize > maxGridSize) ? maxGridSize : defaultGridSize);
|
||||
}
|
||||
|
||||
///////////////////////////////////////////////////////////////////////////////
|
||||
// CPU reference functions
|
||||
///////////////////////////////////////////////////////////////////////////////
|
||||
extern "C" void MonteCarloCPU(TOptionValue &callValue, TOptionData optionData,
|
||||
float *h_Random, int pathN);
|
||||
|
||||
// Black-Scholes formula for call options
|
||||
extern "C" void BlackScholesCall(float &CallResult, TOptionData optionData);
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// GPU-driving host thread
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// Timer
|
||||
StopWatchInterface **hTimer = NULL;
|
||||
|
||||
static CUT_THREADPROC solverThread(TOptionPlan *plan) {
|
||||
// Init GPU
|
||||
checkCudaErrors(cudaSetDevice(plan->device));
|
||||
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, plan->device));
|
||||
|
||||
// Start the timer
|
||||
sdkStartTimer(&hTimer[plan->device]);
|
||||
|
||||
// Allocate intermediate memory for MC integrator and initialize
|
||||
// RNG states
|
||||
initMonteCarloGPU(plan);
|
||||
|
||||
// Main computation
|
||||
MonteCarloGPU(plan);
|
||||
|
||||
checkCudaErrors(cudaDeviceSynchronize());
|
||||
|
||||
// Stop the timer
|
||||
sdkStopTimer(&hTimer[plan->device]);
|
||||
|
||||
// Shut down this GPU
|
||||
closeMonteCarloGPU(plan);
|
||||
|
||||
cudaStreamSynchronize(0);
|
||||
|
||||
printf("solverThread() finished - GPU Device %d: %s\n", plan->device,
|
||||
deviceProp.name);
|
||||
|
||||
CUT_THREADEND;
|
||||
}
|
||||
|
||||
static void multiSolver(TOptionPlan *plan, int nPlans) {
|
||||
// allocate and initialize an array of stream handles
|
||||
cudaStream_t *streams = (cudaStream_t *)malloc(nPlans * sizeof(cudaStream_t));
|
||||
cudaEvent_t *events = (cudaEvent_t *)malloc(nPlans * sizeof(cudaEvent_t));
|
||||
|
||||
for (int i = 0; i < nPlans; i++) {
|
||||
checkCudaErrors(cudaSetDevice(plan[i].device));
|
||||
checkCudaErrors(cudaStreamCreate(&(streams[i])));
|
||||
checkCudaErrors(cudaEventCreate(&(events[i])));
|
||||
}
|
||||
|
||||
// Init Each GPU
|
||||
// In CUDA 4.0 we can call cudaSetDevice multiple times to target each device
|
||||
// Set the device desired, then perform initializations on that device
|
||||
|
||||
for (int i = 0; i < nPlans; i++) {
|
||||
// set the target device to perform initialization on
|
||||
checkCudaErrors(cudaSetDevice(plan[i].device));
|
||||
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(cudaGetDeviceProperties(&deviceProp, plan[i].device));
|
||||
|
||||
// Allocate intermediate memory for MC integrator
|
||||
// and initialize RNG state
|
||||
initMonteCarloGPU(&plan[i]);
|
||||
}
|
||||
|
||||
for (int i = 0; i < nPlans; i++) {
|
||||
checkCudaErrors(cudaSetDevice(plan[i].device));
|
||||
checkCudaErrors(cudaDeviceSynchronize());
|
||||
}
|
||||
|
||||
// Start the timer
|
||||
sdkResetTimer(&hTimer[0]);
|
||||
sdkStartTimer(&hTimer[0]);
|
||||
|
||||
for (int i = 0; i < nPlans; i++) {
|
||||
checkCudaErrors(cudaSetDevice(plan[i].device));
|
||||
|
||||
// Main computations
|
||||
MonteCarloGPU(&plan[i], streams[i]);
|
||||
|
||||
checkCudaErrors(cudaEventRecord(events[i], streams[i]));
|
||||
}
|
||||
|
||||
for (int i = 0; i < nPlans; i++) {
|
||||
checkCudaErrors(cudaSetDevice(plan[i].device));
|
||||
cudaEventSynchronize(events[i]);
|
||||
}
|
||||
|
||||
// Stop the timer
|
||||
sdkStopTimer(&hTimer[0]);
|
||||
|
||||
for (int i = 0; i < nPlans; i++) {
|
||||
checkCudaErrors(cudaSetDevice(plan[i].device));
|
||||
closeMonteCarloGPU(&plan[i]);
|
||||
checkCudaErrors(cudaStreamDestroy(streams[i]));
|
||||
checkCudaErrors(cudaEventDestroy(events[i]));
|
||||
}
|
||||
}
|
||||
|
||||
///////////////////////////////////////////////////////////////////////////////
|
||||
// Main program
|
||||
///////////////////////////////////////////////////////////////////////////////
|
||||
#define DO_CPU
|
||||
#undef DO_CPU
|
||||
|
||||
#define PRINT_RESULTS
|
||||
#undef PRINT_RESULTS
|
||||
|
||||
void usage() {
|
||||
printf("--method=[threaded,streamed] --scaling=[strong,weak] [--help]\n");
|
||||
printf("Method=threaded: 1 CPU thread for each GPU [default]\n");
|
||||
printf(
|
||||
" streamed: 1 CPU thread handles all GPUs (requires CUDA 4.0 or "
|
||||
"newer)\n");
|
||||
printf("Scaling=strong : constant problem size\n");
|
||||
printf(
|
||||
" weak : problem size scales with number of available GPUs "
|
||||
"[default]\n");
|
||||
}
|
||||
|
||||
int main(int argc, char **argv) {
|
||||
char *multiMethodChoice = NULL;
|
||||
char *scalingChoice = NULL;
|
||||
bool use_threads = true;
|
||||
bool bqatest = false;
|
||||
bool strongScaling = false;
|
||||
|
||||
pArgc = &argc;
|
||||
pArgv = argv;
|
||||
|
||||
printf("%s Starting...\n\n", argv[0]);
|
||||
|
||||
if (checkCmdLineFlag(argc, (const char **)argv, "qatest")) {
|
||||
bqatest = true;
|
||||
}
|
||||
|
||||
getCmdLineArgumentString(argc, (const char **)argv, "method",
|
||||
&multiMethodChoice);
|
||||
getCmdLineArgumentString(argc, (const char **)argv, "scaling",
|
||||
&scalingChoice);
|
||||
|
||||
if (checkCmdLineFlag(argc, (const char **)argv, "h") ||
|
||||
checkCmdLineFlag(argc, (const char **)argv, "help")) {
|
||||
usage();
|
||||
exit(EXIT_SUCCESS);
|
||||
}
|
||||
|
||||
if (multiMethodChoice == NULL) {
|
||||
use_threads = false;
|
||||
} else {
|
||||
if (!strcasecmp(multiMethodChoice, "threaded")) {
|
||||
use_threads = true;
|
||||
} else {
|
||||
use_threads = false;
|
||||
}
|
||||
}
|
||||
|
||||
if (use_threads == false) {
|
||||
printf("Using single CPU thread for multiple GPUs\n");
|
||||
}
|
||||
|
||||
if (scalingChoice == NULL) {
|
||||
strongScaling = false;
|
||||
} else {
|
||||
if (!strcasecmp(scalingChoice, "strong")) {
|
||||
strongScaling = true;
|
||||
} else {
|
||||
strongScaling = false;
|
||||
}
|
||||
}
|
||||
|
||||
// GPU number present in the system
|
||||
int GPU_N;
|
||||
checkCudaErrors(cudaGetDeviceCount(&GPU_N));
|
||||
int nOptions = 8 * 1024;
|
||||
|
||||
nOptions = adjustProblemSize(GPU_N, nOptions);
|
||||
|
||||
// select problem size
|
||||
int scale = (strongScaling) ? 1 : GPU_N;
|
||||
int OPT_N = nOptions * scale;
|
||||
int PATH_N = 262144;
|
||||
|
||||
// initialize the timers
|
||||
hTimer = new StopWatchInterface *[GPU_N];
|
||||
|
||||
for (int i = 0; i < GPU_N; i++) {
|
||||
sdkCreateTimer(&hTimer[i]);
|
||||
sdkResetTimer(&hTimer[i]);
|
||||
}
|
||||
|
||||
// Input data array
|
||||
TOptionData *optionData = new TOptionData[OPT_N];
|
||||
// Final GPU MC results
|
||||
TOptionValue *callValueGPU = new TOptionValue[OPT_N];
|
||||
//"Theoretical" call values by Black-Scholes formula
|
||||
float *callValueBS = new float[OPT_N];
|
||||
// Solver config
|
||||
TOptionPlan *optionSolver = new TOptionPlan[GPU_N];
|
||||
// OS thread ID
|
||||
CUTThread *threadID = new CUTThread[GPU_N];
|
||||
|
||||
int gpuBase, gpuIndex;
|
||||
int i;
|
||||
|
||||
float time;
|
||||
|
||||
double delta, ref, sumDelta, sumRef, sumReserve;
|
||||
|
||||
printf("MonteCarloMultiGPU\n");
|
||||
printf("==================\n");
|
||||
printf("Parallelization method = %s\n",
|
||||
use_threads ? "threaded" : "streamed");
|
||||
printf("Problem scaling = %s\n", strongScaling ? "strong" : "weak");
|
||||
printf("Number of GPUs = %d\n", GPU_N);
|
||||
printf("Total number of options = %d\n", OPT_N);
|
||||
printf("Number of paths = %d\n", PATH_N);
|
||||
|
||||
printf("main(): generating input data...\n");
|
||||
srand(123);
|
||||
|
||||
for (i = 0; i < OPT_N; i++) {
|
||||
optionData[i].S = randFloat(5.0f, 50.0f);
|
||||
optionData[i].X = randFloat(10.0f, 25.0f);
|
||||
optionData[i].T = randFloat(1.0f, 5.0f);
|
||||
optionData[i].R = 0.06f;
|
||||
optionData[i].V = 0.10f;
|
||||
callValueGPU[i].Expected = -1.0f;
|
||||
callValueGPU[i].Confidence = -1.0f;
|
||||
}
|
||||
|
||||
printf("main(): starting %i host threads...\n", GPU_N);
|
||||
|
||||
// Get option count for each GPU
|
||||
for (i = 0; i < GPU_N; i++) {
|
||||
optionSolver[i].optionCount = OPT_N / GPU_N;
|
||||
}
|
||||
|
||||
// Take into account cases with "odd" option counts
|
||||
for (i = 0; i < (OPT_N % GPU_N); i++) {
|
||||
optionSolver[i].optionCount++;
|
||||
}
|
||||
|
||||
// Assign GPU option ranges
|
||||
gpuBase = 0;
|
||||
|
||||
for (i = 0; i < GPU_N; i++) {
|
||||
optionSolver[i].device = i;
|
||||
optionSolver[i].optionData = optionData + gpuBase;
|
||||
optionSolver[i].callValue = callValueGPU + gpuBase;
|
||||
optionSolver[i].pathN = PATH_N;
|
||||
optionSolver[i].gridSize =
|
||||
adjustGridSize(optionSolver[i].device, optionSolver[i].optionCount);
|
||||
gpuBase += optionSolver[i].optionCount;
|
||||
}
|
||||
|
||||
if (use_threads || bqatest) {
|
||||
// Start CPU thread for each GPU
|
||||
for (gpuIndex = 0; gpuIndex < GPU_N; gpuIndex++) {
|
||||
threadID[gpuIndex] = cutStartThread((CUT_THREADROUTINE)solverThread,
|
||||
&optionSolver[gpuIndex]);
|
||||
}
|
||||
|
||||
printf("main(): waiting for GPU results...\n");
|
||||
cutWaitForThreads(threadID, GPU_N);
|
||||
|
||||
printf("main(): GPU statistics, threaded\n");
|
||||
|
||||
for (i = 0; i < GPU_N; i++) {
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(
|
||||
cudaGetDeviceProperties(&deviceProp, optionSolver[i].device));
|
||||
printf("GPU Device #%i: %s\n", optionSolver[i].device, deviceProp.name);
|
||||
printf("Options : %i\n", optionSolver[i].optionCount);
|
||||
printf("Simulation paths: %i\n", optionSolver[i].pathN);
|
||||
time = sdkGetTimerValue(&hTimer[i]);
|
||||
printf("Total time (ms.): %f\n", time);
|
||||
printf("Options per sec.: %f\n", OPT_N / (time * 0.001));
|
||||
}
|
||||
|
||||
printf("main(): comparing Monte Carlo and Black-Scholes results...\n");
|
||||
sumDelta = 0;
|
||||
sumRef = 0;
|
||||
sumReserve = 0;
|
||||
|
||||
for (i = 0; i < OPT_N; i++) {
|
||||
BlackScholesCall(callValueBS[i], optionData[i]);
|
||||
delta = fabs(callValueBS[i] - callValueGPU[i].Expected);
|
||||
ref = callValueBS[i];
|
||||
sumDelta += delta;
|
||||
sumRef += fabs(ref);
|
||||
|
||||
if (delta > 1e-6) {
|
||||
sumReserve += callValueGPU[i].Confidence / delta;
|
||||
}
|
||||
|
||||
#ifdef PRINT_RESULTS
|
||||
printf("BS: %f; delta: %E\n", callValueBS[i], delta);
|
||||
#endif
|
||||
}
|
||||
|
||||
sumReserve /= OPT_N;
|
||||
}
|
||||
|
||||
if (!use_threads || bqatest) {
|
||||
multiSolver(optionSolver, GPU_N);
|
||||
|
||||
printf("main(): GPU statistics, streamed\n");
|
||||
|
||||
for (i = 0; i < GPU_N; i++) {
|
||||
cudaDeviceProp deviceProp;
|
||||
checkCudaErrors(
|
||||
cudaGetDeviceProperties(&deviceProp, optionSolver[i].device));
|
||||
printf("GPU Device #%i: %s\n", optionSolver[i].device, deviceProp.name);
|
||||
printf("Options : %i\n", optionSolver[i].optionCount);
|
||||
printf("Simulation paths: %i\n", optionSolver[i].pathN);
|
||||
}
|
||||
|
||||
time = sdkGetTimerValue(&hTimer[0]);
|
||||
printf("\nTotal time (ms.): %f\n", time);
|
||||
printf("\tNote: This is elapsed time for all to compute.\n");
|
||||
printf("Options per sec.: %f\n", OPT_N / (time * 0.001));
|
||||
|
||||
printf("main(): comparing Monte Carlo and Black-Scholes results...\n");
|
||||
sumDelta = 0;
|
||||
sumRef = 0;
|
||||
sumReserve = 0;
|
||||
|
||||
for (i = 0; i < OPT_N; i++) {
|
||||
BlackScholesCall(callValueBS[i], optionData[i]);
|
||||
delta = fabs(callValueBS[i] - callValueGPU[i].Expected);
|
||||
ref = callValueBS[i];
|
||||
sumDelta += delta;
|
||||
sumRef += fabs(ref);
|
||||
|
||||
if (delta > 1e-6) {
|
||||
sumReserve += callValueGPU[i].Confidence / delta;
|
||||
}
|
||||
|
||||
#ifdef PRINT_RESULTS
|
||||
printf("BS: %f; delta: %E\n", callValueBS[i], delta);
|
||||
#endif
|
||||
}
|
||||
|
||||
sumReserve /= OPT_N;
|
||||
}
|
||||
|
||||
#ifdef DO_CPU
|
||||
printf("main(): running CPU MonteCarlo...\n");
|
||||
TOptionValue callValueCPU;
|
||||
sumDelta = 0;
|
||||
sumRef = 0;
|
||||
|
||||
for (i = 0; i < OPT_N; i++) {
|
||||
MonteCarloCPU(callValueCPU, optionData[i], NULL, PATH_N);
|
||||
delta = fabs(callValueCPU.Expected - callValueGPU[i].Expected);
|
||||
ref = callValueCPU.Expected;
|
||||
sumDelta += delta;
|
||||
sumRef += fabs(ref);
|
||||
printf("Exp : %f | %f\t", callValueCPU.Expected, callValueGPU[i].Expected);
|
||||
printf("Conf: %f | %f\n", callValueCPU.Confidence,
|
||||
callValueGPU[i].Confidence);
|
||||
}
|
||||
|
||||
printf("L1 norm: %E\n", sumDelta / sumRef);
|
||||
#endif
|
||||
|
||||
printf("Shutting down...\n");
|
||||
|
||||
for (int i = 0; i < GPU_N; i++) {
|
||||
sdkStartTimer(&hTimer[i]);
|
||||
checkCudaErrors(cudaSetDevice(i));
|
||||
}
|
||||
|
||||
delete[] optionSolver;
|
||||
delete[] callValueBS;
|
||||
delete[] callValueGPU;
|
||||
delete[] optionData;
|
||||
delete[] threadID;
|
||||
delete[] hTimer;
|
||||
|
||||
printf("Test Summary...\n");
|
||||
printf("L1 norm : %E\n", sumDelta / sumRef);
|
||||
printf("Average reserve: %f\n", sumReserve);
|
||||
printf(
|
||||
"\nNOTE: The CUDA Samples are not meant for performance measurements. "
|
||||
"Results may vary when GPU Boost is enabled.\n\n");
|
||||
printf(sumReserve > 1.0f ? "Test passed\n" : "Test failed!\n");
|
||||
exit(sumReserve > 1.0f ? EXIT_SUCCESS : EXIT_FAILURE);
|
||||
}
|
||||
20
Samples/MonteCarloMultiGPU/MonteCarloMultiGPU_vs2017.sln
Normal file
20
Samples/MonteCarloMultiGPU/MonteCarloMultiGPU_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}") = "MonteCarloMultiGPU", "MonteCarloMultiGPU_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
|
||||
119
Samples/MonteCarloMultiGPU/MonteCarloMultiGPU_vs2017.vcxproj
Normal file
119
Samples/MonteCarloMultiGPU/MonteCarloMultiGPU_vs2017.vcxproj
Normal file
@@ -0,0 +1,119 @@
|
||||
<?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>MonteCarloMultiGPU_vs2017</RootNamespace>
|
||||
<ProjectName>MonteCarloMultiGPU</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(WindowsTargetPlatformVersion)'==''">
|
||||
<LatestTargetPlatformVersion>$([Microsoft.Build.Utilities.ToolLocationHelper]::GetLatestSDKTargetPlatformVersion('Windows', '10.0'))</LatestTargetPlatformVersion>
|
||||
<WindowsTargetPlatformVersion Condition="'$(WindowsTargetPlatformVersion)' == ''">$(LatestTargetPlatformVersion)</WindowsTargetPlatformVersion>
|
||||
<TargetPlatformVersion>$(WindowsTargetPlatformVersion)</TargetPlatformVersion>
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v141</PlatformToolset>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<UseDebugLibraries>true</UseDebugLibraries>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Release'">
|
||||
<WholeProgramOptimization>true</WholeProgramOptimization>
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
|
||||
<ImportGroup Label="ExtensionSettings">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.5.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>curand.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/MonteCarloMultiGPU.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;compute_80,sm_80;compute_86,sm_86;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<AdditionalOptions>--threads 0</AdditionalOptions>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<ClCompile Include="MonteCarloMultiGPU.cpp" />
|
||||
<ClCompile Include="MonteCarlo_gold.cpp" />
|
||||
<CudaCompile Include="MonteCarlo_kernel.cu" />
|
||||
<ClCompile Include="multithreading.cpp" />
|
||||
<ClInclude Include="MonteCarlo_common.h" />
|
||||
<ClInclude Include="multithreading.h" />
|
||||
<ClInclude Include="realtype.h" />
|
||||
<None Include="MonteCarlo_reduction.cuh" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.5.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
20
Samples/MonteCarloMultiGPU/MonteCarloMultiGPU_vs2019.sln
Normal file
20
Samples/MonteCarloMultiGPU/MonteCarloMultiGPU_vs2019.sln
Normal file
@@ -0,0 +1,20 @@
|
||||
|
||||
Microsoft Visual Studio Solution File, Format Version 12.00
|
||||
# Visual Studio 2019
|
||||
Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "MonteCarloMultiGPU", "MonteCarloMultiGPU_vs2019.vcxproj", "{997E0757-EA74-4A4E-A0FC-47D8C8831A15}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|x64 = Debug|x64
|
||||
Release|x64 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(ProjectConfigurationPlatforms) = postSolution
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.ActiveCfg = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Debug|x64.Build.0 = Debug|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.ActiveCfg = Release|x64
|
||||
{997E0757-EA74-4A4E-A0FC-47D8C8831A15}.Release|x64.Build.0 = Release|x64
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
EndGlobalSection
|
||||
EndGlobal
|
||||
115
Samples/MonteCarloMultiGPU/MonteCarloMultiGPU_vs2019.vcxproj
Normal file
115
Samples/MonteCarloMultiGPU/MonteCarloMultiGPU_vs2019.vcxproj
Normal file
@@ -0,0 +1,115 @@
|
||||
<?xml version="1.0" encoding="utf-8"?>
|
||||
<Project DefaultTargets="Build" ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
|
||||
<PropertyGroup>
|
||||
<CUDAPropsPath Condition="'$(CUDAPropsPath)'==''">$(VCTargetsPath)\BuildCustomizations</CUDAPropsPath>
|
||||
</PropertyGroup>
|
||||
<ItemGroup Label="ProjectConfigurations">
|
||||
<ProjectConfiguration Include="Debug|x64">
|
||||
<Configuration>Debug</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
<ProjectConfiguration Include="Release|x64">
|
||||
<Configuration>Release</Configuration>
|
||||
<Platform>x64</Platform>
|
||||
</ProjectConfiguration>
|
||||
</ItemGroup>
|
||||
<PropertyGroup Label="Globals">
|
||||
<ProjectGuid>{997E0757-EA74-4A4E-A0FC-47D8C8831A15}</ProjectGuid>
|
||||
<RootNamespace>MonteCarloMultiGPU_vs2019</RootNamespace>
|
||||
<ProjectName>MonteCarloMultiGPU</ProjectName>
|
||||
<CudaToolkitCustomDir />
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
|
||||
<PropertyGroup>
|
||||
<ConfigurationType>Application</ConfigurationType>
|
||||
<CharacterSet>MultiByte</CharacterSet>
|
||||
<PlatformToolset>v142</PlatformToolset>
|
||||
<WindowsTargetPlatformVersion>10.0</WindowsTargetPlatformVersion>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<UseDebugLibraries>true</UseDebugLibraries>
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Configuration)'=='Release'">
|
||||
<WholeProgramOptimization>true</WholeProgramOptimization>
|
||||
</PropertyGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
|
||||
<ImportGroup Label="ExtensionSettings">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.5.props" />
|
||||
</ImportGroup>
|
||||
<ImportGroup Label="PropertySheets">
|
||||
<Import Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" />
|
||||
</ImportGroup>
|
||||
<PropertyGroup Label="UserMacros" />
|
||||
<PropertyGroup>
|
||||
<IntDir>$(Platform)/$(Configuration)/</IntDir>
|
||||
<IncludePath>$(IncludePath)</IncludePath>
|
||||
<CodeAnalysisRuleSet>AllRules.ruleset</CodeAnalysisRuleSet>
|
||||
<CodeAnalysisRules />
|
||||
<CodeAnalysisRuleAssemblies />
|
||||
</PropertyGroup>
|
||||
<PropertyGroup Condition="'$(Platform)'=='x64'">
|
||||
<OutDir>../../bin/win64/$(Configuration)/</OutDir>
|
||||
</PropertyGroup>
|
||||
<ItemDefinitionGroup>
|
||||
<ClCompile>
|
||||
<WarningLevel>Level3</WarningLevel>
|
||||
<PreprocessorDefinitions>WIN32;_MBCS;%(PreprocessorDefinitions)</PreprocessorDefinitions>
|
||||
<AdditionalIncludeDirectories>./;$(CudaToolkitDir)/include;../../Common;$(CudaToolkitIncludeDir);</AdditionalIncludeDirectories>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<SubSystem>Console</SubSystem>
|
||||
<AdditionalDependencies>curand.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
|
||||
<AdditionalLibraryDirectories>$(CudaToolkitLibDir);</AdditionalLibraryDirectories>
|
||||
<OutputFile>$(OutDir)/MonteCarloMultiGPU.exe</OutputFile>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<CodeGeneration>compute_35,sm_35;compute_37,sm_37;compute_50,sm_50;compute_52,sm_52;compute_60,sm_60;compute_61,sm_61;compute_70,sm_70;compute_75,sm_75;compute_80,sm_80;compute_86,sm_86;</CodeGeneration>
|
||||
<AdditionalOptions>-Xcompiler "/wd 4819" %(AdditionalOptions)</AdditionalOptions>
|
||||
<Include>./;../../Common</Include>
|
||||
<Defines>WIN32</Defines>
|
||||
<AdditionalOptions>--threads 0</AdditionalOptions>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
|
||||
<ClCompile>
|
||||
<Optimization>Disabled</Optimization>
|
||||
<RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>true</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>Default</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MTd</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
|
||||
<ClCompile>
|
||||
<Optimization>MaxSpeed</Optimization>
|
||||
<RuntimeLibrary>MultiThreaded</RuntimeLibrary>
|
||||
</ClCompile>
|
||||
<Link>
|
||||
<GenerateDebugInformation>false</GenerateDebugInformation>
|
||||
<LinkTimeCodeGeneration>UseLinkTimeCodeGeneration</LinkTimeCodeGeneration>
|
||||
</Link>
|
||||
<CudaCompile>
|
||||
<Runtime>MT</Runtime>
|
||||
<TargetMachinePlatform>64</TargetMachinePlatform>
|
||||
</CudaCompile>
|
||||
</ItemDefinitionGroup>
|
||||
<ItemGroup>
|
||||
<ClCompile Include="MonteCarloMultiGPU.cpp" />
|
||||
<ClCompile Include="MonteCarlo_gold.cpp" />
|
||||
<CudaCompile Include="MonteCarlo_kernel.cu" />
|
||||
<ClCompile Include="multithreading.cpp" />
|
||||
<ClInclude Include="MonteCarlo_common.h" />
|
||||
<ClInclude Include="multithreading.h" />
|
||||
<ClInclude Include="realtype.h" />
|
||||
<None Include="MonteCarlo_reduction.cuh" />
|
||||
</ItemGroup>
|
||||
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
|
||||
<ImportGroup Label="ExtensionTargets">
|
||||
<Import Project="$(CUDAPropsPath)\CUDA 11.5.targets" />
|
||||
</ImportGroup>
|
||||
</Project>
|
||||
98
Samples/MonteCarloMultiGPU/MonteCarlo_common.h
Normal file
98
Samples/MonteCarloMultiGPU/MonteCarlo_common.h
Normal file
@@ -0,0 +1,98 @@
|
||||
/* Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved.
|
||||
*
|
||||
* Redistribution and use in source and binary forms, with or without
|
||||
* modification, are permitted provided that the following conditions
|
||||
* are met:
|
||||
* * Redistributions of source code must retain the above copyright
|
||||
* notice, this list of conditions and the following disclaimer.
|
||||
* * Redistributions in binary form must reproduce the above copyright
|
||||
* notice, this list of conditions and the following disclaimer in the
|
||||
* documentation and/or other materials provided with the distribution.
|
||||
* * Neither the name of NVIDIA CORPORATION nor the names of its
|
||||
* contributors may be used to endorse or promote products derived
|
||||
* from this software without specific prior written permission.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
||||
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||||
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
||||
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
||||
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||||
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||||
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
||||
* OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
*/
|
||||
|
||||
#ifndef MONTECARLO_COMMON_H
|
||||
#define MONTECARLO_COMMON_H
|
||||
#include "realtype.h"
|
||||
#include "curand_kernel.h"
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// Global types
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
typedef struct {
|
||||
float S;
|
||||
float X;
|
||||
float T;
|
||||
float R;
|
||||
float V;
|
||||
} TOptionData;
|
||||
|
||||
typedef struct
|
||||
//#ifdef __CUDACC__
|
||||
//__align__(8)
|
||||
//#endif
|
||||
{
|
||||
float Expected;
|
||||
float Confidence;
|
||||
} TOptionValue;
|
||||
|
||||
// GPU outputs before CPU postprocessing
|
||||
typedef struct {
|
||||
real Expected;
|
||||
real Confidence;
|
||||
} __TOptionValue;
|
||||
|
||||
typedef struct {
|
||||
// Device ID for multi-GPU version
|
||||
int device;
|
||||
// Option count for this plan
|
||||
int optionCount;
|
||||
|
||||
// Host-side data source and result destination
|
||||
TOptionData *optionData;
|
||||
TOptionValue *callValue;
|
||||
|
||||
// Temporary Host-side pinned memory for async + faster data transfers
|
||||
__TOptionValue *h_CallValue;
|
||||
|
||||
// Device- and host-side option data
|
||||
void *d_OptionData;
|
||||
void *h_OptionData;
|
||||
|
||||
// Device-side option values
|
||||
void *d_CallValue;
|
||||
|
||||
// Intermediate device-side buffers
|
||||
void *d_Buffer;
|
||||
|
||||
// random number generator states
|
||||
curandState *rngStates;
|
||||
|
||||
// Pseudorandom samples count
|
||||
int pathN;
|
||||
|
||||
// Time stamp
|
||||
float time;
|
||||
|
||||
int gridSize;
|
||||
} TOptionPlan;
|
||||
|
||||
extern "C" void initMonteCarloGPU(TOptionPlan *plan);
|
||||
extern "C" void MonteCarloGPU(TOptionPlan *plan, cudaStream_t stream = 0);
|
||||
extern "C" void closeMonteCarloGPU(TOptionPlan *plan);
|
||||
|
||||
#endif
|
||||
140
Samples/MonteCarloMultiGPU/MonteCarlo_gold.cpp
Normal file
140
Samples/MonteCarloMultiGPU/MonteCarlo_gold.cpp
Normal file
@@ -0,0 +1,140 @@
|
||||
/* Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved.
|
||||
*
|
||||
* Redistribution and use in source and binary forms, with or without
|
||||
* modification, are permitted provided that the following conditions
|
||||
* are met:
|
||||
* * Redistributions of source code must retain the above copyright
|
||||
* notice, this list of conditions and the following disclaimer.
|
||||
* * Redistributions in binary form must reproduce the above copyright
|
||||
* notice, this list of conditions and the following disclaimer in the
|
||||
* documentation and/or other materials provided with the distribution.
|
||||
* * Neither the name of NVIDIA CORPORATION nor the names of its
|
||||
* contributors may be used to endorse or promote products derived
|
||||
* from this software without specific prior written permission.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
||||
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||||
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
||||
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
||||
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||||
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||||
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
||||
* OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
*/
|
||||
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <math.h>
|
||||
|
||||
#include <curand.h>
|
||||
|
||||
//#include "curand_kernel.h"
|
||||
#include "helper_cuda.h"
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// Common types
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
#include "MonteCarlo_common.h"
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// Black-Scholes formula for Monte Carlo results validation
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
#define A1 0.31938153
|
||||
#define A2 -0.356563782
|
||||
#define A3 1.781477937
|
||||
#define A4 -1.821255978
|
||||
#define A5 1.330274429
|
||||
#define RSQRT2PI 0.39894228040143267793994605993438
|
||||
|
||||
// Polynomial approximation of
|
||||
// cumulative normal distribution function
|
||||
double CND(double d) {
|
||||
double K = 1.0 / (1.0 + 0.2316419 * fabs(d));
|
||||
|
||||
double cnd = RSQRT2PI * exp(-0.5 * d * d) *
|
||||
(K * (A1 + K * (A2 + K * (A3 + K * (A4 + K * A5)))));
|
||||
|
||||
if (d > 0) cnd = 1.0 - cnd;
|
||||
|
||||
return cnd;
|
||||
}
|
||||
|
||||
// Black-Scholes formula for call value
|
||||
extern "C" void BlackScholesCall(float &callValue, TOptionData optionData) {
|
||||
double S = optionData.S;
|
||||
double X = optionData.X;
|
||||
double T = optionData.T;
|
||||
double R = optionData.R;
|
||||
double V = optionData.V;
|
||||
|
||||
double sqrtT = sqrt(T);
|
||||
double d1 = (log(S / X) + (R + 0.5 * V * V) * T) / (V * sqrtT);
|
||||
double d2 = d1 - V * sqrtT;
|
||||
double CNDD1 = CND(d1);
|
||||
double CNDD2 = CND(d2);
|
||||
double expRT = exp(-R * T);
|
||||
|
||||
callValue = (float)(S * CNDD1 - X * expRT * CNDD2);
|
||||
}
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// CPU Monte Carlo
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
static double endCallValue(double S, double X, double r, double MuByT,
|
||||
double VBySqrtT) {
|
||||
double callValue = S * exp(MuByT + VBySqrtT * r) - X;
|
||||
return (callValue > 0) ? callValue : 0;
|
||||
}
|
||||
|
||||
extern "C" void MonteCarloCPU(TOptionValue &callValue, TOptionData optionData,
|
||||
float *h_Samples, int pathN) {
|
||||
const double S = optionData.S;
|
||||
const double X = optionData.X;
|
||||
const double T = optionData.T;
|
||||
const double R = optionData.R;
|
||||
const double V = optionData.V;
|
||||
const double MuByT = (R - 0.5 * V * V) * T;
|
||||
const double VBySqrtT = V * sqrt(T);
|
||||
|
||||
float *samples;
|
||||
curandGenerator_t gen;
|
||||
|
||||
checkCudaErrors(curandCreateGeneratorHost(&gen, CURAND_RNG_PSEUDO_DEFAULT));
|
||||
unsigned long long seed = 1234ULL;
|
||||
checkCudaErrors(curandSetPseudoRandomGeneratorSeed(gen, seed));
|
||||
|
||||
if (h_Samples != NULL) {
|
||||
samples = h_Samples;
|
||||
} else {
|
||||
samples = (float *)malloc(pathN * sizeof(float));
|
||||
checkCudaErrors(curandGenerateNormal(gen, samples, pathN, 0.0, 1.0));
|
||||
}
|
||||
|
||||
// for(int i=0; i<10; i++) printf("CPU sample = %f\n", samples[i]);
|
||||
|
||||
double sum = 0, sum2 = 0;
|
||||
|
||||
for (int pos = 0; pos < pathN; pos++) {
|
||||
double sample = samples[pos];
|
||||
double callValue = endCallValue(S, X, sample, MuByT, VBySqrtT);
|
||||
sum += callValue;
|
||||
sum2 += callValue * callValue;
|
||||
}
|
||||
|
||||
if (h_Samples == NULL) free(samples);
|
||||
|
||||
checkCudaErrors(curandDestroyGenerator(gen));
|
||||
|
||||
// Derive average from the total sum and discount by riskfree rate
|
||||
callValue.Expected = (float)(exp(-R * T) * sum / (double)pathN);
|
||||
// Standard deviation
|
||||
double stdDev = sqrt(((double)pathN * sum2 - sum * sum) /
|
||||
((double)pathN * (double)(pathN - 1)));
|
||||
// Confidence width; in 95% of all cases theoretical value lies within these
|
||||
// borders
|
||||
callValue.Confidence =
|
||||
(float)(exp(-R * T) * 1.96 * stdDev / sqrt((double)pathN));
|
||||
}
|
||||
226
Samples/MonteCarloMultiGPU/MonteCarlo_kernel.cu
Normal file
226
Samples/MonteCarloMultiGPU/MonteCarlo_kernel.cu
Normal file
@@ -0,0 +1,226 @@
|
||||
/* Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved.
|
||||
*
|
||||
* Redistribution and use in source and binary forms, with or without
|
||||
* modification, are permitted provided that the following conditions
|
||||
* are met:
|
||||
* * Redistributions of source code must retain the above copyright
|
||||
* notice, this list of conditions and the following disclaimer.
|
||||
* * Redistributions in binary form must reproduce the above copyright
|
||||
* notice, this list of conditions and the following disclaimer in the
|
||||
* documentation and/or other materials provided with the distribution.
|
||||
* * Neither the name of NVIDIA CORPORATION nor the names of its
|
||||
* contributors may be used to endorse or promote products derived
|
||||
* from this software without specific prior written permission.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
||||
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||||
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
||||
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
||||
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||||
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||||
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
||||
* OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
*/
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// Global types
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
#include <stdlib.h>
|
||||
#include <stdio.h>
|
||||
#include <cooperative_groups.h>
|
||||
|
||||
namespace cg = cooperative_groups;
|
||||
#include <helper_cuda.h>
|
||||
#include <curand_kernel.h>
|
||||
#include "MonteCarlo_common.h"
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// Helper reduction template
|
||||
// Please see the "reduction" CUDA Sample for more information
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
#include "MonteCarlo_reduction.cuh"
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// Internal GPU-side data structures
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
#define MAX_OPTIONS (1024 * 1024)
|
||||
|
||||
// Preprocessed input option data
|
||||
typedef struct {
|
||||
real S;
|
||||
real X;
|
||||
real MuByT;
|
||||
real VBySqrtT;
|
||||
} __TOptionData;
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// Overloaded shortcut payoff functions for different precision modes
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
__device__ inline float endCallValue(float S, float X, float r, float MuByT,
|
||||
float VBySqrtT) {
|
||||
float callValue = S * __expf(MuByT + VBySqrtT * r) - X;
|
||||
return (callValue > 0.0F) ? callValue : 0.0F;
|
||||
}
|
||||
|
||||
__device__ inline double endCallValue(double S, double X, double r,
|
||||
double MuByT, double VBySqrtT) {
|
||||
double callValue = S * exp(MuByT + VBySqrtT * r) - X;
|
||||
return (callValue > 0.0) ? callValue : 0.0;
|
||||
}
|
||||
|
||||
#define THREAD_N 256
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// This kernel computes the integral over all paths using a single thread block
|
||||
// per option. It is fastest when the number of thread blocks times the work per
|
||||
// block is high enough to keep the GPU busy.
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
static __global__ void MonteCarloOneBlockPerOption(
|
||||
curandState *__restrict rngStates,
|
||||
const __TOptionData *__restrict d_OptionData,
|
||||
__TOptionValue *__restrict d_CallValue, int pathN, int optionN) {
|
||||
// Handle to thread block group
|
||||
cg::thread_block cta = cg::this_thread_block();
|
||||
cg::thread_block_tile<32> tile32 = cg::tiled_partition<32>(cta);
|
||||
|
||||
const int SUM_N = THREAD_N;
|
||||
__shared__ real s_SumCall[SUM_N];
|
||||
__shared__ real s_Sum2Call[SUM_N];
|
||||
|
||||
// determine global thread id
|
||||
int tid = threadIdx.x + blockIdx.x * blockDim.x;
|
||||
|
||||
// Copy random number state to local memory for efficiency
|
||||
curandState localState = rngStates[tid];
|
||||
for (int optionIndex = blockIdx.x; optionIndex < optionN;
|
||||
optionIndex += gridDim.x) {
|
||||
const real S = d_OptionData[optionIndex].S;
|
||||
const real X = d_OptionData[optionIndex].X;
|
||||
const real MuByT = d_OptionData[optionIndex].MuByT;
|
||||
const real VBySqrtT = d_OptionData[optionIndex].VBySqrtT;
|
||||
|
||||
// Cycle through the entire samples array:
|
||||
// derive end stock price for each path
|
||||
// accumulate partial integrals into intermediate shared memory buffer
|
||||
for (int iSum = threadIdx.x; iSum < SUM_N; iSum += blockDim.x) {
|
||||
__TOptionValue sumCall = {0, 0};
|
||||
|
||||
#pragma unroll 8
|
||||
for (int i = iSum; i < pathN; i += SUM_N) {
|
||||
real r = curand_normal(&localState);
|
||||
real callValue = endCallValue(S, X, r, MuByT, VBySqrtT);
|
||||
sumCall.Expected += callValue;
|
||||
sumCall.Confidence += callValue * callValue;
|
||||
}
|
||||
|
||||
s_SumCall[iSum] = sumCall.Expected;
|
||||
s_Sum2Call[iSum] = sumCall.Confidence;
|
||||
}
|
||||
|
||||
// Reduce shared memory accumulators
|
||||
// and write final result to global memory
|
||||
cg::sync(cta);
|
||||
sumReduce<real, SUM_N, THREAD_N>(s_SumCall, s_Sum2Call, cta, tile32,
|
||||
&d_CallValue[optionIndex]);
|
||||
}
|
||||
}
|
||||
|
||||
static __global__ void rngSetupStates(curandState *rngState, int device_id) {
|
||||
// determine global thread id
|
||||
int tid = threadIdx.x + blockIdx.x * blockDim.x;
|
||||
// Each threadblock gets different seed,
|
||||
// Threads within a threadblock get different sequence numbers
|
||||
curand_init(blockIdx.x + gridDim.x * device_id, threadIdx.x, 0,
|
||||
&rngState[tid]);
|
||||
}
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// Host-side interface to GPU Monte Carlo
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
|
||||
extern "C" void initMonteCarloGPU(TOptionPlan *plan) {
|
||||
checkCudaErrors(cudaMalloc(&plan->d_OptionData,
|
||||
sizeof(__TOptionData) * (plan->optionCount)));
|
||||
checkCudaErrors(cudaMalloc(&plan->d_CallValue,
|
||||
sizeof(__TOptionValue) * (plan->optionCount)));
|
||||
checkCudaErrors(cudaMallocHost(&plan->h_OptionData,
|
||||
sizeof(__TOptionData) * (plan->optionCount)));
|
||||
// Allocate internal device memory
|
||||
checkCudaErrors(cudaMallocHost(&plan->h_CallValue,
|
||||
sizeof(__TOptionValue) * (plan->optionCount)));
|
||||
// Allocate states for pseudo random number generators
|
||||
checkCudaErrors(cudaMalloc((void **)&plan->rngStates,
|
||||
plan->gridSize * THREAD_N * sizeof(curandState)));
|
||||
checkCudaErrors(cudaMemset(plan->rngStates, 0,
|
||||
plan->gridSize * THREAD_N * sizeof(curandState)));
|
||||
|
||||
// place each device pathN random numbers apart on the random number sequence
|
||||
rngSetupStates<<<plan->gridSize, THREAD_N>>>(plan->rngStates, plan->device);
|
||||
getLastCudaError("rngSetupStates kernel failed.\n");
|
||||
}
|
||||
|
||||
// Compute statistics and deallocate internal device memory
|
||||
extern "C" void closeMonteCarloGPU(TOptionPlan *plan) {
|
||||
for (int i = 0; i < plan->optionCount; i++) {
|
||||
const double RT = plan->optionData[i].R * plan->optionData[i].T;
|
||||
const double sum = plan->h_CallValue[i].Expected;
|
||||
const double sum2 = plan->h_CallValue[i].Confidence;
|
||||
const double pathN = plan->pathN;
|
||||
// Derive average from the total sum and discount by riskfree rate
|
||||
plan->callValue[i].Expected = (float)(exp(-RT) * sum / pathN);
|
||||
// Standard deviation
|
||||
double stdDev = sqrt((pathN * sum2 - sum * sum) / (pathN * (pathN - 1)));
|
||||
// Confidence width; in 95% of all cases theoretical value lies within these
|
||||
// borders
|
||||
plan->callValue[i].Confidence =
|
||||
(float)(exp(-RT) * 1.96 * stdDev / sqrt(pathN));
|
||||
}
|
||||
|
||||
checkCudaErrors(cudaFree(plan->rngStates));
|
||||
checkCudaErrors(cudaFreeHost(plan->h_CallValue));
|
||||
checkCudaErrors(cudaFreeHost(plan->h_OptionData));
|
||||
checkCudaErrors(cudaFree(plan->d_CallValue));
|
||||
checkCudaErrors(cudaFree(plan->d_OptionData));
|
||||
}
|
||||
|
||||
// Main computations
|
||||
extern "C" void MonteCarloGPU(TOptionPlan *plan, cudaStream_t stream) {
|
||||
__TOptionValue *h_CallValue = plan->h_CallValue;
|
||||
|
||||
if (plan->optionCount <= 0 || plan->optionCount > MAX_OPTIONS) {
|
||||
printf("MonteCarloGPU(): bad option count.\n");
|
||||
return;
|
||||
}
|
||||
|
||||
__TOptionData *h_OptionData = (__TOptionData *)plan->h_OptionData;
|
||||
|
||||
for (int i = 0; i < plan->optionCount; i++) {
|
||||
const double T = plan->optionData[i].T;
|
||||
const double R = plan->optionData[i].R;
|
||||
const double V = plan->optionData[i].V;
|
||||
const double MuByT = (R - 0.5 * V * V) * T;
|
||||
const double VBySqrtT = V * sqrt(T);
|
||||
h_OptionData[i].S = (real)plan->optionData[i].S;
|
||||
h_OptionData[i].X = (real)plan->optionData[i].X;
|
||||
h_OptionData[i].MuByT = (real)MuByT;
|
||||
h_OptionData[i].VBySqrtT = (real)VBySqrtT;
|
||||
}
|
||||
|
||||
checkCudaErrors(cudaMemcpyAsync(plan->d_OptionData, h_OptionData,
|
||||
plan->optionCount * sizeof(__TOptionData),
|
||||
cudaMemcpyHostToDevice, stream));
|
||||
|
||||
MonteCarloOneBlockPerOption<<<plan->gridSize, THREAD_N, 0, stream>>>(
|
||||
plan->rngStates, (__TOptionData *)(plan->d_OptionData),
|
||||
(__TOptionValue *)(plan->d_CallValue), plan->pathN, plan->optionCount);
|
||||
getLastCudaError("MonteCarloOneBlockPerOption() execution failed\n");
|
||||
|
||||
checkCudaErrors(cudaMemcpyAsync(h_CallValue, plan->d_CallValue,
|
||||
plan->optionCount * sizeof(__TOptionValue),
|
||||
cudaMemcpyDeviceToHost, stream));
|
||||
|
||||
// cudaDeviceSynchronize();
|
||||
}
|
||||
79
Samples/MonteCarloMultiGPU/MonteCarlo_reduction.cuh
Normal file
79
Samples/MonteCarloMultiGPU/MonteCarlo_reduction.cuh
Normal file
@@ -0,0 +1,79 @@
|
||||
/* Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved.
|
||||
*
|
||||
* Redistribution and use in source and binary forms, with or without
|
||||
* modification, are permitted provided that the following conditions
|
||||
* are met:
|
||||
* * Redistributions of source code must retain the above copyright
|
||||
* notice, this list of conditions and the following disclaimer.
|
||||
* * Redistributions in binary form must reproduce the above copyright
|
||||
* notice, this list of conditions and the following disclaimer in the
|
||||
* documentation and/or other materials provided with the distribution.
|
||||
* * Neither the name of NVIDIA CORPORATION nor the names of its
|
||||
* contributors may be used to endorse or promote products derived
|
||||
* from this software without specific prior written permission.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
||||
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||||
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
||||
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
||||
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||||
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||||
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
||||
* OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
*/
|
||||
|
||||
#ifndef MONTECARLO_REDUCTION_CUH
|
||||
#define MONTECARLO_REDUCTION_CUH
|
||||
|
||||
#include <cooperative_groups.h>
|
||||
|
||||
namespace cg = cooperative_groups;
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
// This function calculates total sum for each of the two input arrays.
|
||||
// SUM_N must be power of two
|
||||
// Unrolling provides a bit of a performance improvement for small
|
||||
// to medium path counts.
|
||||
////////////////////////////////////////////////////////////////////////////////
|
||||
|
||||
template <class T, int SUM_N, int blockSize>
|
||||
__device__ void sumReduce(T *sum, T *sum2, cg::thread_block &cta,
|
||||
cg::thread_block_tile<32> &tile32,
|
||||
__TOptionValue *d_CallValue) {
|
||||
const int VEC = 32;
|
||||
const int tid = cta.thread_rank();
|
||||
|
||||
T beta = sum[tid];
|
||||
T beta2 = sum2[tid];
|
||||
T temp, temp2;
|
||||
|
||||
for (int i = VEC / 2; i > 0; i >>= 1) {
|
||||
if (tile32.thread_rank() < i) {
|
||||
temp = sum[tid + i];
|
||||
temp2 = sum2[tid + i];
|
||||
beta += temp;
|
||||
beta2 += temp2;
|
||||
sum[tid] = beta;
|
||||
sum2[tid] = beta2;
|
||||
}
|
||||
cg::sync(tile32);
|
||||
}
|
||||
cg::sync(cta);
|
||||
|
||||
if (tid == 0) {
|
||||
beta = 0;
|
||||
beta2 = 0;
|
||||
for (int i = 0; i < blockDim.x; i += VEC) {
|
||||
beta += sum[i];
|
||||
beta2 += sum2[i];
|
||||
}
|
||||
__TOptionValue t = {beta, beta2};
|
||||
*d_CallValue = t;
|
||||
}
|
||||
cg::sync(cta);
|
||||
}
|
||||
|
||||
#endif
|
||||
72
Samples/MonteCarloMultiGPU/NsightEclipse.xml
Normal file
72
Samples/MonteCarloMultiGPU/NsightEclipse.xml
Normal file
@@ -0,0 +1,72 @@
|
||||
<?xml version="1.0" encoding="UTF-8"?>
|
||||
<!DOCTYPE entry SYSTEM "SamplesInfo.dtd">
|
||||
<entry>
|
||||
<name>MonteCarloMultiGPU</name>
|
||||
<description><![CDATA[This sample evaluates fair call price for a given set of European options using the Monte Carlo approach, taking advantage of all CUDA-capable GPUs installed in the system. This sample use double precision hardware if a GTX 200 class GPU is present. The sample also takes advantage of CUDA 4.0 capability to supporting using a single CPU thread to control multiple GPUs]]></description>
|
||||
<devicecompilation>whole</devicecompilation>
|
||||
<includepaths>
|
||||
<path>./</path>
|
||||
<path>../</path>
|
||||
<path>../../Common</path>
|
||||
</includepaths>
|
||||
<keywords>
|
||||
<keyword>CUDA</keyword>
|
||||
<keyword>CURAND</keyword>
|
||||
<keyword>Computational Finance</keyword>
|
||||
<keyword>option pricing</keyword>
|
||||
<keyword>Monte Carlo</keyword>
|
||||
<keyword>multi-GPU support</keyword>
|
||||
</keywords>
|
||||
<libraries>
|
||||
<library>curand</library>
|
||||
</libraries>
|
||||
<librarypaths>
|
||||
</librarypaths>
|
||||
<nsight_eclipse>true</nsight_eclipse>
|
||||
<primary_file>MonteCarloMultiGPU.cpp</primary_file>
|
||||
<required_dependencies>
|
||||
<dependency>CURAND</dependency>
|
||||
</required_dependencies>
|
||||
<scopes>
|
||||
<scope>1:CUDA Advanced Topics</scope>
|
||||
<scope>1:Performance Strategies</scope>
|
||||
<scope>3:Computational Finance</scope>
|
||||
</scopes>
|
||||
<sm-arch>sm35</sm-arch>
|
||||
<sm-arch>sm37</sm-arch>
|
||||
<sm-arch>sm50</sm-arch>
|
||||
<sm-arch>sm52</sm-arch>
|
||||
<sm-arch>sm60</sm-arch>
|
||||
<sm-arch>sm61</sm-arch>
|
||||
<sm-arch>sm70</sm-arch>
|
||||
<sm-arch>sm72</sm-arch>
|
||||
<sm-arch>sm75</sm-arch>
|
||||
<sm-arch>sm80</sm-arch>
|
||||
<sm-arch>sm86</sm-arch>
|
||||
<supported_envs>
|
||||
<env>
|
||||
<arch>x86_64</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
<env>
|
||||
<platform>windows7</platform>
|
||||
</env>
|
||||
<env>
|
||||
<arch>x86_64</arch>
|
||||
<platform>macosx</platform>
|
||||
</env>
|
||||
<env>
|
||||
<arch>arm</arch>
|
||||
</env>
|
||||
<env>
|
||||
<arch>ppc64le</arch>
|
||||
<platform>linux</platform>
|
||||
</env>
|
||||
</supported_envs>
|
||||
<supported_sm_architectures>
|
||||
<include>all</include>
|
||||
</supported_sm_architectures>
|
||||
<title>Monte Carlo Option Pricing with Multi-GPU support</title>
|
||||
<type>exe</type>
|
||||
<whitepaper>doc\MonteCarlo.pdf</whitepaper>
|
||||
</entry>
|
||||
69
Samples/MonteCarloMultiGPU/README.md
Normal file
69
Samples/MonteCarloMultiGPU/README.md
Normal file
@@ -0,0 +1,69 @@
|
||||
# MonteCarloMultiGPU - Monte Carlo Option Pricing with Multi-GPU support
|
||||
|
||||
## Description
|
||||
|
||||
This sample evaluates fair call price for a given set of European options using the Monte Carlo approach, taking advantage of all CUDA-capable GPUs installed in the system. This sample use double precision hardware if a GTX 200 class GPU is present. The sample also takes advantage of CUDA 4.0 capability to supporting using a single CPU thread to control multiple GPUs
|
||||
|
||||
## Key Concepts
|
||||
|
||||
## Supported SM Architectures
|
||||
|
||||
[SM 3.5 ](https://developer.nvidia.com/cuda-gpus) [SM 3.7 ](https://developer.nvidia.com/cuda-gpus) [SM 5.0 ](https://developer.nvidia.com/cuda-gpus) [SM 5.2 ](https://developer.nvidia.com/cuda-gpus) [SM 6.0 ](https://developer.nvidia.com/cuda-gpus) [SM 6.1 ](https://developer.nvidia.com/cuda-gpus) [SM 7.0 ](https://developer.nvidia.com/cuda-gpus) [SM 7.2 ](https://developer.nvidia.com/cuda-gpus) [SM 7.5 ](https://developer.nvidia.com/cuda-gpus) [SM 8.0 ](https://developer.nvidia.com/cuda-gpus) [SM 8.6 ](https://developer.nvidia.com/cuda-gpus)
|
||||
|
||||
## Supported OSes
|
||||
|
||||
Linux, Windows
|
||||
|
||||
## Supported CPU Architecture
|
||||
|
||||
x86_64, ppc64le, armv7l
|
||||
|
||||
## CUDA APIs involved
|
||||
|
||||
## Dependencies needed to build/run
|
||||
[CURAND](../../README.md#curand)
|
||||
|
||||
## Prerequisites
|
||||
|
||||
Download and install the [CUDA Toolkit 11.5](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, armv7l.
|
||||
By default, TARGET_ARCH is set to HOST_ARCH. On a x86_64 machine, not setting TARGET_ARCH is the equivalent of setting TARGET_ARCH=x86_64.<br/>
|
||||
`$ make TARGET_ARCH=x86_64` <br/> `$ make TARGET_ARCH=ppc64le` <br/> `$ make TARGET_ARCH=armv7l` <br/>
|
||||
See [here](http://docs.nvidia.com/cuda/cuda-samples/index.html#cross-samples) for more details.
|
||||
* **dbg=1** - build with debug symbols
|
||||
```
|
||||
$ make dbg=1
|
||||
```
|
||||
* **SMS="A B ..."** - override the SM architectures for which the sample will be built, where `"A B ..."` is a space-delimited list of SM architectures. For example, to generate SASS for SM 50 and SM 60, use `SMS="50 60"`.
|
||||
```
|
||||
$ make SMS="50 60"
|
||||
```
|
||||
|
||||
* **HOST_COMPILER=<host_compiler>** - override the default g++ host compiler. See the [Linux Installation Guide](http://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#system-requirements) for a list of supported host compilers.
|
||||
```
|
||||
$ make HOST_COMPILER=g++
|
||||
```
|
||||
|
||||
## References (for more details)
|
||||
|
||||
BIN
Samples/MonteCarloMultiGPU/doc/MonteCarlo.doc
Normal file
BIN
Samples/MonteCarloMultiGPU/doc/MonteCarlo.doc
Normal file
Binary file not shown.
BIN
Samples/MonteCarloMultiGPU/doc/MonteCarlo.pdf
Normal file
BIN
Samples/MonteCarloMultiGPU/doc/MonteCarlo.pdf
Normal file
Binary file not shown.
69
Samples/MonteCarloMultiGPU/multithreading.cpp
Normal file
69
Samples/MonteCarloMultiGPU/multithreading.cpp
Normal file
@@ -0,0 +1,69 @@
|
||||
/* Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved.
|
||||
*
|
||||
* Redistribution and use in source and binary forms, with or without
|
||||
* modification, are permitted provided that the following conditions
|
||||
* are met:
|
||||
* * Redistributions of source code must retain the above copyright
|
||||
* notice, this list of conditions and the following disclaimer.
|
||||
* * Redistributions in binary form must reproduce the above copyright
|
||||
* notice, this list of conditions and the following disclaimer in the
|
||||
* documentation and/or other materials provided with the distribution.
|
||||
* * Neither the name of NVIDIA CORPORATION nor the names of its
|
||||
* contributors may be used to endorse or promote products derived
|
||||
* from this software without specific prior written permission.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
||||
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||||
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
||||
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
||||
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||||
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||||
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
||||
* OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
*/
|
||||
|
||||
#include <multithreading.h>
|
||||
|
||||
#if defined(WIN32) || defined(_WIN32) || defined(WIN64) || defined(_WIN64)
|
||||
// Create thread
|
||||
CUTThread cutStartThread(CUT_THREADROUTINE func, void *data) {
|
||||
return CreateThread(NULL, 0, (LPTHREAD_START_ROUTINE)func, data, 0, NULL);
|
||||
}
|
||||
|
||||
// Wait for thread to finish
|
||||
void cutEndThread(CUTThread thread) {
|
||||
WaitForSingleObject(thread, INFINITE);
|
||||
CloseHandle(thread);
|
||||
}
|
||||
|
||||
// Wait for multiple threads
|
||||
void cutWaitForThreads(const CUTThread *threads, int num) {
|
||||
WaitForMultipleObjects(num, threads, true, INFINITE);
|
||||
|
||||
for (int i = 0; i < num; i++) {
|
||||
CloseHandle(threads[i]);
|
||||
}
|
||||
}
|
||||
|
||||
#else
|
||||
// Create thread
|
||||
CUTThread cutStartThread(CUT_THREADROUTINE func, void *data) {
|
||||
pthread_t thread;
|
||||
pthread_create(&thread, NULL, func, data);
|
||||
return thread;
|
||||
}
|
||||
|
||||
// Wait for thread to finish
|
||||
void cutEndThread(CUTThread thread) { pthread_join(thread, NULL); }
|
||||
|
||||
// Wait for multiple threads
|
||||
void cutWaitForThreads(const CUTThread *threads, int num) {
|
||||
for (int i = 0; i < num; i++) {
|
||||
cutEndThread(threads[i]);
|
||||
}
|
||||
}
|
||||
|
||||
#endif
|
||||
71
Samples/MonteCarloMultiGPU/multithreading.h
Normal file
71
Samples/MonteCarloMultiGPU/multithreading.h
Normal file
@@ -0,0 +1,71 @@
|
||||
/* Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved.
|
||||
*
|
||||
* Redistribution and use in source and binary forms, with or without
|
||||
* modification, are permitted provided that the following conditions
|
||||
* are met:
|
||||
* * Redistributions of source code must retain the above copyright
|
||||
* notice, this list of conditions and the following disclaimer.
|
||||
* * Redistributions in binary form must reproduce the above copyright
|
||||
* notice, this list of conditions and the following disclaimer in the
|
||||
* documentation and/or other materials provided with the distribution.
|
||||
* * Neither the name of NVIDIA CORPORATION nor the names of its
|
||||
* contributors may be used to endorse or promote products derived
|
||||
* from this software without specific prior written permission.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
||||
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||||
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
||||
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
||||
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||||
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||||
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
||||
* OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
*/
|
||||
|
||||
#ifndef MULTITHREADING_H
|
||||
#define MULTITHREADING_H
|
||||
|
||||
// Simple portable thread library.
|
||||
|
||||
#if defined(WIN32) || defined(_WIN32) || defined(WIN64) || defined(_WIN64)
|
||||
// Windows threads.
|
||||
#include <windows.h>
|
||||
|
||||
typedef HANDLE CUTThread;
|
||||
typedef unsigned(WINAPI *CUT_THREADROUTINE)(void *);
|
||||
|
||||
#define CUT_THREADPROC unsigned WINAPI
|
||||
#define CUT_THREADEND return 0
|
||||
|
||||
#else
|
||||
// POSIX threads.
|
||||
#include <pthread.h>
|
||||
|
||||
typedef pthread_t CUTThread;
|
||||
typedef void *(*CUT_THREADROUTINE)(void *);
|
||||
|
||||
#define CUT_THREADPROC void
|
||||
#define CUT_THREADEND
|
||||
#endif
|
||||
|
||||
#ifdef __cplusplus
|
||||
extern "C" {
|
||||
#endif
|
||||
|
||||
// Create thread.
|
||||
CUTThread cutStartThread(CUT_THREADROUTINE, void *data);
|
||||
|
||||
// Wait for thread to finish.
|
||||
void cutEndThread(CUTThread thread);
|
||||
|
||||
// Wait for multiple threads.
|
||||
void cutWaitForThreads(const CUTThread *threads, int num);
|
||||
|
||||
#ifdef __cplusplus
|
||||
} // extern "C"
|
||||
#endif
|
||||
|
||||
#endif // MULTITHREADING_H
|
||||
39
Samples/MonteCarloMultiGPU/realtype.h
Normal file
39
Samples/MonteCarloMultiGPU/realtype.h
Normal file
@@ -0,0 +1,39 @@
|
||||
/* Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved.
|
||||
*
|
||||
* Redistribution and use in source and binary forms, with or without
|
||||
* modification, are permitted provided that the following conditions
|
||||
* are met:
|
||||
* * Redistributions of source code must retain the above copyright
|
||||
* notice, this list of conditions and the following disclaimer.
|
||||
* * Redistributions in binary form must reproduce the above copyright
|
||||
* notice, this list of conditions and the following disclaimer in the
|
||||
* documentation and/or other materials provided with the distribution.
|
||||
* * Neither the name of NVIDIA CORPORATION nor the names of its
|
||||
* contributors may be used to endorse or promote products derived
|
||||
* from this software without specific prior written permission.
|
||||
*
|
||||
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
|
||||
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
||||
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
|
||||
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
||||
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
||||
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
||||
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
|
||||
* OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
||||
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
||||
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||
*/
|
||||
|
||||
#ifndef REALTYPE_H
|
||||
#define REALTYPE_H
|
||||
|
||||
//#define DOUBLE_PRECISION
|
||||
|
||||
#ifndef DOUBLE_PRECISION
|
||||
typedef float real;
|
||||
#else
|
||||
typedef double real;
|
||||
#endif
|
||||
|
||||
#endif
|
||||
Reference in New Issue
Block a user