CMakeLists.txt

project(OptiX-test)
cmake_minimum_required(VERSION 2.8.8 FATAL_ERROR)

# As of CMake 2.6 policies were introduced in order to provide a mechanism for
# adding backwards compatibility one feature at a time.  We will just specify
# that all policies will use version 2.6 symantics.
cmake_policy(VERSION 2.6)

# Set the default build to Release.  Note this doesn't do anything for the VS
# default build target which defaults to Debug when you first start it.
IF(NOT CMAKE_BUILD_TYPE)
  SET(CMAKE_BUILD_TYPE "Release" CACHE STRING
      "Choose the type of build, options are: Debug Release RelWithDebInfo MinSizeRel."
      FORCE)
ENDIF(NOT CMAKE_BUILD_TYPE)

# Add paths to our CMake code to the module path, so they can be found automatically by
# CMake.
set(CMAKE_MODULE_PATH
  "$ENV{OPTIX_PATH}/SDK/CMake"
  ${CMAKE_MODULE_PATH}
  )

# Tells CMake to build all the libraries as shared libraries by default.  This can be
# overrided by individual libraries later.
set(BUILD_SHARED_LIBS ON)

##########
# Process our custom setup scripts here.

# Include all CMake Macros.
include(Macros)
# Determine information about the compiler
include (CompilerInfo)
# Check for specific machine/compiler options.
include (ConfigCompilerFlags)

# Turn off the warning that NVCC issues when generating PTX from our CUDA samples.  This
# is a custom extension to the FindCUDA code distributed by CMake.
OPTION(CUDA_REMOVE_GLOBAL_MEMORY_SPACE_WARNING "Suppress the \"Advisory: Cannot tell what pointer points to, assuming global memory space\" warning nvcc makes." ON)

# For Xcode 5, gcc is actually clang, so we have to tell CUDA to treat the compiler as
# clang, so that it doesn't mistake it for something else.
if(USING_CLANG_C)
  set(CUDA_HOST_COMPILER "clang" CACHE FILEPATH "Host side compiler used by NVCC")
endif()

# CUDA 8 is broken for generating dependencies during configure
option(CUDA_GENERATE_DEPENDENCIES_DURING_CONFIGURE "Generate dependencies during configure time instead of only during build time." OFF)

# Find at least a 10.0 version of CUDA.
find_package(CUDA 10.0 REQUIRED)

# Present the CUDA_64_BIT_DEVICE_CODE on the default set of options.
mark_as_advanced(CLEAR CUDA_64_BIT_DEVICE_CODE)