
This kernel is compiled with AVX2, FMA3, and BMI compiler flags. At the moment only Intel Haswell benefits from this, but future AMD CPUs will have these instructions as well. Makes rendering on Haswell CPUs a few percent faster, only benchmarked with clang on OS X though. Part of my GSoC 2014.
166 lines
5.0 KiB
CMake
166 lines
5.0 KiB
CMake
# Standalone or with Blender
|
|
if(NOT WITH_BLENDER AND WITH_CYCLES_STANDALONE)
|
|
set(CYCLES_INSTALL_PATH "")
|
|
else()
|
|
set(WITH_CYCLES_BLENDER ON)
|
|
set(CYCLES_INSTALL_PATH "scripts/addons/cycles")
|
|
endif()
|
|
|
|
# External Libraries
|
|
|
|
include(cmake/external_libs.cmake)
|
|
|
|
# Build Flags
|
|
# todo: refactor this code to match scons
|
|
# note: CXX_HAS_SSE is needed in case passing SSE flags fails altogether (gcc-arm)
|
|
|
|
if(WIN32 AND MSVC)
|
|
set(CXX_HAS_SSE TRUE)
|
|
|
|
# /arch:AVX for VC2012 and above
|
|
if(NOT MSVC_VERSION LESS 1700)
|
|
set(CYCLES_AVX_ARCH_FLAGS "/arch:AVX")
|
|
set(CYCLES_AVX2_ARCH_FLAGS "/arch:AVX /arch:AVX2")
|
|
elseif(NOT CMAKE_CL_64)
|
|
set(CYCLES_AVX_ARCH_FLAGS "/arch:SSE2")
|
|
set(CYCLES_AVX2_ARCH_FLAGS "/arch:SSE2")
|
|
endif()
|
|
|
|
# there is no /arch:SSE3, but intrinsics are available anyway
|
|
if(CMAKE_CL_64)
|
|
set(CYCLES_SSE2_KERNEL_FLAGS "/fp:fast -D_CRT_SECURE_NO_WARNINGS /GS-")
|
|
set(CYCLES_SSE3_KERNEL_FLAGS "/fp:fast -D_CRT_SECURE_NO_WARNINGS /GS-")
|
|
set(CYCLES_SSE41_KERNEL_FLAGS "/fp:fast -D_CRT_SECURE_NO_WARNINGS /GS-")
|
|
set(CYCLES_AVX_KERNEL_FLAGS "${CYCLES_AVX_ARCH_FLAGS} /fp:fast -D_CRT_SECURE_NO_WARNINGS /GS-")
|
|
set(CYCLES_AVX2_KERNEL_FLAGS "${CYCLES_AVX2_ARCH_FLAGS} /fp:fast -D_CRT_SECURE_NO_WARNINGS /GS-")
|
|
else()
|
|
set(CYCLES_SSE2_KERNEL_FLAGS "/arch:SSE2 /fp:fast -D_CRT_SECURE_NO_WARNINGS /GS-")
|
|
set(CYCLES_SSE3_KERNEL_FLAGS "/arch:SSE2 /fp:fast -D_CRT_SECURE_NO_WARNINGS /GS-")
|
|
set(CYCLES_SSE41_KERNEL_FLAGS "/arch:SSE2 /fp:fast -D_CRT_SECURE_NO_WARNINGS /GS-")
|
|
set(CYCLES_AVX_KERNEL_FLAGS "${CYCLES_AVX_ARCH_FLAGS} /fp:fast -D_CRT_SECURE_NO_WARNINGS /GS-")
|
|
set(CYCLES_AVX2_KERNEL_FLAGS "${CYCLES_AVX2_ARCH_FLAGS} /fp:fast -D_CRT_SECURE_NO_WARNINGS /GS-")
|
|
endif()
|
|
|
|
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /fp:fast -D_CRT_SECURE_NO_WARNINGS /GS-")
|
|
set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} /Ox")
|
|
set(CMAKE_CXX_FLAGS_RELWITHDEBINFO "${CMAKE_CXX_FLAGS_RELWITHDEBINFO} /Ox")
|
|
set(CMAKE_CXX_FLAGS_MINSIZEREL "${CMAKE_CXX_FLAGS_MINSIZEREL} /Ox")
|
|
elseif(CMAKE_COMPILER_IS_GNUCC)
|
|
check_cxx_compiler_flag(-msse CXX_HAS_SSE)
|
|
if(CXX_HAS_SSE)
|
|
set(CYCLES_SSE2_KERNEL_FLAGS "-ffast-math -msse -msse2 -mfpmath=sse")
|
|
set(CYCLES_SSE3_KERNEL_FLAGS "-ffast-math -msse -msse2 -msse3 -mssse3 -mfpmath=sse")
|
|
set(CYCLES_SSE41_KERNEL_FLAGS "-ffast-math -msse -msse2 -msse3 -mssse3 -msse4.1 -mfpmath=sse")
|
|
set(CYCLES_AVX_KERNEL_FLAGS "-ffast-math -msse -msse2 -msse3 -mssse3 -msse4.1 -mavx -mfpmath=sse")
|
|
set(CYCLES_AVX2_KERNEL_FLAGS "-ffast-math -msse -msse2 -msse3 -mssse3 -msse4.1 -mavx -mavx2 -mfma -mbmi -mbmi2 -mfpmath=sse")
|
|
endif()
|
|
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -ffast-math")
|
|
elseif(CMAKE_CXX_COMPILER_ID MATCHES "Clang")
|
|
check_cxx_compiler_flag(-msse CXX_HAS_SSE)
|
|
if(CXX_HAS_SSE)
|
|
set(CYCLES_SSE2_KERNEL_FLAGS "-ffast-math -msse -msse2")
|
|
set(CYCLES_SSE3_KERNEL_FLAGS "-ffast-math -msse -msse2 -msse3 -mssse3")
|
|
set(CYCLES_SSE41_KERNEL_FLAGS "-ffast-math -msse -msse2 -msse3 -mssse3 -msse4.1")
|
|
set(CYCLES_AVX_KERNEL_FLAGS "-ffast-math -msse -msse2 -msse3 -mssse3 -msse4.1 -mavx")
|
|
set(CYCLES_AVX2_KERNEL_FLAGS "-ffast-math -msse -msse2 -msse3 -mssse3 -msse4.1 -mavx -mavx2 —mfma -mbmi -mbmi2")
|
|
endif()
|
|
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -ffast-math")
|
|
endif()
|
|
|
|
if(CXX_HAS_SSE)
|
|
add_definitions(
|
|
-DWITH_KERNEL_SSE2
|
|
-DWITH_KERNEL_SSE3
|
|
-DWITH_KERNEL_SSE41
|
|
-DWITH_KERNEL_AVX
|
|
-DWITH_KERNEL_AVX2
|
|
)
|
|
endif()
|
|
|
|
# for OSL
|
|
if(WIN32 AND MSVC)
|
|
set(RTTI_DISABLE_FLAGS "/GR- -DBOOST_NO_RTTI -DBOOST_NO_TYPEID")
|
|
elseif(CMAKE_COMPILER_IS_GNUCC OR (CMAKE_C_COMPILER_ID MATCHES "Clang"))
|
|
set(RTTI_DISABLE_FLAGS "-fno-rtti -DBOOST_NO_RTTI -DBOOST_NO_TYPEID")
|
|
endif()
|
|
|
|
# Definitions and Includes
|
|
|
|
add_definitions(
|
|
${BOOST_DEFINITIONS}
|
|
${OPENIMAGEIO_DEFINITIONS}
|
|
)
|
|
|
|
add_definitions(
|
|
-DCCL_NAMESPACE_BEGIN=namespace\ ccl\ {
|
|
-DCCL_NAMESPACE_END=}
|
|
)
|
|
|
|
if(WITH_CYCLES_NETWORK)
|
|
add_definitions(-DWITH_NETWORK)
|
|
endif()
|
|
|
|
if(WITH_CYCLES_STANDALONE_GUI)
|
|
add_definitions(-DWITH_CYCLES_STANDALONE_GUI)
|
|
endif()
|
|
|
|
if(WITH_CYCLES_PTEX)
|
|
add_definitions(-DWITH_PTEX)
|
|
endif()
|
|
|
|
if(WITH_CYCLES_OPENSUBDIV)
|
|
add_definitions(-DWITH_OPENSUBDIV)
|
|
endif()
|
|
|
|
if(WITH_CYCLES_OSL)
|
|
add_definitions(-DWITH_OSL)
|
|
add_definitions(-DOSL_STATIC_LIBRARY)
|
|
include_directories(${OSL_INCLUDES})
|
|
endif()
|
|
|
|
add_definitions(
|
|
-DWITH_OPENCL
|
|
-DWITH_CUDA
|
|
-DWITH_MULTI
|
|
)
|
|
|
|
include_directories(
|
|
SYSTEM
|
|
${BOOST_INCLUDE_DIR}
|
|
${OPENIMAGEIO_INCLUDE_DIRS}
|
|
${OPENIMAGEIO_INCLUDE_DIRS}/OpenImageIO
|
|
${OPENEXR_INCLUDE_DIR}
|
|
${OPENEXR_INCLUDE_DIRS}
|
|
)
|
|
|
|
|
|
# Warnings
|
|
if(CMAKE_COMPILER_IS_GNUCXX)
|
|
ADD_CHECK_CXX_COMPILER_FLAG(CMAKE_CXX_FLAGS _has_cxxflag_float_conversion "-Werror=float-conversion")
|
|
unset(_has_cxxflag_float_conversion)
|
|
endif()
|
|
|
|
|
|
# Subdirectories
|
|
|
|
if(WITH_CYCLES_BLENDER)
|
|
add_subdirectory(blender)
|
|
endif()
|
|
|
|
if(WITH_CYCLES_STANDALONE OR WITH_CYCLES_NETWORK)
|
|
add_subdirectory(app)
|
|
endif()
|
|
|
|
add_subdirectory(bvh)
|
|
add_subdirectory(device)
|
|
add_subdirectory(doc)
|
|
add_subdirectory(kernel)
|
|
add_subdirectory(render)
|
|
add_subdirectory(subd)
|
|
add_subdirectory(util)
|
|
|
|
if(NOT WITH_BLENDER AND WITH_CYCLES_STANDALONE)
|
|
delayed_do_install(${CMAKE_BINARY_DIR}/bin)
|
|
endif()
|
|
|