# Copyright 2011-2020 Blender Foundation
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

remove_extra_strict_flags()

set(INC
  ..
)

set(INC_SYS

)

set(SRC_DEVICE_CPU
  device/cpu/kernel.cpp
  device/cpu/kernel_sse2.cpp
  device/cpu/kernel_sse3.cpp
  device/cpu/kernel_sse41.cpp
  device/cpu/kernel_avx.cpp
  device/cpu/kernel_avx2.cpp
)

set(SRC_DEVICE_CUDA
  device/cuda/kernel.cu
)

set(SRC_DEVICE_HIP
  device/hip/kernel.cpp
)

set(SRC_DEVICE_OPTIX
  device/optix/kernel.cu
  device/optix/kernel_shader_raytrace.cu
)

set(SRC_BVH_HEADERS
  bvh/bvh.h
  bvh/bvh_nodes.h
  bvh/bvh_shadow_all.h
  bvh/bvh_local.h
  bvh/bvh_traversal.h
  bvh/bvh_types.h
  bvh/bvh_util.h
  bvh/bvh_volume.h
  bvh/bvh_volume_all.h
  bvh/bvh_embree.h
)

set(SRC_HEADERS
  kernel_accumulate.h
  kernel_adaptive_sampling.h
  kernel_bake.h
  kernel_camera.h
  kernel_color.h
  kernel_differential.h
  kernel_emission.h
  kernel_film.h
  kernel_id_passes.h
  kernel_jitter.h
  kernel_light.h
  kernel_light_background.h
  kernel_light_common.h
  kernel_lookup_table.h
  kernel_math.h
  kernel_montecarlo.h
  kernel_passes.h
  kernel_path_state.h
  kernel_profiling.h
  kernel_projection.h
  kernel_random.h
  kernel_shader.h
  kernel_shadow_catcher.h
  kernel_textures.h
  kernel_types.h
  kernel_work_stealing.h
  kernel_write_passes.h
)

set(SRC_DEVICE_CPU_HEADERS
  device/cpu/compat.h
  device/cpu/image.h
  device/cpu/globals.h
  device/cpu/kernel.h
  device/cpu/kernel_arch.h
  device/cpu/kernel_arch_impl.h
)
set(SRC_DEVICE_GPU_HEADERS
  device/gpu/image.h
  device/gpu/kernel.h
  device/gpu/parallel_active_index.h
  device/gpu/parallel_prefix_sum.h
  device/gpu/parallel_reduce.h
  device/gpu/parallel_sorted_index.h
)

set(SRC_DEVICE_CUDA_HEADERS
  device/cuda/compat.h
  device/cuda/config.h
  device/cuda/globals.h
)

set(SRC_DEVICE_HIP_HEADERS
  device/hip/compat.h
  device/hip/config.h
  device/hip/globals.h
)

set(SRC_DEVICE_OPTIX_HEADERS
  device/optix/compat.h
  device/optix/globals.h
)

set(SRC_CLOSURE_HEADERS
  closure/alloc.h
  closure/bsdf.h
  closure/bsdf_ashikhmin_velvet.h
  closure/bsdf_diffuse.h
  closure/bsdf_diffuse_ramp.h
  closure/bsdf_microfacet.h
  closure/bsdf_microfacet_multi.h
  closure/bsdf_microfacet_multi_impl.h
  closure/bsdf_oren_nayar.h
  closure/bsdf_phong_ramp.h
  closure/bsdf_reflection.h
  closure/bsdf_refraction.h
  closure/bsdf_toon.h
  closure/bsdf_transparent.h
  closure/bsdf_util.h
  closure/bsdf_ashikhmin_shirley.h
  closure/bsdf_hair.h
  closure/bssrdf.h
  closure/emissive.h
  closure/volume.h
  closure/bsdf_principled_diffuse.h
  closure/bsdf_principled_sheen.h
  closure/bsdf_hair_principled.h
)

set(SRC_SVM_HEADERS
  svm/svm.h
  svm/svm_ao.h
  svm/svm_aov.h
  svm/svm_attribute.h
  svm/svm_bevel.h
  svm/svm_blackbody.h
  svm/svm_bump.h
  svm/svm_camera.h
  svm/svm_clamp.h
  svm/svm_closure.h
  svm/svm_convert.h
  svm/svm_checker.h
  svm/svm_color_util.h
  svm/svm_brick.h
  svm/svm_displace.h
  svm/svm_fresnel.h
  svm/svm_wireframe.h
  svm/svm_wavelength.h
  svm/svm_gamma.h
  svm/svm_brightness.h
  svm/svm_geometry.h
  svm/svm_gradient.h
  svm/svm_hsv.h
  svm/svm_ies.h
  svm/svm_image.h
  svm/svm_invert.h
  svm/svm_light_path.h
  svm/svm_magic.h
  svm/svm_map_range.h
  svm/svm_mapping.h
  svm/svm_mapping_util.h
  svm/svm_math.h
  svm/svm_math_util.h
  svm/svm_mix.h
  svm/svm_musgrave.h
  svm/svm_noise.h
  svm/svm_noisetex.h
  svm/svm_normal.h
  svm/svm_ramp.h
  svm/svm_ramp_util.h
  svm/svm_sepcomb_hsv.h
  svm/svm_sepcomb_vector.h
  svm/svm_sky.h
  svm/svm_tex_coord.h
  svm/svm_fractal_noise.h
  svm/svm_types.h
  svm/svm_value.h
  svm/svm_vector_rotate.h
  svm/svm_vector_transform.h
  svm/svm_voronoi.h
  svm/svm_voxel.h
  svm/svm_wave.h
  svm/svm_white_noise.h
  svm/svm_vertex_color.h
)

set(SRC_GEOM_HEADERS
  geom/geom.h
  geom/geom_attribute.h
  geom/geom_curve.h
  geom/geom_curve_intersect.h
  geom/geom_motion_curve.h
  geom/geom_motion_triangle.h
  geom/geom_motion_triangle_intersect.h
  geom/geom_motion_triangle_shader.h
  geom/geom_object.h
  geom/geom_patch.h
  geom/geom_primitive.h
  geom/geom_shader_data.h
  geom/geom_subd_triangle.h
  geom/geom_triangle.h
  geom/geom_triangle_intersect.h
  geom/geom_volume.h
)

set(SRC_INTEGRATOR_HEADERS
  integrator/integrator_init_from_bake.h
  integrator/integrator_init_from_camera.h
  integrator/integrator_intersect_closest.h
  integrator/integrator_intersect_shadow.h
  integrator/integrator_intersect_subsurface.h
  integrator/integrator_intersect_volume_stack.h
  integrator/integrator_megakernel.h
  integrator/integrator_shade_background.h
  integrator/integrator_shade_light.h
  integrator/integrator_shade_shadow.h
  integrator/integrator_shade_surface.h
  integrator/integrator_shade_volume.h
  integrator/integrator_state.h
  integrator/integrator_state_flow.h
  integrator/integrator_state_template.h
  integrator/integrator_state_util.h
  integrator/integrator_subsurface.h
  integrator/integrator_volume_stack.h
)

set(SRC_UTIL_HEADERS
  ../util/util_atomic.h
  ../util/util_color.h
  ../util/util_defines.h
  ../util/util_half.h
  ../util/util_hash.h
  ../util/util_math.h
  ../util/util_math_fast.h
  ../util/util_math_intersect.h
  ../util/util_math_float2.h
  ../util/util_math_float3.h
  ../util/util_math_float4.h
  ../util/util_math_int2.h
  ../util/util_math_int3.h
  ../util/util_math_int4.h
  ../util/util_math_matrix.h
  ../util/util_projection.h
  ../util/util_rect.h
  ../util/util_static_assert.h
  ../util/util_transform.h
  ../util/util_texture.h
  ../util/util_types.h
  ../util/util_types_float2.h
  ../util/util_types_float2_impl.h
  ../util/util_types_float3.h
  ../util/util_types_float3_impl.h
  ../util/util_types_float4.h
  ../util/util_types_float4_impl.h
  ../util/util_types_float8.h
  ../util/util_types_float8_impl.h
  ../util/util_types_int2.h
  ../util/util_types_int2_impl.h
  ../util/util_types_int3.h
  ../util/util_types_int3_impl.h
  ../util/util_types_int4.h
  ../util/util_types_int4_impl.h
  ../util/util_types_uchar2.h
  ../util/util_types_uchar2_impl.h
  ../util/util_types_uchar3.h
  ../util/util_types_uchar3_impl.h
  ../util/util_types_uchar4.h
  ../util/util_types_uchar4_impl.h
  ../util/util_types_uint2.h
  ../util/util_types_uint2_impl.h
  ../util/util_types_uint3.h
  ../util/util_types_uint3_impl.h
  ../util/util_types_uint4.h
  ../util/util_types_uint4_impl.h
  ../util/util_types_ushort4.h
  ../util/util_types_vector3.h
  ../util/util_types_vector3_impl.h
)

set(LIB

)

# CUDA module

if(WITH_CYCLES_CUDA_BINARIES)
  # 64 bit only
  set(CUDA_BITS 64)

  # CUDA version
  execute_process(COMMAND ${CUDA_NVCC_EXECUTABLE} "--version" OUTPUT_VARIABLE NVCC_OUT)
  string(REGEX REPLACE ".*release ([0-9]+)\\.([0-9]+).*" "\\1" CUDA_VERSION_MAJOR "${NVCC_OUT}")
  string(REGEX REPLACE ".*release ([0-9]+)\\.([0-9]+).*" "\\2" CUDA_VERSION_MINOR "${NVCC_OUT}")
  set(CUDA_VERSION "${CUDA_VERSION_MAJOR}${CUDA_VERSION_MINOR}")

  # warn for other versions
  if((CUDA_VERSION MATCHES "101") OR
     (CUDA_VERSION MATCHES "102") OR
     (CUDA_VERSION MATCHES "111") OR
     (CUDA_VERSION MATCHES "112") OR
     (CUDA_VERSION MATCHES "113") OR
     (CUDA_VERSION MATCHES "114"))
  else()
    message(WARNING
      "CUDA version ${CUDA_VERSION_MAJOR}.${CUDA_VERSION_MINOR} detected, "
      "build may succeed but only CUDA 10.1 to 11.4 are officially supported")
  endif()

  # build for each arch
  set(cuda_sources device/cuda/kernel.cu
    ${SRC_HEADERS}
    ${SRC_DEVICE_GPU_HEADERS}
    ${SRC_DEVICE_CUDA_HEADERS}
    ${SRC_BVH_HEADERS}
    ${SRC_SVM_HEADERS}
    ${SRC_GEOM_HEADERS}
    ${SRC_INTEGRATOR_HEADERS}
    ${SRC_CLOSURE_HEADERS}
    ${SRC_UTIL_HEADERS}
  )
  set(cuda_cubins)

  macro(CYCLES_CUDA_KERNEL_ADD arch prev_arch name flags sources experimental)
    if(${arch} MATCHES "compute_.*")
      set(format "ptx")
    else()
      set(format "cubin")
    endif()
    set(cuda_file ${name}_${arch}.${format})

    set(kernel_sources ${sources})
    if(NOT ${prev_arch} STREQUAL "none")
      if(${prev_arch} MATCHES "compute_.*")
        set(kernel_sources ${kernel_sources} ${name}_${prev_arch}.ptx)
      else()
        set(kernel_sources ${kernel_sources} ${name}_${prev_arch}.cubin)
      endif()
    endif()

    set(cuda_kernel_src "/device/cuda/${name}.cu")

    set(cuda_flags ${flags}
      -D CCL_NAMESPACE_BEGIN=
      -D CCL_NAMESPACE_END=
      -D NVCC
      -m ${CUDA_BITS}
      -I ${CMAKE_CURRENT_SOURCE_DIR}/..
      -I ${CMAKE_CURRENT_SOURCE_DIR}/device/cuda
      --use_fast_math
      -o ${CMAKE_CURRENT_BINARY_DIR}/${cuda_file})

    if(${experimental})
      set(cuda_flags ${cuda_flags} -D __KERNEL_EXPERIMENTAL__)
      set(name ${name}_experimental)
    endif()

    if(WITH_NANOVDB)
      set(cuda_flags ${cuda_flags}
        -D WITH_NANOVDB
        -I "${NANOVDB_INCLUDE_DIR}")
    endif()

    if(WITH_CYCLES_CUBIN_COMPILER)
      string(SUBSTRING ${arch} 3 -1 CUDA_ARCH)

      # Needed to find libnvrtc-builtins.so. Can't do it from inside
      # cycles_cubin_cc since the env variable is read before main()
      if(APPLE)
        set(CUBIN_CC_ENV ${CMAKE_COMMAND}
          -E env DYLD_LIBRARY_PATH="${cuda_toolkit_root_dir}/lib")
      elseif(UNIX)
        set(CUBIN_CC_ENV ${CMAKE_COMMAND}
          -E env LD_LIBRARY_PATH="${cuda_toolkit_root_dir}/lib64")
      endif()

      add_custom_command(
        OUTPUT ${cuda_file}
        COMMAND ${CUBIN_CC_ENV}
            "$<TARGET_FILE:cycles_cubin_cc>"
            -target ${CUDA_ARCH}
            -i ${CMAKE_CURRENT_SOURCE_DIR}${cuda_kernel_src}
            ${cuda_flags}
            -v
            -cuda-toolkit-dir "${cuda_toolkit_root_dir}"
        DEPENDS ${kernel_sources} cycles_cubin_cc)
    else()
      add_custom_command(
        OUTPUT ${cuda_file}
        COMMAND ${cuda_nvcc_executable}
            -arch=${arch}
            ${CUDA_NVCC_FLAGS}
            --${format}
            ${CMAKE_CURRENT_SOURCE_DIR}${cuda_kernel_src}
            --ptxas-options="-v"
            ${cuda_flags}
        DEPENDS ${kernel_sources})
    endif()
    delayed_install("${CMAKE_CURRENT_BINARY_DIR}" "${cuda_file}" ${CYCLES_INSTALL_PATH}/lib)
    list(APPEND cuda_cubins ${cuda_file})

    unset(cuda_debug_flags)
  endmacro()

  set(prev_arch "none")
  foreach(arch ${CYCLES_CUDA_BINARIES_ARCH})
    if(${arch} MATCHES ".*_2.")
      message(STATUS "CUDA binaries for ${arch} are no longer supported, skipped.")
    elseif(${arch} MATCHES ".*_30")
      if(DEFINED CUDA10_NVCC_EXECUTABLE)
        set(cuda_nvcc_executable ${CUDA10_NVCC_EXECUTABLE})
        set(cuda_toolkit_root_dir ${CUDA10_TOOLKIT_ROOT_DIR})
      elseif(${CUDA_VERSION} LESS 110) # Support for sm_30 was removed in CUDA 11
        set(cuda_nvcc_executable ${CUDA_NVCC_EXECUTABLE})
        set(cuda_toolkit_root_dir ${CUDA_TOOLKIT_ROOT_DIR})
      else()
        message(STATUS "CUDA binaries for ${arch} require CUDA 10 or earlier, skipped.")
      endif()
    elseif(${arch} MATCHES ".*_7." AND ${CUDA_VERSION} LESS 100)
      message(STATUS "CUDA binaries for ${arch} require CUDA 10.0+, skipped.")
    elseif(${arch} MATCHES ".*_8.")
      if(DEFINED CUDA11_NVCC_EXECUTABLE)
        set(cuda_nvcc_executable ${CUDA11_NVCC_EXECUTABLE})
        set(cuda_toolkit_root_dir ${CUDA11_TOOLKIT_ROOT_DIR})
      elseif(${CUDA_VERSION} GREATER_EQUAL 111) # Support for sm_86 was introduced in CUDA 11
        set(cuda_nvcc_executable ${CUDA_NVCC_EXECUTABLE})
        set(cuda_toolkit_root_dir ${CUDA_TOOLKIT_ROOT_DIR})
      else()
        message(STATUS "CUDA binaries for ${arch} require CUDA 11.1+, skipped.")
      endif()
    else()
      set(cuda_nvcc_executable ${CUDA_NVCC_EXECUTABLE})
      set(cuda_toolkit_root_dir ${CUDA_TOOLKIT_ROOT_DIR})
    endif()
    if(DEFINED cuda_nvcc_executable AND DEFINED cuda_toolkit_root_dir)
      # Compile regular kernel
      CYCLES_CUDA_KERNEL_ADD(${arch} ${prev_arch} kernel "" "${cuda_sources}" FALSE)

      if(WITH_CYCLES_CUDA_BUILD_SERIAL)
        set(prev_arch ${arch})
      endif()

      unset(cuda_nvcc_executable)
      unset(cuda_toolkit_root_dir)
    endif()
  endforeach()

  add_custom_target(cycles_kernel_cuda ALL DEPENDS ${cuda_cubins})
  cycles_set_solution_folder(cycles_kernel_cuda)
endif()

####################################################### START

# HIP module

if(WITH_CYCLES_HIP_BINARIES)
  # 64 bit only
  set(HIP_BITS 64)

  # HIP version
  execute_process(COMMAND ${HIP_HIPCC_EXECUTABLE} "--version" OUTPUT_VARIABLE HIPCC_OUT)
  string(REGEX REPLACE ".*release ([0-9]+)\\.([0-9]+).*" "\\1" HIP_VERSION_MAJOR "${HIPCC_OUT}")
  string(REGEX REPLACE ".*release ([0-9]+)\\.([0-9]+).*" "\\2" HIP_VERSION_MINOR "${HIPCC_OUT}")
  set(HIP_VERSION "${HIP_VERSION_MAJOR}${HIP_VERSION_MINOR}")


  message(WARNING
    "HIP version ${HIP_VERSION_MAJOR}.${HIP_VERSION_MINOR} detected")

  # build for each arch
  set(hip_sources device/hip/kernel.cpp
    ${SRC_HEADERS}
    ${SRC_DEVICE_HIP_HEADERS}
    ${SRC_BVH_HEADERS}
    ${SRC_SVM_HEADERS}
    ${SRC_GEOM_HEADERS}
    ${SRC_INTEGRATOR_HEADERS}
    ${SRC_CLOSURE_HEADERS}
    ${SRC_UTIL_HEADERS}
  )
  set(hip_fatbins)

  macro(CYCLES_HIP_KERNEL_ADD arch prev_arch name flags sources experimental)
    if(${arch} MATCHES "compute_.*")
      set(format "ptx")
    else()
      set(format "fatbin")
    endif()
    set(hip_file ${name}_${arch}.${format})

    set(kernel_sources ${sources})
    if(NOT ${prev_arch} STREQUAL "none")
      if(${prev_arch} MATCHES "compute_.*")
        set(kernel_sources ${kernel_sources} ${name}_${prev_arch}.ptx)
      else()
        set(kernel_sources ${kernel_sources} ${name}_${prev_arch}.fatbin)
      endif()
    endif()

    set(hip_kernel_src "/device/hip/${name}.cpp")

    set(hip_flags ${flags}
      -D CCL_NAMESPACE_BEGIN=
      -D CCL_NAMESPACE_END=
      -D HIPCC
      -m ${HIP_BITS}
      -I ${CMAKE_CURRENT_SOURCE_DIR}/..
      -I ${CMAKE_CURRENT_SOURCE_DIR}/device/hip
      --use_fast_math
      -o ${CMAKE_CURRENT_BINARY_DIR}/${hip_file})

    if(${experimental})
      set(hip_flags ${hip_flags} -D __KERNEL_EXPERIMENTAL__)
      set(name ${name}_experimental)
    endif()

    if(WITH_CYCLES_DEBUG)
      set(hip_flags ${hip_flags} -D __KERNEL_DEBUG__)
    endif()

    if(WITH_NANOVDB)
      set(hip_flags ${hip_flags}
        -D WITH_NANOVDB
        -I "${NANOVDB_INCLUDE_DIR}")
    endif()


  set(prev_arch "none")
  foreach(arch ${CYCLES_HIP_BINARIES_ARCH})
      set(hip_hipcc_executable ${HIP_HIPCC_EXECUTABLE})
      set(hip_toolkit_root_dir ${HIP_TOOLKIT_ROOT_DIR})
    if(DEFINED hip_hipcc_executable AND DEFINED hip_toolkit_root_dir)
      # Compile regular kernel
      CYCLES_HIP_KERNEL_ADD(${arch} ${prev_arch} kernel "" "${hip_sources}" FALSE)

      if(WITH_CYCLES_HIP_BUILD_SERIAL)
        set(prev_arch ${arch})
      endif()

      unset(hip_hipcc_executable)
      unset(hip_toolkit_root_dir)
    endif()
  endforeach()

  add_custom_target(cycles_kernel_hip ALL DEPENDS ${hip_fatbins})
  cycles_set_solution_folder(cycles_kernel_hip)
endif()

####################################################### END
# OptiX PTX modules

if(WITH_CYCLES_DEVICE_OPTIX AND WITH_CYCLES_CUDA_BINARIES)
  macro(CYCLES_OPTIX_KERNEL_ADD name input flags)
    set(output "${CMAKE_CURRENT_BINARY_DIR}/${name}.ptx")

    set(cuda_flags ${flags}
      -I "${OPTIX_INCLUDE_DIR}"
      -I "${CMAKE_CURRENT_SOURCE_DIR}/.."
      -I "${CMAKE_CURRENT_SOURCE_DIR}/device/cuda"
      --use_fast_math
      -Wno-deprecated-gpu-targets
      -o ${output})

    if(WITH_NANOVDB)
      set(cuda_flags ${cuda_flags}
        -D WITH_NANOVDB
        -I "${NANOVDB_INCLUDE_DIR}")
    endif()

    if(WITH_CYCLES_CUBIN_COMPILER)
      # Needed to find libnvrtc-builtins.so. Can't do it from inside
      # cycles_cubin_cc since the env variable is read before main()
      if(APPLE)
        set(CUBIN_CC_ENV ${CMAKE_COMMAND}
          -E env DYLD_LIBRARY_PATH="${CUDA_TOOLKIT_ROOT_DIR}/lib")
      elseif(UNIX)
        set(CUBIN_CC_ENV ${CMAKE_COMMAND}
          -E env LD_LIBRARY_PATH="${CUDA_TOOLKIT_ROOT_DIR}/lib64")
      endif()

      add_custom_command(
        OUTPUT ${output}
        DEPENDS
          ${input}
          ${SRC_HEADERS}
          ${SRC_DEVICE_GPU_HEADERS}
          ${SRC_DEVICE_CUDA_HEADERS}
          ${SRC_DEVICE_OPTIX_HEADERS}
          ${SRC_BVH_HEADERS}
          ${SRC_SVM_HEADERS}
          ${SRC_GEOM_HEADERS}
          ${SRC_INTEGRATOR_HEADERS}
          ${SRC_CLOSURE_HEADERS}
          ${SRC_UTIL_HEADERS}
        COMMAND ${CUBIN_CC_ENV}
            "$<TARGET_FILE:cycles_cubin_cc>"
            -target 50
            -ptx
            -i ${CMAKE_CURRENT_SOURCE_DIR}/${input}
            ${cuda_flags}
            -v
            -cuda-toolkit-dir "${CUDA_TOOLKIT_ROOT_DIR}"
        DEPENDS ${kernel_sources} cycles_cubin_cc)
    else()
      add_custom_command(
        OUTPUT
          ${output}
        DEPENDS
          ${input}
          ${SRC_HEADERS}
          ${SRC_DEVICE_GPU_HEADERS}
          ${SRC_DEVICE_CUDA_HEADERS}
          ${SRC_DEVICE_OPTIX_HEADERS}
          ${SRC_BVH_HEADERS}
          ${SRC_SVM_HEADERS}
          ${SRC_GEOM_HEADERS}
          ${SRC_INTEGRATOR_HEADERS}
          ${SRC_CLOSURE_HEADERS}
          ${SRC_UTIL_HEADERS}
        COMMAND
          ${CUDA_NVCC_EXECUTABLE}
          --ptx
          -arch=sm_50
          ${cuda_flags}
          ${input}
        WORKING_DIRECTORY
          "${CMAKE_CURRENT_SOURCE_DIR}")
    endif()
    list(APPEND optix_ptx ${output})

    delayed_install("${CMAKE_CURRENT_BINARY_DIR}" "${output}" ${CYCLES_INSTALL_PATH}/lib)
  endmacro()

  CYCLES_OPTIX_KERNEL_ADD(
    kernel_optix
    "device/optix/kernel.cu"
    "")
  CYCLES_OPTIX_KERNEL_ADD(
    kernel_optix_shader_raytrace
    "device/optix/kernel_shader_raytrace.cu"
    "--keep-device-functions")

  add_custom_target(cycles_kernel_optix ALL DEPENDS ${optix_ptx})
  cycles_set_solution_folder(cycles_kernel_optix)
endif()

# OSL module

if(WITH_CYCLES_OSL)
  list(APPEND LIB
    cycles_kernel_osl
  )
  add_subdirectory(osl)
  add_subdirectory(shaders)
endif()

# CPU module

include_directories(${INC})
include_directories(SYSTEM ${INC_SYS})

if(WITH_COMPILER_ASAN)
  if(CMAKE_COMPILER_IS_GNUCC AND (NOT WITH_CYCLES_KERNEL_ASAN))
    # GCC hangs compiling the big kernel files with asan and release, so disable by default.
    string(APPEND CMAKE_CXX_FLAGS_RELWITHDEBINFO " -fno-sanitize=all")
    string(APPEND CMAKE_CXX_FLAGS_DEBUG " -fno-sanitize=vptr")
  elseif(CMAKE_C_COMPILER_ID MATCHES "Clang")
    # With OSL, Cycles disables rtti in some modules, wich then breaks at linking
    # when trying to use vptr sanitizer (included into 'undefined' general option).
    string(APPEND CMAKE_CXX_FLAGS_RELWITHDEBINFO " -fno-sanitize=vptr")
    string(APPEND CMAKE_CXX_FLAGS_DEBUG " -fno-sanitize=vptr")
  endif()
endif()

set_source_files_properties(device/cpu/kernel.cpp PROPERTIES COMPILE_FLAGS "${CYCLES_KERNEL_FLAGS}")

if(CXX_HAS_SSE)
  set_source_files_properties(device/cpu/kernel_sse2.cpp PROPERTIES COMPILE_FLAGS "${CYCLES_SSE2_KERNEL_FLAGS}")
  set_source_files_properties(device/cpu/kernel_sse3.cpp PROPERTIES COMPILE_FLAGS "${CYCLES_SSE3_KERNEL_FLAGS}")
  set_source_files_properties(device/cpu/kernel_sse41.cpp PROPERTIES COMPILE_FLAGS "${CYCLES_SSE41_KERNEL_FLAGS}")
endif()

if(CXX_HAS_AVX)
  set_source_files_properties(device/cpu/kernel_avx.cpp PROPERTIES COMPILE_FLAGS "${CYCLES_AVX_KERNEL_FLAGS}")
endif()

if(CXX_HAS_AVX2)
  set_source_files_properties(device/cpu/kernel_avx2.cpp PROPERTIES COMPILE_FLAGS "${CYCLES_AVX2_KERNEL_FLAGS}")
endif()

cycles_add_library(cycles_kernel "${LIB}"
  ${SRC_DEVICE_CPU}
  ${SRC_DEVICE_CUDA}
  ${SRC_DEVICE_HIP}
  ${SRC_DEVICE_OPTIX}
  ${SRC_HEADERS}
  ${SRC_DEVICE_CPU_HEADERS}
  ${SRC_DEVICE_GPU_HEADERS}
  ${SRC_DEVICE_CUDA_HEADERS}
  ${SRC_DEVICE_HIP_HEADERS}
  ${SRC_DEVICE_OPTIX_HEADERS}
  ${SRC_BVH_HEADERS}
  ${SRC_CLOSURE_HEADERS}
  ${SRC_SVM_HEADERS}
  ${SRC_GEOM_HEADERS}
  ${SRC_INTEGRATOR_HEADERS}
)

source_group("bvh" FILES ${SRC_BVH_HEADERS})
source_group("closure" FILES ${SRC_CLOSURE_HEADERS})
source_group("geom" FILES ${SRC_GEOM_HEADERS})
source_group("integrator" FILES ${SRC_INTEGRATOR_HEADERS})
source_group("kernel" FILES ${SRC_HEADERS})
source_group("device\\cpu" FILES ${SRC_DEVICE_CPU} ${SRC_DEVICE_CPU_HEADERS})
source_group("device\\hip" FILES ${SRC_DEVICE_HIP} ${SRC_DEVICE_HIP_HEADERS})
source_group("device\\gpu" FILES ${SRC_DEVICE_GPU_HEADERS})
source_group("device\\cuda" FILES ${SRC_DEVICE_CUDA} ${SRC_DEVICE_CUDA_HEADERS})
source_group("device\\optix" FILES ${SRC_DEVICE_OPTIX} ${SRC_DEVICE_OPTIX_HEADERS})
source_group("svm" FILES ${SRC_SVM_HEADERS})

if(WITH_CYCLES_CUDA)
  add_dependencies(cycles_kernel cycles_kernel_cuda)
endif()
if(WITH_CYCLES_DEVICE_OPTIX AND WITH_CYCLES_CUDA_BINARIES)
  add_dependencies(cycles_kernel cycles_kernel_optix)
endif()
if(WITH_CYCLES_HIP)
  add_dependencies(cycles_kernel cycles_kernel_hip)
endif()

# Install kernel source for runtime compilation

delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${SRC_DEVICE_CUDA}" ${CYCLES_INSTALL_PATH}/source/kernel/device/cuda)
delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${SRC_DEVICE_HIP}" ${CYCLES_INSTALL_PATH}/source/kernel/device/hip)
delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${SRC_DEVICE_OPTIX}" ${CYCLES_INSTALL_PATH}/source/kernel/device/optix)
delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${SRC_HEADERS}" ${CYCLES_INSTALL_PATH}/source/kernel)
delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${SRC_DEVICE_GPU_HEADERS}" ${CYCLES_INSTALL_PATH}/source/kernel/device/gpu)
delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${SRC_DEVICE_CUDA_HEADERS}" ${CYCLES_INSTALL_PATH}/source/kernel/device/cuda)
delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${SRC_DEVICE_HIP_HEADERS}" ${CYCLES_INSTALL_PATH}/source/kernel/device/hip)
delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${SRC_DEVICE_OPTIX_HEADERS}" ${CYCLES_INSTALL_PATH}/source/kernel/device/optix)
delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${SRC_BVH_HEADERS}" ${CYCLES_INSTALL_PATH}/source/kernel/bvh)
delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${SRC_CLOSURE_HEADERS}" ${CYCLES_INSTALL_PATH}/source/kernel/closure)
delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${SRC_SVM_HEADERS}" ${CYCLES_INSTALL_PATH}/source/kernel/svm)
delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${SRC_GEOM_HEADERS}" ${CYCLES_INSTALL_PATH}/source/kernel/geom)
delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${SRC_INTEGRATOR_HEADERS}" ${CYCLES_INSTALL_PATH}/source/kernel/integrator)
delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${SRC_UTIL_HEADERS}" ${CYCLES_INSTALL_PATH}/source/util)

if(WITH_NANOVDB)
  set(SRC_NANOVDB_HEADERS
    nanovdb/NanoVDB.h
    nanovdb/CNanoVDB.h
  )
  set(SRC_NANOVDB_UTIL_HEADERS
    nanovdb/util/CSampleFromVoxels.h
    nanovdb/util/SampleFromVoxels.h
  )
  delayed_install(${NANOVDB_INCLUDE_DIR} "${SRC_NANOVDB_HEADERS}" ${CYCLES_INSTALL_PATH}/source/nanovdb)
  delayed_install(${NANOVDB_INCLUDE_DIR} "${SRC_NANOVDB_UTIL_HEADERS}" ${CYCLES_INSTALL_PATH}/source/nanovdb/util)
endif()