Bump version to 3.0.1 release

unity launches blender in background mode to do some
file conversions, ever since the launcher got introduced
this process broke.

The root cause here is: Unity looks up the default program
to launch .blend files with, which is now the launcher, then
launches it in background mode with a script to export the data.

The launcher however was designed to exit as quickly as
possible so there would not be an extra background process
lingering. It does not wait for blender to exit and does not
pass back any error codes.

This broke unity's workflow since it assumed if the process
exits and succeeds the data *must* be ready for reading which
no longer holds true.

This change keeps the launcher design as was previously,
*except* when launching in background mode, then it
waits and passes back any error codes, thus restoring
unity's workflow.

Differential Revision: https://developer.blender.org/D13894
Reviewed by: LazyDodo, Brecht

.clang-tidy

@@ -12,8 +12,6 @@ Checks:  >
   -readability-avoid-const-params-in-decls,
   -readability-simplify-boolean-expr,
   -readability-make-member-function-const,
-  -readability-suspicious-call-argument,
-  -readability-redundant-member-init,
 
   -readability-misleading-indentation,
 
@@ -27,8 +25,6 @@ Checks:  >
   -bugprone-branch-clone,
   -bugprone-macro-parentheses,
   -bugprone-reserved-identifier,
-  -bugprone-easily-swappable-parameters,
-  -bugprone-implicit-widening-of-multiplication-result,
 
   -bugprone-sizeof-expression,
   -bugprone-integer-division,
@@ -44,8 +40,7 @@ Checks:  >
   -modernize-pass-by-value,
   # Cannot be enabled yet, because using raw string literals in tests breaks
   # the windows compiler currently.
-  -modernize-raw-string-literal,
+  -modernize-raw-string-literal
-  -modernize-return-braced-init-list
 
 CheckOptions:
   - key: modernize-use-default-member-init.UseAssignment

CMakeLists.txt

@@ -187,13 +187,6 @@ mark_as_advanced(CPACK_OVERRIDE_PACKAGENAME)
 mark_as_advanced(BUILDINFO_OVERRIDE_DATE)
 mark_as_advanced(BUILDINFO_OVERRIDE_TIME)
 
-if(${CMAKE_VERSION} VERSION_GREATER_EQUAL "3.16")
-  option(WITH_UNITY_BUILD   "Enable unity build for modules that support it to improve compile times" ON)
-  mark_as_advanced(WITH_UNITY_BUILD)
-else()
-  set(WITH_UNITY_BUILD OFF)
-endif()
-
 option(WITH_IK_ITASC      "Enable ITASC IK solver (only disable for development & for incompatible C++ compilers)" ON)
 option(WITH_IK_SOLVER     "Enable Legacy IK solver (only disable for development)" ON)
 option(WITH_FFTW3         "Enable FFTW3 support (Used for smoke, ocean sim, and audio effects)" ON)
@@ -418,7 +411,6 @@ option(WITH_CYCLES                   "Enable Cycles Render Engine" ON)
 option(WITH_CYCLES_OSL               "Build Cycles with OpenShadingLanguage support" ON)
 option(WITH_CYCLES_EMBREE            "Build Cycles with Embree support" ON)
 option(WITH_CYCLES_LOGGING           "Build Cycles with logging support" ON)
-option(WITH_CYCLES_DEBUG             "Build Cycles with options useful for debugging (e.g., MIS)" OFF)
 
 option(WITH_CYCLES_STANDALONE        "Build Cycles standalone application" OFF)
 option(WITH_CYCLES_STANDALONE_GUI    "Build Cycles standalone with GUI" OFF)
@@ -433,40 +425,30 @@ mark_as_advanced(WITH_CYCLES_DEBUG_NAN)
 mark_as_advanced(WITH_CYCLES_NATIVE_ONLY)
 
 # NVIDIA CUDA & OptiX
-if(NOT APPLE)
+option(WITH_CYCLES_DEVICE_CUDA       "Enable Cycles NVIDIA CUDA compute support" ON)
-  option(WITH_CYCLES_DEVICE_CUDA       "Enable Cycles NVIDIA CUDA compute support" ON)
+option(WITH_CYCLES_DEVICE_OPTIX      "Enable Cycles NVIDIA OptiX support" ON)
-  option(WITH_CYCLES_DEVICE_OPTIX      "Enable Cycles NVIDIA OptiX support" ON)
+mark_as_advanced(WITH_CYCLES_DEVICE_CUDA)
-  mark_as_advanced(WITH_CYCLES_DEVICE_CUDA)
 
-  option(WITH_CYCLES_CUDA_BINARIES     "Build Cycles NVIDIA CUDA binaries" OFF)
+option(WITH_CYCLES_CUDA_BINARIES     "Build Cycles NVIDIA CUDA binaries" OFF)
-  set(CYCLES_CUDA_BINARIES_ARCH sm_30 sm_35 sm_37 sm_50 sm_52 sm_60 sm_61 sm_70 sm_75 sm_86 compute_75 CACHE STRING "CUDA architectures to build binaries for")
+set(CYCLES_CUDA_BINARIES_ARCH sm_30 sm_35 sm_37 sm_50 sm_52 sm_60 sm_61 sm_70 sm_75 sm_86 compute_75 CACHE STRING "CUDA architectures to build binaries for")
-  option(WITH_CYCLES_CUBIN_COMPILER    "Build cubins with nvrtc based compiler instead of nvcc" OFF)
+option(WITH_CYCLES_CUBIN_COMPILER    "Build cubins with nvrtc based compiler instead of nvcc" OFF)
-  option(WITH_CYCLES_CUDA_BUILD_SERIAL "Build cubins one after another (useful on machines with limited RAM)" OFF)
+option(WITH_CYCLES_CUDA_BUILD_SERIAL "Build cubins one after another (useful on machines with limited RAM)" OFF)
-  option(WITH_CUDA_DYNLOAD             "Dynamically load CUDA libraries at runtime (for developers, makes cuda-gdb work)" ON)
+option(WITH_CUDA_DYNLOAD             "Dynamically load CUDA libraries at runtime (for developers, makes cuda-gdb work)" ON)
-  mark_as_advanced(CYCLES_CUDA_BINARIES_ARCH)
+mark_as_advanced(CYCLES_CUDA_BINARIES_ARCH)
-  mark_as_advanced(WITH_CYCLES_CUBIN_COMPILER)
+mark_as_advanced(WITH_CYCLES_CUBIN_COMPILER)
-  mark_as_advanced(WITH_CYCLES_CUDA_BUILD_SERIAL)
+mark_as_advanced(WITH_CYCLES_CUDA_BUILD_SERIAL)
-  mark_as_advanced(WITH_CUDA_DYNLOAD)
+mark_as_advanced(WITH_CUDA_DYNLOAD)
-endif()
 
 # AMD HIP
-if(NOT APPLE)
+if(WIN32)
-  if(WIN32)
+  option(WITH_CYCLES_DEVICE_HIP        "Enable Cycles AMD HIP support" ON)
-    option(WITH_CYCLES_DEVICE_HIP        "Enable Cycles AMD HIP support" ON)
+else()
-  else()
+  option(WITH_CYCLES_DEVICE_HIP        "Enable Cycles AMD HIP support" OFF)
-    option(WITH_CYCLES_DEVICE_HIP        "Enable Cycles AMD HIP support" OFF)
-  endif()
-
-  option(WITH_CYCLES_HIP_BINARIES      "Build Cycles AMD HIP binaries" OFF)
-  set(CYCLES_HIP_BINARIES_ARCH gfx1010 gfx1011 gfx1012 gfx1030 gfx1031 gfx1032 gfx1034 CACHE STRING "AMD HIP architectures to build binaries for")
-  mark_as_advanced(WITH_CYCLES_DEVICE_HIP)
-  mark_as_advanced(CYCLES_HIP_BINARIES_ARCH)
-endif()
-
-# Apple Metal
-if(APPLE)
-  option(WITH_CYCLES_DEVICE_METAL       "Enable Cycles Apple Metal compute support" ON)
 endif()
+option(WITH_CYCLES_HIP_BINARIES      "Build Cycles AMD HIP binaries" OFF)
+set(CYCLES_HIP_BINARIES_ARCH gfx1010 gfx1011 gfx1012 gfx1030 gfx1031 gfx1032 gfx1034 CACHE STRING "AMD HIP architectures to build binaries for")
+mark_as_advanced(WITH_CYCLES_DEVICE_HIP)
+mark_as_advanced(CYCLES_HIP_BINARIES_ARCH)
 
 # Draw Manager
 option(WITH_DRAW_DEBUG "Add extra debug capabilities to Draw Manager" OFF)
@@ -511,10 +493,12 @@ if(WIN32)
 endif()
 
 # This should be turned off when Blender enter beta/rc/release
-if("${BLENDER_VERSION_CYCLE}" STREQUAL "alpha")
+if("${BLENDER_VERSION_CYCLE}" STREQUAL "release" OR
-  set(WITH_EXPERIMENTAL_FEATURES ON)
+   "${BLENDER_VERSION_CYCLE}" STREQUAL "rc" OR
-else()
+   "${BLENDER_VERSION_CYCLE}" STREQUAL "beta")
   set(WITH_EXPERIMENTAL_FEATURES OFF)
+else()
+  set(WITH_EXPERIMENTAL_FEATURES ON)
 endif()
 
 # Unit testsing
@@ -536,14 +520,12 @@ option(WITH_OPENGL              "When off limits visibility of the opengl header
 option(WITH_GLEW_ES             "Switches to experimental copy of GLEW that has support for OpenGL ES. (temporary option for development purposes)" OFF)
 option(WITH_GL_EGL              "Use the EGL OpenGL system library instead of the platform specific OpenGL system library (CGL, glX, or WGL)"       OFF)
 option(WITH_GL_PROFILE_ES20     "Support using OpenGL ES 2.0. (through either EGL or the AGL/WGL/XGL 'es20' profile)"                               OFF)
-option(WITH_GPU_SHADER_BUILDER  "Shader builder is a developer option enabling linting on GLSL during compilation"                                  OFF)
 
 mark_as_advanced(
   WITH_OPENGL
   WITH_GLEW_ES
   WITH_GL_EGL
   WITH_GL_PROFILE_ES20
-  WITH_GPU_SHADER_BUILDER
 )
 
 if(WIN32)
@@ -561,14 +543,12 @@ if(WIN32)
   set(CPACK_INSTALL_PREFIX ${CMAKE_GENERIC_PROGRAM_FILES}/${})
 endif()
 
-# Compiler tool-chain.
+# Compiler toolchain
-if(UNIX AND NOT APPLE)
+if(CMAKE_COMPILER_IS_GNUCC)
-  if(CMAKE_COMPILER_IS_GNUCC)
+  option(WITH_LINKER_GOLD "Use ld.gold linker which is usually faster than ld.bfd" ON)
-    option(WITH_LINKER_GOLD "Use ld.gold linker which is usually faster than ld.bfd" ON)
+  mark_as_advanced(WITH_LINKER_GOLD)
-    mark_as_advanced(WITH_LINKER_GOLD)
+  option(WITH_LINKER_LLD "Use ld.lld linker which is usually faster than ld.gold" OFF)
-    option(WITH_LINKER_LLD "Use ld.lld linker which is usually faster than ld.gold" OFF)
+  mark_as_advanced(WITH_LINKER_LLD)
-    mark_as_advanced(WITH_LINKER_LLD)
-  endif()
 endif()
 
 option(WITH_COMPILER_ASAN "Build and link against address sanitizer (only for Debug & RelWithDebInfo targets)." OFF)
@@ -860,7 +840,7 @@ if(WITH_AUDASPACE)
 endif()
 
 # Auto-enable CUDA dynload if toolkit is not found.
-if(WITH_CYCLES AND WITH_CYCLES_DEVICE_CUDA AND NOT WITH_CUDA_DYNLOAD)
+if(NOT WITH_CUDA_DYNLOAD)
   find_package(CUDA)
   if(NOT CUDA_FOUND)
     message(STATUS "CUDA toolkit not found, using dynamic runtime loading of libraries (WITH_CUDA_DYNLOAD) instead")
@@ -1090,7 +1070,7 @@ if(MSVC)
   add_definitions(-D__LITTLE_ENDIAN__)
 
   # OSX-Note: as we do cross-compiling with specific set architecture,
-  # endianness-detection and auto-setting is counterproductive
+  # endianess-detection and auto-setting is counterproductive
   # so we just set endianness according CMAKE_OSX_ARCHITECTURES
 
 elseif(CMAKE_OSX_ARCHITECTURES MATCHES i386 OR CMAKE_OSX_ARCHITECTURES MATCHES x86_64 OR CMAKE_OSX_ARCHITECTURES MATCHES arm64)
@@ -1780,7 +1760,7 @@ endif()
 set(CMAKE_CXX_STANDARD 17)
 # If C++17 is not available, downgrading to an earlier standard is NOT OK.
 set(CMAKE_CXX_STANDARD_REQUIRED ON)
-# Do not enable compiler specific language extensions.
+# Do not enable compiler specific language extentions.
 set(CMAKE_CXX_EXTENSIONS OFF)
 
 # Make MSVC properly report the value of the __cplusplus preprocessor macro

GNUmakefile

@@ -51,7 +51,7 @@ Other Convenience Targets
    * config:        Run cmake configuration tool to set build options.
    * deps:          Build library dependencies (intended only for platform maintainers).
 
-                    The existance of locally build dependencies overrides the pre-built dependencies from subversion.
+                    The existance of locally build dependancies overrides the pre-built dependencies from subversion.
                     These must be manually removed from '../lib/' to go back to using the pre-compiled libraries.
 
 Project Files

build_files/build_environment/cmake/gmp.cmake

@@ -38,6 +38,13 @@ elseif(UNIX AND NOT APPLE)
   )
 endif()
 
+if(BLENDER_PLATFORM_ARM)
+  set(GMP_OPTIONS
+    ${GMP_OPTIONS}
+    --disable-assembly
+  )
+endif()
+
 ExternalProject_Add(external_gmp
   URL file://${PACKAGE_DIR}/${GMP_FILE}
   DOWNLOAD_DIR ${DOWNLOAD_DIR}

build_files/build_environment/cmake/harvest.cmake

@@ -17,7 +17,7 @@
 # ***** END GPL LICENSE BLOCK *****
 
 ########################################################################
-# Copy all generated files to the proper structure as blender prefers
+# Copy all generated files to the proper strucure as blender prefers
 ########################################################################
 
 if(NOT DEFINED HARVEST_TARGET)

build_files/build_environment/cmake/options.cmake

@@ -39,7 +39,7 @@ endif()
 set(DOWNLOAD_DIR "${CMAKE_CURRENT_BINARY_DIR}/downloads" CACHE STRING "Path for downloaded files")
 # This path must be hard-coded like this, so that the GNUmakefile knows where it is and can pass it to make_source_archive.py:
 set(PACKAGE_DIR "${CMAKE_CURRENT_BINARY_DIR}/packages")
-option(PACKAGE_USE_UPSTREAM_SOURCES "Use sources upstream to download the package sources, when OFF the blender mirror will be used" ON)
+option(PACKAGE_USE_UPSTREAM_SOURCES "Use soures upstream to download the package sources, when OFF the blender mirror will be used" ON)
 
 file(TO_CMAKE_PATH ${DOWNLOAD_DIR} DOWNLOAD_DIR)
 file(TO_CMAKE_PATH ${PACKAGE_DIR} PACKAGE_DIR)

build_files/build_environment/cmake/package_python.cmake

@@ -24,7 +24,7 @@ if(MSVC)
     add_custom_command(
       OUTPUT ${PYTARGET}/bin/python${PYTHON_POSTFIX}.exe
       COMMAND echo packaging python
-      COMMAND echo this should output at ${PYTARGET}/bin/python${PYTHON_POSTFIX}.exe
+      COMMAND echo this should ouput at ${PYTARGET}/bin/python${PYTHON_POSTFIX}.exe
       COMMAND ${CMAKE_COMMAND} -E make_directory ${PYTARGET}/libs
       COMMAND ${CMAKE_COMMAND} -E copy ${PYSRC}/libs/python${PYTHON_SHORT_VERSION_NO_DOTS}.lib ${PYTARGET}/libs/python${PYTHON_SHORT_VERSION_NO_DOTS}.lib
       COMMAND ${CMAKE_COMMAND} -E copy ${PYSRC}/python.exe ${PYTARGET}/bin/python.exe
@@ -43,7 +43,7 @@ if(MSVC)
     add_custom_command(
       OUTPUT ${PYTARGET}/bin/python${PYTHON_POSTFIX}.exe
       COMMAND echo packaging python
-      COMMAND echo this should output at ${PYTARGET}/bin/python${PYTHON_POSTFIX}.exe
+      COMMAND echo this should ouput at ${PYTARGET}/bin/python${PYTHON_POSTFIX}.exe
       COMMAND ${CMAKE_COMMAND} -E make_directory ${PYTARGET}/libs
       COMMAND ${CMAKE_COMMAND} -E copy ${PYSRC}/libs/python${PYTHON_SHORT_VERSION_NO_DOTS}${PYTHON_POSTFIX}.lib ${PYTARGET}/libs/python${PYTHON_SHORT_VERSION_NO_DOTS}${PYTHON_POSTFIX}.lib
       COMMAND ${CMAKE_COMMAND} -E copy ${PYSRC}/python${PYTHON_POSTFIX}.exe ${PYTARGET}/bin/python${PYTHON_POSTFIX}.exe

build_files/build_environment/cmake/versions.cmake

@@ -474,9 +474,9 @@ set(ISPC_HASH 2e3abedbc0ea9aaec17d6562c632454d)
 set(ISPC_HASH_TYPE MD5)
 set(ISPC_FILE ispc-${ISPC_VERSION}.tar.gz)
 
-set(GMP_VERSION 6.2.1)
+set(GMP_VERSION 6.2.0)
 set(GMP_URI https://gmplib.org/download/gmp/gmp-${GMP_VERSION}.tar.xz)
-set(GMP_HASH 0b82665c4a92fd2ade7440c13fcaa42b)
+set(GMP_HASH a325e3f09e6d91e62101e59f9bda3ec1)
 set(GMP_HASH_TYPE MD5)
 set(GMP_FILE gmp-${GMP_VERSION}.tar.xz)
 

build_files/build_environment/install_deps.sh

@@ -1826,7 +1826,7 @@ compile_OCIO() {
     # Force linking against static libs
     #rm -f $_inst/lib/*.so*
 
-    # Additional dependencies
+    # Additional depencencies
     #cp ext/dist/lib/libtinyxml.a $_inst/lib
     #cp ext/dist/lib/libyaml-cpp.a $_inst/lib
 
@@ -2083,9 +2083,9 @@ compile_OIIO() {
     cmake_d="$cmake_d -D OPENEXR_VERSION=$OPENEXR_VERSION"
 
     if [ "$_with_built_openexr" = true ]; then
-      cmake_d="$cmake_d -D ILMBASE_ROOT=$INST/openexr"
+      cmake_d="$cmake_d -D ILMBASE_HOME=$INST/openexr"
-      cmake_d="$cmake_d -D OPENEXR_ROOT=$INST/openexr"
+      cmake_d="$cmake_d -D OPENEXR_HOME=$INST/openexr"
-      INFO "Ilmbase_ROOT=$INST/openexr"
+      INFO "ILMBASE_HOME=$INST/openexr"
     fi
 
     # ptex is only needed when nicholas bishop is ready
@@ -2374,9 +2374,9 @@ compile_OSL() {
     #~ cmake_d="$cmake_d -D ILMBASE_VERSION=$ILMBASE_VERSION"
 
     if [ "$_with_built_openexr" = true ]; then
-      cmake_d="$cmake_d -D ILMBASE_ROOT=$INST/openexr"
+      INFO "ILMBASE_HOME=$INST/openexr"
-      cmake_d="$cmake_d -D OPENEXR_ROOT=$INST/openexr"
+      cmake_d="$cmake_d -D OPENEXR_ROOT_DIR=$INST/openexr"
-      INFO "Ilmbase_ROOT=$INST/openexr"
+      cmake_d="$cmake_d -D ILMBASE_ROOT_DIR=$INST/openexr"
       # XXX Temp workaround... sigh, ILMBase really messed the things up by defining their custom names ON by default :(
     fi
 
@@ -5801,7 +5801,7 @@ print_info() {
   PRINT "If you're using CMake add this to your configuration flags:"
 
   _buildargs="-U *SNDFILE* -U PYTHON* -U *BOOST* -U *Boost* -U *TBB*"
-  _buildargs="$_buildargs -U *OPENCOLORIO* -U *OPENEXR* -U *OPENIMAGEIO* -U *LLVM* -U *CLANG* -U *CYCLES*"
+  _buildargs="$_buildargs -U *OPENCOLORIO* -U *OPENEXR* -U *OPENIMAGEIO* -U *LLVM* -U *CYCLES*"
   _buildargs="$_buildargs -U *OPENSUBDIV* -U *OPENVDB*  -U *BLOSC* -U *COLLADA* -U *FFMPEG* -U *ALEMBIC* -U *USD*"
   _buildargs="$_buildargs -U *EMBREE* -U *OPENIMAGEDENOISE* -U *OPENXR*"
 

build_files/build_environment/patches/usd.diff

@@ -197,38 +197,3 @@ index 67ec0d15f..6dc3e85a0 100644
  #else
  #error Unknown architecture.
  #endif
-
-diff --git a/pxr/base/arch/demangle.cpp b/pxr/base/arch/demangle.cpp
-index 67ec0d15f..6dc3e85a0 100644
---- a/pxr/base/arch/demangle.cpp
-+++ b/pxr/base/arch/demangle.cpp
-@@ -36,6 +36,7 @@
- #if (ARCH_COMPILER_GCC_MAJOR == 3 && ARCH_COMPILER_GCC_MINOR >= 1) || \
-     ARCH_COMPILER_GCC_MAJOR > 3 || defined(ARCH_COMPILER_CLANG)
- #define _AT_LEAST_GCC_THREE_ONE_OR_CLANG
-+#include <cxxabi.h>
- #endif
- 
- PXR_NAMESPACE_OPEN_SCOPE
-@@ -138,7 +139,6 @@
- #endif
- 
- #if defined(_AT_LEAST_GCC_THREE_ONE_OR_CLANG)
--#include <cxxabi.h>
- 
- /*
-  * This routine doesn't work when you get to gcc3.4.
-
-diff --git a/pxr/base/work/singularTask.h b/pxr/base/work/singularTask.h
-index 67ec0d15f..6dc3e85a0 100644
---- a/pxr/base/work/singularTask.h
-+++ b/pxr/base/work/singularTask.h
-@@ -120,7 +120,7 @@
-                     // case we go again to ensure the task can do whatever it
-                     // was awakened to do.  Once we successfully take the count
-                     // to zero, we stop.
--                    size_t old = count;
-+                    std::size_t old = count;
-                     do { _fn(); } while (
-                         !count.compare_exchange_strong(old, 0));
-                 });

build_files/cmake/Modules/FindOptiX.cmake

@@ -21,7 +21,7 @@ ENDIF()
 
 SET(_optix_SEARCH_DIRS
   ${OPTIX_ROOT_DIR}
-  "$ENV{PROGRAMDATA}/NVIDIA Corporation/OptiX SDK 7.3.0"
+  "$ENV{PROGRAMDATA}/NVIDIA Corporation/OptiX SDK 7.0.0"
 )
 
 FIND_PATH(OPTIX_INCLUDE_DIR

build_files/cmake/cmake_consistency_check.py

@@ -114,7 +114,7 @@ def is_c_header(filename: str) -> bool:
 
 def is_c(filename: str) -> bool:
     ext = splitext(filename)[1]
-    return (ext in {".c", ".cpp", ".cxx", ".m", ".mm", ".rc", ".cc", ".inl", ".metal"})
+    return (ext in {".c", ".cpp", ".cxx", ".m", ".mm", ".rc", ".cc", ".inl"})
 
 
 def is_c_any(filename: str) -> bool:

build_files/cmake/cmake_netbeans_project.py

@@ -180,7 +180,7 @@ def create_nb_project_main():
         f.write('    </logicalFolder>\n')
 
         f.write('  </logicalFolder>\n')
-        # default, but this dir is in fact not in blender dir so we can ignore it
+        # default, but this dir is infact not in blender dir so we can ignore it
         # f.write('  <sourceFolderFilter>^(nbproject)$</sourceFolderFilter>\n')
         f.write(r'  <sourceFolderFilter>^(nbproject|__pycache__|.*\.py|.*\.html|.*\.blend)$</sourceFolderFilter>\n')
 

build_files/cmake/config/blender_lite.cmake

@@ -19,6 +19,9 @@ set(WITH_CODEC_SNDFILE       OFF CACHE BOOL "" FORCE)
 set(WITH_COMPOSITOR          OFF CACHE BOOL "" FORCE)
 set(WITH_COREAUDIO           OFF CACHE BOOL "" FORCE)
 set(WITH_CYCLES              OFF CACHE BOOL "" FORCE)
+set(WITH_CYCLES_DEVICE_OPTIX OFF CACHE BOOL "" FORCE)
+set(WITH_CYCLES_EMBREE       OFF CACHE BOOL "" FORCE)
+set(WITH_CYCLES_OSL          OFF CACHE BOOL "" FORCE)
 set(WITH_DRACO               OFF CACHE BOOL "" FORCE)
 set(WITH_FFTW3               OFF CACHE BOOL "" FORCE)
 set(WITH_FREESTYLE           OFF CACHE BOOL "" FORCE)

build_files/cmake/config/blender_release.cmake

@@ -61,7 +61,6 @@ set(WITH_MEM_JEMALLOC          ON  CACHE BOOL "" FORCE)
 # platform dependent options
 if(APPLE)
   set(WITH_COREAUDIO           ON  CACHE BOOL "" FORCE)
-  set(WITH_CYCLES_DEVICE_METAL ON  CACHE BOOL "" FORCE)
 endif()
 if(NOT WIN32)
   set(WITH_JACK                ON  CACHE BOOL "" FORCE)

build_files/cmake/macros.cmake

@@ -529,7 +529,7 @@ function(SETUP_LIBDIRS)
 
   # NOTE: For all new libraries, use absolute library paths.
   # This should eventually be phased out.
-  # APPLE platform uses full paths for linking libraries, and avoids link_directories.
+  # APPLE plaform uses full paths for linking libraries, and avoids link_directories.
   if(NOT MSVC AND NOT APPLE)
     link_directories(${JPEG_LIBPATH} ${PNG_LIBPATH} ${ZLIB_LIBPATH} ${FREETYPE_LIBPATH})
 
@@ -1292,6 +1292,29 @@ macro(openmp_delayload
     endif()
 endmacro()
 
+macro(blender_precompile_headers target cpp header)
+  if(MSVC)
+    # get the name for the pch output file
+    get_filename_component(pchbase ${cpp} NAME_WE)
+    set(pchfinal "${CMAKE_CURRENT_BINARY_DIR}/${CMAKE_CFG_INTDIR}/${pchbase}.pch")
+
+    # mark the cpp as the one outputting the pch
+    set_property(SOURCE ${cpp} APPEND PROPERTY OBJECT_OUTPUTS "${pchfinal}")
+
+    # get all sources for the target
+    get_target_property(sources ${target} SOURCES)
+
+    # make all sources depend on the pch to enforce the build order
+    foreach(src ${sources})
+      set_property(SOURCE ${src} APPEND PROPERTY OBJECT_DEPENDS "${pchfinal}")
+    endforeach()
+
+    target_sources(${target} PRIVATE ${cpp} ${header})
+    set_target_properties(${target} PROPERTIES COMPILE_FLAGS "/Yu${header} /Fp${pchfinal} /FI${header}")
+    set_source_files_properties(${cpp} PROPERTIES COMPILE_FLAGS "/Yc${header} /Fp${pchfinal}")
+  endif()
+endmacro()
+
 macro(set_and_warn_dependency
   _dependency _setting _val)
   # when $_dependency is disabled, forces $_setting = $_val

build_files/cmake/platform/platform_apple.cmake

@@ -257,6 +257,9 @@ if(WITH_BOOST)
   if(WITH_INTERNATIONAL)
     list(APPEND _boost_FIND_COMPONENTS locale)
   endif()
+  if(WITH_CYCLES_NETWORK)
+    list(APPEND _boost_FIND_COMPONENTS serialization)
+  endif()
   if(WITH_OPENVDB)
     list(APPEND _boost_FIND_COMPONENTS iostreams)
   endif()
@@ -336,7 +339,7 @@ if(WITH_LLVM)
 
 endif()
 
-if(WITH_CYCLES AND WITH_CYCLES_OSL)
+if(WITH_CYCLES_OSL)
   set(CYCLES_OSL ${LIBDIR}/osl)
 
   find_library(OSL_LIB_EXEC NAMES oslexec PATHS ${CYCLES_OSL}/lib)
@@ -356,7 +359,7 @@ if(WITH_CYCLES AND WITH_CYCLES_OSL)
   endif()
 endif()
 
-if(WITH_CYCLES AND WITH_CYCLES_EMBREE)
+if(WITH_CYCLES_EMBREE)
   find_package(Embree 3.8.0 REQUIRED)
   # Increase stack size for Embree, only works for executables.
   if(NOT WITH_PYTHON_MODULE)

build_files/cmake/platform/platform_apple_xcode.cmake

@@ -96,7 +96,7 @@ else()
     # Detect SDK version to use.
     if(NOT DEFINED OSX_SYSTEM)
       execute_process(
-          COMMAND xcrun --sdk macosx --show-sdk-version
+          COMMAND xcrun --show-sdk-version
           OUTPUT_VARIABLE OSX_SYSTEM
           OUTPUT_STRIP_TRAILING_WHITESPACE)
     endif()

build_files/cmake/platform/platform_unix.cmake

@@ -18,7 +18,7 @@
 # All rights reserved.
 # ***** END GPL LICENSE BLOCK *****
 
-# Libraries configuration for any *nix system including Linux and Unix (excluding APPLE).
+# Libraries configuration for any *nix system including Linux and Unix.
 
 # Detect precompiled library directory
 if(NOT DEFINED LIBDIR)
@@ -241,7 +241,7 @@ if(WITH_INPUT_NDOF)
   endif()
 endif()
 
-if(WITH_CYCLES AND WITH_CYCLES_OSL)
+if(WITH_CYCLES_OSL)
   set(CYCLES_OSL ${LIBDIR}/osl CACHE PATH "Path to OpenShadingLanguage installation")
   if(EXISTS ${CYCLES_OSL} AND NOT OSL_ROOT)
     set(OSL_ROOT ${CYCLES_OSL})
@@ -314,7 +314,7 @@ if(WITH_BOOST)
     endif()
     set(Boost_USE_MULTITHREADED ON)
     set(__boost_packages filesystem regex thread date_time)
-    if(WITH_CYCLES AND WITH_CYCLES_OSL)
+    if(WITH_CYCLES_OSL)
       if(NOT (${OSL_LIBRARY_VERSION_MAJOR} EQUAL "1" AND ${OSL_LIBRARY_VERSION_MINOR} LESS "6"))
         list(APPEND __boost_packages wave)
       else()
@@ -323,6 +323,9 @@ if(WITH_BOOST)
     if(WITH_INTERNATIONAL)
       list(APPEND __boost_packages locale)
     endif()
+    if(WITH_CYCLES_NETWORK)
+      list(APPEND __boost_packages serialization)
+    endif()
     if(WITH_OPENVDB)
       list(APPEND __boost_packages iostreams)
     endif()
@@ -400,7 +403,7 @@ if(WITH_OPENCOLORIO)
   endif()
 endif()
 
-if(WITH_CYCLES AND WITH_CYCLES_EMBREE)
+if(WITH_CYCLES_EMBREE)
   find_package(Embree 3.8.0 REQUIRED)
 endif()
 

build_files/cmake/platform/platform_win32.cmake

@@ -477,7 +477,7 @@ if(WITH_PYTHON)
 endif()
 
 if(WITH_BOOST)
-  if(WITH_CYCLES AND WITH_CYCLES_OSL)
+  if(WITH_CYCLES_OSL)
     set(boost_extra_libs wave)
   endif()
   if(WITH_INTERNATIONAL)
@@ -520,7 +520,7 @@ if(WITH_BOOST)
       debug ${BOOST_LIBPATH}/libboost_thread-${BOOST_DEBUG_POSTFIX}
       debug ${BOOST_LIBPATH}/libboost_chrono-${BOOST_DEBUG_POSTFIX}
     )
-    if(WITH_CYCLES AND WITH_CYCLES_OSL)
+    if(WITH_CYCLES_OSL)
       set(BOOST_LIBRARIES ${BOOST_LIBRARIES}
         optimized ${BOOST_LIBPATH}/libboost_wave-${BOOST_POSTFIX}
         debug ${BOOST_LIBPATH}/libboost_wave-${BOOST_DEBUG_POSTFIX})
@@ -708,7 +708,7 @@ if(WITH_CODEC_SNDFILE)
   set(LIBSNDFILE_LIBRARIES ${LIBSNDFILE_LIBPATH}/libsndfile-1.lib)
 endif()
 
-if(WITH_CYCLES AND WITH_CYCLES_OSL)
+if(WITH_CYCLES_OSL)
   set(CYCLES_OSL ${LIBDIR}/osl CACHE PATH "Path to OpenShadingLanguage installation")
   set(OSL_SHADER_DIR ${CYCLES_OSL}/shaders)
   # Shaders have moved around a bit between OSL versions, check multiple locations
@@ -741,7 +741,7 @@ if(WITH_CYCLES AND WITH_CYCLES_OSL)
   endif()
 endif()
 
-if(WITH_CYCLES AND WITH_CYCLES_EMBREE)
+if(WITH_CYCLES_EMBREE)
   windows_find_package(Embree)
   if(NOT EMBREE_FOUND)
     set(EMBREE_INCLUDE_DIRS ${LIBDIR}/embree/include)

build_files/cmake/platform/platform_win32_bundle_crt.cmake

@@ -27,7 +27,7 @@ if(WITH_WINDOWS_BUNDLE_CRT)
   # Install the CRT to the blender.crt Sub folder.
   install(FILES ${CMAKE_INSTALL_SYSTEM_RUNTIME_LIBS} DESTINATION ./blender.crt COMPONENT Libraries)
 
-  # Generating the manifest is a relatively expensive operation since
+  # Generating the manifest is a relativly expensive operation since
   # it is collecting an sha1 hash for every file required. so only do
   # this work when the libs have either changed or the manifest does
   # not exist yet.

build_files/config/pipeline_config.yaml

@@ -5,38 +5,38 @@
 update-code:
     git:
         submodules:
-        -   branch: master
+        -   branch: blender-v3.0-release
             commit_id: HEAD
             path: release/scripts/addons
-        -   branch: master
+        -   branch: blender-v3.0-release
             commit_id: HEAD
             path: release/scripts/addons_contrib
-        -   branch: master
+        -   branch: blender-v3.0-release
             commit_id: HEAD
             path: release/datafiles/locale
-        -   branch: master
+        -   branch: blender-v3.0-release
             commit_id: HEAD
             path: source/tools
     svn:
         libraries:
             darwin-arm64:
-                branch: trunk
+                branch: tags/blender-3.0-release
                 commit_id: HEAD
                 path: lib/darwin_arm64
             darwin-x86_64:
-                branch: trunk
+                branch: tags/blender-3.0-release
                 commit_id: HEAD
                 path: lib/darwin
             linux-x86_64:
-                branch: trunk
+                branch: tags/blender-3.0-release
                 commit_id: HEAD
                 path: lib/linux_centos7_x86_64
             windows-amd64:
-                branch: trunk
+                branch: tags/blender-3.0-release
                 commit_id: HEAD
                 path: lib/win64_vc15
         tests:
-            branch: trunk
+            branch: tags/blender-3.0-release
             commit_id: HEAD
             path: lib/tests
         benchmarks:

doc/doxygen/Doxyfile

@@ -38,7 +38,7 @@ PROJECT_NAME           = Blender
 # could be handy for archiving the generated documentation or if some version
 # control system is used.
 
-PROJECT_NUMBER         = V3.1
+PROJECT_NUMBER         = V3.0
 
 # Using the PROJECT_BRIEF tag one can provide an optional one line description
 # for a project that appears at the top of each page and should give viewer a

doc/doxygen/doxygen.intern.h

@@ -6,87 +6,91 @@
  *  as part of the normal development process.
  */
 
-/* TODO: other modules.
+/** \defgroup MEM Guarded memory (de)allocation
- * - `libmv`
+ *  \ingroup intern
- * - `cycles`
- * - `opencolorio`
- * - `opensubdiv`
- * - `openvdb`
- * - `quadriflow`
  */
 
-/** \defgroup intern_atomic Atomic Operations
+/** \defgroup clog C-Logging (CLOG)
- *  \ingroup intern */
+ *  \ingroup intern
+ */
 
-/** \defgroup intern_clog C-Logging (CLOG)
+/** \defgroup ctr container
- *  \ingroup intern */
+ *  \ingroup intern
+ */
 
-/** \defgroup intern_eigen Eigen
+/** \defgroup iksolver iksolver
- *  \ingroup intern */
+ *  \ingroup intern
+ */
 
-/** \defgroup intern_glew-mx GLEW with Multiple Rendering Context's
+/** \defgroup itasc itasc
- *  \ingroup intern */
+ *  \ingroup intern
+ */
 
-/** \defgroup intern_iksolver Inverse Kinematics (Solver)
+/** \defgroup memutil memutil
- *  \ingroup intern */
+ *  \ingroup intern
+ */
 
-/** \defgroup intern_itasc Inverse Kinematics (ITASC)
+/** \defgroup mikktspace mikktspace
- *  \ingroup intern */
+ *  \ingroup intern
+ */
 
-/** \defgroup intern_libc_compat libc Compatibility For Linux
+/** \defgroup moto moto
- *  \ingroup intern */
+ *  \ingroup intern
+ */
 
-/** \defgroup intern_locale Locale
+/** \defgroup eigen eigen
- *  \ingroup intern */
+ *  \ingroup intern
+ */
 
-/** \defgroup intern_mantaflow Manta-Flow Fluid Simulation
+/** \defgroup smoke smoke
- *  \ingroup intern */
+ *  \ingroup intern
+ */
 
-/** \defgroup intern_mem Guarded Memory (de)allocation
+/** \defgroup string string
- *  \ingroup intern */
+ *  \ingroup intern
-
+ */
-/** \defgroup intern_memutil Memory Utilities (memutil)
- *  \ingroup intern */
-
-/** \defgroup intern_mikktspace MikktSpace
- *  \ingroup intern */
-
-/** \defgroup intern_rigidbody Rigid-Body C-API
- *  \ingroup intern */
-
-/** \defgroup intern_sky_model Sky Model
- *  \ingroup intern */
-
-/** \defgroup intern_utf_conv UTF-8/16 Conversion (utfconv)
- *  \ingroup intern */
 
 /** \defgroup audaspace Audaspace
  *  \ingroup intern undoc
- *  \todo add to doxygen */
+ *  \todo add to doxygen
+ */
 /** \defgroup audcoreaudio Audaspace CoreAudio
- *  \ingroup audaspace */
+ *  \ingroup audaspace
+ */
 /** \defgroup audfx Audaspace FX
- *  \ingroup audaspace */
+ *  \ingroup audaspace
+ */
 /** \defgroup audopenal Audaspace OpenAL
- *  \ingroup audaspace */
+ *  \ingroup audaspace
+ */
 /** \defgroup audpulseaudio Audaspace PulseAudio
- *  \ingroup audaspace */
+ *  \ingroup audaspace
+ */
 /** \defgroup audwasapi Audaspace WASAPI
- *  \ingroup audaspace */
+ *  \ingroup audaspace
+ */
 /** \defgroup audpython Audaspace Python
- *  \ingroup audaspace */
+ *  \ingroup audaspace
+ */
 /** \defgroup audsdl Audaspace SDL
- *  \ingroup audaspace */
+ *  \ingroup audaspace
+ */
 /** \defgroup audsrc Audaspace SRC
- *  \ingroup audaspace */
+ *
+ *  \ingroup audaspace
+ */
 /** \defgroup audffmpeg Audaspace FFMpeg
- *  \ingroup audaspace */
+ *  \ingroup audaspace
+ */
 /** \defgroup audfftw Audaspace FFTW
- *  \ingroup audaspace */
+ *  \ingroup audaspace
+ */
 /** \defgroup audjack Audaspace Jack
- *  \ingroup audaspace */
+ *  \ingroup audaspace
+ */
 /** \defgroup audsndfile Audaspace sndfile
- *  \ingroup audaspace */
+ *  \ingroup audaspace
+ */
 
 /** \defgroup GHOST GHOST API
  * \ingroup intern GUI

doc/doxygen/doxygen.source.h

@@ -5,8 +5,7 @@
 /** \defgroup bmesh BMesh
  *  \ingroup blender
  */
-/** \defgroup compositor Compositing
+/** \defgroup compositor Compositing */
- *  \ingroup blender */
 
 /** \defgroup python Python
  *  \ingroup blender
@@ -79,8 +78,7 @@
  *  \ingroup blender
  */
 
-/** \defgroup data DNA, RNA and .blend access
+/** \defgroup data DNA, RNA and .blend access*/
- *  \ingroup blender */
 
 /** \defgroup gpu GPU
  *  \ingroup blender
@@ -103,12 +101,11 @@
  *   merged in docs.
  */
 
-/**
+/** \defgroup gui GUI */
- * \defgroup gui GUI
- * \ingroup blender */
 
 /** \defgroup wm Window Manager
- *  \ingroup gui */
+ *  \ingroup blender gui
+ */
 
 /* ================================ */
 
@@ -282,8 +279,7 @@
  *  \ingroup gui
  */
 
-/** \defgroup externformats External Formats
+/** \defgroup externformats External Formats */
- *  \ingroup blender */
 
 /** \defgroup collada COLLADA
  *  \ingroup externformats
@@ -312,7 +308,4 @@
 /* ================================ */
 
 /** \defgroup undoc Undocumented
- *
+ *  \brief Modules and libraries that are still undocumented, or lacking proper integration into the doxygen system, are marked in this group. */
- * \brief Modules and libraries that are still undocumented,
- * or lacking proper integration into the doxygen system, are marked in this group.
- */

doc/manpage/blender.1.py

@@ -61,7 +61,7 @@ def blender_extract_info(blender_bin: str) -> Dict[str, str]:
         stdout=subprocess.PIPE,
     ).stdout.decode(encoding="utf-8")
 
-    blender_version_output = subprocess.run(
+    blender_version_ouput = subprocess.run(
         [blender_bin, "--version"],
         env=blender_env,
         check=True,
@@ -73,7 +73,7 @@ def blender_extract_info(blender_bin: str) -> Dict[str, str]:
     # check for each lines prefix to ensure these aren't included.
     blender_version = ""
     blender_date = ""
-    for l in blender_version_output.split("\n"):
+    for l in blender_version_ouput.split("\n"):
         if l.startswith("Blender "):
             # Remove 'Blender' prefix.
             blender_version = l.split(" ", 1)[1].strip()

doc/python_api/examples/bpy.app.timers.5.py

@@ -11,7 +11,7 @@ import queue
 
 execution_queue = queue.Queue()
 
-# This function can safely be called in another thread.
+# This function can savely be called in another thread.
 # The function will be executed when the timer runs the next time.
 def run_in_main_thread(function):
     execution_queue.put(function)

doc/python_api/rst/info_gotcha.rst

@@ -728,7 +728,7 @@ Abusing RNA property callbacks
 ------------------------------
 
 Python-defined RNA properties can have custom callbacks. Trying to perform complex operations
-from there, like calling an operator, may work, but is not officially recommended nor supported.
+from there, like calling an operator, may work, but is not officialy recommended nor supported.
 
 Main reason is that those callback should be very fast, but additionally, it may for example
 create issues with undo/redo system (most operators store an history step, and editing an RNA

doc/python_api/sphinx_doc_gen.py

@@ -1103,7 +1103,6 @@ context_type_map = {
     "selectable_objects": ("Object", True),
     "selected_asset_files": ("FileSelectEntry", True),
     "selected_bones": ("EditBone", True),
-    "selected_editable_actions": ("Action", True),
     "selected_editable_bones": ("EditBone", True),
     "selected_editable_fcurves": ("FCurve", True),
     "selected_editable_keyframes": ("Keyframe", True),
@@ -1119,7 +1118,6 @@ context_type_map = {
     "selected_pose_bones": ("PoseBone", True),
     "selected_pose_bones_from_active_object": ("PoseBone", True),
     "selected_sequences": ("Sequence", True),
-    "selected_visible_actions": ("Action", True),
     "selected_visible_fcurves": ("FCurve", True),
     "sequences": ("Sequence", True),
     "soft_body": ("SoftBodyModifier", False),
@@ -1226,10 +1224,7 @@ def pycontext2sphinx(basepath):
         while char_array[i] is not None:
             member = ctypes.string_at(char_array[i]).decode(encoding="ascii")
             fw(".. data:: %s\n\n" % member)
-            try:
+            member_type, is_seq = context_type_map[member]
-                member_type, is_seq = context_type_map[member]
-            except KeyError:
-                raise SystemExit("Error: context key %r not found in context_type_map; update %s" % (member, __file__)) from None
             fw("   :type: %s :class:`bpy.types.%s`\n\n" % ("sequence of " if is_seq else "", member_type))
             unique.add(member)
             i += 1
@@ -2256,7 +2251,7 @@ def main():
     # First monkey patch to load in fake members.
     setup_monkey_patch()
 
-    # Perform changes to Blender itself.
+    # Perform changes to Blender it's self.
     setup_data = setup_blender()
 
     # eventually, create the dirs

extern/hipew/src/hipew.c

@@ -257,7 +257,7 @@ static int hipewHipInit(void) {
 #endif
   static int initialized = 0;
   static int result = 0;
-  int error;
+  int error, driver_version;
 
   if (initialized) {
     return result;
@@ -565,6 +565,8 @@ int hipewCompilerVersion(void) {
   const char *path = hipewCompilerPath();
   const char *marker = "Hip compilation tools, release ";
   FILE *pipe;
+  int major, minor;
+  char *versionstr;
   char buf[128];
   char output[65536] = "\0";
   char command[65536] = "\0";

intern/CMakeLists.txt

@@ -25,6 +25,7 @@ add_subdirectory(ghost)
 add_subdirectory(guardedalloc)
 add_subdirectory(libmv)
 add_subdirectory(memutil)
+add_subdirectory(numaapi)
 add_subdirectory(opencolorio)
 add_subdirectory(opensubdiv)
 add_subdirectory(mikktspace)

intern/atomic/atomic_ops.h

@@ -45,7 +45,7 @@
  */
 
 /** \file
- * \ingroup intern_atomic
+ * \ingroup Atomic
  *
  * \brief Provides wrapper around system-specific atomic primitives,
  * and some extensions (faked-atomic operations over float numbers).

intern/atomic/intern/atomic_ops_ext.h

@@ -44,10 +44,6 @@
  * The Original Code is: adapted from jemalloc.
  */
 
-/** \file
- * \ingroup intern_atomic
- */
-
 #ifndef __ATOMIC_OPS_EXT_H__
 #define __ATOMIC_OPS_EXT_H__
 

intern/atomic/intern/atomic_ops_msvc.h

@@ -5,7 +5,7 @@
  * All rights reserved.
  * Copyright (C) 2007-2012 Mozilla Foundation.  All rights reserved.
  * Copyright (C) 2009-2013 Facebook, Inc.  All rights reserved.
- *
+
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  * 1. Redistributions of source code must retain the above copyright notice(s),
@@ -13,7 +13,7 @@
  * 2. Redistributions in binary form must reproduce the above copyright notice(s),
  *    this list of conditions and the following disclaimer in the documentation
  *    and/or other materials provided with the distribution.
- *
+
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``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
@@ -26,10 +26,6 @@
  * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-/** \file
- * \ingroup intern_atomic
- */
-
 #ifndef __ATOMIC_OPS_MSVC_H__
 #define __ATOMIC_OPS_MSVC_H__
 

intern/atomic/intern/atomic_ops_unix.h

@@ -44,10 +44,6 @@
  * The Original Code is: adapted from jemalloc.
  */
 
-/** \file
- * \ingroup intern_atomic
- */
-
 #ifndef __ATOMIC_OPS_UNIX_H__
 #define __ATOMIC_OPS_UNIX_H__
 

intern/atomic/intern/atomic_ops_utils.h

@@ -44,10 +44,6 @@
  * The Original Code is: adapted from jemalloc.
  */
 
-/** \file
- * \ingroup intern_atomic
- */
-
 #ifndef __ATOMIC_OPS_UTILS_H__
 #define __ATOMIC_OPS_UTILS_H__
 

intern/clog/CLG_log.h

@@ -14,8 +14,11 @@
  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  */
 
+#ifndef __CLG_LOG_H__
+#define __CLG_LOG_H__
+
 /** \file
- * \ingroup intern_clog
+ * \ingroup clog
  *
  * C Logging Library (clog)
  * ========================
@@ -65,9 +68,6 @@
  * - 4+: May be used for more details than 3, should be avoided but not prevented.
  */
 
-#ifndef __CLG_LOG_H__
-#define __CLG_LOG_H__
-
 #ifdef __cplusplus
 extern "C" {
 #endif /* __cplusplus */

intern/clog/clog.c

@@ -15,7 +15,7 @@
  */
 
 /** \file
- * \ingroup intern_clog
+ * \ingroup clog
  */
 
 #include <assert.h>
@@ -388,7 +388,7 @@ static void clg_ctx_fatal_action(CLogContext *ctx)
 
 static void clg_ctx_backtrace(CLogContext *ctx)
 {
-  /* NOTE: we avoid writing to 'FILE', for back-trace we make an exception,
+  /* Note: we avoid writing to 'FILE', for back-trace we make an exception,
    * if necessary we could have a version of the callback that writes to file
    * descriptor all at once. */
   ctx->callbacks.backtrace_fn(ctx->output_file);

intern/cycles/CMakeLists.txt

@@ -226,9 +226,6 @@ add_definitions(
   -DCCL_NAMESPACE_END=}
 )
 
-if(WITH_CYCLES_DEBUG)
-  add_definitions(-DWITH_CYCLES_DEBUG)
-endif()
 if(WITH_CYCLES_STANDALONE_GUI)
   add_definitions(-DWITH_CYCLES_STANDALONE_GUI)
 endif()
@@ -337,7 +334,7 @@ else()
 endif()
 
 # Warnings
-if(CMAKE_COMPILER_IS_GNUCXX OR CMAKE_C_COMPILER_ID MATCHES "Clang")
+if(CMAKE_COMPILER_IS_GNUCXX)
   ADD_CHECK_CXX_COMPILER_FLAG(CMAKE_CXX_FLAGS _has_cxxflag_float_conversion "-Werror=float-conversion")
   ADD_CHECK_CXX_COMPILER_FLAG(CMAKE_CXX_FLAGS _has_cxxflag_double_promotion "-Werror=double-promotion")
   ADD_CHECK_CXX_COMPILER_FLAG(CMAKE_CXX_FLAGS _has_no_error_unused_macros "-Wno-error=unused-macros")

intern/cycles/blender/CMakeLists.txt

@@ -40,7 +40,6 @@ set(SRC
   object_cull.cpp
   output_driver.cpp
   particles.cpp
-  pointcloud.cpp
   curves.cpp
   logging.cpp
   python.cpp
@@ -88,7 +87,6 @@ endif()
 
 set(ADDON_FILES
   addon/__init__.py
-  addon/camera.py
   addon/engine.py
   addon/operators.py
   addon/osl.py
@@ -103,11 +101,6 @@ add_definitions(${GL_DEFINITIONS})
 if(WITH_CYCLES_DEVICE_HIP)
   add_definitions(-DWITH_HIP)
 endif()
-
-if(WITH_CYCLES_DEVICE_METAL)
-  add_definitions(-DWITH_METAL)
-endif()
-
 if(WITH_MOD_FLUID)
   add_definitions(-DWITH_FLUID)
 endif()
@@ -145,6 +138,11 @@ endif()
 
 blender_add_lib(bf_intern_cycles "${SRC}" "${INC}" "${INC_SYS}" "${LIB}")
 
+# avoid link failure with clang 3.4 debug
+if(CMAKE_C_COMPILER_ID MATCHES "Clang" AND NOT ${CMAKE_C_COMPILER_VERSION} VERSION_LESS '3.4')
+  string(APPEND CMAKE_CXX_FLAGS_DEBUG " -gline-tables-only")
+endif()
+
 add_dependencies(bf_intern_cycles bf_rna)
 
 delayed_install(${CMAKE_CURRENT_SOURCE_DIR} "${ADDON_FILES}" ${CYCLES_INSTALL_PATH})

intern/cycles/blender/addon/camera.py

@@ -1,84 +0,0 @@
-#
-# Copyright 2011-2021 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.
-#
-
-# <pep8 compliant>
-
-# Fit to match default projective camera with focal_length 50 and sensor_width 36.
-default_fisheye_polynomial = [-1.1735143712967577e-05,
-                              -0.019988736953434998,
-                              -3.3525322965709175e-06,
-                              3.099275275886036e-06,
-                              -2.6064646454854524e-08]
-
-# Utilities to generate lens polynomials to match built-in camera types, only here
-# for reference at the moment, not used by the code.
-def create_grid(sensor_height, sensor_width):
-    import numpy as np
-    if sensor_height is None:
-        sensor_height = sensor_width / (16 / 9)  # Default aspect ration 16:9
-    uu, vv = np.meshgrid(np.linspace(0, 1, 100), np.linspace(0, 1, 100))
-    uu = (uu - 0.5) * sensor_width
-    vv = (vv - 0.5) * sensor_height
-    rr = np.sqrt(uu ** 2 + vv ** 2)
-    return rr
-
-
-def fisheye_lens_polynomial_from_projective(focal_length=50, sensor_width=36, sensor_height=None):
-    import numpy as np
-    rr = create_grid(sensor_height, sensor_width)
-    polynomial = np.polyfit(rr.flat, (-np.arctan(rr / focal_length)).flat, 4)
-    return list(reversed(polynomial))
-
-
-def fisheye_lens_polynomial_from_projective_fov(fov, sensor_width=36, sensor_height=None):
-    import numpy as np
-    f = sensor_width / 2 / np.tan(fov / 2)
-    return fisheye_lens_polynomial_from_projective(f, sensor_width, sensor_height)
-
-
-def fisheye_lens_polynomial_from_equisolid(lens=10.5, sensor_width=36, sensor_height=None):
-    import numpy as np
-    rr = create_grid(sensor_height, sensor_width)
-    x = rr.reshape(-1)
-    x = np.stack([x**i for i in [1, 2, 3, 4]])
-    y = (-2 * np.arcsin(rr / (2 * lens))).reshape(-1)
-    polynomial = np.linalg.lstsq(x.T, y.T, rcond=None)[0]
-    return [0] + list(polynomial)
-
-
-def fisheye_lens_polynomial_from_equidistant(fov=180, sensor_width=36, sensor_height=None):
-    import numpy as np
-    return [0, -np.radians(fov) / sensor_width, 0, 0, 0]
-
-
-def fisheye_lens_polynomial_from_distorted_projective_polynomial(k1, k2, k3, focal_length=50, sensor_width=36, sensor_height=None):
-    import numpy as np
-    rr = create_grid(sensor_height, sensor_width)
-    r2 = (rr / focal_length) ** 2
-    r4 = r2 * r2
-    r6 = r4 * r2
-    r_coeff = 1 + k1 * r2 + k2 * r4 + k3 * r6
-    polynomial = np.polyfit(rr.flat, (-np.arctan(rr / focal_length * r_coeff)).flat, 4)
-    return list(reversed(polynomial))
-
-def fisheye_lens_polynomial_from_distorted_projective_divisions(k1, k2, focal_length=50, sensor_width=36, sensor_height=None):
-    import numpy as np
-    rr = create_grid(sensor_height, sensor_width)
-    r2 = (rr / focal_length) ** 2
-    r4 = r2 * r2
-    r_coeff = 1 + k1 * r2 + k2 * r4
-    polynomial = np.polyfit(rr.flat, (-np.arctan(rr / focal_length / r_coeff)).flat, 4)
-    return list(reversed(polynomial))

intern/cycles/blender/addon/engine.py

@@ -28,7 +28,7 @@ def _configure_argument_parser():
                         action='store_true')
     parser.add_argument("--cycles-device",
                         help="Set the device to use for Cycles, overriding user preferences and the scene setting."
-                             "Valid options are 'CPU', 'CUDA', 'OPTIX', 'HIP' or 'METAL'."
+                             "Valid options are 'CPU', 'CUDA', 'OPTIX', or 'HIP'"
                              "Additionally, you can append '+CPU' to any GPU type for hybrid rendering.",
                         default=None)
     return parser
@@ -60,8 +60,9 @@ def init():
 
     path = os.path.dirname(__file__)
     user_path = os.path.dirname(os.path.abspath(bpy.utils.user_resource('CONFIG', path='')))
+    temp_path = bpy.app.tempdir
 
-    _cycles.init(path, user_path, bpy.app.background)
+    _cycles.init(path, user_path, temp_path, bpy.app.background)
     _parse_command_line()
 
 

intern/cycles/blender/addon/properties.py

@@ -33,7 +33,6 @@ from math import pi
 # enums
 
 from . import engine
-from . import camera
 
 enum_devices = (
     ('CPU', "CPU", "Use CPU for rendering"),
@@ -73,8 +72,6 @@ enum_panorama_types = (
     ('FISHEYE_EQUISOLID', "Fisheye Equisolid",
                           "Similar to most fisheye modern lens, takes sensor dimensions into consideration"),
     ('MIRRORBALL', "Mirror Ball", "Uses the mirror ball mapping"),
-    ('FISHEYE_LENS_POLYNOMIAL', "Fisheye Lens Polynomial",
-     "Defines the lens projection as polynomial to allow real world camera lenses to be mimicked."),
 )
 
 enum_curve_shape = (
@@ -114,8 +111,7 @@ enum_device_type = (
     ('CPU', "CPU", "CPU", 0),
     ('CUDA', "CUDA", "CUDA", 1),
     ('OPTIX', "OptiX", "OptiX", 3),
-    ('HIP', "HIP", "HIP", 4),
+    ("HIP", "HIP", "HIP", 4)
-    ('METAL', "Metal", "Metal", 5)
 )
 
 enum_texture_limit = (
@@ -222,12 +218,6 @@ enum_denoising_prefilter = (
     ('ACCURATE', "Accurate", "Prefilter noisy guiding passes before denoising color. Improves quality when guiding passes are noisy using extra processing time", 3),
 )
 
-enum_direct_light_sampling_type = (
-    ('MULTIPLE_IMPORTANCE_SAMPLING', "Multiple Importance Sampling", "Multiple importance sampling is used to combine direct light contributions from next-event estimation and forward path tracing", 0),
-    ('FORWARD_PATH_TRACING', "Forward Path Tracing", "Direct light contributions are only sampled using forward path tracing", 1),
-    ('NEXT_EVENT_ESTIMATION', "Next-Event Estimation", "Direct light contributions are only sampled using next-event estimation", 2),
-)
-
 def update_render_passes(self, context):
     scene = context.scene
     view_layer = context.view_layer
@@ -335,13 +325,6 @@ class CyclesRenderSettings(bpy.types.PropertyGroup):
         default=1024,
     )
 
-    sample_offset: IntProperty(
-        name="Sample Offset",
-        description="Number of samples to skip when starting render",
-        min=0, max=(1 << 24),
-        default=0,
-    )
-
     time_limit: FloatProperty(
         name="Time Limit",
         description="Limit the render time (excluding synchronization time)."
@@ -432,13 +415,6 @@ class CyclesRenderSettings(bpy.types.PropertyGroup):
         default=0,
     )
 
-    direct_light_sampling_type: EnumProperty(
-        name="Direct Light Sampling",
-        description="The type of strategy used for sampling direct light contributions",
-        items=enum_direct_light_sampling_type,
-        default='MULTIPLE_IMPORTANCE_SAMPLING',
-    )
-
     min_light_bounces: IntProperty(
         name="Min Light Bounces",
         description="Minimum number of light bounces. Setting this higher reduces noise in the first bounces, "
@@ -802,7 +778,7 @@ class CyclesRenderSettings(bpy.types.PropertyGroup):
         name="Tile Size",
         default=2048,
         description="",
-        min=8, max=8192,
+        min=8, max=16384,
     )
 
     # Various fine-tuning debug flags
@@ -894,32 +870,6 @@ class CyclesCameraSettings(bpy.types.PropertyGroup):
         default=pi,
     )
 
-    fisheye_polynomial_k0: FloatProperty(
-        name="Fisheye Polynomial K0",
-        description="Coefficient K0 of the lens polinomial",
-        default=camera.default_fisheye_polynomial[0], precision=6, step=0.1, subtype='ANGLE',
-    )
-    fisheye_polynomial_k1: FloatProperty(
-        name="Fisheye Polynomial K1",
-        description="Coefficient K1 of the lens polinomial",
-        default=camera.default_fisheye_polynomial[1], precision=6, step=0.1, subtype='ANGLE',
-    )
-    fisheye_polynomial_k2: FloatProperty(
-        name="Fisheye Polynomial K2",
-        description="Coefficient K2 of the lens polinomial",
-        default=camera.default_fisheye_polynomial[2], precision=6, step=0.1, subtype='ANGLE',
-    )
-    fisheye_polynomial_k3: FloatProperty(
-        name="Fisheye Polynomial K3",
-        description="Coefficient K3 of the lens polinomial",
-        default=camera.default_fisheye_polynomial[3], precision=6, step=0.1, subtype='ANGLE',
-    )
-    fisheye_polynomial_k4: FloatProperty(
-        name="Fisheye Polynomial K4",
-        description="Coefficient K4 of the lens polinomial",
-        default=camera.default_fisheye_polynomial[4], precision=6, step=0.1, subtype='ANGLE',
-    )
-
     @classmethod
     def register(cls):
         bpy.types.Camera.cycles = PointerProperty(
@@ -1342,7 +1292,8 @@ class CyclesPreferences(bpy.types.AddonPreferences):
 
     def get_device_types(self, context):
         import _cycles
-        has_cuda, has_optix, has_hip, has_metal = _cycles.get_device_types()
+        has_cuda, has_optix, has_hip = _cycles.get_device_types()
+
         list = [('NONE', "None", "Don't use compute device", 0)]
         if has_cuda:
             list.append(('CUDA', "CUDA", "Use CUDA for GPU acceleration", 1))
@@ -1350,8 +1301,6 @@ class CyclesPreferences(bpy.types.AddonPreferences):
             list.append(('OPTIX', "OptiX", "Use OptiX for GPU acceleration", 3))
         if has_hip:
             list.append(('HIP', "HIP", "Use HIP for GPU acceleration", 4))
-        if has_metal:
-            list.append(('METAL', "Metal", "Use Metal for GPU acceleration", 5))
 
         return list
 
@@ -1377,7 +1326,7 @@ class CyclesPreferences(bpy.types.AddonPreferences):
 
     def update_device_entries(self, device_list):
         for device in device_list:
-            if not device[1] in {'CUDA', 'OPTIX', 'CPU', 'HIP', 'METAL'}:
+            if not device[1] in {'CUDA', 'OPTIX', 'CPU', 'HIP'}:
                 continue
             # Try to find existing Device entry
             entry = self.find_existing_device_entry(device)
@@ -1421,7 +1370,7 @@ class CyclesPreferences(bpy.types.AddonPreferences):
         import _cycles
         # Ensure `self.devices` is not re-allocated when the second call to
         # get_devices_for_type is made, freeing items from the first list.
-        for device_type in ('CUDA', 'OPTIX', 'HIP', 'METAL'):
+        for device_type in ('CUDA', 'OPTIX', 'HIP'):
             self.update_device_entries(_cycles.available_devices(device_type))
 
     # Deprecated: use refresh_devices instead.
@@ -1473,8 +1422,6 @@ class CyclesPreferences(bpy.types.AddonPreferences):
                 col.label(text="Requires discrete AMD GPU with RDNA architecture", icon='BLANK1')
                 if sys.platform[:3] == "win":
                     col.label(text="and AMD Radeon Pro 21.Q4 driver or newer", icon='BLANK1')
-            elif device_type == 'METAL':
-                col.label(text="Requires Apple Silicon and macOS 12.0 or newer", icon='BLANK1')
             return
 
         for device in devices:

intern/cycles/blender/addon/ui.py

@@ -97,11 +97,6 @@ def use_cpu(context):
     return (get_device_type(context) == 'NONE' or cscene.device == 'CPU')
 
 
-def use_metal(context):
-    cscene = context.scene.cycles
-
-    return (get_device_type(context) == 'METAL' and cscene.device == 'GPU')
-
 def use_cuda(context):
     cscene = context.scene.cycles
 
@@ -295,9 +290,6 @@ class CYCLES_RENDER_PT_sampling_advanced(CyclesButtonsPanel, Panel):
         col.active = not (cscene.use_adaptive_sampling and cscene.use_preview_adaptive_sampling)
         col.prop(cscene, "sampling_pattern", text="Pattern")
 
-        col = layout.column(align=True)
-        col.prop(cscene, "sample_offset")
-
         layout.separator()
 
         heading = layout.column(align=True, heading="Scrambling Distance")
@@ -1020,7 +1012,7 @@ class CYCLES_OBJECT_PT_motion_blur(CyclesButtonsPanel, Panel):
     def poll(cls, context):
         ob = context.object
         if CyclesButtonsPanel.poll(context) and ob:
-            if ob.type in {'MESH', 'CURVE', 'CURVE', 'SURFACE', 'FONT', 'META', 'CAMERA', 'HAIR', 'POINTCLOUD'}:
+            if ob.type in {'MESH', 'CURVE', 'CURVE', 'SURFACE', 'FONT', 'META', 'CAMERA'}:
                 return True
             if ob.instance_type == 'COLLECTION' and ob.instance_collection:
                 return True
@@ -1824,38 +1816,37 @@ class CYCLES_RENDER_PT_debug(CyclesDebugButtonsPanel, Panel):
 
     def draw(self, context):
         layout = self.layout
-        layout.use_property_split = True
-        layout.use_property_decorate = False  # No animation.
 
         scene = context.scene
         cscene = scene.cycles
 
-        col = layout.column(heading="CPU")
+        col = layout.column()
 
+        col.label(text="CPU Flags:")
         row = col.row(align=True)
         row.prop(cscene, "debug_use_cpu_sse2", toggle=True)
         row.prop(cscene, "debug_use_cpu_sse3", toggle=True)
         row.prop(cscene, "debug_use_cpu_sse41", toggle=True)
         row.prop(cscene, "debug_use_cpu_avx", toggle=True)
         row.prop(cscene, "debug_use_cpu_avx2", toggle=True)
-        col.prop(cscene, "debug_bvh_layout", text="BVH")
+        col.prop(cscene, "debug_bvh_layout")
 
         col.separator()
 
-        col = layout.column(heading="CUDA")
+        col = layout.column()
+        col.label(text="CUDA Flags:")
         col.prop(cscene, "debug_use_cuda_adaptive_compile")
-        col = layout.column(heading="OptiX")
-        col.prop(cscene, "debug_use_optix_debug", text="Module Debug")
 
         col.separator()
 
-        col.prop(cscene, "debug_bvh_type", text="Viewport BVH")
+        col = layout.column()
+        col.label(text="OptiX Flags:")
+        col.prop(cscene, "debug_use_optix_debug")
 
         col.separator()
 
-        import _cycles
+        col = layout.column()
-        if _cycles.with_debug:
+        col.prop(cscene, "debug_bvh_type")
-            col.prop(cscene, "direct_light_sampling_type")
 
 
 class CYCLES_RENDER_PT_simplify(CyclesButtonsPanel, Panel):

intern/cycles/blender/camera.cpp

@@ -69,12 +69,6 @@ struct BlenderCamera {
   float pole_merge_angle_from;
   float pole_merge_angle_to;
 
-  float fisheye_polynomial_k0;
-  float fisheye_polynomial_k1;
-  float fisheye_polynomial_k2;
-  float fisheye_polynomial_k3;
-  float fisheye_polynomial_k4;
-
   enum { AUTO, HORIZONTAL, VERTICAL } sensor_fit;
   float sensor_width;
   float sensor_height;
@@ -206,12 +200,6 @@ static void blender_camera_from_object(BlenderCamera *bcam,
     bcam->longitude_min = RNA_float_get(&ccamera, "longitude_min");
     bcam->longitude_max = RNA_float_get(&ccamera, "longitude_max");
 
-    bcam->fisheye_polynomial_k0 = RNA_float_get(&ccamera, "fisheye_polynomial_k0");
-    bcam->fisheye_polynomial_k1 = RNA_float_get(&ccamera, "fisheye_polynomial_k1");
-    bcam->fisheye_polynomial_k2 = RNA_float_get(&ccamera, "fisheye_polynomial_k2");
-    bcam->fisheye_polynomial_k3 = RNA_float_get(&ccamera, "fisheye_polynomial_k3");
-    bcam->fisheye_polynomial_k4 = RNA_float_get(&ccamera, "fisheye_polynomial_k4");
-
     bcam->interocular_distance = b_camera.stereo().interocular_distance();
     if (b_camera.stereo().convergence_mode() == BL::CameraStereoData::convergence_mode_PARALLEL) {
       bcam->convergence_distance = FLT_MAX;
@@ -434,8 +422,7 @@ static void blender_camera_sync(Camera *cam,
   cam->set_full_height(height);
 
   /* panorama sensor */
-  if (bcam->type == CAMERA_PANORAMA && (bcam->panorama_type == PANORAMA_FISHEYE_EQUISOLID ||
+  if (bcam->type == CAMERA_PANORAMA && bcam->panorama_type == PANORAMA_FISHEYE_EQUISOLID) {
-                                        bcam->panorama_type == PANORAMA_FISHEYE_LENS_POLYNOMIAL)) {
     float fit_xratio = (float)bcam->render_width * bcam->pixelaspect.x;
     float fit_yratio = (float)bcam->render_height * bcam->pixelaspect.y;
     bool horizontal_fit;
@@ -478,12 +465,6 @@ static void blender_camera_sync(Camera *cam,
   cam->set_latitude_min(bcam->latitude_min);
   cam->set_latitude_max(bcam->latitude_max);
 
-  cam->set_fisheye_polynomial_k0(bcam->fisheye_polynomial_k0);
-  cam->set_fisheye_polynomial_k1(bcam->fisheye_polynomial_k1);
-  cam->set_fisheye_polynomial_k2(bcam->fisheye_polynomial_k2);
-  cam->set_fisheye_polynomial_k3(bcam->fisheye_polynomial_k3);
-  cam->set_fisheye_polynomial_k4(bcam->fisheye_polynomial_k4);
-
   cam->set_longitude_min(bcam->longitude_min);
   cam->set_longitude_max(bcam->longitude_max);
 

intern/cycles/blender/curves.cpp

@@ -199,7 +199,7 @@ static bool ObtainCacheParticleUV(Hair *hair,
           b_mesh->uv_layers.begin(l);
 
           float2 uv = zero_float2();
-          if (!b_mesh->uv_layers.empty())
+          if (b_mesh->uv_layers.length())
             b_psys.uv_on_emitter(psmd, *b_pa, pa_no, uv_num, &uv.x);
           CData->curve_uv.push_back_slow(uv);
 
@@ -261,7 +261,7 @@ static bool ObtainCacheParticleVcol(Hair *hair,
           b_mesh->vertex_colors.begin(l);
 
           float4 vcol = make_float4(0.0f, 0.0f, 0.0f, 1.0f);
-          if (!b_mesh->vertex_colors.empty())
+          if (b_mesh->vertex_colors.length())
             b_psys.mcol_on_emitter(psmd, *b_pa, pa_no, vcol_num, &vcol.x);
           CData->curve_vcol.push_back_slow(vcol);
 

intern/cycles/blender/device.cpp

@@ -27,7 +27,6 @@ enum ComputeDevice {
   COMPUTE_DEVICE_CUDA = 1,
   COMPUTE_DEVICE_OPTIX = 3,
   COMPUTE_DEVICE_HIP = 4,
-  COMPUTE_DEVICE_METAL = 5,
 
   COMPUTE_DEVICE_NUM
 };
@@ -86,9 +85,6 @@ DeviceInfo blender_device_info(BL::Preferences &b_preferences, BL::Scene &b_scen
       else if (compute_device == COMPUTE_DEVICE_HIP) {
         mask |= DEVICE_MASK_HIP;
       }
-      else if (compute_device == COMPUTE_DEVICE_METAL) {
-        mask |= DEVICE_MASK_METAL;
-      }
       vector<DeviceInfo> devices = Device::available_devices(mask);
 
       /* Match device preferences and available devices. */

intern/cycles/blender/display_driver.cpp

@@ -272,300 +272,12 @@ uint BlenderDisplaySpaceShader::get_shader_program()
   return shader_program_;
 }
 
-/* --------------------------------------------------------------------
- * DrawTile.
- */
-
-/* Higher level representation of a texture from the graphics library. */
-class GLTexture {
- public:
-  /* Global counter for all allocated OpenGL textures used by instances of this class. */
-  static inline std::atomic<int> num_used = 0;
-
-  GLTexture() = default;
-
-  ~GLTexture()
-  {
-    assert(gl_id == 0);
-  }
-
-  GLTexture(const GLTexture &other) = delete;
-  GLTexture &operator=(GLTexture &other) = delete;
-
-  GLTexture(GLTexture &&other) noexcept
-      : gl_id(other.gl_id), width(other.width), height(other.height)
-  {
-    other.reset();
-  }
-
-  GLTexture &operator=(GLTexture &&other)
-  {
-    if (this == &other) {
-      return *this;
-    }
-
-    gl_id = other.gl_id;
-    width = other.width;
-    height = other.height;
-
-    other.reset();
-
-    return *this;
-  }
-
-  bool gl_resources_ensure()
-  {
-    if (gl_id) {
-      return true;
-    }
-
-    /* Create texture. */
-    glGenTextures(1, &gl_id);
-    if (!gl_id) {
-      LOG(ERROR) << "Error creating texture.";
-      return false;
-    }
-
-    /* Configure the texture. */
-    glActiveTexture(GL_TEXTURE0);
-    glBindTexture(GL_TEXTURE_2D, gl_id);
-
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
-
-    /* Clamp to edge so that precision issues when zoomed out (which forces linear interpolation)
-     * does not cause unwanted repetition. */
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
-
-    glBindTexture(GL_TEXTURE_2D, 0);
-
-    ++num_used;
-
-    return true;
-  }
-
-  void gl_resources_destroy()
-  {
-    if (!gl_id) {
-      return;
-    }
-
-    glDeleteTextures(1, &gl_id);
-
-    reset();
-
-    --num_used;
-  }
-
-  /* OpenGL resource IDs of the texture.
-   *
-   * NOTE: Allocated on the render engine's context. */
-  uint gl_id = 0;
-
-  /* Dimensions of the texture in pixels. */
-  int width = 0;
-  int height = 0;
-
- protected:
-  void reset()
-  {
-    gl_id = 0;
-    width = 0;
-    height = 0;
-  }
-};
-
-/* Higher level representation of a Pixel Buffer Object (PBO) from the graphics library. */
-class GLPixelBufferObject {
- public:
-  /* Global counter for all allocated OpenGL PBOs used by instances of this class. */
-  static inline std::atomic<int> num_used = 0;
-
-  GLPixelBufferObject() = default;
-
-  ~GLPixelBufferObject()
-  {
-    assert(gl_id == 0);
-  }
-
-  GLPixelBufferObject(const GLPixelBufferObject &other) = delete;
-  GLPixelBufferObject &operator=(GLPixelBufferObject &other) = delete;
-
-  GLPixelBufferObject(GLPixelBufferObject &&other) noexcept
-      : gl_id(other.gl_id), width(other.width), height(other.height)
-  {
-    other.reset();
-  }
-
-  GLPixelBufferObject &operator=(GLPixelBufferObject &&other)
-  {
-    if (this == &other) {
-      return *this;
-    }
-
-    gl_id = other.gl_id;
-    width = other.width;
-    height = other.height;
-
-    other.reset();
-
-    return *this;
-  }
-
-  bool gl_resources_ensure()
-  {
-    if (gl_id) {
-      return true;
-    }
-
-    glGenBuffers(1, &gl_id);
-    if (!gl_id) {
-      LOG(ERROR) << "Error creating texture pixel buffer object.";
-      return false;
-    }
-
-    ++num_used;
-
-    return true;
-  }
-
-  void gl_resources_destroy()
-  {
-    if (!gl_id) {
-      return;
-    }
-
-    glDeleteBuffers(1, &gl_id);
-
-    reset();
-
-    --num_used;
-  }
-
-  /* OpenGL resource IDs of the PBO.
-   *
-   * NOTE: Allocated on the render engine's context. */
-  uint gl_id = 0;
-
-  /* Dimensions of the PBO. */
-  int width = 0;
-  int height = 0;
-
- protected:
-  void reset()
-  {
-    gl_id = 0;
-    width = 0;
-    height = 0;
-  }
-};
-
-class DrawTile {
- public:
-  DrawTile() = default;
-  ~DrawTile() = default;
-
-  DrawTile(const DrawTile &other) = delete;
-  DrawTile &operator=(const DrawTile &other) = delete;
-
-  DrawTile(DrawTile &&other) noexcept = default;
-
-  DrawTile &operator=(DrawTile &&other) = default;
-
-  bool gl_resources_ensure()
-  {
-    if (!texture.gl_resources_ensure()) {
-      gl_resources_destroy();
-      return false;
-    }
-
-    if (!gl_vertex_buffer) {
-      glGenBuffers(1, &gl_vertex_buffer);
-      if (!gl_vertex_buffer) {
-        LOG(ERROR) << "Error allocating tile VBO.";
-        gl_resources_destroy();
-        return false;
-      }
-    }
-
-    return true;
-  }
-
-  void gl_resources_destroy()
-  {
-    texture.gl_resources_destroy();
-
-    if (gl_vertex_buffer) {
-      glDeleteBuffers(1, &gl_vertex_buffer);
-      gl_vertex_buffer = 0;
-    }
-  }
-
-  inline bool ready_to_draw() const
-  {
-    return texture.gl_id != 0;
-  }
-
-  /* Texture which contains pixels of the tile. */
-  GLTexture texture;
-
-  /* Display parameters the texture of this tile has been updated for. */
-  BlenderDisplayDriver::Params params;
-
-  /* OpenGL resources needed for drawing. */
-  uint gl_vertex_buffer = 0;
-};
-
-class DrawTileAndPBO {
- public:
-  bool gl_resources_ensure()
-  {
-    if (!tile.gl_resources_ensure() || !buffer_object.gl_resources_ensure()) {
-      gl_resources_destroy();
-      return false;
-    }
-
-    return true;
-  }
-
-  void gl_resources_destroy()
-  {
-    tile.gl_resources_destroy();
-    buffer_object.gl_resources_destroy();
-  }
-
-  DrawTile tile;
-  GLPixelBufferObject buffer_object;
-};
-
 /* --------------------------------------------------------------------
  * BlenderDisplayDriver.
  */
 
-struct BlenderDisplayDriver::Tiles {
-  /* Resources of a tile which is being currently rendered. */
-  DrawTileAndPBO current_tile;
-
-  /* All tiles which rendering is finished and which content will not be changed. */
-  struct {
-    vector<DrawTile> tiles;
-
-    void gl_resources_destroy_and_clear()
-    {
-      for (DrawTile &tile : tiles) {
-        tile.gl_resources_destroy();
-      }
-
-      tiles.clear();
-    }
-  } finished_tiles;
-};
-
 BlenderDisplayDriver::BlenderDisplayDriver(BL::RenderEngine &b_engine, BL::Scene &b_scene)
-    : b_engine_(b_engine),
+    : b_engine_(b_engine), display_shader_(BlenderDisplayShader::create(b_engine, b_scene))
-      display_shader_(BlenderDisplayShader::create(b_engine, b_scene)),
-      tiles_(make_unique<Tiles>())
 {
   /* Create context while on the main thread. */
   gl_context_create();
@@ -580,21 +292,6 @@ BlenderDisplayDriver::~BlenderDisplayDriver()
  * Update procedure.
  */
 
-void BlenderDisplayDriver::next_tile_begin()
-{
-  if (!tiles_->current_tile.tile.ready_to_draw()) {
-    LOG(ERROR)
-        << "Unexpectedly moving to the next tile without any data provided for current tile.";
-    return;
-  }
-
-  /* Moving to the next tile without giving render data for the current tile is not an expected
-   * situation. */
-  DCHECK(!need_clear_);
-
-  tiles_->finished_tiles.tiles.emplace_back(std::move(tiles_->current_tile.tile));
-}
-
 bool BlenderDisplayDriver::update_begin(const Params &params,
                                         int texture_width,
                                         int texture_height)
@@ -615,96 +312,58 @@ bool BlenderDisplayDriver::update_begin(const Params &params,
     glWaitSync((GLsync)gl_render_sync_, 0, GL_TIMEOUT_IGNORED);
   }
 
-  DrawTile &current_tile = tiles_->current_tile.tile;
+  if (!gl_texture_resources_ensure()) {
-  GLPixelBufferObject &current_tile_buffer_object = tiles_->current_tile.buffer_object;
-
-  /* Clear storage of all finished tiles when display clear is requested.
-   * Do it when new tile data is provided to handle the display clear flag in a single place.
-   * It also makes the logic reliable from the whether drawing did happen or not point of view. */
-  if (need_clear_) {
-    tiles_->finished_tiles.gl_resources_destroy_and_clear();
-    need_clear_ = false;
-  }
-
-  if (!tiles_->current_tile.gl_resources_ensure()) {
-    tiles_->current_tile.gl_resources_destroy();
     gl_context_disable();
     return false;
   }
 
   /* Update texture dimensions if needed. */
-  if (current_tile.texture.width != texture_width ||
+  if (texture_.width != texture_width || texture_.height != texture_height) {
-      current_tile.texture.height != texture_height) {
     glActiveTexture(GL_TEXTURE0);
-    glBindTexture(GL_TEXTURE_2D, current_tile.texture.gl_id);
+    glBindTexture(GL_TEXTURE_2D, texture_.gl_id);
     glTexImage2D(
         GL_TEXTURE_2D, 0, GL_RGBA16F, texture_width, texture_height, 0, GL_RGBA, GL_HALF_FLOAT, 0);
-    current_tile.texture.width = texture_width;
+    texture_.width = texture_width;
-    current_tile.texture.height = texture_height;
+    texture_.height = texture_height;
     glBindTexture(GL_TEXTURE_2D, 0);
+
+    /* Texture did change, and no pixel storage was provided. Tag for an explicit zeroing out to
+     * avoid undefined content. */
+    texture_.need_clear = true;
   }
 
   /* Update PBO dimensions if needed.
    *
-   * NOTE: Allocate the PBO for the size which will fit the final render resolution (as in,
+   * NOTE: Allocate the PBO for the the size which will fit the final render resolution (as in,
    * at a resolution divider 1. This was we don't need to recreate graphics interoperability
    * objects which are costly and which are tied to the specific underlying buffer size.
    * The downside of this approach is that when graphics interoperability is not used we are
    * sending too much data to GPU when resolution divider is not 1. */
   /* TODO(sergey): Investigate whether keeping the PBO exact size of the texture makes non-interop
    * mode faster. */
-  const int buffer_width = params.size.x;
+  const int buffer_width = params.full_size.x;
-  const int buffer_height = params.size.y;
+  const int buffer_height = params.full_size.y;
-  if (current_tile_buffer_object.width != buffer_width ||
+  if (texture_.buffer_width != buffer_width || texture_.buffer_height != buffer_height) {
-      current_tile_buffer_object.height != buffer_height) {
     const size_t size_in_bytes = sizeof(half4) * buffer_width * buffer_height;
-    glBindBuffer(GL_PIXEL_UNPACK_BUFFER, current_tile_buffer_object.gl_id);
+    glBindBuffer(GL_PIXEL_UNPACK_BUFFER, texture_.gl_pbo_id);
     glBufferData(GL_PIXEL_UNPACK_BUFFER, size_in_bytes, 0, GL_DYNAMIC_DRAW);
     glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
 
-    current_tile_buffer_object.width = buffer_width;
+    texture_.buffer_width = buffer_width;
-    current_tile_buffer_object.height = buffer_height;
+    texture_.buffer_height = buffer_height;
   }
 
-  /* Store an updated parameters of the current tile.
+  /* New content will be provided to the texture in one way or another, so mark this in a
-   * In theory it is only needed once per update of the tile, but doing it on every update is
+   * centralized place. */
-   * the easiest and is not expensive. */
+  texture_.need_update = true;
-  tiles_->current_tile.tile.params = params;
+
+  texture_.params = params;
 
   return true;
 }
 
-static void update_tile_texture_pixels(const DrawTileAndPBO &tile)
-{
-  const GLTexture &texture = tile.tile.texture;
-
-  DCHECK_NE(tile.buffer_object.gl_id, 0);
-
-  glActiveTexture(GL_TEXTURE0);
-  glBindTexture(GL_TEXTURE_2D, texture.gl_id);
-  glBindBuffer(GL_PIXEL_UNPACK_BUFFER, tile.buffer_object.gl_id);
-
-  glTexSubImage2D(
-      GL_TEXTURE_2D, 0, 0, 0, texture.width, texture.height, GL_RGBA, GL_HALF_FLOAT, 0);
-
-  glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
-  glBindTexture(GL_TEXTURE_2D, 0);
-}
-
 void BlenderDisplayDriver::update_end()
 {
-  /* Unpack the PBO into the texture as soon as the new content is provided.
-   *
-   * This allows to ensure that the unpacking happens while resources like graphics interop (which
-   * lifetime is outside of control of the display driver) are still valid, as well as allows to
-   * move the tile from being current to finished immediately after this call.
-   *
-   * One concern with this approach is that if the update happens more often than drawing then
-   * doing the unpack here occupies GPU transfer for no good reason. However, the render scheduler
-   * takes care of ensuring updates don't happen that often. In regular applications redraw will
-   * happen much more often than this update. */
-  update_tile_texture_pixels(tiles_->current_tile);
-
   gl_upload_sync_ = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
   glFlush();
 
@@ -717,11 +376,7 @@ void BlenderDisplayDriver::update_end()
 
 half4 *BlenderDisplayDriver::map_texture_buffer()
 {
-  const uint pbo_gl_id = tiles_->current_tile.buffer_object.gl_id;
+  glBindBuffer(GL_PIXEL_UNPACK_BUFFER, texture_.gl_pbo_id);
-
-  DCHECK_NE(pbo_gl_id, 0);
-
-  glBindBuffer(GL_PIXEL_UNPACK_BUFFER, pbo_gl_id);
 
   half4 *mapped_rgba_pixels = reinterpret_cast<half4 *>(
       glMapBuffer(GL_PIXEL_UNPACK_BUFFER, GL_WRITE_ONLY));
@@ -729,6 +384,15 @@ half4 *BlenderDisplayDriver::map_texture_buffer()
     LOG(ERROR) << "Error mapping BlenderDisplayDriver pixel buffer object.";
   }
 
+  if (texture_.need_clear) {
+    const int64_t texture_width = texture_.width;
+    const int64_t texture_height = texture_.height;
+    memset(reinterpret_cast<void *>(mapped_rgba_pixels),
+           0,
+           texture_width * texture_height * sizeof(half4));
+    texture_.need_clear = false;
+  }
+
   return mapped_rgba_pixels;
 }
 
@@ -747,9 +411,12 @@ BlenderDisplayDriver::GraphicsInterop BlenderDisplayDriver::graphics_interop_get
 {
   GraphicsInterop interop_dst;
 
-  interop_dst.buffer_width = tiles_->current_tile.buffer_object.width;
+  interop_dst.buffer_width = texture_.buffer_width;
-  interop_dst.buffer_height = tiles_->current_tile.buffer_object.height;
+  interop_dst.buffer_height = texture_.buffer_height;
-  interop_dst.opengl_pbo_id = tiles_->current_tile.buffer_object.gl_id;
+  interop_dst.opengl_pbo_id = texture_.gl_pbo_id;
+
+  interop_dst.need_clear = texture_.need_clear;
+  texture_.need_clear = false;
 
   return interop_dst;
 }
@@ -770,7 +437,7 @@ void BlenderDisplayDriver::graphics_interop_deactivate()
 
 void BlenderDisplayDriver::clear()
 {
-  need_clear_ = true;
+  texture_.need_clear = true;
 }
 
 void BlenderDisplayDriver::set_zoom(float zoom_x, float zoom_y)
@@ -778,155 +445,26 @@ void BlenderDisplayDriver::set_zoom(float zoom_x, float zoom_y)
   zoom_ = make_float2(zoom_x, zoom_y);
 }
 
-/* Update vertex buffer with new coordinates of vertex positions and texture coordinates.
- * This buffer is used to render texture in the viewport.
- *
- * NOTE: The buffer needs to be bound. */
-static void vertex_buffer_update(const DisplayDriver::Params &params)
-{
-  const int x = params.full_offset.x;
-  const int y = params.full_offset.y;
-
-  const int width = params.size.x;
-  const int height = params.size.y;
-
-  /* Invalidate old contents - avoids stalling if the buffer is still waiting in queue to be
-   * rendered. */
-  glBufferData(GL_ARRAY_BUFFER, 16 * sizeof(float), NULL, GL_STREAM_DRAW);
-
-  float *vpointer = reinterpret_cast<float *>(glMapBuffer(GL_ARRAY_BUFFER, GL_WRITE_ONLY));
-  if (!vpointer) {
-    return;
-  }
-
-  vpointer[0] = 0.0f;
-  vpointer[1] = 0.0f;
-  vpointer[2] = x;
-  vpointer[3] = y;
-
-  vpointer[4] = 1.0f;
-  vpointer[5] = 0.0f;
-  vpointer[6] = x + width;
-  vpointer[7] = y;
-
-  vpointer[8] = 1.0f;
-  vpointer[9] = 1.0f;
-  vpointer[10] = x + width;
-  vpointer[11] = y + height;
-
-  vpointer[12] = 0.0f;
-  vpointer[13] = 1.0f;
-  vpointer[14] = x;
-  vpointer[15] = y + height;
-
-  glUnmapBuffer(GL_ARRAY_BUFFER);
-}
-
-static void draw_tile(const float2 &zoom,
-                      const int texcoord_attribute,
-                      const int position_attribute,
-                      const DrawTile &draw_tile)
-{
-  if (!draw_tile.ready_to_draw()) {
-    return;
-  }
-
-  const GLTexture &texture = draw_tile.texture;
-
-  DCHECK_NE(texture.gl_id, 0);
-  DCHECK_NE(draw_tile.gl_vertex_buffer, 0);
-
-  glBindBuffer(GL_ARRAY_BUFFER, draw_tile.gl_vertex_buffer);
-
-  /* Draw at the parameters for which the texture has been updated for. This allows to always draw
-   * texture during bordered-rendered camera view without flickering. The validness of the display
-   * parameters for a texture is guaranteed by the initial "clear" state which makes drawing to
-   * have an early output.
-   *
-   * Such approach can cause some extra "jelly" effect during panning, but it is not more jelly
-   * than overlay of selected objects. Also, it's possible to redraw texture at an intersection of
-   * the texture draw parameters and the latest updated draw parameters (although, complexity of
-   * doing it might not worth it. */
-  vertex_buffer_update(draw_tile.params);
-
-  glBindTexture(GL_TEXTURE_2D, texture.gl_id);
-
-  /* Trick to keep sharp rendering without jagged edges on all GPUs.
-   *
-   * The idea here is to enforce driver to use linear interpolation when the image is not zoomed
-   * in.
-   * For the render result with a resolution divider in effect we always use nearest interpolation.
-   *
-   * Use explicit MIN assignment to make sure the driver does not have an undefined behavior at
-   * the zoom level 1. The MAG filter is always NEAREST. */
-  const float zoomed_width = draw_tile.params.size.x * zoom.x;
-  const float zoomed_height = draw_tile.params.size.y * zoom.y;
-  if (texture.width != draw_tile.params.size.x || texture.height != draw_tile.params.size.y) {
-    /* Resolution divider is different from 1, force nearest interpolation. */
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
-  }
-  else if (zoomed_width - draw_tile.params.size.x > 0.5f ||
-           zoomed_height - draw_tile.params.size.y > 0.5f) {
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
-  }
-  else {
-    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
-  }
-
-  glVertexAttribPointer(
-      texcoord_attribute, 2, GL_FLOAT, GL_FALSE, 4 * sizeof(float), (const GLvoid *)0);
-  glVertexAttribPointer(position_attribute,
-                        2,
-                        GL_FLOAT,
-                        GL_FALSE,
-                        4 * sizeof(float),
-                        (const GLvoid *)(sizeof(float) * 2));
-
-  glDrawArrays(GL_TRIANGLE_FAN, 0, 4);
-}
-
-void BlenderDisplayDriver::flush()
-{
-  /* This is called from the render thread that also calls update_begin/end, right before ending
-   * the render loop. We wait for any queued PBO and render commands to be done, before destroying
-   * the render thread and activating the context in the main thread to destroy resources.
-   *
-   * If we don't do this, the NVIDIA driver hangs for a few seconds for when ending 3D viewport
-   * rendering, for unknown reasons. This was found with NVIDIA driver version 470.73 and a Quadro
-   * RTX 6000 on Linux. */
-  if (!gl_context_enable()) {
-    return;
-  }
-
-  if (gl_upload_sync_) {
-    glWaitSync((GLsync)gl_upload_sync_, 0, GL_TIMEOUT_IGNORED);
-  }
-
-  if (gl_render_sync_) {
-    glWaitSync((GLsync)gl_render_sync_, 0, GL_TIMEOUT_IGNORED);
-  }
-
-  gl_context_disable();
-}
-
 void BlenderDisplayDriver::draw(const Params &params)
 {
   /* See do_update_begin() for why no locking is required here. */
   const bool transparent = true;  // TODO(sergey): Derive this from Film.
 
+  if (!gl_draw_resources_ensure()) {
+    return;
+  }
+
   if (use_gl_context_) {
     gl_context_mutex_.lock();
   }
 
-  if (need_clear_) {
+  if (texture_.need_clear) {
     /* Texture is requested to be cleared and was not yet cleared.
      *
      * Do early return which should be equivalent of drawing all-zero texture.
      * Watch out for the lock though so that the clear happening during update is properly
      * synchronized here. */
-    if (use_gl_context_) {
+    gl_context_mutex_.unlock();
-      gl_context_mutex_.unlock();
-    }
     return;
   }
 
@@ -939,37 +477,66 @@ void BlenderDisplayDriver::draw(const Params &params)
     glBlendFunc(GL_ONE, GL_ONE_MINUS_SRC_ALPHA);
   }
 
-  glActiveTexture(GL_TEXTURE0);
+  display_shader_->bind(params.full_size.x, params.full_size.y);
 
-  /* NOTE: The VAO is to be allocated on the drawing context as it is not shared across contexts.
+  glActiveTexture(GL_TEXTURE0);
-   * Simplest is to allocate it on every redraw so that it is possible to destroy it from a
+  glBindTexture(GL_TEXTURE_2D, texture_.gl_id);
-   * correct context. */
+
+  /* Trick to keep sharp rendering without jagged edges on all GPUs.
+   *
+   * The idea here is to enforce driver to use linear interpolation when the image is not zoomed
+   * in.
+   * For the render result with a resolution divider in effect we always use nearest interpolation.
+   *
+   * Use explicit MIN assignment to make sure the driver does not have an undefined behavior at
+   * the zoom level 1. The MAG filter is always NEAREST. */
+  const float zoomed_width = params.size.x * zoom_.x;
+  const float zoomed_height = params.size.y * zoom_.y;
+  if (texture_.width != params.size.x || texture_.height != params.size.y) {
+    /* Resolution divider is different from 1, force nearest interpolation. */
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+  }
+  else if (zoomed_width - params.size.x > 0.5f || zoomed_height - params.size.y > 0.5f) {
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+  }
+  else {
+    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
+  }
+
+  glBindBuffer(GL_ARRAY_BUFFER, vertex_buffer_);
+
+  texture_update_if_needed();
+  vertex_buffer_update(params);
+
+  /* TODO(sergey): Does it make sense/possible to cache/reuse the VAO? */
   GLuint vertex_array_object;
   glGenVertexArrays(1, &vertex_array_object);
   glBindVertexArray(vertex_array_object);
 
-  display_shader_->bind(params.full_size.x, params.full_size.y);
-
   const int texcoord_attribute = display_shader_->get_tex_coord_attrib_location();
   const int position_attribute = display_shader_->get_position_attrib_location();
 
   glEnableVertexAttribArray(texcoord_attribute);
   glEnableVertexAttribArray(position_attribute);
 
-  draw_tile(zoom_, texcoord_attribute, position_attribute, tiles_->current_tile.tile);
+  glVertexAttribPointer(
+      texcoord_attribute, 2, GL_FLOAT, GL_FALSE, 4 * sizeof(float), (const GLvoid *)0);
+  glVertexAttribPointer(position_attribute,
+                        2,
+                        GL_FLOAT,
+                        GL_FALSE,
+                        4 * sizeof(float),
+                        (const GLvoid *)(sizeof(float) * 2));
 
-  for (const DrawTile &tile : tiles_->finished_tiles.tiles) {
+  glDrawArrays(GL_TRIANGLE_FAN, 0, 4);
-    draw_tile(zoom_, texcoord_attribute, position_attribute, tile);
-  }
 
-  display_shader_->unbind();
-
-  glBindTexture(GL_TEXTURE_2D, 0);
-  glBindVertexArray(0);
   glBindBuffer(GL_ARRAY_BUFFER, 0);
+  glBindTexture(GL_TEXTURE_2D, 0);
 
   glDeleteVertexArrays(1, &vertex_array_object);
 
+  display_shader_->unbind();
+
   if (transparent) {
     glDisable(GL_BLEND);
   }
@@ -977,11 +544,6 @@ void BlenderDisplayDriver::draw(const Params &params)
   gl_render_sync_ = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
   glFlush();
 
-  if (VLOG_IS_ON(5)) {
-    VLOG(5) << "Number of textures: " << GLTexture::num_used;
-    VLOG(5) << "Number of PBOs: " << GLPixelBufferObject::num_used;
-  }
-
   if (use_gl_context_) {
     gl_context_mutex_.unlock();
   }
@@ -1056,16 +618,154 @@ void BlenderDisplayDriver::gl_context_dispose()
   }
 }
 
+bool BlenderDisplayDriver::gl_draw_resources_ensure()
+{
+  if (!texture_.gl_id) {
+    /* If there is no texture allocated, there is nothing to draw. Inform the draw call that it can
+     * can not continue. Note that this is not an unrecoverable error, so once the texture is known
+     * we will come back here and create all the GPU resources needed for draw. */
+    return false;
+  }
+
+  if (gl_draw_resource_creation_attempted_) {
+    return gl_draw_resources_created_;
+  }
+  gl_draw_resource_creation_attempted_ = true;
+
+  if (!vertex_buffer_) {
+    glGenBuffers(1, &vertex_buffer_);
+    if (!vertex_buffer_) {
+      LOG(ERROR) << "Error creating vertex buffer.";
+      return false;
+    }
+  }
+
+  gl_draw_resources_created_ = true;
+
+  return true;
+}
+
 void BlenderDisplayDriver::gl_resources_destroy()
 {
   gl_context_enable();
 
-  tiles_->current_tile.gl_resources_destroy();
+  if (vertex_buffer_ != 0) {
-  tiles_->finished_tiles.gl_resources_destroy_and_clear();
+    glDeleteBuffers(1, &vertex_buffer_);
+  }
+
+  if (texture_.gl_pbo_id) {
+    glDeleteBuffers(1, &texture_.gl_pbo_id);
+    texture_.gl_pbo_id = 0;
+  }
+
+  if (texture_.gl_id) {
+    glDeleteTextures(1, &texture_.gl_id);
+    texture_.gl_id = 0;
+  }
 
   gl_context_disable();
 
   gl_context_dispose();
 }
 
+bool BlenderDisplayDriver::gl_texture_resources_ensure()
+{
+  if (texture_.creation_attempted) {
+    return texture_.is_created;
+  }
+  texture_.creation_attempted = true;
+
+  DCHECK(!texture_.gl_id);
+  DCHECK(!texture_.gl_pbo_id);
+
+  /* Create texture. */
+  glGenTextures(1, &texture_.gl_id);
+  if (!texture_.gl_id) {
+    LOG(ERROR) << "Error creating texture.";
+    return false;
+  }
+
+  /* Configure the texture. */
+  glActiveTexture(GL_TEXTURE0);
+  glBindTexture(GL_TEXTURE_2D, texture_.gl_id);
+  glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+  glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+  glBindTexture(GL_TEXTURE_2D, 0);
+
+  /* Create PBO for the texture. */
+  glGenBuffers(1, &texture_.gl_pbo_id);
+  if (!texture_.gl_pbo_id) {
+    LOG(ERROR) << "Error creating texture pixel buffer object.";
+    return false;
+  }
+
+  /* Creation finished with a success. */
+  texture_.is_created = true;
+
+  return true;
+}
+
+void BlenderDisplayDriver::texture_update_if_needed()
+{
+  if (!texture_.need_update) {
+    return;
+  }
+
+  glBindBuffer(GL_PIXEL_UNPACK_BUFFER, texture_.gl_pbo_id);
+  glTexSubImage2D(
+      GL_TEXTURE_2D, 0, 0, 0, texture_.width, texture_.height, GL_RGBA, GL_HALF_FLOAT, 0);
+  glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
+
+  texture_.need_update = false;
+}
+
+void BlenderDisplayDriver::vertex_buffer_update(const Params & /*params*/)
+{
+  /* Draw at the parameters for which the texture has been updated for. This allows to always draw
+   * texture during bordered-rendered camera view without flickering. The validness of the display
+   * parameters for a texture is guaranteed by the initial "clear" state which makes drawing to
+   * have an early output.
+   *
+   * Such approach can cause some extra "jelly" effect during panning, but it is not more jelly
+   * than overlay of selected objects. Also, it's possible to redraw texture at an intersection of
+   * the texture draw parameters and the latest updated draw parameters (although, complexity of
+   * doing it might not worth it. */
+  const int x = texture_.params.full_offset.x;
+  const int y = texture_.params.full_offset.y;
+
+  const int width = texture_.params.size.x;
+  const int height = texture_.params.size.y;
+
+  /* Invalidate old contents - avoids stalling if the buffer is still waiting in queue to be
+   * rendered. */
+  glBufferData(GL_ARRAY_BUFFER, 16 * sizeof(float), NULL, GL_STREAM_DRAW);
+
+  float *vpointer = reinterpret_cast<float *>(glMapBuffer(GL_ARRAY_BUFFER, GL_WRITE_ONLY));
+  if (!vpointer) {
+    return;
+  }
+
+  vpointer[0] = 0.0f;
+  vpointer[1] = 0.0f;
+  vpointer[2] = x;
+  vpointer[3] = y;
+
+  vpointer[4] = 1.0f;
+  vpointer[5] = 0.0f;
+  vpointer[6] = x + width;
+  vpointer[7] = y;
+
+  vpointer[8] = 1.0f;
+  vpointer[9] = 1.0f;
+  vpointer[10] = x + width;
+  vpointer[11] = y + height;
+
+  vpointer[12] = 0.0f;
+  vpointer[13] = 1.0f;
+  vpointer[14] = x;
+  vpointer[15] = y + height;
+
+  glUnmapBuffer(GL_ARRAY_BUFFER);
+}
+
 CCL_NAMESPACE_END

intern/cycles/blender/display_driver.h

@@ -26,7 +26,6 @@
 
 #include "util/thread.h"
 #include "util/unique_ptr.h"
-#include "util/vector.h"
 
 CCL_NAMESPACE_BEGIN
 
@@ -113,8 +112,6 @@ class BlenderDisplayDriver : public DisplayDriver {
   void set_zoom(float zoom_x, float zoom_y);
 
  protected:
-  virtual void next_tile_begin() override;
-
   virtual bool update_begin(const Params &params, int texture_width, int texture_height) override;
   virtual void update_end() override;
 
@@ -125,17 +122,33 @@ class BlenderDisplayDriver : public DisplayDriver {
 
   virtual void draw(const Params &params) override;
 
-  virtual void flush() override;
-
   /* Helper function which allocates new GPU context. */
   void gl_context_create();
   bool gl_context_enable();
   void gl_context_disable();
   void gl_context_dispose();
 
+  /* Make sure texture is allocated and its initial configuration is performed. */
+  bool gl_texture_resources_ensure();
+
+  /* Ensure all runtime GPU resources needed for drawing are allocated.
+   * Returns true if all resources needed for drawing are available. */
+  bool gl_draw_resources_ensure();
+
   /* Destroy all GPU resources which are being used by this object. */
   void gl_resources_destroy();
 
+  /* Update GPU texture dimensions and content if needed (new pixel data was provided).
+   *
+   * NOTE: The texture needs to be bound. */
+  void texture_update_if_needed();
+
+  /* Update vertex buffer with new coordinates of vertex positions and texture coordinates.
+   * This buffer is used to render texture in the viewport.
+   *
+   * NOTE: The buffer needs to be bound. */
+  void vertex_buffer_update(const Params &params);
+
   BL::RenderEngine b_engine_;
 
   /* OpenGL context which is used the render engine doesn't have its own. */
@@ -146,14 +159,50 @@ class BlenderDisplayDriver : public DisplayDriver {
   /* Mutex used to guard the `gl_context_`. */
   thread_mutex gl_context_mutex_;
 
-  /* Content of the display is to be filled with zeroes. */
+  /* Texture which contains pixels of the render result. */
-  std::atomic<bool> need_clear_ = true;
+  struct {
+    /* Indicates whether texture creation was attempted and succeeded.
+     * Used to avoid multiple attempts of texture creation on GPU issues or GPU context
+     * misconfiguration. */
+    bool creation_attempted = false;
+    bool is_created = false;
+
+    /* OpenGL resource IDs of the texture itself and Pixel Buffer Object (PBO) used to write
+     * pixels to it.
+     *
+     * NOTE: Allocated on the engine's context. */
+    uint gl_id = 0;
+    uint gl_pbo_id = 0;
+
+    /* Is true when new data was written to the PBO, meaning, the texture might need to be resized
+     * and new data is to be uploaded to the GPU. */
+    bool need_update = false;
+
+    /* Content of the texture is to be filled with zeroes. */
+    std::atomic<bool> need_clear = true;
+
+    /* Dimensions of the texture in pixels. */
+    int width = 0;
+    int height = 0;
+
+    /* Dimensions of the underlying PBO. */
+    int buffer_width = 0;
+    int buffer_height = 0;
+
+    /* Display parameters the texture has been updated for. */
+    Params params;
+  } texture_;
 
   unique_ptr<BlenderDisplayShader> display_shader_;
 
-  /* Opaque storage for an internal state and data for tiles. */
+  /* Special track of whether GPU resources were attempted to be created, to avoid attempts of
-  struct Tiles;
+   * their re-creation on failure on every redraw. */
-  unique_ptr<Tiles> tiles_;
+  bool gl_draw_resource_creation_attempted_ = false;
+  bool gl_draw_resources_created_ = false;
+
+  /* Vertex buffer which hold vertices of a triangle fan which is textures with the texture
+   * holding the render result. */
+  uint vertex_buffer_ = 0;
 
   void *gl_render_sync_ = nullptr;
   void *gl_upload_sync_ = nullptr;

intern/cycles/blender/geometry.cpp

@@ -19,7 +19,6 @@
 #include "scene/hair.h"
 #include "scene/mesh.h"
 #include "scene/object.h"
-#include "scene/pointcloud.h"
 #include "scene/volume.h"
 
 #include "blender/sync.h"
@@ -40,10 +39,6 @@ static Geometry::Type determine_geom_type(BObjectInfo &b_ob_info, bool use_parti
     return Geometry::HAIR;
   }
 
-  if (b_ob_info.object_data.is_a(&RNA_PointCloud)) {
-    return Geometry::POINTCLOUD;
-  }
-
   if (b_ob_info.object_data.is_a(&RNA_Volume) ||
       (b_ob_info.object_data == b_ob_info.real_object.data() &&
        object_fluid_gas_domain_find(b_ob_info.real_object))) {
@@ -116,9 +111,6 @@ Geometry *BlenderSync::sync_geometry(BL::Depsgraph &b_depsgraph,
     else if (geom_type == Geometry::VOLUME) {
       geom = scene->create_node<Volume>();
     }
-    else if (geom_type == Geometry::POINTCLOUD) {
-      geom = scene->create_node<PointCloud>();
-    }
     else {
       geom = scene->create_node<Mesh>();
     }
@@ -178,10 +170,6 @@ Geometry *BlenderSync::sync_geometry(BL::Depsgraph &b_depsgraph,
       Volume *volume = static_cast<Volume *>(geom);
       sync_volume(b_ob_info, volume);
     }
-    else if (geom_type == Geometry::POINTCLOUD) {
-      PointCloud *pointcloud = static_cast<PointCloud *>(geom);
-      sync_pointcloud(pointcloud, b_ob_info);
-    }
     else {
       Mesh *mesh = static_cast<Mesh *>(geom);
       sync_mesh(b_depsgraph, b_ob_info, mesh);
@@ -243,10 +231,6 @@ void BlenderSync::sync_geometry_motion(BL::Depsgraph &b_depsgraph,
              object_fluid_gas_domain_find(b_ob_info.real_object)) {
       /* No volume motion blur support yet. */
     }
-    else if (b_ob_info.object_data.is_a(&RNA_PointCloud)) {
-      PointCloud *pointcloud = static_cast<PointCloud *>(geom);
-      sync_pointcloud_motion(pointcloud, b_ob_info, motion_step);
-    }
     else {
       Mesh *mesh = static_cast<Mesh *>(geom);
       sync_mesh_motion(b_depsgraph, b_ob_info, mesh, motion_step);

intern/cycles/blender/mesh.cpp

@@ -555,7 +555,7 @@ static void attr_create_vertex_color(Scene *scene, Mesh *mesh, BL::Mesh &b_mesh,
 /* Create uv map attributes. */
 static void attr_create_uv_map(Scene *scene, Mesh *mesh, BL::Mesh &b_mesh)
 {
-  if (!b_mesh.uv_layers.empty()) {
+  if (b_mesh.uv_layers.length() != 0) {
     for (BL::MeshUVLoopLayer &l : b_mesh.uv_layers) {
       const bool active_render = l.active_render();
       AttributeStandard uv_std = (active_render) ? ATTR_STD_UV : ATTR_STD_NONE;
@@ -619,7 +619,7 @@ static void attr_create_uv_map(Scene *scene, Mesh *mesh, BL::Mesh &b_mesh)
 
 static void attr_create_subd_uv_map(Scene *scene, Mesh *mesh, BL::Mesh &b_mesh, bool subdivide_uvs)
 {
-  if (!b_mesh.uv_layers.empty()) {
+  if (b_mesh.uv_layers.length() != 0) {
     BL::Mesh::uv_layers_iterator l;
     int i = 0;
 
@@ -951,7 +951,7 @@ static void create_mesh(Scene *scene,
   N = attr_N->data_float3();
 
   /* create generated coordinates from undeformed coordinates */
-  const bool need_default_tangent = (subdivision == false) && (b_mesh.uv_layers.empty()) &&
+  const bool need_default_tangent = (subdivision == false) && (b_mesh.uv_layers.length() == 0) &&
                                     (mesh->need_attribute(scene, ATTR_STD_UV_TANGENT));
   if (mesh->need_attribute(scene, ATTR_STD_GENERATED) || need_default_tangent) {
     Attribute *attr = attributes.add(ATTR_STD_GENERATED);
@@ -1086,6 +1086,40 @@ static void create_subd_mesh(Scene *scene,
 
 /* Sync */
 
+/* Check whether some of "built-in" motion-related attributes are needed to be exported (includes
+ * things like velocity from cache modifier, fluid simulation).
+ *
+ * NOTE: This code is run prior to object motion blur initialization. so can not access properties
+ * set by `sync_object_motion_init()`. */
+static bool mesh_need_motion_attribute(BObjectInfo &b_ob_info, Scene *scene)
+{
+  const Scene::MotionType need_motion = scene->need_motion();
+  if (need_motion == Scene::MOTION_NONE) {
+    /* Simple case: neither motion pass nor motion blur is needed, no need in the motion related
+     * attributes. */
+    return false;
+  }
+
+  if (need_motion == Scene::MOTION_BLUR) {
+    /* A bit tricky and implicit case:
+     * - Motion blur is enabled in the scene, which implies specific number of time steps for
+     *   objects.
+     * - If the object has motion blur disabled on it, it will have 0 time steps.
+     * - Motion attribute expects non-zero time steps.
+     *
+     * Avoid adding motion attributes if the motion blur will enforce 0 motion steps. */
+    PointerRNA cobject = RNA_pointer_get(&b_ob_info.real_object.ptr, "cycles");
+    const bool use_motion = get_boolean(cobject, "use_motion_blur");
+    if (!use_motion) {
+      return false;
+    }
+  }
+
+  /* Motion pass which implies 3 motion steps, or motion blur which is not disabled on object
+   * level. */
+  return true;
+}
+
 void BlenderSync::sync_mesh(BL::Depsgraph b_depsgraph, BObjectInfo &b_ob_info, Mesh *mesh)
 {
   /* make a copy of the shaders as the caller in the main thread still need them for syncing the
@@ -1110,7 +1144,7 @@ void BlenderSync::sync_mesh(BL::Depsgraph b_depsgraph, BObjectInfo &b_ob_info, M
 
     if (b_mesh) {
       /* Motion blur attribute is relative to seconds, we need it relative to frames. */
-      const bool need_motion = object_need_motion_attribute(b_ob_info, scene);
+      const bool need_motion = mesh_need_motion_attribute(b_ob_info, scene);
       const float motion_scale = (need_motion) ?
                                      scene->motion_shutter_time() /
                                          (b_scene.render().fps() / b_scene.render().fps_base()) :

intern/cycles/blender/object.cpp

@@ -72,8 +72,7 @@ bool BlenderSync::object_is_geometry(BObjectInfo &b_ob_info)
 
   BL::Object::type_enum type = b_ob_info.iter_object.type();
 
-  if (type == BL::Object::type_VOLUME || type == BL::Object::type_HAIR ||
+  if (type == BL::Object::type_VOLUME || type == BL::Object::type_HAIR) {
-      type == BL::Object::type_POINTCLOUD) {
     /* Will be exported attached to mesh. */
     return true;
   }
@@ -207,7 +206,7 @@ Object *BlenderSync::sync_object(BL::Depsgraph &b_depsgraph,
     return NULL;
   }
 
-  /* only interested in object that we can create geometry from */
+  /* only interested in object that we can create meshes from */
   if (!object_is_geometry(b_ob_info)) {
     return NULL;
   }

intern/cycles/blender/output_driver.cpp

@@ -66,7 +66,7 @@ bool BlenderOutputDriver::read_render_tile(const Tile &tile)
 
 bool BlenderOutputDriver::update_render_tile(const Tile &tile)
 {
-  /* Use final write for preview renders, otherwise render result wouldn't be updated
+  /* Use final write for preview renders, otherwise render result wouldn't be be updated
    * quickly on Blender side. For all other cases we use the display driver. */
   if (b_engine_.is_preview()) {
     write_render_tile(tile);

intern/cycles/blender/pointcloud.cpp

@@ -1,303 +0,0 @@
-/*
- * Copyright 2011-2013 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.
- */
-
-#include "scene/pointcloud.h"
-#include "scene/attribute.h"
-#include "scene/scene.h"
-
-#include "blender/sync.h"
-#include "blender/util.h"
-
-#include "util/foreach.h"
-#include "util/hash.h"
-
-CCL_NAMESPACE_BEGIN
-
-template<typename TypeInCycles, typename GetValueAtIndex>
-static void fill_generic_attribute(BL::PointCloud &b_pointcloud,
-                                   TypeInCycles *data,
-                                   const GetValueAtIndex &get_value_at_index)
-{
-  const int num_points = b_pointcloud.points.length();
-  for (int i = 0; i < num_points; i++) {
-    data[i] = get_value_at_index(i);
-  }
-}
-
-static void attr_create_motion(PointCloud *pointcloud,
-                               BL::Attribute &b_attribute,
-                               const float motion_scale)
-{
-  if (!(b_attribute.domain() == BL::Attribute::domain_POINT) &&
-      (b_attribute.data_type() == BL::Attribute::data_type_FLOAT_VECTOR)) {
-    return;
-  }
-
-  BL::FloatVectorAttribute b_vector_attribute(b_attribute);
-  const int num_points = pointcloud->get_points().size();
-
-  /* Find or add attribute */
-  float3 *P = &pointcloud->get_points()[0];
-  Attribute *attr_mP = pointcloud->attributes.find(ATTR_STD_MOTION_VERTEX_POSITION);
-
-  if (!attr_mP) {
-    attr_mP = pointcloud->attributes.add(ATTR_STD_MOTION_VERTEX_POSITION);
-  }
-
-  /* Only export previous and next frame, we don't have any in between data. */
-  float motion_times[2] = {-1.0f, 1.0f};
-  for (int step = 0; step < 2; step++) {
-    const float relative_time = motion_times[step] * 0.5f * motion_scale;
-    float3 *mP = attr_mP->data_float3() + step * num_points;
-
-    for (int i = 0; i < num_points; i++) {
-      mP[i] = P[i] + get_float3(b_vector_attribute.data[i].vector()) * relative_time;
-    }
-  }
-}
-
-static void copy_attributes(PointCloud *pointcloud,
-                            BL::PointCloud b_pointcloud,
-                            const bool need_motion,
-                            const float motion_scale)
-{
-  AttributeSet &attributes = pointcloud->attributes;
-  static const ustring u_velocity("velocity");
-  for (BL::Attribute &b_attribute : b_pointcloud.attributes) {
-    const ustring name{b_attribute.name().c_str()};
-
-    if (need_motion && name == u_velocity) {
-      attr_create_motion(pointcloud, b_attribute, motion_scale);
-    }
-
-    if (attributes.find(name)) {
-      continue;
-    }
-
-    const AttributeElement element = ATTR_ELEMENT_VERTEX;
-    const BL::Attribute::data_type_enum b_data_type = b_attribute.data_type();
-    switch (b_data_type) {
-      case BL::Attribute::data_type_FLOAT: {
-        BL::FloatAttribute b_float_attribute{b_attribute};
-        Attribute *attr = attributes.add(name, TypeFloat, element);
-        float *data = attr->data_float();
-        fill_generic_attribute(
-            b_pointcloud, data, [&](int i) { return b_float_attribute.data[i].value(); });
-        break;
-      }
-      case BL::Attribute::data_type_BOOLEAN: {
-        BL::BoolAttribute b_bool_attribute{b_attribute};
-        Attribute *attr = attributes.add(name, TypeFloat, element);
-        float *data = attr->data_float();
-        fill_generic_attribute(
-            b_pointcloud, data, [&](int i) { return (float)b_bool_attribute.data[i].value(); });
-        break;
-      }
-      case BL::Attribute::data_type_INT: {
-        BL::IntAttribute b_int_attribute{b_attribute};
-        Attribute *attr = attributes.add(name, TypeFloat, element);
-        float *data = attr->data_float();
-        fill_generic_attribute(
-            b_pointcloud, data, [&](int i) { return (float)b_int_attribute.data[i].value(); });
-        break;
-      }
-      case BL::Attribute::data_type_FLOAT_VECTOR: {
-        BL::FloatVectorAttribute b_vector_attribute{b_attribute};
-        Attribute *attr = attributes.add(name, TypeVector, element);
-        float3 *data = attr->data_float3();
-        fill_generic_attribute(b_pointcloud, data, [&](int i) {
-          BL::Array<float, 3> v = b_vector_attribute.data[i].vector();
-          return make_float3(v[0], v[1], v[2]);
-        });
-        break;
-      }
-      case BL::Attribute::data_type_FLOAT_COLOR: {
-        BL::FloatColorAttribute b_color_attribute{b_attribute};
-        Attribute *attr = attributes.add(name, TypeRGBA, element);
-        float4 *data = attr->data_float4();
-        fill_generic_attribute(b_pointcloud, data, [&](int i) {
-          BL::Array<float, 4> v = b_color_attribute.data[i].color();
-          return make_float4(v[0], v[1], v[2], v[3]);
-        });
-        break;
-      }
-      case BL::Attribute::data_type_FLOAT2: {
-        BL::Float2Attribute b_float2_attribute{b_attribute};
-        Attribute *attr = attributes.add(name, TypeFloat2, element);
-        float2 *data = attr->data_float2();
-        fill_generic_attribute(b_pointcloud, data, [&](int i) {
-          BL::Array<float, 2> v = b_float2_attribute.data[i].vector();
-          return make_float2(v[0], v[1]);
-        });
-        break;
-      }
-      default:
-        /* Not supported. */
-        break;
-    }
-  }
-}
-
-static void export_pointcloud(Scene *scene,
-                              PointCloud *pointcloud,
-                              BL::PointCloud b_pointcloud,
-                              const bool need_motion,
-                              const float motion_scale)
-{
-  /* TODO: optimize so we can straight memcpy arrays from Blender? */
-
-  /* Add requested attributes. */
-  Attribute *attr_random = NULL;
-  if (pointcloud->need_attribute(scene, ATTR_STD_POINT_RANDOM)) {
-    attr_random = pointcloud->attributes.add(ATTR_STD_POINT_RANDOM);
-  }
-
-  /* Reserve memory. */
-  const int num_points = b_pointcloud.points.length();
-  pointcloud->reserve(num_points);
-
-  /* Export points. */
-  BL::PointCloud::points_iterator b_point_iter;
-  for (b_pointcloud.points.begin(b_point_iter); b_point_iter != b_pointcloud.points.end();
-       ++b_point_iter) {
-    BL::Point b_point = *b_point_iter;
-    const float3 co = get_float3(b_point.co());
-    const float radius = b_point.radius();
-    pointcloud->add_point(co, radius);
-
-    /* Random number per point. */
-    if (attr_random != NULL) {
-      attr_random->add(hash_uint2_to_float(b_point.index(), 0));
-    }
-  }
-
-  /* Export attributes */
-  copy_attributes(pointcloud, b_pointcloud, need_motion, motion_scale);
-}
-
-static void export_pointcloud_motion(PointCloud *pointcloud,
-                                     BL::PointCloud b_pointcloud,
-                                     int motion_step)
-{
-  /* Find or add attribute. */
-  Attribute *attr_mP = pointcloud->attributes.find(ATTR_STD_MOTION_VERTEX_POSITION);
-  bool new_attribute = false;
-
-  if (!attr_mP) {
-    attr_mP = pointcloud->attributes.add(ATTR_STD_MOTION_VERTEX_POSITION);
-    new_attribute = true;
-  }
-
-  /* Export motion points. */
-  const int num_points = pointcloud->num_points();
-  float3 *mP = attr_mP->data_float3() + motion_step * num_points;
-  bool have_motion = false;
-  int num_motion_points = 0;
-  const array<float3> &pointcloud_points = pointcloud->get_points();
-
-  BL::PointCloud::points_iterator b_point_iter;
-  for (b_pointcloud.points.begin(b_point_iter); b_point_iter != b_pointcloud.points.end();
-       ++b_point_iter) {
-    BL::Point b_point = *b_point_iter;
-
-    if (num_motion_points < num_points) {
-      float3 P = get_float3(b_point.co());
-      P.w = b_point.radius();
-      mP[num_motion_points] = P;
-      have_motion = have_motion || (P != pointcloud_points[num_motion_points]);
-      num_motion_points++;
-    }
-  }
-
-  /* In case of new attribute, we verify if there really was any motion. */
-  if (new_attribute) {
-    if (num_motion_points != num_points || !have_motion) {
-      pointcloud->attributes.remove(ATTR_STD_MOTION_VERTEX_POSITION);
-    }
-    else if (motion_step > 0) {
-      /* Motion, fill up previous steps that we might have skipped because
-       * they had no motion, but we need them anyway now. */
-      for (int step = 0; step < motion_step; step++) {
-        pointcloud->copy_center_to_motion_step(step);
-      }
-    }
-  }
-
-  /* Export attributes */
-  copy_attributes(pointcloud, b_pointcloud, false, 0.0f);
-}
-
-void BlenderSync::sync_pointcloud(PointCloud *pointcloud, BObjectInfo &b_ob_info)
-{
-  size_t old_numpoints = pointcloud->num_points();
-
-  array<Node *> used_shaders = pointcloud->get_used_shaders();
-
-  PointCloud new_pointcloud;
-  new_pointcloud.set_used_shaders(used_shaders);
-
-  /* TODO: add option to filter out points in the view layer. */
-  BL::PointCloud b_pointcloud(b_ob_info.object_data);
-  /* Motion blur attribute is relative to seconds, we need it relative to frames. */
-  const bool need_motion = object_need_motion_attribute(b_ob_info, scene);
-  const float motion_scale = (need_motion) ?
-                                 scene->motion_shutter_time() /
-                                     (b_scene.render().fps() / b_scene.render().fps_base()) :
-                                 0.0f;
-  export_pointcloud(scene, &new_pointcloud, b_pointcloud, need_motion, motion_scale);
-
-  /* update original sockets */
-  for (const SocketType &socket : new_pointcloud.type->inputs) {
-    /* Those sockets are updated in sync_object, so do not modify them. */
-    if (socket.name == "use_motion_blur" || socket.name == "motion_steps" ||
-        socket.name == "used_shaders") {
-      continue;
-    }
-    pointcloud->set_value(socket, new_pointcloud, socket);
-  }
-
-  pointcloud->attributes.clear();
-  foreach (Attribute &attr, new_pointcloud.attributes.attributes) {
-    pointcloud->attributes.attributes.push_back(std::move(attr));
-  }
-
-  /* tag update */
-  const bool rebuild = (pointcloud && old_numpoints != pointcloud->num_points());
-  pointcloud->tag_update(scene, rebuild);
-}
-
-void BlenderSync::sync_pointcloud_motion(PointCloud *pointcloud,
-                                         BObjectInfo &b_ob_info,
-                                         int motion_step)
-{
-  /* Skip if nothing exported. */
-  if (pointcloud->num_points() == 0) {
-    return;
-  }
-
-  /* Export deformed coordinates. */
-  if (ccl::BKE_object_is_deform_modified(b_ob_info, b_scene, preview)) {
-    /* PointCloud object. */
-    BL::PointCloud b_pointcloud(b_ob_info.object_data);
-    export_pointcloud_motion(pointcloud, b_pointcloud, motion_step);
-  }
-  else {
-    /* No deformation on this frame, copy coordinates if other frames did have it. */
-    pointcloud->copy_center_to_motion_step(motion_step);
-  }
-}
-
-CCL_NAMESPACE_END

intern/cycles/blender/python.cpp

@@ -138,18 +138,20 @@ static const char *PyC_UnicodeAsByte(PyObject *py_str, PyObject **coerce)
 
 static PyObject *init_func(PyObject * /*self*/, PyObject *args)
 {
-  PyObject *path, *user_path;
+  PyObject *path, *user_path, *temp_path;
   int headless;
 
-  if (!PyArg_ParseTuple(args, "OOi", &path, &user_path, &headless)) {
+  if (!PyArg_ParseTuple(args, "OOOi", &path, &user_path, &temp_path, &headless)) {
     return nullptr;
   }
 
-  PyObject *path_coerce = nullptr, *user_path_coerce = nullptr;
+  PyObject *path_coerce = nullptr, *user_path_coerce = nullptr, *temp_path_coerce = nullptr;
   path_init(PyC_UnicodeAsByte(path, &path_coerce),
-            PyC_UnicodeAsByte(user_path, &user_path_coerce));
+            PyC_UnicodeAsByte(user_path, &user_path_coerce),
+            PyC_UnicodeAsByte(temp_path, &temp_path_coerce));
   Py_XDECREF(path_coerce);
   Py_XDECREF(user_path_coerce);
+  Py_XDECREF(temp_path_coerce);
 
   BlenderSession::headless = headless;
 
@@ -733,20 +735,27 @@ static bool image_parse_filepaths(PyObject *pyfilepaths, vector<string> &filepat
 
 static PyObject *denoise_func(PyObject * /*self*/, PyObject *args, PyObject *keywords)
 {
+#if 1
+  (void)args;
+  (void)keywords;
+#else
   static const char *keyword_list[] = {
-      "preferences", "scene", "view_layer", "input", "output", NULL};
+      "preferences", "scene", "view_layer", "input", "output", "tile_size", "samples", NULL};
   PyObject *pypreferences, *pyscene, *pyviewlayer;
   PyObject *pyinput, *pyoutput = NULL;
+  int tile_size = 0, samples = 0;
 
   if (!PyArg_ParseTupleAndKeywords(args,
                                    keywords,
-                                   "OOOO|O",
+                                   "OOOO|Oii",
                                    (char **)keyword_list,
                                    &pypreferences,
                                    &pyscene,
                                    &pyviewlayer,
                                    &pyinput,
-                                   &pyoutput)) {
+                                   &pyoutput,
+                                   &tile_size,
+                                   &samples)) {
     return NULL;
   }
 
@@ -768,10 +777,14 @@ static PyObject *denoise_func(PyObject * /*self*/, PyObject *args, PyObject *key
                      &RNA_ViewLayer,
                      PyLong_AsVoidPtr(pyviewlayer),
                      &viewlayerptr);
-  BL::ViewLayer b_view_layer(viewlayerptr);
+  PointerRNA cviewlayer = RNA_pointer_get(&viewlayerptr, "cycles");
 
-  DenoiseParams params = BlenderSync::get_denoise_params(b_scene, b_view_layer, true);
+  DenoiseParams params;
-  params.use = true;
+  params.radius = get_int(cviewlayer, "denoising_radius");
+  params.strength = get_float(cviewlayer, "denoising_strength");
+  params.feature_strength = get_float(cviewlayer, "denoising_feature_strength");
+  params.relative_pca = get_boolean(cviewlayer, "denoising_relative_pca");
+  params.neighbor_frames = get_int(cviewlayer, "denoising_neighbor_frames");
 
   /* Parse file paths list. */
   vector<string> input, output;
@@ -799,15 +812,24 @@ static PyObject *denoise_func(PyObject * /*self*/, PyObject *args, PyObject *key
   }
 
   /* Create denoiser. */
-  DenoiserPipeline denoiser(device, params);
+  DenoiserPipeline denoiser(device);
+  denoiser.params = params;
   denoiser.input = input;
   denoiser.output = output;
 
+  if (tile_size > 0) {
+    denoiser.tile_size = make_int2(tile_size, tile_size);
+  }
+  if (samples > 0) {
+    denoiser.samples_override = samples;
+  }
+
   /* Run denoiser. */
   if (!denoiser.run()) {
     PyErr_SetString(PyExc_ValueError, denoiser.error.c_str());
     return NULL;
   }
+#endif
 
   Py_RETURN_NONE;
 }
@@ -884,18 +906,16 @@ static PyObject *enable_print_stats_func(PyObject * /*self*/, PyObject * /*args*
 static PyObject *get_device_types_func(PyObject * /*self*/, PyObject * /*args*/)
 {
   vector<DeviceType> device_types = Device::available_types();
-  bool has_cuda = false, has_optix = false, has_hip = false, has_metal = false;
+  bool has_cuda = false, has_optix = false, has_hip = false;
   foreach (DeviceType device_type, device_types) {
     has_cuda |= (device_type == DEVICE_CUDA);
     has_optix |= (device_type == DEVICE_OPTIX);
     has_hip |= (device_type == DEVICE_HIP);
-    has_metal |= (device_type == DEVICE_METAL);
   }
-  PyObject *list = PyTuple_New(4);
+  PyObject *list = PyTuple_New(3);
   PyTuple_SET_ITEM(list, 0, PyBool_FromLong(has_cuda));
   PyTuple_SET_ITEM(list, 1, PyBool_FromLong(has_optix));
   PyTuple_SET_ITEM(list, 2, PyBool_FromLong(has_hip));
-  PyTuple_SET_ITEM(list, 3, PyBool_FromLong(has_metal));
   return list;
 }
 
@@ -924,9 +944,6 @@ static PyObject *set_device_override_func(PyObject * /*self*/, PyObject *arg)
   else if (override == "HIP") {
     BlenderSession::device_override = DEVICE_MASK_HIP;
   }
-  else if (override == "METAL") {
-    BlenderSession::device_override = DEVICE_MASK_METAL;
-  }
   else {
     printf("\nError: %s is not a valid Cycles device.\n", override.c_str());
     Py_RETURN_FALSE;
@@ -1037,13 +1054,5 @@ void *CCL_python_module_init()
     Py_INCREF(Py_False);
   }
 
-#ifdef WITH_CYCLES_DEBUG
-  PyModule_AddObject(mod, "with_debug", Py_True);
-  Py_INCREF(Py_True);
-#else  /* WITH_CYCLES_DEBUG */
-  PyModule_AddObject(mod, "with_debug", Py_False);
-  Py_INCREF(Py_False);
-#endif /* WITH_CYCLES_DEBUG */
-
   return (void *)mod;
 }

intern/cycles/blender/session.cpp

@@ -396,13 +396,6 @@ void BlenderSession::render(BL::Depsgraph &b_depsgraph_)
     /* set the current view */
     b_engine.active_view_set(b_rview_name.c_str());
 
-    /* Force update in this case, since the camera transform on each frame changes
-     * in different views. This could be optimized by somehow storing the animated
-     * camera transforms separate from the fixed stereo transform. */
-    if ((scene->need_motion() != Scene::MOTION_NONE) && view_index > 0) {
-      sync->tag_update();
-    }
-
     /* update scene */
     BL::Object b_camera_override(b_engine.camera_override());
     sync->sync_camera(b_render, b_camera_override, width, height, b_rview_name.c_str());
@@ -502,15 +495,10 @@ void BlenderSession::render_frame_finish()
     path_remove(filename);
   }
 
-  /* Clear output driver. */
+  /* Clear driver. */
   session->set_output_driver(nullptr);
   session->full_buffer_written_cb = function_null;
 
-  /* The display driver holds OpenGL resources which belong to an OpenGL context held by the render
-   * engine on Blender side. Force destruction of those resources. */
-  display_driver_ = nullptr;
-  session->set_display_driver(nullptr);
-
   /* All the files are handled.
    * Clear the list so that this session can be re-used by Persistent Data. */
   full_buffer_files_.clear();
@@ -641,7 +629,7 @@ void BlenderSession::bake(BL::Depsgraph &b_depsgraph_,
       integrator->set_use_emission((bake_filter & BL::BakeSettings::pass_filter_EMIT) != 0);
     }
 
-    /* Always use transparent background for baking. */
+    /* Always use transpanent background for baking. */
     scene->background->set_transparent(true);
 
     /* Load built-in images from Blender. */

intern/cycles/blender/shader.cpp

@@ -378,19 +378,10 @@ static ShaderNode *add_node(Scene *scene,
   }
   else if (b_node.is_a(&RNA_ShaderNodeMapRange)) {
     BL::ShaderNodeMapRange b_map_range_node(b_node);
-    if (b_map_range_node.data_type() == BL::ShaderNodeMapRange::data_type_FLOAT_VECTOR) {
+    MapRangeNode *map_range_node = graph->create_node<MapRangeNode>();
-      VectorMapRangeNode *vector_map_range_node = graph->create_node<VectorMapRangeNode>();
+    map_range_node->set_clamp(b_map_range_node.clamp());
-      vector_map_range_node->set_use_clamp(b_map_range_node.clamp());
+    map_range_node->set_range_type((NodeMapRangeType)b_map_range_node.interpolation_type());
-      vector_map_range_node->set_range_type(
+    node = map_range_node;
-          (NodeMapRangeType)b_map_range_node.interpolation_type());
-      node = vector_map_range_node;
-    }
-    else {
-      MapRangeNode *map_range_node = graph->create_node<MapRangeNode>();
-      map_range_node->set_clamp(b_map_range_node.clamp());
-      map_range_node->set_range_type((NodeMapRangeType)b_map_range_node.interpolation_type());
-      node = map_range_node;
-    }
   }
   else if (b_node.is_a(&RNA_ShaderNodeClamp)) {
     BL::ShaderNodeClamp b_clamp_node(b_node);
@@ -776,7 +767,7 @@ static ShaderNode *add_node(Scene *scene,
       }
       else {
         ustring filename = ustring(
-            image_user_file_path(b_image_user, b_image, b_scene.frame_current()));
+            image_user_file_path(b_image_user, b_image, b_scene.frame_current(), true));
         image->set_filename(filename);
       }
     }
@@ -813,7 +804,7 @@ static ShaderNode *add_node(Scene *scene,
       }
       else {
         env->set_filename(
-            ustring(image_user_file_path(b_image_user, b_image, b_scene.frame_current())));
+            ustring(image_user_file_path(b_image_user, b_image, b_scene.frame_current(), false)));
       }
     }
     node = env;

intern/cycles/blender/sync.cpp

@@ -95,11 +95,6 @@ void BlenderSync::reset(BL::BlendData &b_data, BL::Scene &b_scene)
   this->b_scene = b_scene;
 }
 
-void BlenderSync::tag_update()
-{
-  has_updates_ = true;
-}
-
 /* Sync */
 
 void BlenderSync::sync_recalc(BL::Depsgraph &b_depsgraph, BL::SpaceView3D &b_v3d)
@@ -397,12 +392,6 @@ void BlenderSync::sync_integrator(BL::ViewLayer &b_view_layer, bool background)
     integrator->set_ao_bounces(0);
   }
 
-#ifdef WITH_CYCLES_DEBUG
-  DirectLightSamplingType direct_light_sampling_type = (DirectLightSamplingType)get_enum(
-      cscene, "direct_light_sampling_type", DIRECT_LIGHT_SAMPLING_NUM, DIRECT_LIGHT_SAMPLING_MIS);
-  integrator->set_direct_light_sampling_type(direct_light_sampling_type);
-#endif
-
   const DenoiseParams denoise_params = get_denoise_params(b_scene, b_view_layer, background);
   integrator->set_use_denoise(denoise_params.use);
 
@@ -832,14 +821,6 @@ SessionParams BlenderSync::get_session_params(BL::RenderEngine &b_engine,
   SessionParams params;
   PointerRNA cscene = RNA_pointer_get(&b_scene.ptr, "cycles");
 
-  if (background && !b_engine.is_preview()) {
-    /* Viewport and preview renders do not require temp directory and do request session
-     * parameters more often than the background render.
-     * Optimize RNA-C++ usage and memory allocation a bit by saving string access which we know is
-     * not needed for viewport render. */
-    params.temp_dir = b_engine.temporary_directory();
-  }
-
   /* feature set */
   params.experimental = (get_enum(cscene, "feature_set") != 0);
 
@@ -854,25 +835,18 @@ SessionParams BlenderSync::get_session_params(BL::RenderEngine &b_engine,
   /* samples */
   int samples = get_int(cscene, "samples");
   int preview_samples = get_int(cscene, "preview_samples");
-  int sample_offset = get_int(cscene, "sample_offset");
 
   if (background) {
     params.samples = samples;
-    params.sample_offset = sample_offset;
   }
   else {
     params.samples = preview_samples;
-    if (params.samples == 0) {
+    if (params.samples == 0)
       params.samples = INT_MAX;
-    }
-    params.sample_offset = 0;
   }
 
-  /* Clamp sample offset. */
-  params.sample_offset = clamp(params.sample_offset, 0, Integrator::MAX_SAMPLES);
-
   /* Clamp samples. */
-  params.samples = clamp(params.samples, 0, Integrator::MAX_SAMPLES - params.sample_offset);
+  params.samples = min(params.samples, Integrator::MAX_SAMPLES);
 
   /* Viewport Performance */
   params.pixel_size = b_engine.get_preview_pixel_size(b_scene);
@@ -891,7 +865,7 @@ SessionParams BlenderSync::get_session_params(BL::RenderEngine &b_engine,
 
   /* Time limit. */
   if (background) {
-    params.time_limit = (double)get_float(cscene, "time_limit");
+    params.time_limit = get_float(cscene, "time_limit");
   }
   else {
     /* For the viewport it kind of makes more sense to think in terms of the noise floor, which is

intern/cycles/blender/sync.h

@@ -66,8 +66,6 @@ class BlenderSync {
 
   void reset(BL::BlendData &b_data, BL::Scene &b_scene);
 
-  void tag_update();
-
   /* sync */
   void sync_recalc(BL::Depsgraph &b_depsgraph, BL::SpaceView3D &b_v3d);
   void sync_data(BL::RenderSettings &b_render,
@@ -105,11 +103,11 @@ class BlenderSync {
   static BufferParams get_buffer_params(
       BL::SpaceView3D &b_v3d, BL::RegionView3D &b_rv3d, Camera *cam, int width, int height);
 
+ private:
   static DenoiseParams get_denoise_params(BL::Scene &b_scene,
                                           BL::ViewLayer &b_view_layer,
                                           bool background);
 
- private:
   /* sync */
   void sync_lights(BL::Depsgraph &b_depsgraph, bool update_all);
   void sync_materials(BL::Depsgraph &b_depsgraph, bool update_all);
@@ -169,16 +167,12 @@ class BlenderSync {
       Hair *hair, BL::Mesh &b_mesh, BObjectInfo &b_ob_info, bool motion, int motion_step = 0);
   bool object_has_particle_hair(BL::Object b_ob);
 
-  /* Point Cloud */
-  void sync_pointcloud(PointCloud *pointcloud, BObjectInfo &b_ob_info);
-  void sync_pointcloud_motion(PointCloud *pointcloud, BObjectInfo &b_ob_info, int motion_step = 0);
-
   /* Camera */
   void sync_camera_motion(
       BL::RenderSettings &b_render, BL::Object &b_ob, int width, int height, float motion_time);
 
   /* Geometry */
-  Geometry *sync_geometry(BL::Depsgraph &b_depsgraph,
+  Geometry *sync_geometry(BL::Depsgraph &b_depsgrpah,
                           BObjectInfo &b_ob_info,
                           bool object_updated,
                           bool use_particle_hair,
@@ -273,6 +267,7 @@ class BlenderSync {
 
   Progress &progress;
 
+ protected:
   /* Indicates that `sync_recalc()` detected changes in the scene.
    * If this flag is false then the data is considered to be up-to-date and will not be
    * synchronized at all. */

intern/cycles/blender/util.h

@@ -18,7 +18,6 @@
 #define __BLENDER_UTIL_H__
 
 #include "scene/mesh.h"
-#include "scene/scene.h"
 
 #include "util/algorithm.h"
 #include "util/array.h"
@@ -34,7 +33,7 @@
 
 extern "C" {
 void BKE_image_user_frame_calc(void *ima, void *iuser, int cfra);
-void BKE_image_user_file_path_ex(void *iuser, void *ima, char *path, bool resolve_udim);
+void BKE_image_user_file_path(void *iuser, void *ima, char *path);
 unsigned char *BKE_image_get_pixels_for_frame(void *image, int frame, int tile);
 float *BKE_image_get_float_pixels_for_frame(void *image, int frame, int tile);
 }
@@ -291,14 +290,25 @@ static inline int render_resolution_y(BL::RenderSettings &b_render)
   return b_render.resolution_y() * b_render.resolution_percentage() / 100;
 }
 
-static inline string image_user_file_path(BL::ImageUser &iuser, BL::Image &ima, int cfra)
+static inline string image_user_file_path(BL::ImageUser &iuser,
+                                          BL::Image &ima,
+                                          int cfra,
+                                          bool load_tiled)
 {
   char filepath[1024];
   iuser.tile(0);
   BKE_image_user_frame_calc(ima.ptr.data, iuser.ptr.data, cfra);
-  BKE_image_user_file_path_ex(iuser.ptr.data, ima.ptr.data, filepath, false);
+  BKE_image_user_file_path(iuser.ptr.data, ima.ptr.data, filepath);
 
-  return string(filepath);
+  string filepath_str = string(filepath);
+  if (load_tiled && ima.source() == BL::Image::source_TILED) {
+    string udim;
+    if (ima.tiles.length() > 0) {
+      udim = to_string(ima.tiles[0].number());
+    }
+    string_replace(filepath_str, udim, "<UDIM>");
+  }
+  return filepath_str;
 }
 
 static inline int image_user_frame_number(BL::ImageUser &iuser, BL::Image &ima, int cfra)
@@ -637,7 +647,7 @@ static inline Mesh::SubdivisionType object_subdivision_type(BL::Object &b_ob,
 {
   PointerRNA cobj = RNA_pointer_get(&b_ob.ptr, "cycles");
 
-  if (cobj.data && !b_ob.modifiers.empty() && experimental) {
+  if (cobj.data && b_ob.modifiers.length() > 0 && experimental) {
     BL::Modifier mod = b_ob.modifiers[b_ob.modifiers.length() - 1];
     bool enabled = preview ? mod.show_viewport() : mod.show_render();
 
@@ -671,40 +681,6 @@ static inline uint object_ray_visibility(BL::Object &b_ob)
   return flag;
 }
 
-/* Check whether some of "built-in" motion-related attributes are needed to be exported (includes
- * things like velocity from cache modifier, fluid simulation).
- *
- * NOTE: This code is run prior to object motion blur initialization. so can not access properties
- * set by `sync_object_motion_init()`. */
-static inline bool object_need_motion_attribute(BObjectInfo &b_ob_info, Scene *scene)
-{
-  const Scene::MotionType need_motion = scene->need_motion();
-  if (need_motion == Scene::MOTION_NONE) {
-    /* Simple case: neither motion pass nor motion blur is needed, no need in the motion related
-     * attributes. */
-    return false;
-  }
-
-  if (need_motion == Scene::MOTION_BLUR) {
-    /* A bit tricky and implicit case:
-     * - Motion blur is enabled in the scene, which implies specific number of time steps for
-     *   objects.
-     * - If the object has motion blur disabled on it, it will have 0 time steps.
-     * - Motion attribute expects non-zero time steps.
-     *
-     * Avoid adding motion attributes if the motion blur will enforce 0 motion steps. */
-    PointerRNA cobject = RNA_pointer_get(&b_ob_info.real_object.ptr, "cycles");
-    const bool use_motion = get_boolean(cobject, "use_motion_blur");
-    if (!use_motion) {
-      return false;
-    }
-  }
-
-  /* Motion pass which implies 3 motion steps, or motion blur which is not disabled on object
-   * level. */
-  return true;
-}
-
 class EdgeMap {
  public:
   EdgeMap()

intern/cycles/bvh/CMakeLists.txt

@@ -33,17 +33,6 @@ set(SRC
   unaligned.cpp
 )
 
-set(SRC_METAL
-  metal.mm
-)
-
-if(WITH_CYCLES_DEVICE_METAL)
-  list(APPEND SRC
-    ${SRC_METAL}
-  )
-  add_definitions(-DWITH_METAL)
-endif()
-
 set(SRC_HEADERS
   bvh.h
   bvh2.h
@@ -57,7 +46,6 @@ set(SRC_HEADERS
   sort.h
   split.h
   unaligned.h
-  metal.h
 )
 
 set(LIB

intern/cycles/bvh/build.cpp

@@ -26,7 +26,6 @@
 #include "scene/hair.h"
 #include "scene/mesh.h"
 #include "scene/object.h"
-#include "scene/pointcloud.h"
 #include "scene/scene.h"
 
 #include "util/algorithm.h"
@@ -114,9 +113,9 @@ void BVHBuild::add_reference_triangles(BoundBox &root,
     else {
       /* Motion triangles, trace optimized case:  we split triangle
        * primitives into separate nodes for each of the time steps.
-       * This way we minimize overlap of neighbor triangle primitives.
+       * This way we minimize overlap of neighbor curve primitives.
        */
-      const int num_bvh_steps = params.num_motion_triangle_steps * 2 + 1;
+      const int num_bvh_steps = params.num_motion_curve_steps * 2 + 1;
       const float num_bvh_steps_inv_1 = 1.0f / (num_bvh_steps - 1);
       const size_t num_verts = mesh->verts.size();
       const size_t num_steps = mesh->motion_steps;
@@ -270,101 +269,6 @@ void BVHBuild::add_reference_curves(BoundBox &root, BoundBox &center, Hair *hair
   }
 }
 
-void BVHBuild::add_reference_points(BoundBox &root,
-                                    BoundBox &center,
-                                    PointCloud *pointcloud,
-                                    int i)
-{
-  const Attribute *point_attr_mP = NULL;
-  if (pointcloud->has_motion_blur()) {
-    point_attr_mP = pointcloud->attributes.find(ATTR_STD_MOTION_VERTEX_POSITION);
-  }
-
-  const float3 *points_data = &pointcloud->points[0];
-  const float *radius_data = &pointcloud->radius[0];
-  const size_t num_points = pointcloud->num_points();
-  const float3 *motion_data = (point_attr_mP) ? point_attr_mP->data_float3() : NULL;
-  const size_t num_steps = pointcloud->get_motion_steps();
-
-  if (point_attr_mP == NULL) {
-    /* Really simple logic for static points. */
-    for (uint j = 0; j < num_points; j++) {
-      const PointCloud::Point point = pointcloud->get_point(j);
-      BoundBox bounds = BoundBox::empty;
-      point.bounds_grow(points_data, radius_data, bounds);
-      if (bounds.valid()) {
-        references.push_back(BVHReference(bounds, j, i, PRIMITIVE_POINT));
-        root.grow(bounds);
-        center.grow(bounds.center2());
-      }
-    }
-  }
-  else if (params.num_motion_point_steps == 0 || params.use_spatial_split) {
-    /* Simple case of motion points: single node for the whole
-     * shutter time. Lowest memory usage but less optimal
-     * rendering.
-     */
-    /* TODO(sergey): Support motion steps for spatially split BVH. */
-    for (uint j = 0; j < num_points; j++) {
-      const PointCloud::Point point = pointcloud->get_point(j);
-      BoundBox bounds = BoundBox::empty;
-      point.bounds_grow(points_data, radius_data, bounds);
-      for (size_t step = 0; step < num_steps - 1; step++) {
-        point.bounds_grow(motion_data + step * num_points, radius_data, bounds);
-      }
-      if (bounds.valid()) {
-        references.push_back(BVHReference(bounds, j, i, PRIMITIVE_MOTION_POINT));
-        root.grow(bounds);
-        center.grow(bounds.center2());
-      }
-    }
-  }
-  else {
-    /* Motion points, trace optimized case:  we split point
-     * primitives into separate nodes for each of the time steps.
-     * This way we minimize overlap of neighbor point primitives.
-     */
-    const int num_bvh_steps = params.num_motion_point_steps * 2 + 1;
-    const float num_bvh_steps_inv_1 = 1.0f / (num_bvh_steps - 1);
-
-    for (uint j = 0; j < num_points; j++) {
-      const PointCloud::Point point = pointcloud->get_point(j);
-      const size_t num_steps = pointcloud->get_motion_steps();
-      const float3 *point_steps = point_attr_mP->data_float3();
-
-      /* Calculate bounding box of the previous time step.
-       * Will be reused later to avoid duplicated work on
-       * calculating BVH time step boundbox.
-       */
-      float4 prev_key = point.motion_key(
-          points_data, radius_data, point_steps, num_points, num_steps, 0.0f, j);
-      BoundBox prev_bounds = BoundBox::empty;
-      point.bounds_grow(prev_key, prev_bounds);
-      /* Create all primitive time steps, */
-      for (int bvh_step = 1; bvh_step < num_bvh_steps; ++bvh_step) {
-        const float curr_time = (float)(bvh_step)*num_bvh_steps_inv_1;
-        float4 curr_key = point.motion_key(
-            points_data, radius_data, point_steps, num_points, num_steps, curr_time, j);
-        BoundBox curr_bounds = BoundBox::empty;
-        point.bounds_grow(curr_key, curr_bounds);
-        BoundBox bounds = prev_bounds;
-        bounds.grow(curr_bounds);
-        if (bounds.valid()) {
-          const float prev_time = (float)(bvh_step - 1) * num_bvh_steps_inv_1;
-          references.push_back(
-              BVHReference(bounds, j, i, PRIMITIVE_MOTION_POINT, prev_time, curr_time));
-          root.grow(bounds);
-          center.grow(bounds.center2());
-        }
-        /* Current time boundbox becomes previous one for the
-         * next time step.
-         */
-        prev_bounds = curr_bounds;
-      }
-    }
-  }
-}
-
 void BVHBuild::add_reference_geometry(BoundBox &root,
                                       BoundBox &center,
                                       Geometry *geom,
@@ -378,10 +282,6 @@ void BVHBuild::add_reference_geometry(BoundBox &root,
     Hair *hair = static_cast<Hair *>(geom);
     add_reference_curves(root, center, hair, object_index);
   }
-  else if (geom->geometry_type == Geometry::POINTCLOUD) {
-    PointCloud *pointcloud = static_cast<PointCloud *>(geom);
-    add_reference_points(root, center, pointcloud, object_index);
-  }
 }
 
 void BVHBuild::add_reference_object(BoundBox &root, BoundBox &center, Object *ob, int i)
@@ -411,10 +311,6 @@ static size_t count_primitives(Geometry *geom)
     Hair *hair = static_cast<Hair *>(geom);
     return count_curve_segments(hair);
   }
-  else if (geom->geometry_type == Geometry::POINTCLOUD) {
-    PointCloud *pointcloud = static_cast<PointCloud *>(geom);
-    return pointcloud->num_points();
-  }
 
   return 0;
 }
@@ -432,9 +328,8 @@ void BVHBuild::add_references(BVHRange &root)
       if (!ob->get_geometry()->is_instanced()) {
         num_alloc_references += count_primitives(ob->get_geometry());
       }
-      else {
+      else
         num_alloc_references++;
-      }
     }
     else {
       num_alloc_references += count_primitives(ob->get_geometry());
@@ -499,7 +394,7 @@ BVHNode *BVHBuild::run()
   spatial_min_overlap = root.bounds().safe_area() * params.spatial_split_alpha;
   spatial_free_index = 0;
 
-  need_prim_time = params.use_motion_steps();
+  need_prim_time = params.num_motion_curve_steps > 0 || params.num_motion_triangle_steps > 0;
 
   /* init progress updates */
   double build_start_time;
@@ -640,8 +535,7 @@ bool BVHBuild::range_within_max_leaf_size(const BVHRange &range,
                                           const vector<BVHReference> &references) const
 {
   size_t size = range.size();
-  size_t max_leaf_size = max(max(params.max_triangle_leaf_size, params.max_curve_leaf_size),
+  size_t max_leaf_size = max(params.max_triangle_leaf_size, params.max_curve_leaf_size);
-                             params.max_point_leaf_size);
 
   if (size > max_leaf_size)
     return false;
@@ -650,44 +544,32 @@ bool BVHBuild::range_within_max_leaf_size(const BVHRange &range,
   size_t num_motion_triangles = 0;
   size_t num_curves = 0;
   size_t num_motion_curves = 0;
-  size_t num_points = 0;
-  size_t num_motion_points = 0;
 
   for (int i = 0; i < size; i++) {
     const BVHReference &ref = references[range.start() + i];
 
-    if (ref.prim_type() & PRIMITIVE_CURVE) {
+    if (ref.prim_type() & PRIMITIVE_ALL_CURVE) {
-      if (ref.prim_type() & PRIMITIVE_MOTION) {
+      if (ref.prim_type() & PRIMITIVE_ALL_MOTION) {
         num_motion_curves++;
       }
       else {
         num_curves++;
       }
     }
-    else if (ref.prim_type() & PRIMITIVE_TRIANGLE) {
+    else if (ref.prim_type() & PRIMITIVE_ALL_TRIANGLE) {
-      if (ref.prim_type() & PRIMITIVE_MOTION) {
+      if (ref.prim_type() & PRIMITIVE_ALL_MOTION) {
         num_motion_triangles++;
       }
       else {
         num_triangles++;
       }
     }
-    else if (ref.prim_type() & PRIMITIVE_POINT) {
-      if (ref.prim_type() & PRIMITIVE_MOTION) {
-        num_motion_points++;
-      }
-      else {
-        num_points++;
-      }
-    }
   }
 
   return (num_triangles <= params.max_triangle_leaf_size) &&
          (num_motion_triangles <= params.max_motion_triangle_leaf_size) &&
          (num_curves <= params.max_curve_leaf_size) &&
-         (num_motion_curves <= params.max_motion_curve_leaf_size) &&
+         (num_motion_curves <= params.max_motion_curve_leaf_size);
-         (num_points <= params.max_point_leaf_size) &&
-         (num_motion_points <= params.max_motion_point_leaf_size);
 }
 
 /* multithreaded binning builder */
@@ -973,7 +855,7 @@ BVHNode *BVHBuild::create_leaf_node(const BVHRange &range, const vector<BVHRefer
   for (int i = 0; i < range.size(); i++) {
     const BVHReference &ref = references[range.start() + i];
     if (ref.prim_index() != -1) {
-      uint32_t type_index = PRIMITIVE_INDEX(ref.prim_type() & PRIMITIVE_ALL);
+      uint32_t type_index = bitscan((uint32_t)(ref.prim_type() & PRIMITIVE_ALL));
       p_ref[type_index].push_back(ref);
       p_type[type_index].push_back(ref.prim_type());
       p_index[type_index].push_back(ref.prim_index());

intern/cycles/bvh/build.h

@@ -39,7 +39,6 @@ class Geometry;
 class Hair;
 class Mesh;
 class Object;
-class PointCloud;
 class Progress;
 
 /* BVH Builder */
@@ -69,7 +68,6 @@ class BVHBuild {
   /* Adding references. */
   void add_reference_triangles(BoundBox &root, BoundBox &center, Mesh *mesh, int i);
   void add_reference_curves(BoundBox &root, BoundBox &center, Hair *hair, int i);
-  void add_reference_points(BoundBox &root, BoundBox &center, PointCloud *pointcloud, int i);
   void add_reference_geometry(BoundBox &root, BoundBox &center, Geometry *geom, int i);
   void add_reference_object(BoundBox &root, BoundBox &center, Object *ob, int i);
   void add_references(BVHRange &root);

intern/cycles/bvh/bvh.cpp

@@ -19,7 +19,6 @@
 
 #include "bvh/bvh2.h"
 #include "bvh/embree.h"
-#include "bvh/metal.h"
 #include "bvh/multi.h"
 #include "bvh/optix.h"
 
@@ -41,12 +40,8 @@ const char *bvh_layout_name(BVHLayout layout)
       return "EMBREE";
     case BVH_LAYOUT_OPTIX:
       return "OPTIX";
-    case BVH_LAYOUT_METAL:
-      return "METAL";
     case BVH_LAYOUT_MULTI_OPTIX:
-    case BVH_LAYOUT_MULTI_METAL:
     case BVH_LAYOUT_MULTI_OPTIX_EMBREE:
-    case BVH_LAYOUT_MULTI_METAL_EMBREE:
       return "MULTI";
     case BVH_LAYOUT_ALL:
       return "ALL";
@@ -107,18 +102,9 @@ BVH *BVH::create(const BVHParams &params,
 #else
       (void)device;
       break;
-#endif
-    case BVH_LAYOUT_METAL:
-#ifdef WITH_METAL
-      return bvh_metal_create(params, geometry, objects, device);
-#else
-      (void)device;
-      break;
 #endif
     case BVH_LAYOUT_MULTI_OPTIX:
-    case BVH_LAYOUT_MULTI_METAL:
     case BVH_LAYOUT_MULTI_OPTIX_EMBREE:
-    case BVH_LAYOUT_MULTI_METAL_EMBREE:
       return new BVHMulti(params, geometry, objects);
     case BVH_LAYOUT_NONE:
     case BVH_LAYOUT_ALL:

intern/cycles/bvh/bvh2.cpp

@@ -20,7 +20,6 @@
 #include "scene/hair.h"
 #include "scene/mesh.h"
 #include "scene/object.h"
-#include "scene/pointcloud.h"
 
 #include "bvh/build.h"
 #include "bvh/node.h"
@@ -387,7 +386,7 @@ void BVH2::refit_primitives(int start, int end, BoundBox &bbox, uint &visibility
     }
     else {
       /* Primitives. */
-      if (pack.prim_type[prim] & PRIMITIVE_CURVE) {
+      if (pack.prim_type[prim] & PRIMITIVE_ALL_CURVE) {
         /* Curves. */
         const Hair *hair = static_cast<const Hair *>(ob->get_geometry());
         int prim_offset = (params.top_level) ? hair->prim_offset : 0;
@@ -410,30 +409,6 @@ void BVH2::refit_primitives(int start, int end, BoundBox &bbox, uint &visibility
           }
         }
       }
-      else if (pack.prim_type[prim] & PRIMITIVE_POINT) {
-        /* Points. */
-        const PointCloud *pointcloud = static_cast<const PointCloud *>(ob->get_geometry());
-        int prim_offset = (params.top_level) ? pointcloud->prim_offset : 0;
-        const float3 *points = &pointcloud->points[0];
-        const float *radius = &pointcloud->radius[0];
-        PointCloud::Point point = pointcloud->get_point(pidx - prim_offset);
-
-        point.bounds_grow(points, radius, bbox);
-
-        /* Motion points. */
-        if (pointcloud->get_use_motion_blur()) {
-          Attribute *attr = pointcloud->attributes.find(ATTR_STD_MOTION_VERTEX_POSITION);
-
-          if (attr) {
-            size_t pointcloud_size = pointcloud->points.size();
-            size_t steps = pointcloud->get_motion_steps() - 1;
-            float3 *point_steps = attr->data_float3();
-
-            for (size_t i = 0; i < steps; i++)
-              point.bounds_grow(point_steps + i * pointcloud_size, radius, bbox);
-          }
-        }
-      }
       else {
         /* Triangles. */
         const Mesh *mesh = static_cast<const Mesh *>(ob->get_geometry());
@@ -530,8 +505,7 @@ void BVH2::pack_instances(size_t nodes_size, size_t leaf_nodes_size)
   pack.leaf_nodes.resize(leaf_nodes_size);
   pack.object_node.resize(objects.size());
 
-  if (params.num_motion_curve_steps > 0 || params.num_motion_triangle_steps > 0 ||
+  if (params.num_motion_curve_steps > 0 || params.num_motion_triangle_steps > 0) {
-      params.num_motion_point_steps > 0) {
     pack.prim_time.resize(prim_index_size);
   }
 
@@ -590,7 +564,13 @@ void BVH2::pack_instances(size_t nodes_size, size_t leaf_nodes_size)
       float2 *bvh_prim_time = bvh->pack.prim_time.size() ? &bvh->pack.prim_time[0] : NULL;
 
       for (size_t i = 0; i < bvh_prim_index_size; i++) {
-        pack_prim_index[pack_prim_index_offset] = bvh_prim_index[i] + geom_prim_offset;
+        if (bvh->pack.prim_type[i] & PRIMITIVE_ALL_CURVE) {
+          pack_prim_index[pack_prim_index_offset] = bvh_prim_index[i] + geom_prim_offset;
+        }
+        else {
+          pack_prim_index[pack_prim_index_offset] = bvh_prim_index[i] + geom_prim_offset;
+        }
+
         pack_prim_type[pack_prim_index_offset] = bvh_prim_type[i];
         pack_prim_visibility[pack_prim_index_offset] = bvh_prim_visibility[i];
         pack_prim_object[pack_prim_index_offset] = 0;  // unused for instances

intern/cycles/bvh/embree.cpp

@@ -45,7 +45,6 @@
 #  include "scene/hair.h"
 #  include "scene/mesh.h"
 #  include "scene/object.h"
-#  include "scene/pointcloud.h"
 
 #  include "util/foreach.h"
 #  include "util/log.h"
@@ -91,7 +90,7 @@ static void rtc_filter_occluded_func(const RTCFilterFunctionNArguments *args)
       ++ctx->num_hits;
 
       /* Always use baked shadow transparency for curves. */
-      if (current_isect.type & PRIMITIVE_CURVE) {
+      if (current_isect.type & PRIMITIVE_ALL_CURVE) {
         ctx->throughput *= intersection_curve_shadow_transparency(
             kg, current_isect.object, current_isect.prim, current_isect.u);
 
@@ -246,7 +245,7 @@ static void rtc_filter_occluded_func(const RTCFilterFunctionNArguments *args)
   }
 }
 
-static void rtc_filter_func_backface_cull(const RTCFilterFunctionNArguments *args)
+static void rtc_filter_func_thick_curve(const RTCFilterFunctionNArguments *args)
 {
   const RTCRay *ray = (RTCRay *)args->ray;
   RTCHit *hit = (RTCHit *)args->hit;
@@ -259,7 +258,7 @@ static void rtc_filter_func_backface_cull(const RTCFilterFunctionNArguments *arg
   }
 }
 
-static void rtc_filter_occluded_func_backface_cull(const RTCFilterFunctionNArguments *args)
+static void rtc_filter_occluded_func_thick_curve(const RTCFilterFunctionNArguments *args)
 {
   const RTCRay *ray = (RTCRay *)args->ray;
   RTCHit *hit = (RTCHit *)args->hit;
@@ -304,7 +303,7 @@ static void rtc_error_func(void *, enum RTCError, const char *str)
   VLOG(1) << str;
 }
 
-static double progress_start_time = 0.0;
+static double progress_start_time = 0.0f;
 
 static bool rtc_progress_func(void *user_ptr, const double n)
 {
@@ -411,12 +410,6 @@ void BVHEmbree::add_object(Object *ob, int i)
       add_curves(ob, hair, i);
     }
   }
-  else if (geom->geometry_type == Geometry::POINTCLOUD) {
-    PointCloud *pointcloud = static_cast<PointCloud *>(geom);
-    if (pointcloud->num_points() > 0) {
-      add_points(ob, pointcloud, i);
-    }
-  }
 }
 
 void BVHEmbree::add_instance(Object *ob, int i)
@@ -631,89 +624,6 @@ void BVHEmbree::set_curve_vertex_buffer(RTCGeometry geom_id, const Hair *hair, c
   }
 }
 
-void BVHEmbree::set_point_vertex_buffer(RTCGeometry geom_id,
-                                        const PointCloud *pointcloud,
-                                        const bool update)
-{
-  const Attribute *attr_mP = NULL;
-  size_t num_motion_steps = 1;
-  if (pointcloud->has_motion_blur()) {
-    attr_mP = pointcloud->attributes.find(ATTR_STD_MOTION_VERTEX_POSITION);
-    if (attr_mP) {
-      num_motion_steps = pointcloud->get_motion_steps();
-    }
-  }
-
-  const size_t num_points = pointcloud->num_points();
-
-  /* Copy the point data to Embree */
-  const int t_mid = (num_motion_steps - 1) / 2;
-  const float *radius = pointcloud->get_radius().data();
-  for (int t = 0; t < num_motion_steps; ++t) {
-    const float3 *verts;
-    if (t == t_mid || attr_mP == NULL) {
-      verts = pointcloud->get_points().data();
-    }
-    else {
-      int t_ = (t > t_mid) ? (t - 1) : t;
-      verts = &attr_mP->data_float3()[t_ * num_points];
-    }
-
-    float4 *rtc_verts = (update) ? (float4 *)rtcGetGeometryBufferData(
-                                       geom_id, RTC_BUFFER_TYPE_VERTEX, t) :
-                                   (float4 *)rtcSetNewGeometryBuffer(geom_id,
-                                                                     RTC_BUFFER_TYPE_VERTEX,
-                                                                     t,
-                                                                     RTC_FORMAT_FLOAT4,
-                                                                     sizeof(float) * 4,
-                                                                     num_points);
-
-    assert(rtc_verts);
-    if (rtc_verts) {
-      for (size_t j = 0; j < num_points; ++j) {
-        rtc_verts[j] = float3_to_float4(verts[j]);
-        rtc_verts[j].w = radius[j];
-      }
-    }
-
-    if (update) {
-      rtcUpdateGeometryBuffer(geom_id, RTC_BUFFER_TYPE_VERTEX, t);
-    }
-  }
-}
-
-void BVHEmbree::add_points(const Object *ob, const PointCloud *pointcloud, int i)
-{
-  size_t prim_offset = pointcloud->prim_offset;
-
-  const Attribute *attr_mP = NULL;
-  size_t num_motion_steps = 1;
-  if (pointcloud->has_motion_blur()) {
-    attr_mP = pointcloud->attributes.find(ATTR_STD_MOTION_VERTEX_POSITION);
-    if (attr_mP) {
-      num_motion_steps = pointcloud->get_motion_steps();
-    }
-  }
-
-  enum RTCGeometryType type = RTC_GEOMETRY_TYPE_SPHERE_POINT;
-
-  RTCGeometry geom_id = rtcNewGeometry(rtc_device, type);
-
-  rtcSetGeometryBuildQuality(geom_id, build_quality);
-  rtcSetGeometryTimeStepCount(geom_id, num_motion_steps);
-
-  set_point_vertex_buffer(geom_id, pointcloud, false);
-
-  rtcSetGeometryUserData(geom_id, (void *)prim_offset);
-  rtcSetGeometryIntersectFilterFunction(geom_id, rtc_filter_func_backface_cull);
-  rtcSetGeometryOccludedFilterFunction(geom_id, rtc_filter_occluded_func_backface_cull);
-  rtcSetGeometryMask(geom_id, ob->visibility_for_tracing());
-
-  rtcCommitGeometry(geom_id);
-  rtcAttachGeometryByID(scene, geom_id, i * 2);
-  rtcReleaseGeometry(geom_id);
-}
-
 void BVHEmbree::add_curves(const Object *ob, const Hair *hair, int i)
 {
   size_t prim_offset = hair->curve_segment_offset;
@@ -768,8 +678,8 @@ void BVHEmbree::add_curves(const Object *ob, const Hair *hair, int i)
     rtcSetGeometryOccludedFilterFunction(geom_id, rtc_filter_occluded_func);
   }
   else {
-    rtcSetGeometryIntersectFilterFunction(geom_id, rtc_filter_func_backface_cull);
+    rtcSetGeometryIntersectFilterFunction(geom_id, rtc_filter_func_thick_curve);
-    rtcSetGeometryOccludedFilterFunction(geom_id, rtc_filter_occluded_func_backface_cull);
+    rtcSetGeometryOccludedFilterFunction(geom_id, rtc_filter_occluded_func_thick_curve);
   }
   rtcSetGeometryMask(geom_id, ob->visibility_for_tracing());
 
@@ -806,14 +716,6 @@ void BVHEmbree::refit(Progress &progress)
           rtcCommitGeometry(geom);
         }
       }
-      else if (geom->geometry_type == Geometry::POINTCLOUD) {
-        PointCloud *pointcloud = static_cast<PointCloud *>(geom);
-        if (pointcloud->num_points() > 0) {
-          RTCGeometry geom = rtcGetGeometry(scene, geom_id);
-          set_point_vertex_buffer(geom, pointcloud, true);
-          rtcCommitGeometry(geom);
-        }
-      }
     }
     geom_id += 2;
   }

intern/cycles/bvh/embree.h

@@ -33,7 +33,6 @@ CCL_NAMESPACE_BEGIN
 
 class Hair;
 class Mesh;
-class PointCloud;
 
 class BVHEmbree : public BVH {
  public:
@@ -52,15 +51,11 @@ class BVHEmbree : public BVH {
   void add_object(Object *ob, int i);
   void add_instance(Object *ob, int i);
   void add_curves(const Object *ob, const Hair *hair, int i);
-  void add_points(const Object *ob, const PointCloud *pointcloud, int i);
   void add_triangles(const Object *ob, const Mesh *mesh, int i);
 
  private:
   void set_tri_vertex_buffer(RTCGeometry geom_id, const Mesh *mesh, const bool update);
   void set_curve_vertex_buffer(RTCGeometry geom_id, const Hair *hair, const bool update);
-  void set_point_vertex_buffer(RTCGeometry geom_id,
-                               const PointCloud *pointcloud,
-                               const bool update);
 
   RTCDevice rtc_device;
   enum RTCBuildQuality build_quality;

intern/cycles/bvh/metal.h

@@ -1,35 +0,0 @@
-/*
- * Copyright 2021 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.
- */
-
-#ifndef __BVH_METAL_H__
-#define __BVH_METAL_H__
-
-#ifdef WITH_METAL
-
-#  include "bvh/bvh.h"
-
-CCL_NAMESPACE_BEGIN
-
-BVH *bvh_metal_create(const BVHParams &params,
-                      const vector<Geometry *> &geometry,
-                      const vector<Object *> &objects,
-                      Device *device);
-
-CCL_NAMESPACE_END
-
-#endif /* WITH_METAL */
-
-#endif /* __BVH_METAL_H__ */

intern/cycles/bvh/metal.mm

@@ -1,33 +0,0 @@
-/*
- * Copyright 2021 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.
- */
-
-#ifdef WITH_METAL
-
-#  include "device/metal/bvh.h"
-
-CCL_NAMESPACE_BEGIN
-
-BVH *bvh_metal_create(const BVHParams &params,
-                      const vector<Geometry *> &geometry,
-                      const vector<Object *> &objects,
-                      Device *device)
-{
-  return new BVHMetal(params, geometry, objects, device);
-}
-
-CCL_NAMESPACE_END
-
-#endif /* WITH_METAL */

intern/cycles/bvh/node.cpp

@@ -153,7 +153,7 @@ void BVHNode::update_time()
 namespace {
 
 struct DumpTraversalContext {
-  /* Descriptor of while where writing is happening. */
+  /* Descriptor of wile where writing is happening. */
   FILE *stream;
   /* Unique identifier of the node current. */
   int id;

intern/cycles/bvh/node.h

@@ -178,7 +178,7 @@ class InnerNode : public BVHNode {
     reset_unused_children();
   }
 
-  /* NOTE: This function is only used during binary BVH builder, and it's
+  /* NOTE: This function is only used during binary BVH builder, and it
    * supposed to be configured to have 2 children which will be filled-in in a
    * bit. But this is important to have children reset to NULL. */
   explicit InnerNode(const BoundBox &bounds) : BVHNode(bounds), num_children_(0)

intern/cycles/bvh/params.h

@@ -83,8 +83,6 @@ class BVHParams {
   int max_motion_triangle_leaf_size;
   int max_curve_leaf_size;
   int max_motion_curve_leaf_size;
-  int max_point_leaf_size;
-  int max_motion_point_leaf_size;
 
   /* object or mesh level bvh */
   bool top_level;
@@ -100,13 +98,13 @@ class BVHParams {
   /* Split time range to this number of steps and create leaf node for each
    * of this time steps.
    *
-   * Speeds up rendering of motion primitives in the cost of higher memory usage.
+   * Speeds up rendering of motion curve primitives in the cost of higher
+   * memory usage.
    */
+  int num_motion_curve_steps;
 
   /* Same as above, but for triangle primitives. */
   int num_motion_triangle_steps;
-  int num_motion_curve_steps;
-  int num_motion_point_steps;
 
   /* Same as in SceneParams. */
   int bvh_type;
@@ -134,8 +132,6 @@ class BVHParams {
     max_motion_triangle_leaf_size = 8;
     max_curve_leaf_size = 1;
     max_motion_curve_leaf_size = 4;
-    max_point_leaf_size = 8;
-    max_motion_point_leaf_size = 8;
 
     top_level = false;
     bvh_layout = BVH_LAYOUT_BVH2;
@@ -143,7 +139,6 @@ class BVHParams {
 
     num_motion_curve_steps = 0;
     num_motion_triangle_steps = 0;
-    num_motion_point_steps = 0;
 
     bvh_type = 0;
 
@@ -171,12 +166,6 @@ class BVHParams {
     return (size <= min_leaf_size || level >= MAX_DEPTH);
   }
 
-  bool use_motion_steps()
-  {
-    return num_motion_curve_steps > 0 || num_motion_triangle_steps > 0 ||
-           num_motion_point_steps > 0;
-  }
-
   /* Gets best matching BVH.
    *
    * If the requested layout is supported by the device, it will be used.

intern/cycles/bvh/split.cpp

@@ -23,7 +23,6 @@
 #include "scene/hair.h"
 #include "scene/mesh.h"
 #include "scene/object.h"
-#include "scene/pointcloud.h"
 
 #include "util/algorithm.h"
 
@@ -427,32 +426,6 @@ void BVHSpatialSplit::split_curve_primitive(const Hair *hair,
   }
 }
 
-void BVHSpatialSplit::split_point_primitive(const PointCloud *pointcloud,
-                                            const Transform *tfm,
-                                            int prim_index,
-                                            int dim,
-                                            float pos,
-                                            BoundBox &left_bounds,
-                                            BoundBox &right_bounds)
-{
-  /* No real splitting support for points, assume they are small enough for it
-   * not to matter. */
-  float3 point = pointcloud->get_points()[prim_index];
-
-  if (tfm != NULL) {
-    point = transform_point(tfm, point);
-  }
-  point = get_unaligned_point(point);
-
-  if (point[dim] <= pos) {
-    left_bounds.grow(point);
-  }
-
-  if (point[dim] >= pos) {
-    right_bounds.grow(point);
-  }
-}
-
 void BVHSpatialSplit::split_triangle_reference(const BVHReference &ref,
                                                const Mesh *mesh,
                                                int dim,
@@ -480,16 +453,6 @@ void BVHSpatialSplit::split_curve_reference(const BVHReference &ref,
                         right_bounds);
 }
 
-void BVHSpatialSplit::split_point_reference(const BVHReference &ref,
-                                            const PointCloud *pointcloud,
-                                            int dim,
-                                            float pos,
-                                            BoundBox &left_bounds,
-                                            BoundBox &right_bounds)
-{
-  split_point_primitive(pointcloud, NULL, ref.prim_index(), dim, pos, left_bounds, right_bounds);
-}
-
 void BVHSpatialSplit::split_object_reference(
     const Object *object, int dim, float pos, BoundBox &left_bounds, BoundBox &right_bounds)
 {
@@ -512,13 +475,6 @@ void BVHSpatialSplit::split_object_reference(
       }
     }
   }
-  else if (geom->geometry_type == Geometry::POINTCLOUD) {
-    PointCloud *pointcloud = static_cast<PointCloud *>(geom);
-    for (int point_idx = 0; point_idx < pointcloud->num_points(); ++point_idx) {
-      split_point_primitive(
-          pointcloud, &object->get_tfm(), point_idx, dim, pos, left_bounds, right_bounds);
-    }
-  }
 }
 
 void BVHSpatialSplit::split_reference(const BVHBuild &builder,
@@ -535,18 +491,14 @@ void BVHSpatialSplit::split_reference(const BVHBuild &builder,
   /* loop over vertices/edges. */
   const Object *ob = builder.objects[ref.prim_object()];
 
-  if (ref.prim_type() & PRIMITIVE_TRIANGLE) {
+  if (ref.prim_type() & PRIMITIVE_ALL_TRIANGLE) {
     Mesh *mesh = static_cast<Mesh *>(ob->get_geometry());
     split_triangle_reference(ref, mesh, dim, pos, left_bounds, right_bounds);
   }
-  else if (ref.prim_type() & PRIMITIVE_CURVE) {
+  else if (ref.prim_type() & PRIMITIVE_ALL_CURVE) {
     Hair *hair = static_cast<Hair *>(ob->get_geometry());
     split_curve_reference(ref, hair, dim, pos, left_bounds, right_bounds);
   }
-  else if (ref.prim_type() & PRIMITIVE_POINT) {
-    PointCloud *pointcloud = static_cast<PointCloud *>(ob->get_geometry());
-    split_point_reference(ref, pointcloud, dim, pos, left_bounds, right_bounds);
-  }
   else {
     split_object_reference(ob, dim, pos, left_bounds, right_bounds);
   }

intern/cycles/bvh/split.h

@@ -26,7 +26,6 @@ CCL_NAMESPACE_BEGIN
 class BVHBuild;
 class Hair;
 class Mesh;
-class PointCloud;
 struct Transform;
 
 /* Object Split */
@@ -124,13 +123,6 @@ class BVHSpatialSplit {
                              float pos,
                              BoundBox &left_bounds,
                              BoundBox &right_bounds);
-  void split_point_primitive(const PointCloud *pointcloud,
-                             const Transform *tfm,
-                             int prim_index,
-                             int dim,
-                             float pos,
-                             BoundBox &left_bounds,
-                             BoundBox &right_bounds);
 
   /* Lower-level functions which calculates boundaries of left and right nodes
    * needed for spatial split.
@@ -149,12 +141,6 @@ class BVHSpatialSplit {
                              float pos,
                              BoundBox &left_bounds,
                              BoundBox &right_bounds);
-  void split_point_reference(const BVHReference &ref,
-                             const PointCloud *pointcloud,
-                             int dim,
-                             float pos,
-                             BoundBox &left_bounds,
-                             BoundBox &right_bounds);
   void split_object_reference(
       const Object *object, int dim, float pos, BoundBox &left_bounds, BoundBox &right_bounds);
 

intern/cycles/bvh/unaligned.cpp

@@ -69,7 +69,7 @@ bool BVHUnaligned::compute_aligned_space(const BVHReference &ref, Transform *ali
   const int packed_type = ref.prim_type();
   const int type = (packed_type & PRIMITIVE_ALL);
   /* No motion blur curves here, we can't fit them to aligned boxes well. */
-  if ((type & PRIMITIVE_CURVE) && !(type & PRIMITIVE_MOTION)) {
+  if (type & (PRIMITIVE_CURVE_RIBBON | PRIMITIVE_CURVE_THICK)) {
     const int curve_index = ref.prim_index();
     const int segment = PRIMITIVE_UNPACK_SEGMENT(packed_type);
     const Hair *hair = static_cast<const Hair *>(object->get_geometry());
@@ -95,7 +95,7 @@ BoundBox BVHUnaligned::compute_aligned_prim_boundbox(const BVHReference &prim,
   const int packed_type = prim.prim_type();
   const int type = (packed_type & PRIMITIVE_ALL);
   /* No motion blur curves here, we can't fit them to aligned boxes well. */
-  if ((type & PRIMITIVE_CURVE) && !(type & PRIMITIVE_MOTION)) {
+  if (type & (PRIMITIVE_CURVE_RIBBON | PRIMITIVE_CURVE_THICK)) {
     const int curve_index = prim.prim_index();
     const int segment = PRIMITIVE_UNPACK_SEGMENT(packed_type);
     const Hair *hair = static_cast<const Hair *>(object->get_geometry());

intern/cycles/cmake/external_libs.cmake

@@ -551,23 +551,4 @@ if(NOT WITH_HIP_DYNLOAD)
   set(WITH_HIP_DYNLOAD ON)
 endif()
 
-###########################################################################
-# Metal
-###########################################################################
-
-if(WITH_CYCLES_DEVICE_METAL)
-  find_library(METAL_LIBRARY Metal)
-
-  # This file was added in the 12.0 SDK, use it as a way to detect the version.
-  if (METAL_LIBRARY AND NOT EXISTS "${METAL_LIBRARY}/Headers/MTLFunctionStitching.h")
-    message(STATUS "Metal version too old, must be SDK 12.0 or newer, disabling WITH_CYCLES_DEVICE_METAL")
-    set(WITH_CYCLES_DEVICE_METAL OFF)
-  elseif (NOT METAL_LIBRARY)
-    message(STATUS "Metal not found, disabling WITH_CYCLES_DEVICE_METAL")
-    set(WITH_CYCLES_DEVICE_METAL OFF)
-  else()
-    message(STATUS "Found Metal: ${METAL_LIBRARY}")
-  endif()
-endif()
-
 unset(_cycles_lib_dir)

intern/cycles/cmake/macros.cmake

@@ -88,7 +88,7 @@ endmacro()
 
 function(cycles_link_directories)
   if(APPLE)
-    # APPLE platform uses full paths for linking libraries, and avoids link_directories.
+    # APPLE plaform uses full paths for linking libraries, and avoids link_directories.
     return()
   endif()
 
@@ -168,6 +168,12 @@ macro(cycles_target_link_libraries target)
     target_link_libraries(${target} extern_hipew)
   endif()
 
+  if(CYCLES_STANDALONE_REPOSITORY)
+    target_link_libraries(${target} extern_numaapi)
+  else()
+    target_link_libraries(${target} bf_intern_numaapi)
+  endif()
+
   if(UNIX AND NOT APPLE)
     if(CYCLES_STANDALONE_REPOSITORY)
       target_link_libraries(${target} extern_libc_compat)

intern/cycles/device/CMakeLists.txt

@@ -43,7 +43,7 @@ if(WITH_CYCLES_DEVICE_HIP AND WITH_HIP_DYNLOAD)
   add_definitions(-DWITH_HIP_DYNLOAD)
 endif()
 
-set(SRC_BASE
+set(SRC
   device.cpp
   denoise.cpp
   graphics_interop.cpp
@@ -104,21 +104,6 @@ set(SRC_MULTI
   multi/device.h
 )
 
-set(SRC_METAL
-  metal/bvh.mm
-  metal/bvh.h
-  metal/device.mm
-  metal/device.h
-  metal/device_impl.mm
-  metal/device_impl.h
-  metal/kernel.mm
-  metal/kernel.h
-  metal/queue.mm
-  metal/queue.h
-  metal/util.mm
-  metal/util.h
-)
-
 set(SRC_OPTIX
   optix/device.cpp
   optix/device.h
@@ -138,17 +123,6 @@ set(SRC_HEADERS
   queue.h
 )
 
-set(SRC
-  ${SRC_BASE}
-  ${SRC_CPU}
-  ${SRC_CUDA}
-  ${SRC_HIP}
-  ${SRC_DUMMY}
-  ${SRC_MULTI}
-  ${SRC_OPTIX}
-  ${SRC_HEADERS}
-)
-
 set(LIB
   cycles_kernel
   cycles_util
@@ -184,15 +158,6 @@ endif()
 if(WITH_CYCLES_DEVICE_OPTIX)
   add_definitions(-DWITH_OPTIX)
 endif()
-if(WITH_CYCLES_DEVICE_METAL)
-  list(APPEND LIB
-    ${METAL_LIBRARY}
-  )
-  add_definitions(-DWITH_METAL)
-  list(APPEND SRC
-    ${SRC_METAL}
-  )
-endif()
 
 if(WITH_OPENIMAGEDENOISE)
   list(APPEND LIB
@@ -203,12 +168,20 @@ endif()
 include_directories(${INC})
 include_directories(SYSTEM ${INC_SYS})
 
-cycles_add_library(cycles_device "${LIB}" ${SRC})
+cycles_add_library(cycles_device "${LIB}"
+  ${SRC}
+  ${SRC_CPU}
+  ${SRC_CUDA}
+  ${SRC_HIP}
+  ${SRC_DUMMY}
+  ${SRC_MULTI}
+  ${SRC_OPTIX}
+  ${SRC_HEADERS}
+)
 
 source_group("cpu" FILES ${SRC_CPU})
 source_group("cuda" FILES ${SRC_CUDA})
 source_group("dummy" FILES ${SRC_DUMMY})
 source_group("multi" FILES ${SRC_MULTI})
-source_group("metal" FILES ${SRC_METAL})
 source_group("optix" FILES ${SRC_OPTIX})
 source_group("common" FILES ${SRC} ${SRC_HEADERS})

intern/cycles/device/cpu/device.cpp

@@ -38,6 +38,7 @@ void device_cpu_info(vector<DeviceInfo> &devices)
   info.id = "CPU";
   info.num = 0;
   info.has_osl = true;
+  info.has_half_images = true;
   info.has_nanovdb = true;
   info.has_profiling = true;
   if (openimagedenoise_supported()) {

intern/cycles/device/cpu/device_impl.cpp

@@ -93,6 +93,11 @@ CPUDevice::~CPUDevice()
   texture_info.free();
 }
 
+bool CPUDevice::show_samples() const
+{
+  return (info.cpu_threads == 1);
+}
+
 BVHLayoutMask CPUDevice::get_bvh_layout_mask() const
 {
   BVHLayoutMask bvh_layout_mask = BVH_LAYOUT_BVH2;
@@ -273,8 +278,7 @@ void CPUDevice::build_bvh(BVH *bvh, Progress &progress, bool refit)
 {
 #ifdef WITH_EMBREE
   if (bvh->params.bvh_layout == BVH_LAYOUT_EMBREE ||
-      bvh->params.bvh_layout == BVH_LAYOUT_MULTI_OPTIX_EMBREE ||
+      bvh->params.bvh_layout == BVH_LAYOUT_MULTI_OPTIX_EMBREE) {
-      bvh->params.bvh_layout == BVH_LAYOUT_MULTI_METAL_EMBREE) {
     BVHEmbree *const bvh_embree = static_cast<BVHEmbree *>(bvh);
     if (refit) {
       bvh_embree->refit(progress);

intern/cycles/device/cpu/device_impl.h

@@ -60,6 +60,8 @@ class CPUDevice : public Device {
   CPUDevice(const DeviceInfo &info_, Stats &stats_, Profiler &profiler_);
   ~CPUDevice();
 
+  virtual bool show_samples() const override;
+
   virtual BVHLayoutMask get_bvh_layout_mask() const override;
 
   /* Returns true if the texture info was copied to the device (meaning, some more

intern/cycles/device/cuda/device.cpp

@@ -144,6 +144,7 @@ void device_cuda_info(vector<DeviceInfo> &devices)
     info.description = string(name);
     info.num = num;
 
+    info.has_half_images = (major >= 3);
     info.has_nanovdb = true;
     info.denoisers = 0;
 

intern/cycles/device/cuda/device_impl.cpp

@@ -46,6 +46,12 @@ bool CUDADevice::have_precompiled_kernels()
   return path_exists(cubins_path);
 }
 
+bool CUDADevice::show_samples() const
+{
+  /* The CUDADevice only processes one tile at a time, so showing samples is fine. */
+  return true;
+}
+
 BVHLayoutMask CUDADevice::get_bvh_layout_mask() const
 {
   return BVH_LAYOUT_BVH2;
@@ -236,10 +242,6 @@ string CUDADevice::compile_kernel_get_common_cflags(const uint kernel_features)
   cflags += " -DWITH_NANOVDB";
 #  endif
 
-#  ifdef WITH_CYCLES_DEBUG
-  cflags += " -DWITH_CYCLES_DEBUG";
-#  endif
-
   return cflags;
 }
 
@@ -477,10 +479,10 @@ void CUDADevice::reserve_local_memory(const uint kernel_features)
      * still to make it faster. */
     CUDADeviceQueue queue(this);
 
-    device_ptr d_path_index = 0;
+    void *d_path_index = nullptr;
-    device_ptr d_render_buffer = 0;
+    void *d_render_buffer = nullptr;
     int d_work_size = 0;
-    DeviceKernelArguments args(&d_path_index, &d_render_buffer, &d_work_size);
+    void *args[] = {&d_path_index, &d_render_buffer, &d_work_size};
 
     queue.init_execution();
     queue.enqueue(test_kernel, 1, args);
@@ -936,6 +938,7 @@ void CUDADevice::tex_alloc(device_texture &mem)
 {
   CUDAContextScope scope(this);
 
+  /* General variables for both architectures */
   string bind_name = mem.name;
   size_t dsize = datatype_size(mem.data_type);
   size_t size = mem.memory_size();
@@ -1098,6 +1101,7 @@ void CUDADevice::tex_alloc(device_texture &mem)
 
   if (mem.info.data_type != IMAGE_DATA_TYPE_NANOVDB_FLOAT &&
       mem.info.data_type != IMAGE_DATA_TYPE_NANOVDB_FLOAT3) {
+    /* Kepler+, bindless textures. */
     CUDA_RESOURCE_DESC resDesc;
     memset(&resDesc, 0, sizeof(resDesc));
 

intern/cycles/device/cuda/device_impl.h

@@ -76,6 +76,8 @@ class CUDADevice : public Device {
 
   static bool have_precompiled_kernels();
 
+  virtual bool show_samples() const override;
+
   virtual BVHLayoutMask get_bvh_layout_mask() const override;
 
   void set_error(const string &error) override;

intern/cycles/device/cuda/graphics_interop.cpp

@@ -45,10 +45,8 @@ void CUDADeviceGraphicsInterop::set_display_interop(
 
   need_clear_ = display_interop.need_clear;
 
-  if (!display_interop.need_recreate) {
+  if (opengl_pbo_id_ == display_interop.opengl_pbo_id && buffer_area_ == new_buffer_area) {
-    if (opengl_pbo_id_ == display_interop.opengl_pbo_id && buffer_area_ == new_buffer_area) {
+    return;
-      return;
-    }
   }
 
   CUDAContextScope scope(device_);

intern/cycles/device/cuda/queue.cpp

@@ -89,9 +89,7 @@ bool CUDADeviceQueue::kernel_available(DeviceKernel kernel) const
   return cuda_device_->kernels.available(kernel);
 }
 
-bool CUDADeviceQueue::enqueue(DeviceKernel kernel,
+bool CUDADeviceQueue::enqueue(DeviceKernel kernel, const int work_size, void *args[])
-                              const int work_size,
-                              DeviceKernelArguments const &args)
 {
   if (cuda_device_->have_error()) {
     return false;
@@ -135,7 +133,7 @@ bool CUDADeviceQueue::enqueue(DeviceKernel kernel,
                                 1,
                                 shared_mem_bytes,
                                 cuda_stream_,
-                                const_cast<void **>(args.values),
+                                args,
                                 0),
                  "enqueue");
 

intern/cycles/device/cuda/queue.h

@@ -42,9 +42,7 @@ class CUDADeviceQueue : public DeviceQueue {
 
   virtual bool kernel_available(DeviceKernel kernel) const override;
 
-  virtual bool enqueue(DeviceKernel kernel,
+  virtual bool enqueue(DeviceKernel kernel, const int work_size, void *args[]) override;
-                       const int work_size,
-                       DeviceKernelArguments const &args) override;
 
   virtual bool synchronize() override;
 

intern/cycles/device/denoise.cpp

@@ -76,8 +76,6 @@ NODE_DEFINE(DenoiseParams)
   SOCKET_BOOLEAN(use_pass_albedo, "Use Pass Albedo", true);
   SOCKET_BOOLEAN(use_pass_normal, "Use Pass Normal", false);
 
-  SOCKET_BOOLEAN(temporally_stable, "Temporally Stable", false);
-
   SOCKET_ENUM(prefilter, "Prefilter", *prefilter_enum, DENOISER_PREFILTER_FAST);
 
   return type;

intern/cycles/device/denoise.h

@@ -72,9 +72,6 @@ class DenoiseParams : public Node {
   bool use_pass_albedo = true;
   bool use_pass_normal = true;
 
-  /* Configure the denoiser to use motion vectors, previous image and a temporally stable model. */
-  bool temporally_stable = false;
-
   DenoiserPrefilter prefilter = DENOISER_PREFILTER_FAST;
 
   static const NodeEnum *get_type_enum();
@@ -86,8 +83,7 @@ class DenoiseParams : public Node {
   {
     return !(use == other.use && type == other.type && start_sample == other.start_sample &&
              use_pass_albedo == other.use_pass_albedo &&
-             use_pass_normal == other.use_pass_normal &&
+             use_pass_normal == other.use_pass_normal && prefilter == other.prefilter);
-             temporally_stable == other.temporally_stable && prefilter == other.prefilter);
   }
 };
 

intern/cycles/device/device.cpp

@@ -27,7 +27,6 @@
 #include "device/cuda/device.h"
 #include "device/dummy/device.h"
 #include "device/hip/device.h"
-#include "device/metal/device.h"
 #include "device/multi/device.h"
 #include "device/optix/device.h"
 
@@ -37,7 +36,6 @@
 #include "util/math.h"
 #include "util/string.h"
 #include "util/system.h"
-#include "util/task.h"
 #include "util/time.h"
 #include "util/types.h"
 #include "util/vector.h"
@@ -51,7 +49,6 @@ vector<DeviceInfo> Device::cuda_devices;
 vector<DeviceInfo> Device::optix_devices;
 vector<DeviceInfo> Device::cpu_devices;
 vector<DeviceInfo> Device::hip_devices;
-vector<DeviceInfo> Device::metal_devices;
 uint Device::devices_initialized_mask = 0;
 
 /* Device */
@@ -108,12 +105,6 @@ Device *Device::create(const DeviceInfo &info, Stats &stats, Profiler &profiler)
       break;
 #endif
 
-#ifdef WITH_METAL
-    case DEVICE_METAL:
-      if (device_metal_init())
-        device = device_metal_create(info, stats, profiler);
-      break;
-#endif
     default:
       break;
   }
@@ -137,8 +128,6 @@ DeviceType Device::type_from_string(const char *name)
     return DEVICE_MULTI;
   else if (strcmp(name, "HIP") == 0)
     return DEVICE_HIP;
-  else if (strcmp(name, "METAL") == 0)
-    return DEVICE_METAL;
 
   return DEVICE_NONE;
 }
@@ -155,8 +144,6 @@ string Device::string_from_type(DeviceType type)
     return "MULTI";
   else if (type == DEVICE_HIP)
     return "HIP";
-  else if (type == DEVICE_METAL)
-    return "METAL";
 
   return "";
 }
@@ -174,9 +161,7 @@ vector<DeviceType> Device::available_types()
 #ifdef WITH_HIP
   types.push_back(DEVICE_HIP);
 #endif
-#ifdef WITH_METAL
+
-  types.push_back(DEVICE_METAL);
-#endif
   return types;
 }
 
@@ -242,20 +227,6 @@ vector<DeviceInfo> Device::available_devices(uint mask)
     }
   }
 
-#ifdef WITH_METAL
-  if (mask & DEVICE_MASK_METAL) {
-    if (!(devices_initialized_mask & DEVICE_MASK_METAL)) {
-      if (device_metal_init()) {
-        device_metal_info(metal_devices);
-      }
-      devices_initialized_mask |= DEVICE_MASK_METAL;
-    }
-    foreach (DeviceInfo &info, metal_devices) {
-      devices.push_back(info);
-    }
-  }
-#endif
-
   return devices;
 }
 
@@ -295,15 +266,6 @@ string Device::device_capabilities(uint mask)
   }
 #endif
 
-#ifdef WITH_METAL
-  if (mask & DEVICE_MASK_METAL) {
-    if (device_metal_init()) {
-      capabilities += "\nMetal device capabilities:\n";
-      capabilities += device_metal_capabilities();
-    }
-  }
-#endif
-
   return capabilities;
 }
 
@@ -324,6 +286,7 @@ DeviceInfo Device::get_multi_device(const vector<DeviceInfo> &subdevices,
   info.description = "Multi Device";
   info.num = 0;
 
+  info.has_half_images = true;
   info.has_nanovdb = true;
   info.has_osl = true;
   info.has_profiling = true;
@@ -334,7 +297,7 @@ DeviceInfo Device::get_multi_device(const vector<DeviceInfo> &subdevices,
     /* Ensure CPU device does not slow down GPU. */
     if (device.type == DEVICE_CPU && subdevices.size() > 1) {
       if (background) {
-        int orig_cpu_threads = (threads) ? threads : TaskScheduler::num_threads();
+        int orig_cpu_threads = (threads) ? threads : system_cpu_thread_count();
         int cpu_threads = max(orig_cpu_threads - (subdevices.size() - 1), 0);
 
         VLOG(1) << "CPU render threads reduced from " << orig_cpu_threads << " to " << cpu_threads
@@ -370,6 +333,7 @@ DeviceInfo Device::get_multi_device(const vector<DeviceInfo> &subdevices,
     }
 
     /* Accumulate device info. */
+    info.has_half_images &= device.has_half_images;
     info.has_nanovdb &= device.has_nanovdb;
     info.has_osl &= device.has_osl;
     info.has_profiling &= device.has_profiling;
@@ -392,7 +356,6 @@ void Device::free_memory()
   optix_devices.free_memory();
   hip_devices.free_memory();
   cpu_devices.free_memory();
-  metal_devices.free_memory();
 }
 
 unique_ptr<DeviceQueue> Device::gpu_queue_create()

intern/cycles/device/device.h

@@ -52,7 +52,6 @@ enum DeviceType {
   DEVICE_MULTI,
   DEVICE_OPTIX,
   DEVICE_HIP,
-  DEVICE_METAL,
   DEVICE_DUMMY,
 };
 
@@ -61,7 +60,6 @@ enum DeviceTypeMask {
   DEVICE_MASK_CUDA = (1 << DEVICE_CUDA),
   DEVICE_MASK_OPTIX = (1 << DEVICE_OPTIX),
   DEVICE_MASK_HIP = (1 << DEVICE_HIP),
-  DEVICE_MASK_METAL = (1 << DEVICE_METAL),
   DEVICE_MASK_ALL = ~0
 };
 
@@ -75,6 +73,7 @@ class DeviceInfo {
   int num;
   bool display_device;        /* GPU is used as a display device. */
   bool has_nanovdb;           /* Support NanoVDB volumes. */
+  bool has_half_images;       /* Support half-float textures. */
   bool has_osl;               /* Support Open Shading Language. */
   bool has_profiling;         /* Supports runtime collection of profiling info. */
   bool has_peer_memory;       /* GPU has P2P access to memory of another GPU. */
@@ -91,6 +90,7 @@ class DeviceInfo {
     num = 0;
     cpu_threads = 0;
     display_device = false;
+    has_half_images = false;
     has_nanovdb = false;
     has_osl = false;
     has_profiling = false;
@@ -151,6 +151,10 @@ class Device {
     fprintf(stderr, "%s\n", error.c_str());
     fflush(stderr);
   }
+  virtual bool show_samples() const
+  {
+    return false;
+  }
   virtual BVHLayoutMask get_bvh_layout_mask() const = 0;
 
   /* statistics */
@@ -283,7 +287,6 @@ class Device {
   static vector<DeviceInfo> optix_devices;
   static vector<DeviceInfo> cpu_devices;
   static vector<DeviceInfo> hip_devices;
-  static vector<DeviceInfo> metal_devices;
   static uint devices_initialized_mask;
 };
 

intern/cycles/device/hip/device.cpp

@@ -148,6 +148,7 @@ void device_hip_info(vector<DeviceInfo> &devices)
     info.description = string(name);
     info.num = num;
 
+    info.has_half_images = true;
     info.has_nanovdb = true;
     info.denoisers = 0;
 

intern/cycles/device/hip/device_impl.cpp

@@ -47,6 +47,12 @@ bool HIPDevice::have_precompiled_kernels()
   return path_exists(fatbins_path);
 }
 
+bool HIPDevice::show_samples() const
+{
+  /* The HIPDevice only processes one tile at a time, so showing samples is fine. */
+  return true;
+}
+
 BVHLayoutMask HIPDevice::get_bvh_layout_mask() const
 {
   return BVH_LAYOUT_BVH2;
@@ -237,7 +243,7 @@ string HIPDevice::compile_kernel(const uint kernel_features, const char *name, c
   hipGetDeviceProperties(&props, hipDevId);
 
   /* gcnArchName can contain tokens after the arch name with features, ie.
-   * `gfx1010:sramecc-:xnack-` so we tokenize it to get the first part. */
+    "gfx1010:sramecc-:xnack-" so we tokenize it to get the first part. */
   char *arch = strtok(props.gcnArchName, ":");
   if (arch == NULL) {
     arch = props.gcnArchName;
@@ -368,9 +374,10 @@ string HIPDevice::compile_kernel(const uint kernel_features, const char *name, c
 
 bool HIPDevice::load_kernels(const uint kernel_features)
 {
-  /* TODO(sergey): Support kernels re-load for HIP devices adaptive compile.
+  /* TODO(sergey): Support kernels re-load for CUDA devices adaptive compile.
    *
-   * Currently re-loading kernels will invalidate memory pointers.
+   * Currently re-loading kernel will invalidate memory pointers,
+   * causing problems in cuCtxSynchronize.
    */
   if (hipModule) {
     if (use_adaptive_compilation()) {
@@ -440,10 +447,10 @@ void HIPDevice::reserve_local_memory(const uint kernel_features)
      * still to make it faster. */
     HIPDeviceQueue queue(this);
 
-    device_ptr d_path_index = 0;
+    void *d_path_index = nullptr;
-    device_ptr d_render_buffer = 0;
+    void *d_render_buffer = nullptr;
     int d_work_size = 0;
-    DeviceKernelArguments args(&d_path_index, &d_render_buffer, &d_work_size);
+    void *args[] = {&d_path_index, &d_render_buffer, &d_work_size};
 
     queue.init_execution();
     queue.enqueue(test_kernel, 1, args);
@@ -892,6 +899,7 @@ void HIPDevice::tex_alloc(device_texture &mem)
 {
   HIPContextScope scope(this);
 
+  /* General variables for both architectures */
   string bind_name = mem.name;
   size_t dsize = datatype_size(mem.data_type);
   size_t size = mem.memory_size();

intern/cycles/device/hip/device_impl.h

@@ -75,6 +75,8 @@ class HIPDevice : public Device {
 
   static bool have_precompiled_kernels();
 
+  virtual bool show_samples() const override;
+
   virtual BVHLayoutMask get_bvh_layout_mask() const override;
 
   void set_error(const string &error) override;
@@ -91,7 +93,9 @@ class HIPDevice : public Device {
 
   virtual string compile_kernel_get_common_cflags(const uint kernel_features);
 
-  string compile_kernel(const uint kernel_features, const char *name, const char *base = "hip");
+  string compile_kernel(const uint kernel_features,
+                        const char *name,
+                        const char *base = "hip");
 
   virtual bool load_kernels(const uint kernel_features) override;
   void reserve_local_memory(const uint kernel_features);

intern/cycles/device/hip/graphics_interop.h

@@ -48,7 +48,7 @@ class HIPDeviceGraphicsInterop : public DeviceGraphicsInterop {
   HIPDeviceQueue *queue_ = nullptr;
   HIPDevice *device_ = nullptr;
 
-  /* OpenGL PBO which is currently registered as the destination for the HIP buffer. */
+  /* OpenGL PBO which is currently registered as the destination for the CUDA buffer. */
   uint opengl_pbo_id_ = 0;
   /* Buffer area in pixels of the corresponding PBO. */
   int64_t buffer_area_ = 0;

intern/cycles/device/hip/queue.cpp

@@ -89,9 +89,7 @@ bool HIPDeviceQueue::kernel_available(DeviceKernel kernel) const
   return hip_device_->kernels.available(kernel);
 }
 
-bool HIPDeviceQueue::enqueue(DeviceKernel kernel,
+bool HIPDeviceQueue::enqueue(DeviceKernel kernel, const int work_size, void *args[])
-                             const int work_size,
-                             DeviceKernelArguments const &args)
 {
   if (hip_device_->have_error()) {
     return false;
@@ -134,7 +132,7 @@ bool HIPDeviceQueue::enqueue(DeviceKernel kernel,
                                        1,
                                        shared_mem_bytes,
                                        hip_stream_,
-                                       const_cast<void **>(args.values),
+                                       args,
                                        0),
                  "enqueue");
 

intern/cycles/device/hip/queue.h

@@ -42,9 +42,7 @@ class HIPDeviceQueue : public DeviceQueue {
 
   virtual bool kernel_available(DeviceKernel kernel) const override;
 
-  virtual bool enqueue(DeviceKernel kernel,
+  virtual bool enqueue(DeviceKernel kernel, const int work_size, void *args[]) override;
-                       const int work_size,
-                       DeviceKernelArguments const &args) override;
 
   virtual bool synchronize() override;
 

intern/cycles/device/memory.cpp

@@ -23,7 +23,7 @@ CCL_NAMESPACE_BEGIN
 
 device_memory::device_memory(Device *device, const char *name, MemoryType type)
     : data_type(device_type_traits<uchar>::data_type),
-      data_elements(device_type_traits<uchar>::num_elements),
+      data_elements(device_type_traits<uchar>::num_elements_cpu),
       data_size(0),
       device_size(0),
       data_width(0),

intern/cycles/device/memory.h

@@ -81,140 +81,155 @@ static constexpr size_t datatype_size(DataType datatype)
 
 template<typename T> struct device_type_traits {
   static const DataType data_type = TYPE_UNKNOWN;
-  static const size_t num_elements = sizeof(T);
+  static const size_t num_elements_cpu = sizeof(T);
+  static const size_t num_elements_gpu = sizeof(T);
 };
 
 template<> struct device_type_traits<uchar> {
   static const DataType data_type = TYPE_UCHAR;
-  static const size_t num_elements = 1;
+  static const size_t num_elements_cpu = 1;
-  static_assert(sizeof(uchar) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 1;
+  static_assert(sizeof(uchar) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<uchar2> {
   static const DataType data_type = TYPE_UCHAR;
-  static const size_t num_elements = 2;
+  static const size_t num_elements_cpu = 2;
-  static_assert(sizeof(uchar2) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 2;
+  static_assert(sizeof(uchar2) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<uchar3> {
   static const DataType data_type = TYPE_UCHAR;
-  static const size_t num_elements = 3;
+  static const size_t num_elements_cpu = 3;
-  static_assert(sizeof(uchar3) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 3;
+  static_assert(sizeof(uchar3) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<uchar4> {
   static const DataType data_type = TYPE_UCHAR;
-  static const size_t num_elements = 4;
+  static const size_t num_elements_cpu = 4;
-  static_assert(sizeof(uchar4) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 4;
+  static_assert(sizeof(uchar4) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<uint> {
   static const DataType data_type = TYPE_UINT;
-  static const size_t num_elements = 1;
+  static const size_t num_elements_cpu = 1;
-  static_assert(sizeof(uint) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 1;
+  static_assert(sizeof(uint) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<uint2> {
   static const DataType data_type = TYPE_UINT;
-  static const size_t num_elements = 2;
+  static const size_t num_elements_cpu = 2;
-  static_assert(sizeof(uint2) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 2;
+  static_assert(sizeof(uint2) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<uint3> {
   static const DataType data_type = TYPE_UINT;
-  static const size_t num_elements = 3;
+  static const size_t num_elements_cpu = 3;
-  static_assert(sizeof(uint3) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 3;
+  static_assert(sizeof(uint3) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<uint4> {
   static const DataType data_type = TYPE_UINT;
-  static const size_t num_elements = 4;
+  static const size_t num_elements_cpu = 4;
-  static_assert(sizeof(uint4) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 4;
+  static_assert(sizeof(uint4) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<int> {
   static const DataType data_type = TYPE_INT;
-  static const size_t num_elements = 1;
+  static const size_t num_elements_cpu = 1;
-  static_assert(sizeof(int) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 1;
+  static_assert(sizeof(int) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<int2> {
   static const DataType data_type = TYPE_INT;
-  static const size_t num_elements = 2;
+  static const size_t num_elements_cpu = 2;
-  static_assert(sizeof(int2) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 2;
+  static_assert(sizeof(int2) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<int3> {
   static const DataType data_type = TYPE_INT;
-  static const size_t num_elements = 4;
+  static const size_t num_elements_cpu = 4;
-  static_assert(sizeof(int3) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 3;
+  static_assert(sizeof(int3) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<int4> {
   static const DataType data_type = TYPE_INT;
-  static const size_t num_elements = 4;
+  static const size_t num_elements_cpu = 4;
-  static_assert(sizeof(int4) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 4;
+  static_assert(sizeof(int4) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<float> {
   static const DataType data_type = TYPE_FLOAT;
-  static const size_t num_elements = 1;
+  static const size_t num_elements_cpu = 1;
-  static_assert(sizeof(float) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 1;
+  static_assert(sizeof(float) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<float2> {
   static const DataType data_type = TYPE_FLOAT;
-  static const size_t num_elements = 2;
+  static const size_t num_elements_cpu = 2;
-  static_assert(sizeof(float2) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 2;
+  static_assert(sizeof(float2) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<float3> {
-  /* float3 has different size depending on the device, can't use it for interchanging
-   * memory between CPU and GPU.
-   *
-   * Leave body empty to trigger a compile error if used. */
-};
-
-template<> struct device_type_traits<packed_float3> {
   static const DataType data_type = TYPE_FLOAT;
-  static const size_t num_elements = 3;
+  static const size_t num_elements_cpu = 4;
-  static_assert(sizeof(packed_float3) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 3;
+  static_assert(sizeof(float3) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<float4> {
   static const DataType data_type = TYPE_FLOAT;
-  static const size_t num_elements = 4;
+  static const size_t num_elements_cpu = 4;
-  static_assert(sizeof(float4) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 4;
+  static_assert(sizeof(float4) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<half> {
   static const DataType data_type = TYPE_HALF;
-  static const size_t num_elements = 1;
+  static const size_t num_elements_cpu = 1;
-  static_assert(sizeof(half) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 1;
+  static_assert(sizeof(half) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<ushort4> {
   static const DataType data_type = TYPE_UINT16;
-  static const size_t num_elements = 4;
+  static const size_t num_elements_cpu = 4;
-  static_assert(sizeof(ushort4) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 4;
+  static_assert(sizeof(ushort4) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<uint16_t> {
   static const DataType data_type = TYPE_UINT16;
-  static const size_t num_elements = 1;
+  static const size_t num_elements_cpu = 1;
-  static_assert(sizeof(uint16_t) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 1;
+  static_assert(sizeof(uint16_t) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<half4> {
   static const DataType data_type = TYPE_HALF;
-  static const size_t num_elements = 4;
+  static const size_t num_elements_cpu = 4;
-  static_assert(sizeof(half4) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 4;
+  static_assert(sizeof(half4) == num_elements_cpu * datatype_size(data_type));
 };
 
 template<> struct device_type_traits<uint64_t> {
   static const DataType data_type = TYPE_UINT64;
-  static const size_t num_elements = 1;
+  static const size_t num_elements_cpu = 1;
-  static_assert(sizeof(uint64_t) == num_elements * datatype_size(data_type));
+  static const size_t num_elements_gpu = 1;
+  static_assert(sizeof(uint64_t) == num_elements_cpu * datatype_size(data_type));
 };
 
 /* Device Memory
@@ -263,7 +278,6 @@ class device_memory {
   friend class CUDADevice;
   friend class OptiXDevice;
   friend class HIPDevice;
-  friend class MetalDevice;
 
   /* Only create through subclasses. */
   device_memory(Device *device, const char *name, MemoryType type);
@@ -311,7 +325,9 @@ template<typename T> class device_only_memory : public device_memory {
       : device_memory(device, name, allow_host_memory_fallback ? MEM_READ_WRITE : MEM_DEVICE_ONLY)
   {
     data_type = device_type_traits<T>::data_type;
-    data_elements = max(device_type_traits<T>::num_elements, 1);
+    data_elements = max(device_is_cpu() ? device_type_traits<T>::num_elements_cpu :
+                                          device_type_traits<T>::num_elements_gpu,
+                        1);
   }
 
   device_only_memory(device_only_memory &&other) noexcept : device_memory(std::move(other))
@@ -367,11 +383,15 @@ template<typename T> class device_only_memory : public device_memory {
 
 template<typename T> class device_vector : public device_memory {
  public:
+  /* Can only use this for types that have the same size on CPU and GPU. */
+  static_assert(device_type_traits<T>::num_elements_cpu ==
+                device_type_traits<T>::num_elements_gpu);
+
   device_vector(Device *device, const char *name, MemoryType type)
       : device_memory(device, name, type)
   {
     data_type = device_type_traits<T>::data_type;
-    data_elements = device_type_traits<T>::num_elements;
+    data_elements = device_type_traits<T>::num_elements_cpu;
     modified = true;
     need_realloc_ = true;
 
@@ -582,7 +602,7 @@ template<typename T> class device_vector : public device_memory {
  * from an already allocated base memory. It is freed automatically when it
  * goes out of scope, which should happen before base memory is freed.
  *
- * NOTE: some devices require offset and size of the sub_ptr to be properly
+ * Note: some devices require offset and size of the sub_ptr to be properly
  * aligned to device->mem_address_alingment(). */
 
 class device_sub_ptr {

intern/cycles/device/metal/bvh.h

@@ -1,66 +0,0 @@
-/*
- * Copyright 2021 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.
- */
-
-#pragma once
-
-#ifdef WITH_METAL
-
-#  include "bvh/bvh.h"
-#  include "bvh/params.h"
-#  include "device/memory.h"
-
-#  include <Metal/Metal.h>
-
-CCL_NAMESPACE_BEGIN
-
-class BVHMetal : public BVH {
- public:
-  API_AVAILABLE(macos(11.0))
-  id<MTLAccelerationStructure> accel_struct = nil;
-  bool accel_struct_building = false;
-
-  API_AVAILABLE(macos(11.0))
-  vector<id<MTLAccelerationStructure>> blas_array;
-
-  bool motion_blur = false;
-
-  Stats &stats;
-
-  bool build(Progress &progress, id<MTLDevice> device, id<MTLCommandQueue> queue, bool refit);
-
-  BVHMetal(const BVHParams &params,
-           const vector<Geometry *> &geometry,
-           const vector<Object *> &objects,
-           Device *device);
-  virtual ~BVHMetal();
-
-  bool build_BLAS(Progress &progress, id<MTLDevice> device, id<MTLCommandQueue> queue, bool refit);
-  bool build_BLAS_mesh(Progress &progress,
-                       id<MTLDevice> device,
-                       id<MTLCommandQueue> queue,
-                       Geometry *const geom,
-                       bool refit);
-  bool build_BLAS_hair(Progress &progress,
-                       id<MTLDevice> device,
-                       id<MTLCommandQueue> queue,
-                       Geometry *const geom,
-                       bool refit);
-  bool build_TLAS(Progress &progress, id<MTLDevice> device, id<MTLCommandQueue> queue, bool refit);
-};
-
-CCL_NAMESPACE_END
-
-#endif /* WITH_METAL */