test using compiled function in math node

CMakeLists.txt

@@ -536,14 +536,12 @@ 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)
@@ -561,18 +559,12 @@ if(WIN32)
   set(CPACK_INSTALL_PREFIX ${CMAKE_GENERIC_PROGRAM_FILES}/${})
 endif()
 
-# Compiler tool-chain.
-if(UNIX AND NOT APPLE)
-  if(CMAKE_COMPILER_IS_GNUCC)
-    option(WITH_LINKER_GOLD "Use ld.gold linker which is usually faster than ld.bfd" ON)
-    mark_as_advanced(WITH_LINKER_GOLD)
-    option(WITH_LINKER_LLD "Use ld.lld linker which is usually faster than ld.gold" OFF)
-    mark_as_advanced(WITH_LINKER_LLD)
-  endif()
-  if(CMAKE_COMPILER_IS_GNUCC OR CMAKE_C_COMPILER_ID MATCHES "Clang")
-    option(WITH_LINKER_MOLD "Use ld.mold linker which is usually faster than ld.gold & ld.lld." OFF)
-    mark_as_advanced(WITH_LINKER_MOLD)
-  endif()
+# Compiler toolchain
+if(CMAKE_COMPILER_IS_GNUCC)
+  option(WITH_LINKER_GOLD "Use ld.gold linker which is usually faster than ld.bfd" ON)
+  mark_as_advanced(WITH_LINKER_GOLD)
+  option(WITH_LINKER_LLD "Use ld.lld linker which is usually faster than ld.gold" OFF)
+  mark_as_advanced(WITH_LINKER_LLD)
 endif()
 
 option(WITH_COMPILER_ASAN "Build and link against address sanitizer (only for Debug & RelWithDebInfo targets)." OFF)
@@ -708,12 +700,9 @@ if(UNIX AND NOT APPLE)
 endif()
 
 # Installation process.
-set(POSTINSTALL_SCRIPT "" CACHE FILEPATH "Run given CMake script after installation process")
+option(POSTINSTALL_SCRIPT "Run given CMake script after installation process" OFF)
 mark_as_advanced(POSTINSTALL_SCRIPT)
 
-set(POSTCONFIGURE_SCRIPT "" CACHE FILEPATH "Run given CMake script as the last step of CMake configuration")
-mark_as_advanced(POSTCONFIGURE_SCRIPT)
-
 # end option(...)
 
 
@@ -2077,8 +2066,3 @@ endif()
 if(0)
   print_all_vars()
 endif()
-
-# Should be the last step of configuration.
-if(POSTCONFIGURE_SCRIPT)
-  include(${POSTCONFIGURE_SCRIPT})
-endif()

build_files/build_environment/CMakeLists.txt

@@ -63,7 +63,6 @@ include(cmake/jpeg.cmake)
 include(cmake/blosc.cmake)
 include(cmake/pthreads.cmake)
 include(cmake/openexr.cmake)
-include(cmake/brotli.cmake)
 include(cmake/freetype.cmake)
 include(cmake/freeglut.cmake)
 include(cmake/glew.cmake)

build_files/build_environment/cmake/brotli.cmake

@@ -1,38 +0,0 @@
-# ***** BEGIN GPL LICENSE BLOCK *****
-#
-# This program is free software; you can redistribute it and/or
-# modify it under the terms of the GNU General Public License
-# as published by the Free Software Foundation; either version 2
-# of the License, or (at your option) any later version.
-#
-# This program is distributed in the hope that it will be useful,
-# but WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-# GNU General Public License for more details.
-#
-# You should have received a copy of the GNU General Public License
-# along with this program; if not, write to the Free Software Foundation,
-# Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
-#
-# ***** END GPL LICENSE BLOCK *****
-
-set(BROTLI_EXTRA_ARGS
-)
-
-ExternalProject_Add(external_brotli
-  URL file://${PACKAGE_DIR}/${BROTLI_FILE}
-  DOWNLOAD_DIR ${DOWNLOAD_DIR}
-  URL_HASH ${BROTLI_HASH_TYPE}=${BROTLI_HASH}
-  PREFIX ${BUILD_DIR}/brotli
-  CMAKE_ARGS -DCMAKE_INSTALL_PREFIX=${LIBDIR}/brotli ${DEFAULT_CMAKE_FLAGS} ${BROTLI_EXTRA_ARGS}
-  INSTALL_DIR ${LIBDIR}/brotli
-)
-
-if(BUILD_MODE STREQUAL Release AND WIN32)
-  ExternalProject_Add_Step(external_brotli after_install
-    COMMAND ${CMAKE_COMMAND} -E copy_directory ${LIBDIR}/brotli/include ${HARVEST_TARGET}/brotli/include
-    COMMAND ${CMAKE_COMMAND} -E copy ${LIBDIR}/brotli/lib/brotlidec-static${LIBEXT} ${HARVEST_TARGET}/brotli/lib/brotlidec-static${LIBEXT}
-    COMMAND ${CMAKE_COMMAND} -E copy ${LIBDIR}/brotli/lib/brotlicommon-static${LIBEXT} ${HARVEST_TARGET}/brotli/lib/brotlicommon-static${LIBEXT}
-    DEPENDEES install
-  )
-endif()

build_files/build_environment/cmake/download.cmake

@@ -94,4 +94,3 @@ download_source(POTRACE)
 download_source(HARU)
 download_source(ZSTD)
 download_source(FLEX)
-download_source(BROTLI)

build_files/build_environment/cmake/freetype.cmake

@@ -23,12 +23,9 @@ set(FREETYPE_EXTRA_ARGS
   -DWITH_HarfBuzz=OFF
   -DFT_WITH_HARFBUZZ=OFF
   -DFT_WITH_BZIP2=OFF
-  -DFT_WITH_BROTLI=ON
   -DCMAKE_DISABLE_FIND_PACKAGE_HarfBuzz=TRUE
   -DCMAKE_DISABLE_FIND_PACKAGE_BZip2=TRUE
-  -DPC_BROTLIDEC_INCLUDEDIR=${LIBDIR}/brotli/include
-  -DPC_BROTLIDEC_LIBDIR=${LIBDIR}/brotli/lib
-  )
+  -DCMAKE_DISABLE_FIND_PACKAGE_BrotliDec=TRUE)
 
 ExternalProject_Add(external_freetype
   URL file://${PACKAGE_DIR}/${FREETYPE_FILE}
@@ -39,11 +36,6 @@ ExternalProject_Add(external_freetype
   INSTALL_DIR ${LIBDIR}/freetype
 )
 
-add_dependencies(
-  external_freetype
-  external_brotli
-)
-
 if(BUILD_MODE STREQUAL Release AND WIN32)
   ExternalProject_Add_Step(external_freetype after_install
     COMMAND ${CMAKE_COMMAND} -E copy_directory ${LIBDIR}/freetype ${HARVEST_TARGET}/freetype

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

@@ -79,8 +79,6 @@ endfunction()
 harvest(alembic/include alembic/include "*.h")
 harvest(alembic/lib/libAlembic.a alembic/lib/libAlembic.a)
 harvest(alembic/bin alembic/bin "*")
-harvest(brotli/include brotli/include "*.h")
-harvest(brotli/lib brotli/lib "*.a")
 harvest(boost/include boost/include "*")
 harvest(boost/lib boost/lib "*.a")
 harvest(ffmpeg/include ffmpeg/include "*.h")

build_files/build_environment/cmake/versions.cmake

@@ -83,9 +83,9 @@ else()
   set(OPENEXR_VERSION_POSTFIX)
 endif()
 
-set(FREETYPE_VERSION 2.11.0)
+set(FREETYPE_VERSION 2.10.2)
 set(FREETYPE_URI http://prdownloads.sourceforge.net/freetype/freetype-${FREETYPE_VERSION}.tar.gz)
-set(FREETYPE_HASH cf09172322f6b50cf8f568bf8fe14bde)
+set(FREETYPE_HASH b1cb620e4c875cd4d1bfa04945400945)
 set(FREETYPE_HASH_TYPE MD5)
 set(FREETYPE_FILE freetype-${FREETYPE_VERSION}.tar.gz)
 
@@ -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)
 
@@ -500,10 +500,3 @@ set(ZSTD_FILE zstd-${ZSTD_VERSION}.tar.gz)
 
 set(SSE2NEON_GIT https://github.com/DLTcollab/sse2neon.git)
 set(SSE2NEON_GIT_HASH fe5ff00bb8d19b327714a3c290f3e2ce81ba3525)
-
-set(BROTLI_VERSION v1.0.9)
-set(BROTLI_URI https://github.com/google/brotli/archive/refs/tags/${BROTLI_VERSION}.tar.gz)
-set(BROTLI_HASH f9e8d81d0405ba66d181529af42a3354f838c939095ff99930da6aa9cdf6fe46)
-set(BROTLI_HASH_TYPE SHA256)
-set(BROTLI_FILE brotli-${BROTLI_VERSION}.tar.gz)
-

build_files/build_environment/install_deps.sh

@@ -492,7 +492,7 @@ OIIO_SKIP=false
 LLVM_VERSION="12.0.0"
 LLVM_VERSION_SHORT="12.0"
 LLVM_VERSION_MIN="11.0"
-LLVM_VERSION_MEX="14.0"
+LLVM_VERSION_MEX="13.0"
 LLVM_VERSION_FOUND=""
 LLVM_FORCE_BUILD=false
 LLVM_FORCE_REBUILD=false
@@ -3620,8 +3620,8 @@ compile_FFmpeg() {
     fi
 
     ./configure --cc="gcc -Wl,--as-needed" \
-        --extra-ldflags="-pthread" \
-        --prefix=$_inst --enable-shared \
+        --extra-ldflags="-pthread -static-libgcc" \
+        --prefix=$_inst --enable-static \
         --disable-ffplay --disable-doc \
         --enable-gray \
         --enable-avfilter --disable-vdpau \
@@ -5721,6 +5721,76 @@ install_OTHER() {
 # ----------------------------------------------------------------------------
 # Printing User Info
 
+print_info_ffmpeglink_DEB() {
+  dpkg -L $_packages | grep -e ".*\/lib[^\/]\+\.so" | gawk '{ printf(nlines ? "'"$_ffmpeg_list_sep"'%s" : "%s", gensub(/.*lib([^\/]+)\.so/, "\\1", "g", $0)); nlines++ }'
+}
+
+print_info_ffmpeglink_RPM() {
+  rpm -ql $_packages | grep -e ".*\/lib[^\/]\+\.so" | gawk '{ printf(nlines ? "'"$_ffmpeg_list_sep"'%s" : "%s", gensub(/.*lib([^\/]+)\.so/, "\\1", "g", $0)); nlines++ }'
+}
+
+print_info_ffmpeglink_ARCH() {
+  pacman -Ql $_packages | grep -e ".*\/lib[^\/]\+\.so$" | gawk '{ printf(nlines ? "'"$_ffmpeg_list_sep"'%s" : "%s", gensub(/.*lib([^\/]+)\.so/, "\\1", "g", $0)); nlines++ }'
+}
+
+print_info_ffmpeglink() {
+  # This func must only print a ';'-separated list of libs...
+  if [ -z "$DISTRO" ]; then
+    ERROR "Failed to detect distribution type"
+    exit 1
+  fi
+
+  # Create list of packages from which to get libs names...
+  _packages=""
+
+  if [ "$THEORA_USE" = true ]; then
+    _packages="$_packages $THEORA_DEV"
+  fi
+
+  if [ "$VORBIS_USE" = true ]; then
+    _packages="$_packages $VORBIS_DEV"
+  fi
+
+  if [ "$OGG_USE" = true ]; then
+    _packages="$_packages $OGG_DEV"
+  fi
+
+  if [ "$XVID_USE" = true ]; then
+    _packages="$_packages $XVID_DEV"
+  fi
+
+  if [ "$VPX_USE" = true ]; then
+    _packages="$_packages $VPX_DEV"
+  fi
+
+  if [ "$OPUS_USE" = true ]; then
+    _packages="$_packages $OPUS_DEV"
+  fi
+
+  if [ "$MP3LAME_USE" = true ]; then
+    _packages="$_packages $MP3LAME_DEV"
+  fi
+
+  if [ "$X264_USE" = true ]; then
+    _packages="$_packages $X264_DEV"
+  fi
+
+  if [ "$OPENJPEG_USE" = true ]; then
+    _packages="$_packages $OPENJPEG_DEV"
+  fi
+
+  if [ "$DISTRO" = "DEB" ]; then
+    print_info_ffmpeglink_DEB
+  elif [ "$DISTRO" = "RPM" ]; then
+    print_info_ffmpeglink_RPM
+  elif [ "$DISTRO" = "ARCH" ]; then
+    print_info_ffmpeglink_ARCH
+  # XXX TODO!
+  else
+    PRINT "<Could not determine additional link libraries needed for ffmpeg, replace this by valid list of libs...>"
+  fi
+}
+
 print_info() {
   PRINT ""
   PRINT ""
@@ -5731,7 +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*"
 
@@ -5932,10 +6002,12 @@ print_info() {
 
   if [ "$FFMPEG_SKIP" = false ]; then
     _1="-D WITH_CODEC_FFMPEG=ON"
+    _2="-D FFMPEG_LIBRARIES='avformat;avcodec;avutil;avdevice;swscale;swresample;lzma;rt;`print_info_ffmpeglink`'"
     PRINT "  $_1"
-    _buildargs="$_buildargs $_1"
+    PRINT "  $_2"
+    _buildargs="$_buildargs $_1 $_2"
     if [ -d $INST/ffmpeg ]; then
-      _1="-D FFMPEG_ROOT_DIR=$INST/ffmpeg"
+      _1="-D FFMPEG=$INST/ffmpeg"
       PRINT "  $_1"
       _buildargs="$_buildargs $_1"
     fi

build_files/cmake/Modules/FindFFmpeg.cmake

@@ -33,8 +33,6 @@ if(NOT FFMPEG_FIND_COMPONENTS)
     avfilter
     avformat
     avutil
-    swscale
-    swresample
   )
 endif()
 
@@ -52,9 +50,9 @@ foreach(_component ${FFMPEG_FIND_COMPONENTS})
   string(TOUPPER ${_component} _upper_COMPONENT)
   find_library(FFMPEG_${_upper_COMPONENT}_LIBRARY
     NAMES
-      ${_component}
+      ${_upper_COMPONENT}
     HINTS
-      ${_ffmpeg_SEARCH_DIRS}
+      ${LIBDIR}/ffmpeg
     PATH_SUFFIXES
       lib64 lib
   )

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/macros.cmake

@@ -488,6 +488,7 @@ function(blender_add_test_executable
 
   include_directories(${includes})
   include_directories(${includes_sys})
+  setup_libdirs()
 
   BLENDER_SRC_GTEST_EX(
     NAME ${name}
@@ -524,6 +525,83 @@ function(setup_heavy_lib_pool)
   endif()
 endfunction()
 
+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.
+  if(NOT MSVC AND NOT APPLE)
+    link_directories(${JPEG_LIBPATH} ${PNG_LIBPATH} ${ZLIB_LIBPATH} ${FREETYPE_LIBPATH})
+
+    if(WITH_PYTHON)  #  AND NOT WITH_PYTHON_MODULE  # WIN32 needs
+      link_directories(${PYTHON_LIBPATH})
+    endif()
+    if(WITH_SDL AND NOT WITH_SDL_DYNLOAD)
+      link_directories(${SDL_LIBPATH})
+    endif()
+    if(WITH_CODEC_FFMPEG)
+      link_directories(${FFMPEG_LIBPATH})
+    endif()
+    if(WITH_IMAGE_OPENEXR)
+      link_directories(${OPENEXR_LIBPATH})
+    endif()
+    if(WITH_IMAGE_TIFF)
+      link_directories(${TIFF_LIBPATH})
+    endif()
+    if(WITH_BOOST)
+      link_directories(${BOOST_LIBPATH})
+    endif()
+    if(WITH_OPENIMAGEIO)
+      link_directories(${OPENIMAGEIO_LIBPATH})
+    endif()
+    if(WITH_OPENIMAGEDENOISE)
+      link_directories(${OPENIMAGEDENOISE_LIBPATH})
+    endif()
+    if(WITH_OPENCOLORIO)
+      link_directories(${OPENCOLORIO_LIBPATH})
+    endif()
+    if(WITH_OPENVDB)
+      link_directories(${OPENVDB_LIBPATH})
+    endif()
+    if(WITH_OPENAL)
+      link_directories(${OPENAL_LIBPATH})
+    endif()
+    if(WITH_JACK AND NOT WITH_JACK_DYNLOAD)
+      link_directories(${JACK_LIBPATH})
+    endif()
+    if(WITH_PULSEAUDIO AND NOT WITH_PULSEAUDIO_DYNLOAD)
+      link_directories(${LIBPULSE_LIBPATH})
+    endif()
+    if(WITH_CODEC_SNDFILE)
+      link_directories(${LIBSNDFILE_LIBPATH})
+    endif()
+    if(WITH_FFTW3)
+      link_directories(${FFTW3_LIBPATH})
+    endif()
+    if(WITH_OPENCOLLADA)
+      link_directories(${OPENCOLLADA_LIBPATH})
+      # # Never set
+      # link_directories(${PCRE_LIBPATH})
+      # link_directories(${EXPAT_LIBPATH})
+    endif()
+    if(WITH_LLVM)
+      link_directories(${LLVM_LIBPATH})
+    endif()
+
+    if(WITH_ALEMBIC)
+      link_directories(${ALEMBIC_LIBPATH})
+    endif()
+
+    if(WITH_GMP)
+      link_directories(${GMP_LIBPATH})
+    endif()
+
+    if(WIN32 AND NOT UNIX)
+      link_directories(${PTHREADS_LIBPATH})
+    endif()
+  endif()
+endfunction()
+
 # Platform specific linker flags for targets.
 function(setup_platform_linker_flags
   target)
@@ -1214,6 +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

@@ -166,18 +166,13 @@ if(WITH_FFTW3)
   find_package(Fftw3)
 endif()
 
-# FreeType compiled with Brotli compression for woff2.
 find_package(Freetype REQUIRED)
-list(APPEND FREETYPE_LIBRARIES
-  ${LIBDIR}/brotli/lib/libbrotlicommon-static.a
-  ${LIBDIR}/brotli/lib/libbrotlidec-static.a)
 
 if(WITH_IMAGE_OPENEXR)
   find_package(OpenEXR)
 endif()
 
 if(WITH_CODEC_FFMPEG)
-  set(FFMPEG_ROOT_DIR ${LIBDIR}/ffmpeg)
   set(FFMPEG_FIND_COMPONENTS
     avcodec avdevice avformat avutil
     mp3lame ogg opus swresample swscale
@@ -484,11 +479,8 @@ string(APPEND PLATFORM_LINKFLAGS " -stdlib=libc++")
 # Suppress ranlib "has no symbols" warnings (workaround for T48250)
 set(CMAKE_C_ARCHIVE_CREATE   "<CMAKE_AR> Scr <TARGET> <LINK_FLAGS> <OBJECTS>")
 set(CMAKE_CXX_ARCHIVE_CREATE "<CMAKE_AR> Scr <TARGET> <LINK_FLAGS> <OBJECTS>")
-# llvm-ranlib doesn't support this flag. Xcode's libtool does.
-if(NOT ${CMAKE_RANLIB} MATCHES ".*llvm-ranlib$")
-  set(CMAKE_C_ARCHIVE_FINISH   "<CMAKE_RANLIB> -no_warning_for_no_symbols -c <TARGET>")
-  set(CMAKE_CXX_ARCHIVE_FINISH "<CMAKE_RANLIB> -no_warning_for_no_symbols -c <TARGET>")
-endif()
+set(CMAKE_C_ARCHIVE_FINISH   "<CMAKE_RANLIB> -no_warning_for_no_symbols -c <TARGET>")
+set(CMAKE_CXX_ARCHIVE_FINISH "<CMAKE_RANLIB> -no_warning_for_no_symbols -c <TARGET>")
 
 if(WITH_COMPILER_CCACHE)
   if(NOT CMAKE_GENERATOR STREQUAL "Xcode")
@@ -515,6 +507,3 @@ list(APPEND CMAKE_BUILD_RPATH "${OpenMP_LIBRARY_DIR}")
 
 set(CMAKE_SKIP_INSTALL_RPATH FALSE)
 list(APPEND CMAKE_INSTALL_RPATH "@loader_path/../Resources/${BLENDER_VERSION}/lib")
-
-# Same as `CFBundleIdentifier` in Info.plist.
-set(CMAKE_XCODE_ATTRIBUTE_PRODUCT_BUNDLE_IDENTIFIER "org.blenderfoundation.blender")

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)
@@ -178,30 +178,26 @@ endif()
 
 if(WITH_CODEC_FFMPEG)
   if(EXISTS ${LIBDIR})
-    set(FFMPEG_ROOT_DIR ${LIBDIR}/ffmpeg)
-    # Override FFMPEG components to also include static library dependencies
-    # included with precompiled libraries, and to ensure correct link order.
-    set(FFMPEG_FIND_COMPONENTS
-      avformat avcodec avdevice avutil swresample swscale
-      sndfile
-      FLAC
-      mp3lame
-      opus
-      theora theoradec theoraenc
-      vorbis vorbisenc vorbisfile ogg
-      vpx
-      x264
-      xvidcore)
-  elseif(FFMPEG)
-    # Old cache variable used for root dir, convert to new standard.
-    set(FFMPEG_ROOT_DIR ${FFMPEG})
+    # For precompiled lib directory, all ffmpeg dependencies are in the same folder
+    file(GLOB ffmpeg_libs ${LIBDIR}/ffmpeg/lib/*.a ${LIBDIR}/sndfile/lib/*.a)
+    set(FFMPEG ${LIBDIR}/ffmpeg CACHE PATH "FFMPEG Directory")
+    set(FFMPEG_LIBRARIES ${ffmpeg_libs} ${ffmpeg_libs} CACHE STRING "FFMPEG Libraries")
+  else()
+    set(FFMPEG /usr CACHE PATH "FFMPEG Directory")
+    set(FFMPEG_LIBRARIES avformat avcodec avutil avdevice swscale CACHE STRING "FFMPEG Libraries")
   endif()
-  find_package(FFmpeg)
 
-  if(NOT FFMPEG_FOUND)
-    set(WITH_CODEC_FFMPEG OFF)
-    message(STATUS "FFmpeg not found, disabling it")
+  mark_as_advanced(FFMPEG)
+
+  # lame, but until we have proper find module for ffmpeg
+  set(FFMPEG_INCLUDE_DIRS ${FFMPEG}/include)
+  if(EXISTS "${FFMPEG}/include/ffmpeg/")
+    list(APPEND FFMPEG_INCLUDE_DIRS "${FFMPEG}/include/ffmpeg")
   endif()
+  # end lameness
+
+  mark_as_advanced(FFMPEG_LIBRARIES)
+  set(FFMPEG_LIBPATH ${FFMPEG}/lib)
 endif()
 
 if(WITH_FFTW3)
@@ -648,9 +644,6 @@ endif()
 # ----------------------------------------------------------------------------
 # Compilers
 
-# Only set the linker once.
-set(_IS_LINKER_DEFAULT ON)
-
 # GNU Compiler
 if(CMAKE_COMPILER_IS_GNUCC)
   # ffp-contract=off:
@@ -669,89 +662,26 @@ if(CMAKE_COMPILER_IS_GNUCC)
   string(PREPEND CMAKE_CXX_FLAGS_RELWITHDEBINFO "${GCC_EXTRA_FLAGS_RELEASE} ")
   unset(GCC_EXTRA_FLAGS_RELEASE)
 
-  # NOTE(@campbellbarton): Eventually mold will be able to use `-fuse-ld=mold`,
-  # however at the moment this only works for GCC 12.1+ (unreleased at time of writing).
-  # So a workaround is used here "-B" which points to another path to find system commands
-  # such as `ld`.
-  if(WITH_LINKER_MOLD AND _IS_LINKER_DEFAULT)
-    find_program(MOLD_BIN "mold")
-    mark_as_advanced(MOLD_BIN)
-    if(NOT MOLD_BIN)
-      message(STATUS "The \"mold\" binary could not be found, using system linker.")
-      set(WITH_LINKER_MOLD OFF)
-    else()
-      # By default mold installs the binary to:
-      # - `{PREFIX}/bin/mold` as well as a symbolic-link in...
-      # - `{PREFIX}/lib/mold/ld`.
-      # (where `PREFIX` is typically `/usr/`).
-      #
-      # This block of code finds `{PREFIX}/lib/mold` from the `mold` binary.
-      # Other methods of searching for the path could also be made to work,
-      # we could even make our own directory and symbolic-link, however it's more
-      # convenient to use the one provided by mold.
-      #
-      # Use the binary path to "mold", to find the common prefix which contains "lib/mold".
-      # The parent directory: e.g. `/usr/bin/mold` -> `/usr/bin/`.
-      get_filename_component(MOLD_PREFIX "${MOLD_BIN}" DIRECTORY)
-      # The common prefix path: e.g. `/usr/bin/` -> `/usr/` to use as a hint.
-      get_filename_component(MOLD_PREFIX "${MOLD_PREFIX}" DIRECTORY)
-      # Find `{PREFIX}/lib/mold/ld`, store the directory component (without the `ld`).
-      # Then pass `-B {PREFIX}/lib/mold` to GCC so the `ld` located there overrides the default.
-      find_path(
-        MOLD_BIN_DIR "ld"
-        HINTS "${MOLD_PREFIX}"
-        # The default path is `libexec`, Arch Linux for e.g.
-        # replaces this with `lib` so check both.
-        PATH_SUFFIXES "libexec/mold" "lib/mold" "lib64/mold"
-        NO_DEFAULT_PATH
-        NO_CACHE
-      )
-      if(NOT MOLD_BIN_DIR)
-        message(STATUS
-          "The mold linker could not find the directory containing the linker command "
-          "(typically "
-          "\"${MOLD_PREFIX}/libexec/mold/ld\") or "
-          "\"${MOLD_PREFIX}/lib/mold/ld\") using system linker.")
-        set(WITH_LINKER_MOLD OFF)
-      endif()
-      unset(MOLD_PREFIX)
-    endif()
-
-    if(WITH_LINKER_MOLD)
-      # GCC will search for `ld` in this directory first.
-      string(APPEND CMAKE_EXE_LINKER_FLAGS    " -B \"${MOLD_BIN_DIR}\"")
-      string(APPEND CMAKE_SHARED_LINKER_FLAGS " -B \"${MOLD_BIN_DIR}\"")
-      string(APPEND CMAKE_MODULE_LINKER_FLAGS " -B \"${MOLD_BIN_DIR}\"")
-      set(_IS_LINKER_DEFAULT OFF)
-    endif()
-    unset(MOLD_BIN)
-    unset(MOLD_BIN_DIR)
-  endif()
-
-  if(WITH_LINKER_GOLD AND _IS_LINKER_DEFAULT)
+  if(WITH_LINKER_GOLD)
     execute_process(
       COMMAND ${CMAKE_C_COMPILER} -fuse-ld=gold -Wl,--version
       ERROR_QUIET OUTPUT_VARIABLE LD_VERSION)
     if("${LD_VERSION}" MATCHES "GNU gold")
-      string(APPEND CMAKE_EXE_LINKER_FLAGS    " -fuse-ld=gold")
-      string(APPEND CMAKE_SHARED_LINKER_FLAGS " -fuse-ld=gold")
-      string(APPEND CMAKE_MODULE_LINKER_FLAGS " -fuse-ld=gold")
-      set(_IS_LINKER_DEFAULT OFF)
+      string(APPEND CMAKE_C_FLAGS " -fuse-ld=gold")
+      string(APPEND CMAKE_CXX_FLAGS " -fuse-ld=gold")
     else()
       message(STATUS "GNU gold linker isn't available, using the default system linker.")
     endif()
     unset(LD_VERSION)
   endif()
 
-  if(WITH_LINKER_LLD AND _IS_LINKER_DEFAULT)
+  if(WITH_LINKER_LLD)
     execute_process(
       COMMAND ${CMAKE_C_COMPILER} -fuse-ld=lld -Wl,--version
       ERROR_QUIET OUTPUT_VARIABLE LD_VERSION)
     if("${LD_VERSION}" MATCHES "LLD")
-      string(APPEND CMAKE_EXE_LINKER_FLAGS    " -fuse-ld=lld")
-      string(APPEND CMAKE_SHARED_LINKER_FLAGS " -fuse-ld=lld")
-      string(APPEND CMAKE_MODULE_LINKER_FLAGS " -fuse-ld=lld")
-      set(_IS_LINKER_DEFAULT OFF)
+      string(APPEND CMAKE_C_FLAGS " -fuse-ld=lld")
+      string(APPEND CMAKE_CXX_FLAGS " -fuse-ld=lld")
     else()
       message(STATUS "LLD linker isn't available, using the default system linker.")
     endif()
@@ -761,28 +691,6 @@ if(CMAKE_COMPILER_IS_GNUCC)
 # CLang is the same as GCC for now.
 elseif(CMAKE_C_COMPILER_ID MATCHES "Clang")
   set(PLATFORM_CFLAGS "-pipe -fPIC -funsigned-char -fno-strict-aliasing")
-
-  if(WITH_LINKER_MOLD AND _IS_LINKER_DEFAULT)
-    find_program(MOLD_BIN "mold")
-    mark_as_advanced(MOLD_BIN)
-    if(NOT MOLD_BIN)
-      message(STATUS "The \"mold\" binary could not be found, using system linker.")
-      set(WITH_LINKER_MOLD OFF)
-    else()
-      if(CMAKE_C_COMPILER_VERSION VERSION_GREATER_EQUAL 12.0)
-        string(APPEND CMAKE_EXE_LINKER_FLAGS    " --ld-path=\"${MOLD_BIN}\"")
-        string(APPEND CMAKE_SHARED_LINKER_FLAGS " --ld-path=\"${MOLD_BIN}\"")
-        string(APPEND CMAKE_MODULE_LINKER_FLAGS " --ld-path=\"${MOLD_BIN}\"")
-      else()
-        string(APPEND CMAKE_EXE_LINKER_FLAGS    " -fuse-ld=\"${MOLD_BIN}\"")
-        string(APPEND CMAKE_SHARED_LINKER_FLAGS " -fuse-ld=\"${MOLD_BIN}\"")
-        string(APPEND CMAKE_MODULE_LINKER_FLAGS " -fuse-ld=\"${MOLD_BIN}\"")
-      endif()
-      set(_IS_LINKER_DEFAULT OFF)
-    endif()
-    unset(MOLD_BIN)
-  endif()
-
 # Intel C++ Compiler
 elseif(CMAKE_C_COMPILER_ID MATCHES "Intel")
   # think these next two are broken
@@ -806,8 +714,6 @@ elseif(CMAKE_C_COMPILER_ID MATCHES "Intel")
   string(APPEND PLATFORM_LINKFLAGS " -static-intel")
 endif()
 
-unset(_IS_LINKER_DEFAULT)
-
 # Avoid conflicts with Mesa llvmpipe, Luxrender, and other plug-ins that may
 # use the same libraries as Blender with a different version or build options.
 set(PLATFORM_LINKFLAGS

build_files/cmake/platform/platform_win32.cmake

@@ -347,11 +347,7 @@ set(FREETYPE_INCLUDE_DIRS
   ${LIBDIR}/freetype/include
   ${LIBDIR}/freetype/include/freetype2
 )
-set(FREETYPE_LIBRARIES
-  ${LIBDIR}/freetype/lib/freetype2ST.lib
-  ${LIBDIR}/brotli/lib/brotlidec-static.lib
-  ${LIBDIR}/brotli/lib/brotlicommon-static.lib
-)
+set(FREETYPE_LIBRARY ${LIBDIR}/freetype/lib/freetype2ST.lib)
 windows_find_package(freetype REQUIRED)
 
 if(WITH_FFTW3)

doc/doxygen/doxygen.intern.h

@@ -51,6 +51,9 @@
 /** \defgroup intern_mikktspace MikktSpace
  *  \ingroup intern */
 
+/** \defgroup intern_numaapi NUMA (Non Uniform Memory Architecture)
+ *  \ingroup intern */
+
 /** \defgroup intern_rigidbody Rigid-Body C-API
  *  \ingroup intern */
 

doc/python_api/examples/bpy.types.Bone.convert_local_to_pose.py

@@ -8,42 +8,27 @@ def set_pose_matrices(obj, matrix_map):
     "Assign pose space matrices of all bones at once, ignoring constraints."
 
     def rec(pbone, parent_matrix):
-        if pbone.name in matrix_map:
-            matrix = matrix_map[pbone.name]
+        matrix = matrix_map[pbone.name]
 
-            ## Instead of:
-            # pbone.matrix = matrix
-            # bpy.context.view_layer.update()
+        ## Instead of:
+        # pbone.matrix = matrix
+        # bpy.context.view_layer.update()
 
-            # Compute and assign local matrix, using the new parent matrix
-            if pbone.parent:
-                pbone.matrix_basis = pbone.bone.convert_local_to_pose(
-                    matrix,
-                    pbone.bone.matrix_local,
-                    parent_matrix=parent_matrix,
-                    parent_matrix_local=pbone.parent.bone.matrix_local,
-                    invert=True
-                )
-            else:
-                pbone.matrix_basis = pbone.bone.convert_local_to_pose(
-                    matrix,
-                    pbone.bone.matrix_local,
-                    invert=True
-                )
+        # Compute and assign local matrix, using the new parent matrix
+        if pbone.parent:
+            pbone.matrix_basis = pbone.bone.convert_local_to_pose(
+                matrix,
+                pbone.bone.matrix_local,
+                parent_matrix=parent_matrix,
+                parent_matrix_local=pbone.parent.bone.matrix_local,
+                invert=True
+            )
         else:
-            # Compute the updated pose matrix from local and new parent matrix
-            if pbone.parent:
-                matrix = pbone.bone.convert_local_to_pose(
-                    pbone.matrix_basis,
-                    pbone.bone.matrix_local,
-                    parent_matrix=parent_matrix,
-                    parent_matrix_local=pbone.parent.bone.matrix_local,
-                )
-            else:
-                matrix = pbone.bone.convert_local_to_pose(
-                    pbone.matrix_basis,
-                    pbone.bone.matrix_local,
-                )
+            pbone.matrix_basis = pbone.bone.convert_local_to_pose(
+                matrix,
+                pbone.bone.matrix_local,
+                invert=True
+            )
 
         # Recursively process children, passing the new matrix through
         for child in pbone.children:

doc/python_api/sphinx_doc_gen.py

@@ -1762,7 +1762,6 @@ except ModuleNotFoundError:
     fw("html_show_sphinx = False\n")
     fw("html_baseurl = 'https://docs.blender.org/api/current/'\n")
     fw("html_use_opensearch = 'https://docs.blender.org/api/current'\n")
-    fw("html_show_search_summary = True\n")
     fw("html_split_index = True\n")
     fw("html_static_path = ['static']\n")
     fw("html_extra_path = ['static/favicon.ico', 'static/blender_logo.svg']\n")

extern/audaspace/CMakeLists.txt

@@ -1092,12 +1092,12 @@ if(WITH_PYTHON)
 		configure_file(${PYTHON_SOURCE_DIRECTORY}/setup.py.in ${CMAKE_CURRENT_BINARY_DIR}/setup.py ESCAPE_QUOTES @ONLY)
 
 		if(APPLE)
-			add_custom_command(OUTPUT build COMMAND MACOSX_DEPLOYMENT_TARGET=${CMAKE_OSX_DEPLOYMENT_TARGET} ${PYTHON_EXECUTABLE} setup.py build DEPENDS ${PYTHON_SRC} ${PYTHON_HDR} setup.py)
+			add_custom_command(OUTPUT build COMMAND MACOSX_DEPLOYMENT_TARGET=${CMAKE_OSX_DEPLOYMENT_TARGET} ${PYTHON_EXECUTABLE} setup.py build DEPENDS ${PYTHON_SRC} ${PYTHON_HDR})
 		elseif(WIN32)
 			set(ENV{VS100COMNTOOLS} $ENV{VS120COMNTOOLS})
-			add_custom_command(OUTPUT build COMMAND ${PYTHON_EXECUTABLE} setup.py build DEPENDS ${PYTHON_SRC} ${PYTHON_HDR} setup.py)
+			add_custom_command(OUTPUT build COMMAND ${PYTHON_EXECUTABLE} setup.py build DEPENDS ${PYTHON_SRC} ${PYTHON_HDR})
 		else()
-			add_custom_command(OUTPUT build COMMAND ${PYTHON_EXECUTABLE} setup.py build DEPENDS ${PYTHON_SRC} ${PYTHON_HDR} setup.py)
+			add_custom_command(OUTPUT build COMMAND ${PYTHON_EXECUTABLE} setup.py build DEPENDS ${PYTHON_SRC} ${PYTHON_HDR})
 		endif()
 		add_custom_target(pythonmodule ALL DEPENDS build SOURCES ${PYTHON_SOURCE_DIRECTORY}/setup.py.in ${PYTHON_SRC} ${PYTHON_HDR})
 		add_dependencies(pythonmodule audaspace)

extern/audaspace/bindings/python/setup.py.in

@@ -8,20 +8,20 @@ import numpy
 from distutils.core import setup, Extension
 
 if len(sys.argv) > 2 and sys.argv[1] == '--build-docs':
-    import subprocess
-    from distutils.core import Distribution
-    from distutils.command.build import build
+	import subprocess
+	from distutils.core import Distribution
+	from distutils.command.build import build
 
-    dist = Distribution()
-    cmd = build(dist)
-    cmd.finalize_options()
-    #print(cmd.build_platlib)
+	dist = Distribution()
+	cmd = build(dist)
+	cmd.finalize_options()
+	#print(cmd.build_platlib)
 
-    os.environ['PYTHONPATH'] = os.path.join(os.getcwd(), cmd.build_platlib)
-    os.environ['LD_LIBRARY_PATH'] = os.getcwd()
+	os.environ['PYTHONPATH'] = os.path.join(os.getcwd(), cmd.build_platlib)
+	os.environ['LD_LIBRARY_PATH'] = os.getcwd()
 
-    ret = subprocess.call(sys.argv[2:])
-    sys.exit(ret)
+	ret = subprocess.call(sys.argv[2:])
+	sys.exit(ret)
 
 
 # the following line is not working due to https://bugs.python.org/issue9023
@@ -43,8 +43,7 @@ audaspace = Extension(
                       library_dirs = ['.', 'Release', 'Debug'],
                       language = 'c++',
                       extra_compile_args = extra_args,
-                      define_macros = [('WITH_CONVOLUTION', None)] if '@WITH_FFTW@' == 'ON' else [],
-                      sources = [os.path.join(source_directory, file) for file in ['PyAPI.cpp', 'PyDevice.cpp', 'PyHandle.cpp', 'PySound.cpp', 'PySequenceEntry.cpp', 'PySequence.cpp', 'PyPlaybackManager.cpp', 'PyDynamicMusic.cpp', 'PyThreadPool.cpp', 'PySource.cpp'] + (['PyImpulseResponse.cpp', 'PyHRTF.cpp'] if '@WITH_FFTW@' == 'ON' else [])]
+					  sources = [os.path.join(source_directory, file) for file in ['PyAPI.cpp', 'PyDevice.cpp', 'PyHandle.cpp', 'PySound.cpp', 'PySequenceEntry.cpp', 'PySequence.cpp', 'PyPlaybackManager.cpp', 'PyDynamicMusic.cpp', 'PyThreadPool.cpp', 'PySource.cpp'] + (['PyImpulseResponse.cpp', 'PyHRTF.cpp'] if '@WITH_FFTW@' == 'ON' else [])]
 )
 
 setup(
@@ -57,6 +56,6 @@ setup(
       license = 'Apache License 2.0',
       long_description = codecs.open(os.path.join(source_directory, '../../README.md'), 'r', 'utf-8').read(),
       ext_modules = [audaspace],
-      headers = [os.path.join(source_directory, file) for file in ['PyAPI.h', 'PyDevice.h', 'PyHandle.h', 'PySound.h', 'PySequenceEntry.h', 'PySequence.h', 'PyPlaybackManager.h', 'PyDynamicMusic.h', 'PyThreadPool.h', 'PySource.h'] + (['PyImpulseResponse.h', 'PyHRTF.h'] if '@WITH_FFTW@' == 'ON' else [])] + ['Audaspace.h']
+	  headers = [os.path.join(source_directory, file) for file in ['PyAPI.h', 'PyDevice.h', 'PyHandle.h', 'PySound.h', 'PySequenceEntry.h', 'PySequence.h', 'PyPlaybackManager.h', 'PyDynamicMusic.h', 'PyThreadPool.h', 'PySource.h'] + (['PyImpulseResponse.h', 'PyHRTF.h'] if '@WITH_FFTW@' == 'ON' else [])] + ['Audaspace.h']
 )
 

extern/audaspace/plugins/wasapi/WASAPIDevice.cpp

@@ -95,13 +95,6 @@ void WASAPIDevice::runMixingThread()
 				sleep_duration = std::chrono::milliseconds(buffer_size * 1000 / int(m_specs.rate) / 2);
 			}
 
-			if(m_default_device_changed)
-			{
-				m_default_device_changed = false;
-				result = AUDCLNT_E_DEVICE_INVALIDATED;
-				goto stop_thread;
-			}
-
 			if(FAILED(result = m_audio_client->GetCurrentPadding(&padding)))
 				goto stop_thread;
 
@@ -303,78 +296,13 @@ bool WASAPIDevice::setupDevice(DeviceSpecs &specs)
 	return true;
 }
 
-ULONG WASAPIDevice::AddRef()
-{
-	return InterlockedIncrement(&m_reference_count);
-}
-
-ULONG WASAPIDevice::Release()
-{
-	ULONG reference_count = InterlockedDecrement(&m_reference_count);
-
-	if(0 == reference_count)
-		delete this;
-
-	return reference_count;
-}
-
-HRESULT WASAPIDevice::QueryInterface(REFIID riid, void **ppvObject)
-{
-	if(riid == __uuidof(IMMNotificationClient))
-	{
-		*ppvObject = reinterpret_cast<IMMNotificationClient*>(this);
-		AddRef();
-	}
-	else if(riid == IID_IUnknown)
-	{
-		*ppvObject = reinterpret_cast<IUnknown*>(this);
-		AddRef();
-	}
-	else
-	{
-		*ppvObject = nullptr;
-		return E_NOINTERFACE;
-	}
-
-	return S_OK;
-}
-
-HRESULT WASAPIDevice::OnDeviceStateChanged(LPCWSTR pwstrDeviceId, DWORD dwNewState)
-{
-	return S_OK;
-}
-
-HRESULT WASAPIDevice::OnDeviceAdded(LPCWSTR pwstrDeviceId)
-{
-	return S_OK;
-}
-
-HRESULT WASAPIDevice::OnDeviceRemoved(LPCWSTR pwstrDeviceId)
-{
-	return S_OK;
-}
-
-HRESULT WASAPIDevice::OnDefaultDeviceChanged(EDataFlow flow, ERole role, LPCWSTR pwstrDeviceId)
-{
-	if(flow != EDataFlow::eCapture)
-		m_default_device_changed = true;
-
-	return S_OK;
-}
-
-HRESULT WASAPIDevice::OnPropertyValueChanged(LPCWSTR pwstrDeviceId, const PROPERTYKEY key)
-{
-	return S_OK;
-}
-
 WASAPIDevice::WASAPIDevice(DeviceSpecs specs, int buffersize) :
 	m_buffersize(buffersize),
 	m_imm_device_enumerator(nullptr),
 	m_imm_device(nullptr),
 	m_audio_client(nullptr),
-	m_wave_format_extensible({}),
-	m_default_device_changed(false),
-	m_reference_count(1)
+
+	m_wave_format_extensible({})
 {
 	// initialize COM if it hasn't happened yet
 	CoInitializeEx(nullptr, COINIT_MULTITHREADED);
@@ -399,8 +327,6 @@ WASAPIDevice::WASAPIDevice(DeviceSpecs specs, int buffersize) :
 
 	create();
 
-	m_imm_device_enumerator->RegisterEndpointNotificationCallback(this);
-
 	return;
 
 	error:
@@ -414,8 +340,6 @@ WASAPIDevice::~WASAPIDevice()
 {
 	stopMixingThread();
 
-	m_imm_device_enumerator->UnregisterEndpointNotificationCallback(this);
-
 	SafeRelease(&m_audio_client);
 	SafeRelease(&m_imm_device);
 	SafeRelease(&m_imm_device_enumerator);

extern/audaspace/plugins/wasapi/WASAPIDevice.h

@@ -40,7 +40,7 @@ AUD_NAMESPACE_BEGIN
 /**
  * This device plays back through WASAPI, the Windows audio API.
  */
-class AUD_PLUGIN_API WASAPIDevice : IMMNotificationClient, public ThreadedDevice
+class AUD_PLUGIN_API WASAPIDevice : public ThreadedDevice
 {
 private:
 	int m_buffersize;
@@ -48,8 +48,6 @@ private:
 	IMMDevice* m_imm_device;
 	IAudioClient* m_audio_client;
 	WAVEFORMATEXTENSIBLE m_wave_format_extensible;
-	bool m_default_device_changed;
-	LONG m_reference_count;
 
 	AUD_LOCAL HRESULT setupRenderClient(IAudioRenderClient*& render_client, UINT32& buffer_size);
 
@@ -60,17 +58,6 @@ private:
 
 	AUD_LOCAL bool setupDevice(DeviceSpecs& specs);
 
-	// IUnknown implementation
-	ULONG STDMETHODCALLTYPE AddRef();
-	ULONG STDMETHODCALLTYPE Release();
-	HRESULT STDMETHODCALLTYPE QueryInterface(REFIID riid, void **ppvObject);
-	// IMMNotificationClient implementation
-	HRESULT STDMETHODCALLTYPE OnDeviceStateChanged(LPCWSTR pwstrDeviceId, DWORD dwNewState);
-	HRESULT STDMETHODCALLTYPE OnDeviceAdded(LPCWSTR pwstrDeviceId);
-	HRESULT STDMETHODCALLTYPE OnDeviceRemoved(LPCWSTR pwstrDeviceId);
-	HRESULT STDMETHODCALLTYPE OnDefaultDeviceChanged(EDataFlow flow, ERole role, LPCWSTR pwstrDeviceId);
-	HRESULT STDMETHODCALLTYPE OnPropertyValueChanged(LPCWSTR pwstrDeviceId, const PROPERTYKEY key);
-
 	// delete copy constructor and operator=
 	WASAPIDevice(const WASAPIDevice&) = delete;
 	WASAPIDevice& operator=(const WASAPIDevice&) = delete;

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/cycles/app/CMakeLists.txt

@@ -51,6 +51,8 @@ list(APPEND LIBRARIES ${CYCLES_GL_LIBRARIES})
 
 # Common configuration.
 
+cycles_link_directories()
+
 add_definitions(${GL_DEFINITIONS})
 
 include_directories(${INC})

intern/cycles/app/cycles_standalone.cpp

@@ -82,7 +82,7 @@ static void session_print_status()
   string status, substatus;
 
   /* get status */
-  double progress = options.session->progress.get_progress();
+  float progress = options.session->progress.get_progress();
   options.session->progress.get_status(status, substatus);
 
   if (substatus != "")
@@ -183,7 +183,7 @@ static void display_info(Progress &progress)
 
   progress.get_time(total_time, sample_time);
   progress.get_status(status, substatus);
-  double progress_val = progress.get_progress();
+  float progress_val = progress.get_progress();
 
   if (substatus != "")
     status += ": " + substatus;

intern/cycles/blender/addon/engine.py

@@ -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

@@ -802,7 +802,7 @@ class CyclesRenderSettings(bpy.types.PropertyGroup):
         name="Tile Size",
         default=2048,
         description="",
-        min=8, max=8192,
+        min=8, max=16384,
     )
 
     # Various fine-tuning debug flags

intern/cycles/blender/addon/ui.py

@@ -119,6 +119,12 @@ def use_optix(context):
     return (get_device_type(context) == 'OPTIX' and cscene.device == 'GPU')
 
 
+def use_sample_all_lights(context):
+    cscene = context.scene.cycles
+
+    return cscene.sample_all_lights_direct or cscene.sample_all_lights_indirect
+
+
 def show_device_active(context):
     cscene = context.scene.cycles
     if cscene.device != 'GPU':
@@ -1797,45 +1803,18 @@ class CYCLES_RENDER_PT_bake_output(CyclesButtonsPanel, Panel):
         rd = scene.render
 
         if rd.use_bake_multires:
+            layout.prop(rd, "bake_margin")
             layout.prop(rd, "use_bake_clear", text="Clear Image")
+
             if rd.bake_type == 'DISPLACEMENT':
                 layout.prop(rd, "use_bake_lores_mesh")
         else:
             layout.prop(cbk, "target")
+
             if cbk.target == 'IMAGE_TEXTURES':
+                layout.prop(cbk, "margin")
                 layout.prop(cbk, "use_clear", text="Clear Image")
 
-class CYCLES_RENDER_PT_bake_output_margin(CyclesButtonsPanel, Panel):
-    bl_label = "Margin"
-    bl_context = "render"
-    bl_parent_id = "CYCLES_RENDER_PT_bake_output"
-    COMPAT_ENGINES = {'CYCLES'}
-
-    @classmethod
-    def poll(cls, context):
-        scene = context.scene
-        cbk = scene.render.bake
-        return cbk.target == 'IMAGE_TEXTURES'
-
-    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
-        cbk = scene.render.bake
-        rd = scene.render
-
-        if rd.use_bake_multires:
-            layout.prop(rd, "bake_margin_type", text="Type")
-            layout.prop(rd, "bake_margin", text="Size")
-        else:
-            if cbk.target == 'IMAGE_TEXTURES':
-                layout.prop(cbk, "margin_type", text="Type")
-                layout.prop(cbk, "margin", text="Size")
-
-
 
 class CYCLES_RENDER_PT_debug(CyclesDebugButtonsPanel, Panel):
     bl_label = "Debug"
@@ -2204,7 +2183,6 @@ classes = (
     CYCLES_RENDER_PT_bake_influence,
     CYCLES_RENDER_PT_bake_selected_to_active,
     CYCLES_RENDER_PT_bake_output,
-    CYCLES_RENDER_PT_bake_output_margin,
     CYCLES_RENDER_PT_debug,
     node_panel(CYCLES_MATERIAL_PT_settings),
     node_panel(CYCLES_MATERIAL_PT_settings_surface),

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),
-      display_shader_(BlenderDisplayShader::create(b_engine, b_scene)),
-      tiles_(make_unique<Tiles>())
+    : b_engine_(b_engine), display_shader_(BlenderDisplayShader::create(b_engine, b_scene))
 {
   /* 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)
@@ -615,33 +312,24 @@ bool BlenderDisplayDriver::update_begin(const Params &params,
     glWaitSync((GLsync)gl_render_sync_, 0, GL_TIMEOUT_IGNORED);
   }
 
-  DrawTile &current_tile = tiles_->current_tile.tile;
-  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();
+  if (!gl_texture_resources_ensure()) {
     gl_context_disable();
     return false;
   }
 
   /* Update texture dimensions if needed. */
-  if (current_tile.texture.width != texture_width ||
-      current_tile.texture.height != texture_height) {
+  if (texture_.width != texture_width || 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;
-    current_tile.texture.height = texture_height;
+    texture_.width = texture_width;
+    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.
@@ -653,58 +341,29 @@ bool BlenderDisplayDriver::update_begin(const Params &params,
    * 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_height = params.size.y;
-  if (current_tile_buffer_object.width != buffer_width ||
-      current_tile_buffer_object.height != buffer_height) {
+  const int buffer_width = params.full_size.x;
+  const int buffer_height = params.full_size.y;
+  if (texture_.buffer_width != buffer_width || texture_.buffer_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;
-    current_tile_buffer_object.height = buffer_height;
+    texture_.buffer_width = buffer_width;
+    texture_.buffer_height = buffer_height;
   }
 
-  /* Store an updated parameters of the current tile.
-   * In theory it is only needed once per update of the tile, but doing it on every update is
-   * the easiest and is not expensive. */
-  tiles_->current_tile.tile.params = params;
+  /* New content will be provided to the texture in one way or another, so mark this in a
+   * centralized place. */
+  texture_.need_update = true;
+
+  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;
-
-  DCHECK_NE(pbo_gl_id, 0);
-
-  glBindBuffer(GL_PIXEL_UNPACK_BUFFER, pbo_gl_id);
+  glBindBuffer(GL_PIXEL_UNPACK_BUFFER, texture_.gl_pbo_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_height = tiles_->current_tile.buffer_object.height;
-  interop_dst.opengl_pbo_id = tiles_->current_tile.buffer_object.gl_id;
+  interop_dst.buffer_width = texture_.buffer_width;
+  interop_dst.buffer_height = texture_.buffer_height;
+  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.
-   * Simplest is to allocate it on every redraw so that it is possible to destroy it from a
-   * correct context. */
+  glActiveTexture(GL_TEXTURE0);
+  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 = 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) {
-    draw_tile(zoom_, texcoord_attribute, position_attribute, tile);
-  }
+  glDrawArrays(GL_TRIANGLE_FAN, 0, 4);
 
-  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();
-  tiles_->finished_tiles.gl_resources_destroy_and_clear();
+  if (vertex_buffer_ != 0) {
+    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. */
-  std::atomic<bool> need_clear_ = true;
+  /* Texture which contains pixels of the render result. */
+  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. */
-  struct Tiles;
-  unique_ptr<Tiles> tiles_;
+  /* Special track of whether GPU resources were attempted to be created, to avoid attempts of
+   * their re-creation on failure on every redraw. */
+  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/mesh.cpp

@@ -1071,15 +1071,7 @@ static void create_subd_mesh(Scene *scene,
 
   for (BL::MeshEdge &e : b_mesh.edges) {
     if (e.crease() != 0.0f) {
-      mesh->add_edge_crease(e.vertices()[0], e.vertices()[1], e.crease());
-    }
-  }
-
-  for (BL::MeshVertexCreaseLayer &c : b_mesh.vertex_creases) {
-    for (int i = 0; i < c.data.length(); ++i) {
-      if (c.data[i].value() != 0.0f) {
-        mesh->add_vertex_crease(i, c.data[i].value());
-      }
+      mesh->add_crease(e.vertices()[0], e.vertices()[1], e.crease());
     }
   }
 
@@ -1094,6 +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
@@ -1118,7 +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

@@ -529,17 +529,6 @@ void BlenderSync::sync_procedural(BL::Object &b_ob,
   string absolute_path = blender_absolute_path(b_data, b_ob, b_mesh_cache.cache_file().filepath());
   procedural->set_filepath(ustring(absolute_path));
 
-  array<ustring> layers;
-  for (BL::CacheFileLayer &layer : cache_file.layers) {
-    if (layer.hide_layer()) {
-      continue;
-    }
-
-    absolute_path = blender_absolute_path(b_data, b_ob, layer.filepath());
-    layers.push_back_slow(ustring(absolute_path));
-  }
-  procedural->set_layers(layers);
-
   procedural->set_scale(cache_file.scale());
 
   procedural->set_use_prefetch(cache_file.use_prefetch());

intern/cycles/blender/output_driver.cpp

@@ -51,7 +51,7 @@ bool BlenderOutputDriver::read_render_tile(const Tile &tile)
 
   BL::RenderLayer b_rlay = *b_single_rlay;
 
-  vector<float> pixels(static_cast<size_t>(tile.size.x) * tile.size.y * 4);
+  vector<float> pixels(tile.size.x * tile.size.y * 4);
 
   /* Copy each pass.
    * TODO:copy only the required ones for better performance? */
@@ -109,7 +109,7 @@ void BlenderOutputDriver::write_render_tile(const Tile &tile)
 
   BL::RenderLayer b_rlay = *b_single_rlay;
 
-  vector<float> pixels(static_cast<size_t>(tile.size.x) * tile.size.y * 4);
+  vector<float> pixels(tile.size.x * tile.size.y * 4);
 
   /* Copy each pass. */
   for (BL::RenderPass &b_pass : b_rlay.passes) {
@@ -120,7 +120,7 @@ void BlenderOutputDriver::write_render_tile(const Tile &tile)
     b_pass.rect(&pixels[0]);
   }
 
-  b_engine_.end_result(b_rr, false, false, true);
+  b_engine_.end_result(b_rr, true, false, true);
 }
 
 CCL_NAMESPACE_END

intern/cycles/blender/pointcloud.cpp

@@ -37,52 +37,12 @@ static void fill_generic_attribute(BL::PointCloud &b_pointcloud,
   }
 }
 
-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)
+static void copy_attributes(PointCloud *pointcloud, BL::PointCloud b_pointcloud)
 {
   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;
     }
@@ -151,11 +111,7 @@ static void copy_attributes(PointCloud *pointcloud,
   }
 }
 
-static void export_pointcloud(Scene *scene,
-                              PointCloud *pointcloud,
-                              BL::PointCloud b_pointcloud,
-                              const bool need_motion,
-                              const float motion_scale)
+static void export_pointcloud(Scene *scene, PointCloud *pointcloud, BL::PointCloud b_pointcloud)
 {
   /* TODO: optimize so we can straight memcpy arrays from Blender? */
 
@@ -185,7 +141,7 @@ static void export_pointcloud(Scene *scene,
   }
 
   /* Export attributes */
-  copy_attributes(pointcloud, b_pointcloud, need_motion, motion_scale);
+  copy_attributes(pointcloud, b_pointcloud);
 }
 
 static void export_pointcloud_motion(PointCloud *pointcloud,
@@ -237,7 +193,7 @@ static void export_pointcloud_motion(PointCloud *pointcloud,
   }
 
   /* Export attributes */
-  copy_attributes(pointcloud, b_pointcloud, false, 0.0f);
+  copy_attributes(pointcloud, b_pointcloud);
 }
 
 void BlenderSync::sync_pointcloud(PointCloud *pointcloud, BObjectInfo &b_ob_info)
@@ -251,13 +207,7 @@ void BlenderSync::sync_pointcloud(PointCloud *pointcloud, BObjectInfo &b_ob_info
 
   /* 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);
+  export_pointcloud(scene, &new_pointcloud, b_pointcloud);
 
   /* update original sockets */
   for (const SocketType &socket : new_pointcloud.type->inputs) {

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);
-  params.use = true;
+  DenoiseParams params;
+  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;
 }

intern/cycles/blender/session.cpp

@@ -502,15 +502,10 @@ 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();

intern/cycles/blender/shader.cpp

@@ -776,7 +776,7 @@ 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 +813,7 @@ 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

@@ -832,14 +832,6 @@ 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);
 

intern/cycles/blender/sync.h

@@ -105,11 +105,11 @@ 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);

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.empty()) {
+      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)
@@ -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/build.cpp

@@ -935,7 +935,7 @@ BVHNode *BVHBuild::create_object_leaf_nodes(const BVHReference *ref, int start,
 
 BVHNode *BVHBuild::create_leaf_node(const BVHRange &range, const vector<BVHReference> &references)
 {
-  /* This is a bit over-allocating here (considering leaf size into account),
+  /* This is a bit overallocating here (considering leaf size into account),
    * but chunk-based re-allocation in vector makes it difficult to use small
    * size of stack storage here. Some tweaks are possible tho.
    *

intern/cycles/cmake/macros.cmake

@@ -84,6 +84,39 @@ macro(cycles_add_library target library_deps)
   cycles_set_solution_folder(${target})
 endmacro()
 
+# Cycles library dependencies common to all executables
+
+function(cycles_link_directories)
+  if(APPLE)
+    # APPLE platform uses full paths for linking libraries, and avoids link_directories.
+    return()
+  endif()
+
+  if(WITH_OPENCOLORIO)
+    link_directories(${OPENCOLORIO_LIBPATH})
+  endif()
+  if(WITH_OPENVDB)
+    link_directories(${OPENVDB_LIBPATH} ${BLOSC_LIBPATH})
+  endif()
+  if(WITH_OPENSUBDIV)
+    link_directories(${OPENSUBDIV_LIBPATH})
+  endif()
+  if(WITH_OPENIMAGEDENOISE)
+    link_directories(${OPENIMAGEDENOISE_LIBPATH})
+  endif()
+
+  link_directories(
+    ${OPENIMAGEIO_LIBPATH}
+    ${BOOST_LIBPATH}
+    ${PNG_LIBPATH}
+    ${JPEG_LIBPATH}
+    ${ZLIB_LIBPATH}
+    ${TIFF_LIBPATH}
+    ${OPENEXR_LIBPATH}
+    ${OPENJPEG_LIBPATH}
+  )
+endfunction()
+
 macro(cycles_target_link_libraries target)
   if(WITH_CYCLES_LOGGING)
     target_link_libraries(${target} ${GLOG_LIBRARIES} ${GFLAGS_LIBRARIES})
@@ -135,6 +168,12 @@ macro(cycles_target_link_libraries target)
     target_link_libraries(${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/cpu/device_impl.cpp

@@ -72,7 +72,7 @@ CPUDevice::CPUDevice(const DeviceInfo &info_, Stats &stats_, Profiler &profiler_
           << " CPU kernels.";
 
   if (info.cpu_threads == 0) {
-    info.cpu_threads = TaskScheduler::max_concurrency();
+    info.cpu_threads = TaskScheduler::num_threads();
   }
 
 #ifdef WITH_OSL

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) {
-      return;
-    }
+  if (opengl_pbo_id_ == display_interop.opengl_pbo_id && buffer_area_ == new_buffer_area) {
+    return;
   }
 
   CUDAContextScope scope(device_);

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 &&
-             temporally_stable == other.temporally_stable && prefilter == other.prefilter);
+             use_pass_normal == other.use_pass_normal && prefilter == other.prefilter);
   }
 };
 

intern/cycles/device/device.cpp

@@ -37,7 +37,6 @@
 #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"
@@ -334,7 +333,7 @@ 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::max_concurrency();
+        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

intern/cycles/device/metal/bvh.h

@@ -58,11 +58,6 @@ class BVHMetal : public BVH {
                        id<MTLCommandQueue> queue,
                        Geometry *const geom,
                        bool refit);
-  bool build_BLAS_pointcloud(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);
 };
 

intern/cycles/device/metal/bvh.mm

@@ -19,7 +19,6 @@
 #  include "scene/hair.h"
 #  include "scene/mesh.h"
 #  include "scene/object.h"
-#  include "scene/pointcloud.h"
 
 #  include "util/progress.h"
 
@@ -476,220 +475,6 @@ bool BVHMetal::build_BLAS_hair(Progress &progress,
   return false;
 }
 
-bool BVHMetal::build_BLAS_pointcloud(Progress &progress,
-                                     id<MTLDevice> device,
-                                     id<MTLCommandQueue> queue,
-                                     Geometry *const geom,
-                                     bool refit)
-{
-  if (@available(macos 12.0, *)) {
-    /* Build BLAS for point cloud */
-    PointCloud *pointcloud = static_cast<PointCloud *>(geom);
-    if (pointcloud->num_points() == 0) {
-      return false;
-    }
-
-    /*------------------------------------------------*/
-    BVH_status("Building pointcloud BLAS | %7d points | %s",
-               (int)pointcloud->num_points(),
-               geom->name.c_str());
-    /*------------------------------------------------*/
-
-    const size_t num_points = pointcloud->get_points().size();
-    const float3 *points = pointcloud->get_points().data();
-    const float *radius = pointcloud->get_radius().data();
-
-    const bool use_fast_trace_bvh = (params.bvh_type == BVH_TYPE_STATIC);
-
-    size_t num_motion_steps = 1;
-    Attribute *motion_keys = pointcloud->attributes.find(ATTR_STD_MOTION_VERTEX_POSITION);
-    if (motion_blur && pointcloud->get_use_motion_blur() && motion_keys) {
-      num_motion_steps = pointcloud->get_motion_steps();
-    }
-
-    const size_t num_aabbs = num_motion_steps;
-
-    MTLResourceOptions storage_mode;
-    if (device.hasUnifiedMemory) {
-      storage_mode = MTLResourceStorageModeShared;
-    }
-    else {
-      storage_mode = MTLResourceStorageModeManaged;
-    }
-
-    /* Allocate a GPU buffer for the AABB data and populate it */
-    id<MTLBuffer> aabbBuf = [device
-        newBufferWithLength:num_aabbs * sizeof(MTLAxisAlignedBoundingBox)
-                    options:storage_mode];
-    MTLAxisAlignedBoundingBox *aabb_data = (MTLAxisAlignedBoundingBox *)[aabbBuf contents];
-
-    /* Get AABBs for each motion step */
-    size_t center_step = (num_motion_steps - 1) / 2;
-    for (size_t step = 0; step < num_motion_steps; ++step) {
-      /* The center step for motion vertices is not stored in the attribute */
-      if (step != center_step) {
-        size_t attr_offset = (step > center_step) ? step - 1 : step;
-        points = motion_keys->data_float3() + attr_offset * num_points;
-      }
-
-      for (size_t j = 0; j < num_points; ++j) {
-        const PointCloud::Point point = pointcloud->get_point(j);
-        BoundBox bounds = BoundBox::empty;
-        point.bounds_grow(points, radius, bounds);
-
-        const size_t index = step * num_points + j;
-        aabb_data[index].min = (MTLPackedFloat3 &)bounds.min;
-        aabb_data[index].max = (MTLPackedFloat3 &)bounds.max;
-      }
-    }
-
-    if (storage_mode == MTLResourceStorageModeManaged) {
-      [aabbBuf didModifyRange:NSMakeRange(0, aabbBuf.length)];
-    }
-
-#  if 0
-    for (size_t i=0; i<num_aabbs && i < 400; i++) {
-      MTLAxisAlignedBoundingBox& bb = aabb_data[i];
-      printf("  %d:   %.1f,%.1f,%.1f -- %.1f,%.1f,%.1f\n", int(i), bb.min.x, bb.min.y, bb.min.z, bb.max.x, bb.max.y, bb.max.z);
-    }
-#  endif
-
-    MTLAccelerationStructureGeometryDescriptor *geomDesc;
-    if (motion_blur) {
-      std::vector<MTLMotionKeyframeData *> aabb_ptrs;
-      aabb_ptrs.reserve(num_motion_steps);
-      for (size_t step = 0; step < num_motion_steps; ++step) {
-        MTLMotionKeyframeData *k = [MTLMotionKeyframeData data];
-        k.buffer = aabbBuf;
-        k.offset = step * num_points * sizeof(MTLAxisAlignedBoundingBox);
-        aabb_ptrs.push_back(k);
-      }
-
-      MTLAccelerationStructureMotionBoundingBoxGeometryDescriptor *geomDescMotion =
-          [MTLAccelerationStructureMotionBoundingBoxGeometryDescriptor descriptor];
-      geomDescMotion.boundingBoxBuffers = [NSArray arrayWithObjects:aabb_ptrs.data()
-                                                              count:aabb_ptrs.size()];
-      geomDescMotion.boundingBoxCount = num_points;
-      geomDescMotion.boundingBoxStride = sizeof(aabb_data[0]);
-      geomDescMotion.intersectionFunctionTableOffset = 2;
-
-      /* Force a single any-hit call, so shadow record-all behavior works correctly */
-      /* (Match optix behavior: unsigned int build_flags =
-       * OPTIX_GEOMETRY_FLAG_REQUIRE_SINGLE_ANYHIT_CALL;) */
-      geomDescMotion.allowDuplicateIntersectionFunctionInvocation = false;
-      geomDescMotion.opaque = true;
-      geomDesc = geomDescMotion;
-    }
-    else {
-      MTLAccelerationStructureBoundingBoxGeometryDescriptor *geomDescNoMotion =
-          [MTLAccelerationStructureBoundingBoxGeometryDescriptor descriptor];
-      geomDescNoMotion.boundingBoxBuffer = aabbBuf;
-      geomDescNoMotion.boundingBoxBufferOffset = 0;
-      geomDescNoMotion.boundingBoxCount = int(num_aabbs);
-      geomDescNoMotion.boundingBoxStride = sizeof(aabb_data[0]);
-      geomDescNoMotion.intersectionFunctionTableOffset = 2;
-
-      /* Force a single any-hit call, so shadow record-all behavior works correctly */
-      /* (Match optix behavior: unsigned int build_flags =
-       * OPTIX_GEOMETRY_FLAG_REQUIRE_SINGLE_ANYHIT_CALL;) */
-      geomDescNoMotion.allowDuplicateIntersectionFunctionInvocation = false;
-      geomDescNoMotion.opaque = true;
-      geomDesc = geomDescNoMotion;
-    }
-
-    MTLPrimitiveAccelerationStructureDescriptor *accelDesc =
-        [MTLPrimitiveAccelerationStructureDescriptor descriptor];
-    accelDesc.geometryDescriptors = @[ geomDesc ];
-
-    if (motion_blur) {
-      accelDesc.motionStartTime = 0.0f;
-      accelDesc.motionEndTime = 1.0f;
-      accelDesc.motionStartBorderMode = MTLMotionBorderModeVanish;
-      accelDesc.motionEndBorderMode = MTLMotionBorderModeVanish;
-      accelDesc.motionKeyframeCount = num_motion_steps;
-    }
-
-    if (!use_fast_trace_bvh) {
-      accelDesc.usage |= (MTLAccelerationStructureUsageRefit |
-                          MTLAccelerationStructureUsagePreferFastBuild);
-    }
-
-    MTLAccelerationStructureSizes accelSizes = [device
-        accelerationStructureSizesWithDescriptor:accelDesc];
-    id<MTLAccelerationStructure> accel_uncompressed = [device
-        newAccelerationStructureWithSize:accelSizes.accelerationStructureSize];
-    id<MTLBuffer> scratchBuf = [device newBufferWithLength:accelSizes.buildScratchBufferSize
-                                                   options:MTLResourceStorageModePrivate];
-    id<MTLBuffer> sizeBuf = [device newBufferWithLength:8 options:MTLResourceStorageModeShared];
-    id<MTLCommandBuffer> accelCommands = [queue commandBuffer];
-    id<MTLAccelerationStructureCommandEncoder> accelEnc =
-        [accelCommands accelerationStructureCommandEncoder];
-    if (refit) {
-      [accelEnc refitAccelerationStructure:accel_struct
-                                descriptor:accelDesc
-                               destination:accel_uncompressed
-                             scratchBuffer:scratchBuf
-                       scratchBufferOffset:0];
-    }
-    else {
-      [accelEnc buildAccelerationStructure:accel_uncompressed
-                                descriptor:accelDesc
-                             scratchBuffer:scratchBuf
-                       scratchBufferOffset:0];
-    }
-    if (use_fast_trace_bvh) {
-      [accelEnc writeCompactedAccelerationStructureSize:accel_uncompressed
-                                               toBuffer:sizeBuf
-                                                 offset:0
-                                           sizeDataType:MTLDataTypeULong];
-    }
-    [accelEnc endEncoding];
-    [accelCommands addCompletedHandler:^(id<MTLCommandBuffer> command_buffer) {
-      /* free temp resources */
-      [scratchBuf release];
-      [aabbBuf release];
-
-      if (use_fast_trace_bvh) {
-        /* Compact the accel structure */
-        uint64_t compressed_size = *(uint64_t *)sizeBuf.contents;
-
-        dispatch_async(dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0), ^{
-          id<MTLCommandBuffer> accelCommands = [queue commandBuffer];
-          id<MTLAccelerationStructureCommandEncoder> accelEnc =
-              [accelCommands accelerationStructureCommandEncoder];
-          id<MTLAccelerationStructure> accel = [device
-              newAccelerationStructureWithSize:compressed_size];
-          [accelEnc copyAndCompactAccelerationStructure:accel_uncompressed
-                                toAccelerationStructure:accel];
-          [accelEnc endEncoding];
-          [accelCommands addCompletedHandler:^(id<MTLCommandBuffer> command_buffer) {
-            uint64_t allocated_size = [accel allocatedSize];
-            stats.mem_alloc(allocated_size);
-            accel_struct = accel;
-            [accel_uncompressed release];
-            accel_struct_building = false;
-          }];
-          [accelCommands commit];
-        });
-      }
-      else {
-        /* set our acceleration structure to the uncompressed structure */
-        accel_struct = accel_uncompressed;
-
-        uint64_t allocated_size = [accel_struct allocatedSize];
-        stats.mem_alloc(allocated_size);
-        accel_struct_building = false;
-      }
-      [sizeBuf release];
-    }];
-
-    accel_struct_building = true;
-    [accelCommands commit];
-    return true;
-  }
-  return false;
-}
-
 bool BVHMetal::build_BLAS(Progress &progress,
                           id<MTLDevice> device,
                           id<MTLCommandQueue> queue,
@@ -706,8 +491,6 @@ bool BVHMetal::build_BLAS(Progress &progress,
         return build_BLAS_mesh(progress, device, queue, geom, refit);
       case Geometry::HAIR:
         return build_BLAS_hair(progress, device, queue, geom, refit);
-      case Geometry::POINTCLOUD:
-        return build_BLAS_pointcloud(progress, device, queue, geom, refit);
       default:
         return false;
     }

intern/cycles/device/metal/device_impl.h

@@ -115,8 +115,6 @@ class MetalDevice : public Device {
 
   void load_texture_info();
 
-  void erase_allocation(device_memory &mem);
-
   virtual bool should_use_graphics_interop() override;
 
   virtual unique_ptr<DeviceQueue> gpu_queue_create() override;

intern/cycles/device/metal/device_impl.mm

@@ -87,14 +87,17 @@ MetalDevice::MetalDevice(const DeviceInfo &info, Stats &stats, Profiler &profile
     default:
       break;
     case METAL_GPU_INTEL: {
+      use_metalrt = false;
       max_threads_per_threadgroup = 64;
       break;
     }
     case METAL_GPU_AMD: {
+      use_metalrt = false;
       max_threads_per_threadgroup = 128;
       break;
     }
     case METAL_GPU_APPLE: {
+      use_metalrt = true;
       max_threads_per_threadgroup = 512;
       break;
     }
@@ -429,25 +432,6 @@ void MetalDevice::load_texture_info()
   }
 }
 
-void MetalDevice::erase_allocation(device_memory &mem)
-{
-  stats.mem_free(mem.device_size);
-  mem.device_pointer = 0;
-  mem.device_size = 0;
-
-  auto it = metal_mem_map.find(&mem);
-  if (it != metal_mem_map.end()) {
-    MetalMem *mmem = it->second.get();
-
-    /* blank out reference to MetalMem* in the launch params (fixes crash T94736) */
-    if (mmem->pointer_index >= 0) {
-      device_ptr *pointers = (device_ptr *)&launch_params;
-      pointers[mmem->pointer_index] = 0;
-    }
-    metal_mem_map.erase(it);
-  }
-}
-
 MetalDevice::MetalMem *MetalDevice::generic_alloc(device_memory &mem)
 {
   size_t size = mem.memory_size();
@@ -577,7 +561,11 @@ void MetalDevice::generic_free(device_memory &mem)
       mmem.mtlBuffer = nil;
     }
 
-    erase_allocation(mem);
+    stats.mem_free(mem.device_size);
+    mem.device_pointer = 0;
+    mem.device_size = 0;
+
+    metal_mem_map.erase(&mem);
   }
 }
 
@@ -966,7 +954,10 @@ void MetalDevice::tex_free(device_texture &mem)
       delayed_free_list.push_back(mmem.mtlTexture);
       mmem.mtlTexture = nil;
     }
-    erase_allocation(mem);
+    stats.mem_free(mem.device_size);
+    mem.device_pointer = 0;
+    mem.device_size = 0;
+    metal_mem_map.erase(&mem);
   }
 }
 

intern/cycles/device/metal/kernel.h

@@ -36,8 +36,6 @@ enum {
   METALRT_FUNC_CURVE_RIBBON_SHADOW,
   METALRT_FUNC_CURVE_ALL,
   METALRT_FUNC_CURVE_ALL_SHADOW,
-  METALRT_FUNC_POINT,
-  METALRT_FUNC_POINT_SHADOW,
   METALRT_FUNC_NUM
 };
 

intern/cycles/device/metal/kernel.mm

@@ -358,8 +358,6 @@ bool MetalDeviceKernels::load(MetalDevice *device, int kernel_type)
           "__intersection__curve_ribbon_shadow",
           "__intersection__curve_all",
           "__intersection__curve_all_shadow",
-          "__intersection__point",
-          "__intersection__point_shadow",
       };
       assert(sizeof(function_names) / sizeof(function_names[0]) == METALRT_FUNC_NUM);
 
@@ -402,50 +400,36 @@ bool MetalDeviceKernels::load(MetalDevice *device, int kernel_type)
     NSArray *function_list = nil;
 
     if (device->use_metalrt) {
-      id<MTLFunction> curve_intersect_default = nil;
-      id<MTLFunction> curve_intersect_shadow = nil;
-      id<MTLFunction> point_intersect_default = nil;
-      id<MTLFunction> point_intersect_shadow = nil;
+      id<MTLFunction> box_intersect_default = nil;
+      id<MTLFunction> box_intersect_shadow = nil;
       if (device->kernel_features & KERNEL_FEATURE_HAIR) {
         /* Add curve intersection programs. */
         if (device->kernel_features & KERNEL_FEATURE_HAIR_THICK) {
           /* Slower programs for thick hair since that also slows down ribbons.
            * Ideally this should not be needed. */
-          curve_intersect_default = rt_intersection_funcs[kernel_type][METALRT_FUNC_CURVE_ALL];
-          curve_intersect_shadow =
-              rt_intersection_funcs[kernel_type][METALRT_FUNC_CURVE_ALL_SHADOW];
+          box_intersect_default = rt_intersection_funcs[kernel_type][METALRT_FUNC_CURVE_ALL];
+          box_intersect_shadow = rt_intersection_funcs[kernel_type][METALRT_FUNC_CURVE_ALL_SHADOW];
         }
         else {
-          curve_intersect_default = rt_intersection_funcs[kernel_type][METALRT_FUNC_CURVE_RIBBON];
-          curve_intersect_shadow =
+          box_intersect_default = rt_intersection_funcs[kernel_type][METALRT_FUNC_CURVE_RIBBON];
+          box_intersect_shadow =
               rt_intersection_funcs[kernel_type][METALRT_FUNC_CURVE_RIBBON_SHADOW];
         }
       }
-      if (device->kernel_features & KERNEL_FEATURE_POINTCLOUD) {
-        point_intersect_default = rt_intersection_funcs[kernel_type][METALRT_FUNC_POINT];
-        point_intersect_shadow = rt_intersection_funcs[kernel_type][METALRT_FUNC_POINT_SHADOW];
-      }
       table_functions[METALRT_TABLE_DEFAULT] = [NSArray
           arrayWithObjects:rt_intersection_funcs[kernel_type][METALRT_FUNC_DEFAULT_TRI],
-                           curve_intersect_default ?
-                               curve_intersect_default :
-                               rt_intersection_funcs[kernel_type][METALRT_FUNC_DEFAULT_BOX],
-                           point_intersect_default ?
-                               point_intersect_default :
+                           box_intersect_default ?
+                               box_intersect_default :
                                rt_intersection_funcs[kernel_type][METALRT_FUNC_DEFAULT_BOX],
                            nil];
       table_functions[METALRT_TABLE_SHADOW] = [NSArray
           arrayWithObjects:rt_intersection_funcs[kernel_type][METALRT_FUNC_SHADOW_TRI],
-                           curve_intersect_shadow ?
-                               curve_intersect_shadow :
-                               rt_intersection_funcs[kernel_type][METALRT_FUNC_SHADOW_BOX],
-                           point_intersect_shadow ?
-                               point_intersect_shadow :
+                           box_intersect_shadow ?
+                               box_intersect_shadow :
                                rt_intersection_funcs[kernel_type][METALRT_FUNC_SHADOW_BOX],
                            nil];
       table_functions[METALRT_TABLE_LOCAL] = [NSArray
           arrayWithObjects:rt_intersection_funcs[kernel_type][METALRT_FUNC_LOCAL_TRI],
-                           rt_intersection_funcs[kernel_type][METALRT_FUNC_LOCAL_BOX],
                            rt_intersection_funcs[kernel_type][METALRT_FUNC_LOCAL_BOX],
                            nil];
 

intern/cycles/device/optix/device_impl.cpp

@@ -566,19 +566,6 @@ class OptiXDevice::DenoiseContext {
       }
     }
 
-    if (denoise_params.temporally_stable) {
-      prev_output.device_pointer = render_buffers->buffer.device_pointer;
-
-      prev_output.offset = buffer_params.get_pass_offset(PASS_DENOISING_PREVIOUS);
-
-      prev_output.stride = buffer_params.stride;
-      prev_output.pass_stride = buffer_params.pass_stride;
-
-      num_input_passes += 1;
-      use_pass_flow = true;
-      pass_motion = buffer_params.get_pass_offset(PASS_MOTION);
-    }
-
     use_guiding_passes = (num_input_passes - 1) > 0;
 
     if (use_guiding_passes) {
@@ -587,7 +574,6 @@ class OptiXDevice::DenoiseContext {
 
         guiding_params.pass_albedo = pass_denoising_albedo;
         guiding_params.pass_normal = pass_denoising_normal;
-        guiding_params.pass_flow = pass_motion;
 
         guiding_params.stride = buffer_params.stride;
         guiding_params.pass_stride = buffer_params.pass_stride;
@@ -602,10 +588,6 @@ class OptiXDevice::DenoiseContext {
           guiding_params.pass_normal = guiding_params.pass_stride;
           guiding_params.pass_stride += 3;
         }
-        if (use_pass_flow) {
-          guiding_params.pass_flow = guiding_params.pass_stride;
-          guiding_params.pass_stride += 2;
-        }
 
         guiding_params.stride = buffer_params.width;
 
@@ -623,16 +605,6 @@ class OptiXDevice::DenoiseContext {
   RenderBuffers *render_buffers = nullptr;
   const BufferParams &buffer_params;
 
-  /* Previous output. */
-  struct {
-    device_ptr device_pointer = 0;
-
-    int offset = PASS_UNUSED;
-
-    int stride = -1;
-    int pass_stride = -1;
-  } prev_output;
-
   /* Device-side storage of the guiding passes. */
   device_only_memory<float> guiding_buffer;
 
@@ -642,7 +614,6 @@ class OptiXDevice::DenoiseContext {
     /* NOTE: Are only initialized when the corresponding guiding pass is enabled. */
     int pass_albedo = PASS_UNUSED;
     int pass_normal = PASS_UNUSED;
-    int pass_flow = PASS_UNUSED;
 
     int stride = -1;
     int pass_stride = -1;
@@ -653,7 +624,6 @@ class OptiXDevice::DenoiseContext {
   bool use_guiding_passes = false;
   bool use_pass_albedo = false;
   bool use_pass_normal = false;
-  bool use_pass_flow = false;
 
   int num_samples = 0;
 
@@ -662,7 +632,6 @@ class OptiXDevice::DenoiseContext {
   /* NOTE: Are only initialized when the corresponding guiding pass is enabled. */
   int pass_denoising_albedo = PASS_UNUSED;
   int pass_denoising_normal = PASS_UNUSED;
-  int pass_motion = PASS_UNUSED;
 
   /* For passes which don't need albedo channel for denoising we replace the actual albedo with
    * the (0.5, 0.5, 0.5). This flag indicates that the real albedo pass has been replaced with
@@ -733,7 +702,6 @@ bool OptiXDevice::denoise_filter_guiding_preprocess(DenoiseContext &context)
                              &context.guiding_params.pass_stride,
                              &context.guiding_params.pass_albedo,
                              &context.guiding_params.pass_normal,
-                             &context.guiding_params.pass_flow,
                              &context.render_buffers->buffer.device_pointer,
                              &buffer_params.offset,
                              &buffer_params.stride,
@@ -741,7 +709,6 @@ bool OptiXDevice::denoise_filter_guiding_preprocess(DenoiseContext &context)
                              &context.pass_sample_count,
                              &context.pass_denoising_albedo,
                              &context.pass_denoising_normal,
-                             &context.pass_motion,
                              &buffer_params.full_x,
                              &buffer_params.full_y,
                              &buffer_params.width,
@@ -914,8 +881,7 @@ bool OptiXDevice::denoise_create_if_needed(DenoiseContext &context)
 {
   const bool recreate_denoiser = (denoiser_.optix_denoiser == nullptr) ||
                                  (denoiser_.use_pass_albedo != context.use_pass_albedo) ||
-                                 (denoiser_.use_pass_normal != context.use_pass_normal) ||
-                                 (denoiser_.use_pass_flow != context.use_pass_flow);
+                                 (denoiser_.use_pass_normal != context.use_pass_normal);
   if (!recreate_denoiser) {
     return true;
   }
@@ -929,14 +895,8 @@ bool OptiXDevice::denoise_create_if_needed(DenoiseContext &context)
   OptixDenoiserOptions denoiser_options = {};
   denoiser_options.guideAlbedo = context.use_pass_albedo;
   denoiser_options.guideNormal = context.use_pass_normal;
-
-  OptixDenoiserModelKind model = OPTIX_DENOISER_MODEL_KIND_HDR;
-  if (context.use_pass_flow) {
-    model = OPTIX_DENOISER_MODEL_KIND_TEMPORAL;
-  }
-
   const OptixResult result = optixDenoiserCreate(
-      this->context, model, &denoiser_options, &denoiser_.optix_denoiser);
+      this->context, OPTIX_DENOISER_MODEL_KIND_HDR, &denoiser_options, &denoiser_.optix_denoiser);
 
   if (result != OPTIX_SUCCESS) {
     set_error("Failed to create OptiX denoiser");
@@ -946,7 +906,6 @@ bool OptiXDevice::denoise_create_if_needed(DenoiseContext &context)
   /* OptiX denoiser handle was created with the requested number of input passes. */
   denoiser_.use_pass_albedo = context.use_pass_albedo;
   denoiser_.use_pass_normal = context.use_pass_normal;
-  denoiser_.use_pass_flow = context.use_pass_flow;
 
   /* OptiX denoiser has been created, but it needs configuration. */
   denoiser_.is_configured = false;
@@ -1006,10 +965,8 @@ bool OptiXDevice::denoise_run(DenoiseContext &context, const DenoisePass &pass)
   OptixImage2D color_layer = {0};
   OptixImage2D albedo_layer = {0};
   OptixImage2D normal_layer = {0};
-  OptixImage2D flow_layer = {0};
 
   OptixImage2D output_layer = {0};
-  OptixImage2D prev_output_layer = {0};
 
   /* Color pass. */
   {
@@ -1025,19 +982,6 @@ bool OptiXDevice::denoise_run(DenoiseContext &context, const DenoisePass &pass)
     color_layer.format = OPTIX_PIXEL_FORMAT_FLOAT3;
   }
 
-  /* Previous output. */
-  if (context.prev_output.offset != PASS_UNUSED) {
-    const int64_t pass_stride_in_bytes = context.prev_output.pass_stride * sizeof(float);
-
-    prev_output_layer.data = context.prev_output.device_pointer +
-                             context.prev_output.offset * sizeof(float);
-    prev_output_layer.width = width;
-    prev_output_layer.height = height;
-    prev_output_layer.rowStrideInBytes = pass_stride_in_bytes * context.prev_output.stride;
-    prev_output_layer.pixelStrideInBytes = pass_stride_in_bytes;
-    prev_output_layer.format = OPTIX_PIXEL_FORMAT_FLOAT3;
-  }
-
   /* Optional albedo and color passes. */
   if (context.num_input_passes > 1) {
     const device_ptr d_guiding_buffer = context.guiding_params.device_pointer;
@@ -1061,32 +1005,21 @@ bool OptiXDevice::denoise_run(DenoiseContext &context, const DenoisePass &pass)
       normal_layer.pixelStrideInBytes = pixel_stride_in_bytes;
       normal_layer.format = OPTIX_PIXEL_FORMAT_FLOAT3;
     }
-
-    if (context.use_pass_flow) {
-      flow_layer.data = d_guiding_buffer + context.guiding_params.pass_flow * sizeof(float);
-      flow_layer.width = width;
-      flow_layer.height = height;
-      flow_layer.rowStrideInBytes = row_stride_in_bytes;
-      flow_layer.pixelStrideInBytes = pixel_stride_in_bytes;
-      flow_layer.format = OPTIX_PIXEL_FORMAT_FLOAT2;
-    }
   }
 
   /* Denoise in-place of the noisy input in the render buffers. */
   output_layer = color_layer;
 
-  OptixDenoiserGuideLayer guide_layers = {};
-  guide_layers.albedo = albedo_layer;
-  guide_layers.normal = normal_layer;
-  guide_layers.flow = flow_layer;
+  /* Finally run denoising. */
+  OptixDenoiserParams params = {}; /* All parameters are disabled/zero. */
 
   OptixDenoiserLayer image_layers = {};
   image_layers.input = color_layer;
-  image_layers.previousOutput = prev_output_layer;
   image_layers.output = output_layer;
 
-  /* Finally run denoising. */
-  OptixDenoiserParams params = {}; /* All parameters are disabled/zero. */
+  OptixDenoiserGuideLayer guide_layers = {};
+  guide_layers.albedo = albedo_layer;
+  guide_layers.normal = normal_layer;
 
   optix_assert(optixUtilDenoiserInvokeTiled(denoiser_.optix_denoiser,
                                             denoiser_.queue.stream(),

intern/cycles/device/optix/device_impl.h

@@ -104,7 +104,6 @@ class OptiXDevice : public CUDADevice {
 
     bool use_pass_albedo = false;
     bool use_pass_normal = false;
-    bool use_pass_flow = false;
   };
   Denoiser denoiser_;
 

intern/cycles/device/queue.h

@@ -19,7 +19,6 @@
 #include "device/kernel.h"
 
 #include "device/graphics_interop.h"
-#include "util/debug.h"
 #include "util/log.h"
 #include "util/map.h"
 #include "util/string.h"
@@ -43,7 +42,7 @@ struct DeviceKernelArguments {
     KERNEL_FILM_CONVERT,
   };
 
-  static const int MAX_ARGS = 18;
+  static const int MAX_ARGS = 16;
   Type types[MAX_ARGS];
   void *values[MAX_ARGS];
   size_t sizes[MAX_ARGS];
@@ -86,8 +85,6 @@ struct DeviceKernelArguments {
   }
   void add(const Type type, const void *value, size_t size)
   {
-    assert(count < MAX_ARGS);
-
     types[count] = type;
     values[count] = (void *)value;
     sizes[count] = size;

intern/cycles/integrator/pass_accessor.cpp

@@ -141,7 +141,6 @@ bool PassAccessor::get_render_tile_pixels(const RenderBuffers *render_buffers,
   const PassType type = pass_access_info_.type;
   const PassMode mode = pass_access_info_.mode;
   const PassInfo pass_info = Pass::get_info(type, pass_access_info_.include_albedo);
-  int num_written_components = pass_info.num_components;
 
   if (pass_info.num_components == 1) {
     /* Single channel passes. */
@@ -189,10 +188,8 @@ bool PassAccessor::get_render_tile_pixels(const RenderBuffers *render_buffers,
     else if ((pass_info.divide_type != PASS_NONE || pass_info.direct_type != PASS_NONE ||
               pass_info.indirect_type != PASS_NONE) &&
              mode != PassMode::DENOISED) {
-      /* RGB lighting passes that need to divide out color and/or sum direct and indirect.
-       * These can also optionally write alpha like the combined pass. */
+      /* RGB lighting passes that need to divide out color and/or sum direct and indirect. */
       get_pass_light_path(render_buffers, buffer_params, destination);
-      num_written_components = 4;
     }
     else {
       /* Passes that need no special computation, or denoised passes that already
@@ -218,7 +215,7 @@ bool PassAccessor::get_render_tile_pixels(const RenderBuffers *render_buffers,
     }
   }
 
-  pad_pixels(buffer_params, destination, num_written_components);
+  pad_pixels(buffer_params, destination, pass_info.num_components);
 
   return true;
 }

intern/cycles/integrator/path_trace.cpp

@@ -115,9 +115,7 @@ bool PathTrace::ready_to_reset()
   return false;
 }
 
-void PathTrace::reset(const BufferParams &full_params,
-                      const BufferParams &big_tile_params,
-                      const bool reset_rendering)
+void PathTrace::reset(const BufferParams &full_params, const BufferParams &big_tile_params)
 {
   if (big_tile_params_.modified(big_tile_params)) {
     big_tile_params_ = big_tile_params;
@@ -130,7 +128,7 @@ void PathTrace::reset(const BufferParams &full_params,
    * It is requires to inform about reset whenever it happens, so that the redraw state tracking is
    * properly updated. */
   if (display_) {
-    display_->reset(big_tile_params, reset_rendering);
+    display_->reset(full_params);
   }
 
   render_state_.has_denoised_result = false;
@@ -596,15 +594,6 @@ void PathTrace::draw()
   did_draw_after_reset_ |= display_->draw();
 }
 
-void PathTrace::flush_display()
-{
-  if (!display_) {
-    return;
-  }
-
-  display_->flush();
-}
-
 void PathTrace::update_display(const RenderWork &render_work)
 {
   if (!render_work.display.update) {
@@ -633,8 +622,9 @@ void PathTrace::update_display(const RenderWork &render_work)
   if (display_) {
     VLOG(3) << "Perform copy to GPUDisplay work.";
 
-    const int texture_width = render_state_.effective_big_tile_params.window_width;
-    const int texture_height = render_state_.effective_big_tile_params.window_height;
+    const int resolution_divider = render_work.resolution_divider;
+    const int texture_width = max(1, full_params_.width / resolution_divider);
+    const int texture_height = max(1, full_params_.height / resolution_divider);
     if (!display_->update_begin(texture_width, texture_height)) {
       LOG(ERROR) << "Error beginning GPUDisplay update.";
       return;

intern/cycles/integrator/path_trace.h

@@ -72,9 +72,7 @@ class PathTrace {
    * render result. */
   bool ready_to_reset();
 
-  void reset(const BufferParams &full_params,
-             const BufferParams &big_tile_params,
-             bool reset_rendering);
+  void reset(const BufferParams &full_params, const BufferParams &big_tile_params);
 
   void device_free();
 
@@ -114,9 +112,6 @@ class PathTrace {
   /* Perform drawing of the current state of the DisplayDriver. */
   void draw();
 
-  /* Flush outstanding display commands before ending the render loop. */
-  void flush_display();
-
   /* Cancel rendering process as soon as possible, without waiting for full tile to be sampled.
    * Used in cases like reset of render session.
    *

intern/cycles/integrator/path_trace_display.cpp

@@ -26,20 +26,15 @@ PathTraceDisplay::PathTraceDisplay(unique_ptr<DisplayDriver> driver) : driver_(m
 {
 }
 
-void PathTraceDisplay::reset(const BufferParams &buffer_params, const bool reset_rendering)
+void PathTraceDisplay::reset(const BufferParams &buffer_params)
 {
   thread_scoped_lock lock(mutex_);
 
-  params_.full_offset = make_int2(buffer_params.full_x + buffer_params.window_x,
-                                  buffer_params.full_y + buffer_params.window_y);
+  params_.full_offset = make_int2(buffer_params.full_x, buffer_params.full_y);
   params_.full_size = make_int2(buffer_params.full_width, buffer_params.full_height);
-  params_.size = make_int2(buffer_params.window_width, buffer_params.window_height);
+  params_.size = make_int2(buffer_params.width, buffer_params.height);
 
   texture_state_.is_outdated = true;
-
-  if (!reset_rendering) {
-    driver_->next_tile_begin();
-  }
 }
 
 void PathTraceDisplay::mark_texture_updated()
@@ -253,9 +248,4 @@ bool PathTraceDisplay::draw()
   return !is_outdated;
 }
 
-void PathTraceDisplay::flush()
-{
-  driver_->flush();
-}
-
 CCL_NAMESPACE_END

intern/cycles/integrator/path_trace_display.h

@@ -38,17 +38,14 @@ class BufferParams;
 
 class PathTraceDisplay {
  public:
-  explicit PathTraceDisplay(unique_ptr<DisplayDriver> driver);
+  PathTraceDisplay(unique_ptr<DisplayDriver> driver);
   virtual ~PathTraceDisplay() = default;
 
   /* Reset the display for the new state of render session. Is called whenever session is reset,
    * which happens on changes like viewport navigation or viewport dimension change.
    *
-   * This call will configure parameters for a changed buffer and reset the texture state.
-   *
-   * When the `reset_rendering` a complete display reset happens. When it is false reset happens
-   * for a new state of the buffer parameters which is assumed to correspond to the next tile. */
-  void reset(const BufferParams &buffer_params, bool reset_rendering);
+   * This call will configure parameters for a changed buffer and reset the texture state. */
+  void reset(const BufferParams &buffer_params);
 
   /* --------------------------------------------------------------------
    * Update procedure.
@@ -154,9 +151,6 @@ class PathTraceDisplay {
    * Returns true if this call did draw an updated state of the texture. */
   bool draw();
 
-  /* Flush outstanding display commands before ending the render loop. */
-  void flush();
-
  private:
   /* Display driver implemented by the host application. */
   unique_ptr<DisplayDriver> driver_;

intern/cycles/integrator/path_trace_work.cpp

@@ -194,10 +194,10 @@ PassAccessor::Destination PathTraceWork::get_display_destination_template(
   PassAccessor::Destination destination(film_->get_display_pass());
 
   const int2 display_texture_size = display->get_texture_size();
-  const int texture_x = effective_buffer_params_.full_x - effective_big_tile_params_.full_x +
-                        effective_buffer_params_.window_x - effective_big_tile_params_.window_x;
-  const int texture_y = effective_buffer_params_.full_y - effective_big_tile_params_.full_y +
-                        effective_buffer_params_.window_y - effective_big_tile_params_.window_y;
+  const int texture_x = effective_buffer_params_.full_x - effective_full_params_.full_x +
+                        effective_buffer_params_.window_x;
+  const int texture_y = effective_buffer_params_.full_y - effective_full_params_.full_y +
+                        effective_buffer_params_.window_y;
 
   destination.offset = texture_y * display_texture_size.x + texture_x;
   destination.stride = display_texture_size.x;

intern/cycles/integrator/path_trace_work_gpu.cpp

@@ -875,10 +875,8 @@ void PathTraceWorkGPU::copy_to_display_naive(PathTraceDisplay *display,
   const int final_width = buffers_->params.window_width;
   const int final_height = buffers_->params.window_height;
 
-  const int texture_x = full_x - effective_big_tile_params_.full_x +
-                        effective_buffer_params_.window_x - effective_big_tile_params_.window_x;
-  const int texture_y = full_y - effective_big_tile_params_.full_y +
-                        effective_buffer_params_.window_y - effective_big_tile_params_.window_y;
+  const int texture_x = full_x - effective_full_params_.full_x + effective_buffer_params_.window_x;
+  const int texture_y = full_y - effective_full_params_.full_y + effective_buffer_params_.window_y;
 
   /* Re-allocate display memory if needed, and make sure the device pointer is allocated.
    *

intern/cycles/integrator/render_scheduler.cpp

@@ -406,6 +406,9 @@ bool RenderScheduler::set_postprocess_render_work(RenderWork *render_work)
     any_scheduled = true;
   }
 
+  /* Force update. */
+  any_scheduled = true;
+
   if (any_scheduled) {
     render_work->display.update = true;
   }

intern/cycles/integrator/render_scheduler.h

@@ -283,7 +283,7 @@ class RenderScheduler {
   /* Check whether timing report about the given work need to reset accumulated average time. */
   bool work_report_reset_average(const RenderWork &render_work);
 
-  /* Check whether render time limit has been reached (or exceeded), and if so store related
+  /* CHeck whether render time limit has been reached (or exceeded), and if so store related
    * information in the state so that rendering is considered finished, and is possible to report
    * average render time information. */
   void check_time_limit_reached();

intern/cycles/kernel/bvh/local.h

@@ -148,23 +148,12 @@ ccl_device_inline
             /* intersect ray against primitive */
             for (; prim_addr < prim_addr2; prim_addr++) {
               kernel_assert(kernel_tex_fetch(__prim_type, prim_addr) == type);
-
-              /* Only intersect with matching object, for instanced objects we
-               * already know we are only intersecting the right object. */
-              if (object == OBJECT_NONE) {
-                if (kernel_tex_fetch(__prim_object, prim_addr) != local_object) {
-                  continue;
-                }
-              }
-
-              const int prim = kernel_tex_fetch(__prim_index, prim_addr);
-
               if (triangle_intersect_local(kg,
                                            local_isect,
                                            P,
                                            dir,
+                                           object,
                                            local_object,
-                                           prim,
                                            prim_addr,
                                            isect_t,
                                            lcg_state,
@@ -179,24 +168,13 @@ ccl_device_inline
             /* intersect ray against primitive */
             for (; prim_addr < prim_addr2; prim_addr++) {
               kernel_assert(kernel_tex_fetch(__prim_type, prim_addr) == type);
-
-              /* Only intersect with matching object, for instanced objects we
-               * already know we are only intersecting the right object. */
-              if (object == OBJECT_NONE) {
-                if (kernel_tex_fetch(__prim_object, prim_addr) != local_object) {
-                  continue;
-                }
-              }
-
-              const int prim = kernel_tex_fetch(__prim_index, prim_addr);
-
               if (motion_triangle_intersect_local(kg,
                                                   local_isect,
                                                   P,
                                                   dir,
                                                   ray->time,
+                                                  object,
                                                   local_object,
-                                                  prim,
                                                   prim_addr,
                                                   isect_t,
                                                   lcg_state,

intern/cycles/kernel/bvh/shadow_all.h

@@ -146,7 +146,7 @@ ccl_device_inline
           --stack_ptr;
 
           /* primitive intersection */
-          for (; prim_addr < prim_addr2; prim_addr++) {
+          while (prim_addr < prim_addr2) {
             kernel_assert((kernel_tex_fetch(__prim_type, prim_addr) & PRIMITIVE_ALL) ==
                           (type & PRIMITIVE_ALL));
             bool hit;
@@ -156,29 +156,16 @@ ccl_device_inline
              * might give a few % performance improvement */
             Intersection isect ccl_optional_struct_init;
 
-            const int prim_object = (object == OBJECT_NONE) ?
-                                        kernel_tex_fetch(__prim_object, prim_addr) :
-                                        object;
-            const int prim = kernel_tex_fetch(__prim_index, prim_addr);
-
             switch (type & PRIMITIVE_ALL) {
               case PRIMITIVE_TRIANGLE: {
                 hit = triangle_intersect(
-                    kg, &isect, P, dir, t_max_current, visibility, prim_object, prim, prim_addr);
+                    kg, &isect, P, dir, t_max_current, visibility, object, prim_addr);
                 break;
               }
 #if BVH_FEATURE(BVH_MOTION)
               case PRIMITIVE_MOTION_TRIANGLE: {
-                hit = motion_triangle_intersect(kg,
-                                                &isect,
-                                                P,
-                                                dir,
-                                                t_max_current,
-                                                ray->time,
-                                                visibility,
-                                                prim_object,
-                                                prim,
-                                                prim_addr);
+                hit = motion_triangle_intersect(
+                    kg, &isect, P, dir, t_max_current, ray->time, visibility, object, prim_addr);
                 break;
               }
 #endif
@@ -195,9 +182,20 @@ ccl_device_inline
                   }
                 }
 
+                const int curve_object = (object == OBJECT_NONE) ?
+                                             kernel_tex_fetch(__prim_object, prim_addr) :
+                                             object;
                 const int curve_type = kernel_tex_fetch(__prim_type, prim_addr);
-                hit = curve_intersect(
-                    kg, &isect, P, dir, t_max_current, prim_object, prim, ray->time, curve_type);
+                const int curve_prim = kernel_tex_fetch(__prim_index, prim_addr);
+                hit = curve_intersect(kg,
+                                      &isect,
+                                      P,
+                                      dir,
+                                      t_max_current,
+                                      curve_object,
+                                      curve_prim,
+                                      ray->time,
+                                      curve_type);
 
                 break;
               }
@@ -213,9 +211,20 @@ ccl_device_inline
                   }
                 }
 
+                const int point_object = (object == OBJECT_NONE) ?
+                                             kernel_tex_fetch(__prim_object, prim_addr) :
+                                             object;
+                const int point_prim = kernel_tex_fetch(__prim_index, prim_addr);
                 const int point_type = kernel_tex_fetch(__prim_type, prim_addr);
-                hit = point_intersect(
-                    kg, &isect, P, dir, t_max_current, prim_object, prim, ray->time, point_type);
+                hit = point_intersect(kg,
+                                      &isect,
+                                      P,
+                                      dir,
+                                      t_max_current,
+                                      point_object,
+                                      point_prim,
+                                      ray->time,
+                                      point_type);
                 break;
               }
 #endif /* BVH_FEATURE(BVH_POINTCLOUD) */
@@ -292,6 +301,8 @@ ccl_device_inline
                 integrator_state_write_shadow_isect(state, &isect, record_index);
               }
             }
+
+            prim_addr++;
           }
         }
         else {

intern/cycles/kernel/bvh/traversal.h

@@ -137,14 +137,8 @@ ccl_device_noinline bool BVH_FUNCTION_FULL_NAME(BVH)(KernelGlobals kg,
             case PRIMITIVE_TRIANGLE: {
               for (; prim_addr < prim_addr2; prim_addr++) {
                 kernel_assert(kernel_tex_fetch(__prim_type, prim_addr) == type);
-
-                const int prim_object = (object == OBJECT_NONE) ?
-                                            kernel_tex_fetch(__prim_object, prim_addr) :
-                                            object;
-                const int prim = kernel_tex_fetch(__prim_index, prim_addr);
-
                 if (triangle_intersect(
-                        kg, isect, P, dir, isect->t, visibility, prim_object, prim, prim_addr)) {
+                        kg, isect, P, dir, isect->t, visibility, object, prim_addr)) {
                   /* shadow ray early termination */
                   if (visibility & PATH_RAY_SHADOW_OPAQUE)
                     return true;
@@ -156,22 +150,8 @@ ccl_device_noinline bool BVH_FUNCTION_FULL_NAME(BVH)(KernelGlobals kg,
             case PRIMITIVE_MOTION_TRIANGLE: {
               for (; prim_addr < prim_addr2; prim_addr++) {
                 kernel_assert(kernel_tex_fetch(__prim_type, prim_addr) == type);
-
-                const int prim_object = (object == OBJECT_NONE) ?
-                                            kernel_tex_fetch(__prim_object, prim_addr) :
-                                            object;
-                const int prim = kernel_tex_fetch(__prim_index, prim_addr);
-
-                if (motion_triangle_intersect(kg,
-                                              isect,
-                                              P,
-                                              dir,
-                                              isect->t,
-                                              ray->time,
-                                              visibility,
-                                              prim_object,
-                                              prim,
-                                              prim_addr)) {
+                if (motion_triangle_intersect(
+                        kg, isect, P, dir, isect->t, ray->time, visibility, object, prim_addr)) {
                   /* shadow ray early termination */
                   if (visibility & PATH_RAY_SHADOW_OPAQUE)
                     return true;
@@ -193,14 +173,13 @@ ccl_device_noinline bool BVH_FUNCTION_FULL_NAME(BVH)(KernelGlobals kg,
                   }
                 }
 
-                const int prim_object = (object == OBJECT_NONE) ?
-                                            kernel_tex_fetch(__prim_object, prim_addr) :
-                                            object;
-                const int prim = kernel_tex_fetch(__prim_index, prim_addr);
-
+                const int curve_object = (object == OBJECT_NONE) ?
+                                             kernel_tex_fetch(__prim_object, prim_addr) :
+                                             object;
+                const int curve_prim = kernel_tex_fetch(__prim_index, prim_addr);
                 const int curve_type = kernel_tex_fetch(__prim_type, prim_addr);
                 const bool hit = curve_intersect(
-                    kg, isect, P, dir, isect->t, prim_object, prim, ray->time, curve_type);
+                    kg, isect, P, dir, isect->t, curve_object, curve_prim, ray->time, curve_type);
                 if (hit) {
                   /* shadow ray early termination */
                   if (visibility & PATH_RAY_SHADOW_OPAQUE)
@@ -221,14 +200,13 @@ ccl_device_noinline bool BVH_FUNCTION_FULL_NAME(BVH)(KernelGlobals kg,
                   }
                 }
 
-                const int prim_object = (object == OBJECT_NONE) ?
-                                            kernel_tex_fetch(__prim_object, prim_addr) :
-                                            object;
-                const int prim = kernel_tex_fetch(__prim_index, prim_addr);
-
+                const int point_object = (object == OBJECT_NONE) ?
+                                             kernel_tex_fetch(__prim_object, prim_addr) :
+                                             object;
+                const int point_prim = kernel_tex_fetch(__prim_index, prim_addr);
                 const int point_type = kernel_tex_fetch(__prim_type, prim_addr);
                 const bool hit = point_intersect(
-                    kg, isect, P, dir, isect->t, prim_object, prim, ray->time, point_type);
+                    kg, isect, P, dir, isect->t, point_object, point_prim, ray->time, point_type);
                 if (hit) {
                   /* shadow ray early termination */
                   if (visibility & PATH_RAY_SHADOW_OPAQUE)

intern/cycles/kernel/bvh/volume.h

@@ -140,17 +140,14 @@ ccl_device_inline
               for (; prim_addr < prim_addr2; prim_addr++) {
                 kernel_assert(kernel_tex_fetch(__prim_type, prim_addr) == type);
                 /* only primitives from volume object */
-                const int prim_object = (object == OBJECT_NONE) ?
-                                            kernel_tex_fetch(__prim_object, prim_addr) :
-                                            object;
-                const int prim = kernel_tex_fetch(__prim_index, prim_addr);
-
-                int object_flag = kernel_tex_fetch(__object_flag, prim_object);
+                uint tri_object = (object == OBJECT_NONE) ?
+                                      kernel_tex_fetch(__prim_object, prim_addr) :
+                                      object;
+                int object_flag = kernel_tex_fetch(__object_flag, tri_object);
                 if ((object_flag & SD_OBJECT_HAS_VOLUME) == 0) {
                   continue;
                 }
-                triangle_intersect(
-                    kg, isect, P, dir, isect->t, visibility, prim_object, prim, prim_addr);
+                triangle_intersect(kg, isect, P, dir, isect->t, visibility, object, prim_addr);
               }
               break;
             }
@@ -160,24 +157,15 @@ ccl_device_inline
               for (; prim_addr < prim_addr2; prim_addr++) {
                 kernel_assert(kernel_tex_fetch(__prim_type, prim_addr) == type);
                 /* only primitives from volume object */
-                const int prim_object = (object == OBJECT_NONE) ?
-                                            kernel_tex_fetch(__prim_object, prim_addr) :
-                                            object;
-                const int prim = kernel_tex_fetch(__prim_index, prim_addr);
-                int object_flag = kernel_tex_fetch(__object_flag, prim_object);
+                uint tri_object = (object == OBJECT_NONE) ?
+                                      kernel_tex_fetch(__prim_object, prim_addr) :
+                                      object;
+                int object_flag = kernel_tex_fetch(__object_flag, tri_object);
                 if ((object_flag & SD_OBJECT_HAS_VOLUME) == 0) {
                   continue;
                 }
-                motion_triangle_intersect(kg,
-                                          isect,
-                                          P,
-                                          dir,
-                                          isect->t,
-                                          ray->time,
-                                          visibility,
-                                          prim_object,
-                                          prim,
-                                          prim_addr);
+                motion_triangle_intersect(
+                    kg, isect, P, dir, isect->t, ray->time, visibility, object, prim_addr);
               }
               break;
             }

intern/cycles/kernel/bvh/volume_all.h

@@ -143,16 +143,15 @@ ccl_device_inline
               for (; prim_addr < prim_addr2; prim_addr++) {
                 kernel_assert(kernel_tex_fetch(__prim_type, prim_addr) == type);
                 /* only primitives from volume object */
-                const int prim_object = (object == OBJECT_NONE) ?
-                                            kernel_tex_fetch(__prim_object, prim_addr) :
-                                            object;
-                const int prim = kernel_tex_fetch(__prim_index, prim_addr);
-                int object_flag = kernel_tex_fetch(__object_flag, prim_object);
+                uint tri_object = (object == OBJECT_NONE) ?
+                                      kernel_tex_fetch(__prim_object, prim_addr) :
+                                      object;
+                int object_flag = kernel_tex_fetch(__object_flag, tri_object);
                 if ((object_flag & SD_OBJECT_HAS_VOLUME) == 0) {
                   continue;
                 }
                 hit = triangle_intersect(
-                    kg, isect_array, P, dir, isect_t, visibility, prim_object, prim, prim_addr);
+                    kg, isect_array, P, dir, isect_t, visibility, object, prim_addr);
                 if (hit) {
                   /* Move on to next entry in intersections array. */
                   isect_array++;
@@ -184,24 +183,15 @@ ccl_device_inline
               for (; prim_addr < prim_addr2; prim_addr++) {
                 kernel_assert(kernel_tex_fetch(__prim_type, prim_addr) == type);
                 /* only primitives from volume object */
-                const int prim_object = (object == OBJECT_NONE) ?
-                                            kernel_tex_fetch(__prim_object, prim_addr) :
-                                            object;
-                const int prim = kernel_tex_fetch(__prim_index, prim_addr);
-                int object_flag = kernel_tex_fetch(__object_flag, prim_object);
+                uint tri_object = (object == OBJECT_NONE) ?
+                                      kernel_tex_fetch(__prim_object, prim_addr) :
+                                      object;
+                int object_flag = kernel_tex_fetch(__object_flag, tri_object);
                 if ((object_flag & SD_OBJECT_HAS_VOLUME) == 0) {
                   continue;
                 }
-                hit = motion_triangle_intersect(kg,
-                                                isect_array,
-                                                P,
-                                                dir,
-                                                isect_t,
-                                                ray->time,
-                                                visibility,
-                                                prim_object,
-                                                prim,
-                                                prim_addr);
+                hit = motion_triangle_intersect(
+                    kg, isect_array, P, dir, isect_t, ray->time, visibility, object, prim_addr);
                 if (hit) {
                   /* Move on to next entry in intersections array. */
                   isect_array++;

intern/cycles/kernel/closure/volume.h

@@ -199,18 +199,22 @@ ccl_device int volume_sample_channel(float3 albedo,
    *  Tracing". Matt Jen-Yuan Chiang, Peter Kutz, Brent Burley. SIGGRAPH 2016. */
   float3 weights = fabs(throughput * albedo);
   float sum_weights = weights.x + weights.y + weights.z;
+  float3 weights_pdf;
 
   if (sum_weights > 0.0f) {
-    *pdf = weights / sum_weights;
+    weights_pdf = weights / sum_weights;
   }
   else {
-    *pdf = make_float3(1.0f / 3.0f, 1.0f / 3.0f, 1.0f / 3.0f);
+    weights_pdf = make_float3(1.0f / 3.0f, 1.0f / 3.0f, 1.0f / 3.0f);
   }
 
-  if (rand < pdf->x) {
+  *pdf = weights_pdf;
+
+  /* OpenCL does not support -> on float3, so don't use pdf->x. */
+  if (rand < weights_pdf.x) {
     return 0;
   }
-  else if (rand < pdf->x + pdf->y) {
+  else if (rand < weights_pdf.x + weights_pdf.y) {
     return 1;
   }
   else {

intern/cycles/kernel/device/gpu/kernel.h

@@ -243,10 +243,6 @@ ccl_gpu_kernel(GPU_KERNEL_BLOCK_NUM_THREADS, GPU_KERNEL_MAX_REGISTERS)
   }
 }
 
-#ifdef __KERNEL_METAL__
-constant int __dummy_constant [[function_constant(0)]];
-#endif
-
 ccl_gpu_kernel(GPU_KERNEL_BLOCK_NUM_THREADS, GPU_KERNEL_MAX_REGISTERS)
     ccl_gpu_kernel_signature(integrator_shade_surface_raytrace,
                              ccl_global const int *path_index_array,
@@ -257,16 +253,7 @@ ccl_gpu_kernel(GPU_KERNEL_BLOCK_NUM_THREADS, GPU_KERNEL_MAX_REGISTERS)
 
   if (global_index < work_size) {
     const int state = (path_index_array) ? path_index_array[global_index] : global_index;
-
-#ifdef __KERNEL_METAL__
-    KernelGlobals kg = NULL;
-    /* Workaround Ambient Occlusion and Bevel nodes not working with Metal.
-     * Dummy offset should not affect result, but somehow fixes bug! */
-    kg += __dummy_constant;
-    ccl_gpu_kernel_call(integrator_shade_surface_raytrace(kg, state, render_buffer));
-#else
     ccl_gpu_kernel_call(integrator_shade_surface_raytrace(NULL, state, render_buffer));
-#endif
   }
 }
 
@@ -769,7 +756,6 @@ ccl_gpu_kernel(GPU_KERNEL_BLOCK_NUM_THREADS, GPU_KERNEL_MAX_REGISTERS)
                              int guiding_pass_stride,
                              int guiding_pass_albedo,
                              int guiding_pass_normal,
-                             int guiding_pass_flow,
                              ccl_global const float *render_buffer,
                              int render_offset,
                              int render_stride,
@@ -777,7 +763,6 @@ ccl_gpu_kernel(GPU_KERNEL_BLOCK_NUM_THREADS, GPU_KERNEL_MAX_REGISTERS)
                              int render_pass_sample_count,
                              int render_pass_denoising_albedo,
                              int render_pass_denoising_normal,
-                             int render_pass_motion,
                              int full_x,
                              int full_y,
                              int width,
@@ -829,17 +814,6 @@ ccl_gpu_kernel(GPU_KERNEL_BLOCK_NUM_THREADS, GPU_KERNEL_MAX_REGISTERS)
     normal_out[1] = normal_in[1] * pixel_scale;
     normal_out[2] = normal_in[2] * pixel_scale;
   }
-
-  /* Flow pass. */
-  if (guiding_pass_flow != PASS_UNUSED) {
-    kernel_assert(render_pass_motion != PASS_UNUSED);
-
-    ccl_global const float *motion_in = buffer + render_pass_motion;
-    ccl_global float *flow_out = guiding_pixel + guiding_pass_flow;
-
-    flow_out[0] = -motion_in[0] * pixel_scale;
-    flow_out[1] = -motion_in[1] * pixel_scale;
-  }
 }
 
 ccl_gpu_kernel(GPU_KERNEL_BLOCK_NUM_THREADS, GPU_KERNEL_MAX_REGISTERS)
@@ -925,6 +899,7 @@ ccl_gpu_kernel(GPU_KERNEL_BLOCK_NUM_THREADS, GPU_KERNEL_MAX_REGISTERS)
   else {
     /* Assigning to zero since this is a default alpha value for 3-component passes, and it
      * is an opaque pixel for 4 component passes. */
+
     denoised_pixel[3] = 0;
   }
 }

intern/cycles/kernel/device/metal/compat.h

@@ -98,12 +98,8 @@ using namespace metal::raytracing;
 #define FN14(p1, p2, p3, p4, p5, p6, p7, p8, p9, p10, p11, p12, p13, p14) p1; p2; p3; p4; p5; p6; p7; p8; p9; p10; p11; p12; p13; p14;
 #define FN15(p1, p2, p3, p4, p5, p6, p7, p8, p9, p10, p11, p12, p13, p14, p15) p1; p2; p3; p4; p5; p6; p7; p8; p9; p10; p11; p12; p13; p14; p15;
 #define FN16(p1, p2, p3, p4, p5, p6, p7, p8, p9, p10, p11, p12, p13, p14, p15, p16) p1; p2; p3; p4; p5; p6; p7; p8; p9; p10; p11; p12; p13; p14; p15; p16;
-#define FN17(p1, p2, p3, p4, p5, p6, p7, p8, p9, p10, p11, p12, p13, p14, p15, p16, p17) p1; p2; p3; p4; p5; p6; p7; p8; p9; p10; p11; p12; p13; p14; p15; p16; p17;
-#define FN18(p1, p2, p3, p4, p5, p6, p7, p8, p9, p10, p11, p12, p13, p14, p15, p16, p17, p18) p1; p2; p3; p4; p5; p6; p7; p8; p9; p10; p11; p12; p13; p14; p15; p16; p17; p18;
-#define FN19(p1, p2, p3, p4, p5, p6, p7, p8, p9, p10, p11, p12, p13, p14, p15, p16, p17, p18, p19) p1; p2; p3; p4; p5; p6; p7; p8; p9; p10; p11; p12; p13; p14; p15; p16; p17; p18; p19;
-#define FN20(p1, p2, p3, p4, p5, p6, p7, p8, p9, p10, p11, p12, p13, p14, p15, p16, p17, p18, p19, p20) p1; p2; p3; p4; p5; p6; p7; p8; p9; p10; p11; p12; p13; p14; p15; p16; p17; p18; p19; p20;
-#define GET_LAST_ARG(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, p10, p11, p12, p13, p14, p15, p16, p17, p18, p19, p20, ...) p20
-#define PARAMS_MAKER(...) GET_LAST_ARG(__VA_ARGS__, FN20, FN19, FN18, FN17, FN16, FN15, FN14, FN13, FN12, FN11, FN10, FN9, FN8, FN7, FN6, FN5, FN4, FN3, FN2, FN1, FN0)
+#define GET_LAST_ARG(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, p10, p11, p12, p13, p14, p15, p16, ...) p16
+#define PARAMS_MAKER(...) GET_LAST_ARG(__VA_ARGS__, FN16, FN15, FN14, FN13, FN12, FN11, FN10, FN9, FN8, FN7, FN6, FN5, FN4, FN3, FN2, FN1, FN0)
 
 /* Generate a struct containing the entry-point parameters and a "run"
  * method which can access them implicitly via this-> */

intern/cycles/kernel/device/metal/kernel.metal

@@ -576,150 +576,6 @@ __intersection__curve_all_shadow(constant KernelParamsMetal &launch_params_metal
 
   return result;
 }
+
 #endif /* __HAIR__ */
-
-#ifdef __POINTCLOUD__
-ccl_device_inline
-void metalrt_intersection_point(constant KernelParamsMetal &launch_params_metal,
-                                ray_data MetalKernelContext::MetalRTIntersectionPayload &payload,
-                                const uint object,
-                                const uint prim,
-                                const uint type,
-                                const float3 ray_origin,
-                                const float3 ray_direction,
-                                float time,
-                                const float ray_tmax,
-                                thread BoundingBoxIntersectionResult &result)
-{
-#  ifdef __VISIBILITY_FLAG__
-  const uint visibility = payload.visibility;
-  if ((kernel_tex_fetch(__objects, object).visibility & visibility) == 0) {
-    return;
-  }
-#  endif
-
-  float3 P = ray_origin;
-  float3 dir = ray_direction;
-
-  /* The direction is not normalized by default, but the point intersection routine expects that */
-  float len;
-  dir = normalize_len(dir, &len);
-
-  Intersection isect;
-  isect.t = ray_tmax;
-  /* Transform maximum distance into object space. */
-  if (isect.t != FLT_MAX)
-    isect.t *= len;
-
-  MetalKernelContext context(launch_params_metal);
-  if (context.point_intersect(NULL, &isect, P, dir, isect.t, object, prim, time, type)) {
-    result = metalrt_visibility_test<BoundingBoxIntersectionResult, METALRT_HIT_BOUNDING_BOX>(
-                  launch_params_metal, payload, object, prim, isect.u);
-    if (result.accept) {
-      result.distance = isect.t / len;
-      payload.u = isect.u;
-      payload.v = isect.v;
-      payload.prim = prim;
-      payload.type = type;
-    }
-  }
-}
-
-ccl_device_inline
-void metalrt_intersection_point_shadow(constant KernelParamsMetal &launch_params_metal,
-                                       ray_data MetalKernelContext::MetalRTIntersectionShadowPayload &payload,
-                                       const uint object,
-                                       const uint prim,
-                                       const uint type,
-                                       const float3 ray_origin,
-                                       const float3 ray_direction,
-                                       float time,
-                                       const float ray_tmax,
-                                       thread BoundingBoxIntersectionResult &result)
-{
-  const uint visibility = payload.visibility;
-
-  float3 P = ray_origin;
-  float3 dir = ray_direction;
-
-  /* The direction is not normalized by default, but the point intersection routine expects that */
-  float len;
-  dir = normalize_len(dir, &len);
-
-  Intersection isect;
-  isect.t = ray_tmax;
-  /* Transform maximum distance into object space */
-  if (isect.t != FLT_MAX)
-    isect.t *= len;
-
-  MetalKernelContext context(launch_params_metal);
-  if (context.point_intersect(NULL, &isect, P, dir, isect.t, object, prim, time, type)) {
-    result.continue_search = metalrt_shadow_all_hit<METALRT_HIT_BOUNDING_BOX>(
-                launch_params_metal, payload, object, prim, float2(isect.u, isect.v), ray_tmax);
-    result.accept = !result.continue_search;
-
-    if (result.accept) {
-      result.distance = isect.t / len;
-    }
-  }
-}
-
-[[intersection(bounding_box, triangle_data, METALRT_TAGS)]]
-BoundingBoxIntersectionResult
-__intersection__point(constant KernelParamsMetal &launch_params_metal [[buffer(1)]],
-                             ray_data MetalKernelContext::MetalRTIntersectionPayload &payload [[payload]],
-                             const uint object [[user_instance_id]],
-                             const uint primitive_id [[primitive_id]],
-                             const float3 ray_origin [[origin]],
-                             const float3 ray_direction [[direction]],
-                             const float ray_tmax [[max_distance]])
-{
-  const uint prim = primitive_id + kernel_tex_fetch(__object_prim_offset, object);
-  const int type = kernel_tex_fetch(__objects, object).primitive_type;
-
-  BoundingBoxIntersectionResult result;
-  result.accept = false;
-  result.continue_search = true;
-  result.distance = ray_tmax;
-
-  metalrt_intersection_point(launch_params_metal, payload, object, prim, type, ray_origin, ray_direction,
-#  if defined(__METALRT_MOTION__)
-                             payload.time,
-#  else
-                             0.0f,
-#  endif
-                             ray_tmax, result);
-
-  return result;
-}
-
-[[intersection(bounding_box, triangle_data, METALRT_TAGS)]]
-BoundingBoxIntersectionResult
-__intersection__point_shadow(constant KernelParamsMetal &launch_params_metal [[buffer(1)]],
-                                    ray_data MetalKernelContext::MetalRTIntersectionShadowPayload &payload [[payload]],
-                                    const uint object [[user_instance_id]],
-                                    const uint primitive_id [[primitive_id]],
-                                    const float3 ray_origin [[origin]],
-                                    const float3 ray_direction [[direction]],
-                                    const float ray_tmax [[max_distance]])
-{
-  const uint prim = primitive_id + kernel_tex_fetch(__object_prim_offset, object);
-  const int type = kernel_tex_fetch(__objects, object).primitive_type;
-
-  BoundingBoxIntersectionResult result;
-  result.accept = false;
-  result.continue_search = true;
-  result.distance = ray_tmax;
-
-  metalrt_intersection_point_shadow(launch_params_metal, payload, object, prim, type, ray_origin, ray_direction,
-#  if defined(__METALRT_MOTION__)
-                             payload.time,
-#  else
-                             0.0f,
-#  endif
-                             ray_tmax, result);
-
-  return result;
-}
-#endif /* __POINTCLOUD__ */
 #endif /* __METALRT__ */

intern/cycles/kernel/film/passes.h

@@ -92,14 +92,6 @@ ccl_device_forceinline void kernel_write_denoising_features_surface(
     else if (sc->type == CLOSURE_BSDF_HAIR_PRINCIPLED_ID) {
       closure_albedo *= bsdf_principled_hair_albedo(sc);
     }
-    else if (sc->type == CLOSURE_BSDF_PRINCIPLED_DIFFUSE_ID) {
-      /* BSSRDF already accounts for weight, retro-reflection would double up. */
-      ccl_private const PrincipledDiffuseBsdf *bsdf = (ccl_private const PrincipledDiffuseBsdf *)
-          sc;
-      if (bsdf->components == PRINCIPLED_DIFFUSE_RETRO_REFLECTION) {
-        continue;
-      }
-    }
 
     if (bsdf_get_specular_roughness_squared(sc) > sqr(0.075f)) {
       diffuse_albedo += closure_albedo;

intern/cycles/kernel/film/read.h

@@ -214,21 +214,6 @@ ccl_device_inline void film_get_pass_pixel_light_path(
   pixel[0] = f.x;
   pixel[1] = f.y;
   pixel[2] = f.z;
-
-  /* Optional alpha channel. */
-  if (kfilm_convert->num_components >= 4) {
-    if (kfilm_convert->pass_combined != PASS_UNUSED) {
-      float scale, scale_exposure;
-      film_get_scale_and_scale_exposure(kfilm_convert, buffer, &scale, &scale_exposure);
-
-      ccl_global const float *in_combined = buffer + kfilm_convert->pass_combined;
-      const float alpha = in_combined[3] * scale;
-      pixel[3] = film_transparency_to_alpha(alpha);
-    }
-    else {
-      pixel[3] = 1.0f;
-    }
-  }
 }
 
 ccl_device_inline void film_get_pass_pixel_float3(ccl_global const KernelFilmConvert *ccl_restrict

intern/cycles/kernel/geom/motion_triangle.h

@@ -116,52 +116,6 @@ ccl_device_inline void motion_triangle_vertices(
   verts[2] = (1.0f - t) * verts[2] + t * next_verts[2];
 }
 
-ccl_device_inline void motion_triangle_vertices_and_normals(
-    KernelGlobals kg, int object, int prim, float time, float3 verts[3], float3 normals[3])
-{
-  /* get motion info */
-  int numsteps, numverts;
-  object_motion_info(kg, object, &numsteps, &numverts, NULL);
-
-  /* Figure out which steps we need to fetch and their interpolation factor. */
-  int maxstep = numsteps * 2;
-  int step = min((int)(time * maxstep), maxstep - 1);
-  float t = time * maxstep - step;
-
-  /* Find attribute. */
-  int offset = intersection_find_attribute(kg, object, ATTR_STD_MOTION_VERTEX_POSITION);
-  kernel_assert(offset != ATTR_STD_NOT_FOUND);
-
-  /* Fetch vertex coordinates. */
-  float3 next_verts[3];
-  uint4 tri_vindex = kernel_tex_fetch(__tri_vindex, prim);
-
-  motion_triangle_verts_for_step(kg, tri_vindex, offset, numverts, numsteps, step, verts);
-  motion_triangle_verts_for_step(kg, tri_vindex, offset, numverts, numsteps, step + 1, next_verts);
-
-  /* Interpolate between steps. */
-  verts[0] = (1.0f - t) * verts[0] + t * next_verts[0];
-  verts[1] = (1.0f - t) * verts[1] + t * next_verts[1];
-  verts[2] = (1.0f - t) * verts[2] + t * next_verts[2];
-
-  /* Compute smooth normal. */
-
-  /* Find attribute. */
-  offset = intersection_find_attribute(kg, object, ATTR_STD_MOTION_VERTEX_NORMAL);
-  kernel_assert(offset != ATTR_STD_NOT_FOUND);
-
-  /* Fetch vertex coordinates. */
-  float3 next_normals[3];
-  motion_triangle_normals_for_step(kg, tri_vindex, offset, numverts, numsteps, step, normals);
-  motion_triangle_normals_for_step(
-      kg, tri_vindex, offset, numverts, numsteps, step + 1, next_normals);
-
-  /* Interpolate between steps. */
-  normals[0] = (1.0f - t) * normals[0] + t * next_normals[0];
-  normals[1] = (1.0f - t) * normals[1] + t * next_normals[1];
-  normals[2] = (1.0f - t) * normals[2] + t * next_normals[2];
-}
-
 ccl_device_inline float3 motion_triangle_smooth_normal(
     KernelGlobals kg, float3 Ng, int object, int prim, float u, float v, float time)
 {

intern/cycles/kernel/geom/motion_triangle_intersect.h

@@ -153,12 +153,14 @@ ccl_device_inline bool motion_triangle_intersect(KernelGlobals kg,
                                                  float time,
                                                  uint visibility,
                                                  int object,
-                                                 int prim,
                                                  int prim_addr)
 {
+  /* Primitive index for vertex location lookup. */
+  int prim = kernel_tex_fetch(__prim_index, prim_addr);
+  int fobject = (object == OBJECT_NONE) ? kernel_tex_fetch(__prim_object, prim_addr) : object;
   /* Get vertex locations for intersection. */
   float3 verts[3];
-  motion_triangle_vertices(kg, object, prim, time, verts);
+  motion_triangle_vertices(kg, fobject, prim, time, verts);
   /* Ray-triangle intersection, unoptimized. */
   float t, u, v;
   if (ray_triangle_intersect(P, dir, tmax, verts[0], verts[1], verts[2], &u, &v, &t)) {
@@ -173,7 +175,8 @@ ccl_device_inline bool motion_triangle_intersect(KernelGlobals kg,
       isect->u = u;
       isect->v = v;
       isect->prim = prim;
-      isect->object = object;
+      isect->object = (object == OBJECT_NONE) ? kernel_tex_fetch(__prim_object, prim_addr) :
+                                                object;
       isect->type = PRIMITIVE_MOTION_TRIANGLE;
       return true;
     }
@@ -193,15 +196,25 @@ ccl_device_inline bool motion_triangle_intersect_local(KernelGlobals kg,
                                                        float3 dir,
                                                        float time,
                                                        int object,
-                                                       int prim,
+                                                       int local_object,
                                                        int prim_addr,
                                                        float tmax,
                                                        ccl_private uint *lcg_state,
                                                        int max_hits)
 {
+  /* Only intersect with matching object, for instanced objects we
+   * already know we are only intersecting the right object. */
+  if (object == OBJECT_NONE) {
+    if (kernel_tex_fetch(__prim_object, prim_addr) != local_object) {
+      return false;
+    }
+  }
+
+  /* Primitive index for vertex location lookup. */
+  int prim = kernel_tex_fetch(__prim_index, prim_addr);
   /* Get vertex locations for intersection. */
   float3 verts[3];
-  motion_triangle_vertices(kg, object, prim, time, verts);
+  motion_triangle_vertices(kg, local_object, prim, time, verts);
   /* Ray-triangle intersection, unoptimized. */
   float t, u, v;
   if (!ray_triangle_intersect(P, dir, tmax, verts[0], verts[1], verts[2], &u, &v, &t)) {
@@ -253,7 +266,7 @@ ccl_device_inline bool motion_triangle_intersect_local(KernelGlobals kg,
   isect->u = u;
   isect->v = v;
   isect->prim = prim;
-  isect->object = object;
+  isect->object = local_object;
   isect->type = PRIMITIVE_MOTION_TRIANGLE;
 
   /* Record geometric normal. */

intern/cycles/kernel/geom/triangle_intersect.h

@@ -33,9 +33,9 @@ ccl_device_inline bool triangle_intersect(KernelGlobals kg,
                                           float tmax,
                                           uint visibility,
                                           int object,
-                                          int prim,
                                           int prim_addr)
 {
+  const int prim = kernel_tex_fetch(__prim_index, prim_addr);
   const uint tri_vindex = kernel_tex_fetch(__tri_vindex, prim).w;
   const float3 tri_a = kernel_tex_fetch(__tri_verts, tri_vindex + 0),
                tri_b = kernel_tex_fetch(__tri_verts, tri_vindex + 1),
@@ -49,7 +49,8 @@ ccl_device_inline bool triangle_intersect(KernelGlobals kg,
     if (kernel_tex_fetch(__prim_visibility, prim_addr) & visibility)
 #endif
     {
-      isect->object = object;
+      isect->object = (object == OBJECT_NONE) ? kernel_tex_fetch(__prim_object, prim_addr) :
+                                                object;
       isect->prim = prim;
       isect->type = PRIMITIVE_TRIANGLE;
       isect->u = u;
@@ -73,12 +74,21 @@ ccl_device_inline bool triangle_intersect_local(KernelGlobals kg,
                                                 float3 P,
                                                 float3 dir,
                                                 int object,
-                                                int prim,
+                                                int local_object,
                                                 int prim_addr,
                                                 float tmax,
                                                 ccl_private uint *lcg_state,
                                                 int max_hits)
 {
+  /* Only intersect with matching object, for instanced objects we
+   * already know we are only intersecting the right object. */
+  if (object == OBJECT_NONE) {
+    if (kernel_tex_fetch(__prim_object, prim_addr) != local_object) {
+      return false;
+    }
+  }
+
+  const int prim = kernel_tex_fetch(__prim_index, prim_addr);
   const uint tri_vindex = kernel_tex_fetch(__tri_vindex, prim).w;
   const float3 tri_a = kernel_tex_fetch(__tri_verts, tri_vindex + 0),
                tri_b = kernel_tex_fetch(__tri_verts, tri_vindex + 1),
@@ -129,7 +139,7 @@ ccl_device_inline bool triangle_intersect_local(KernelGlobals kg,
   /* Record intersection. */
   ccl_private Intersection *isect = &local_isect->hits[hit];
   isect->prim = prim;
-  isect->object = object;
+  isect->object = local_object;
   isect->type = PRIMITIVE_TRIANGLE;
   isect->u = u;
   isect->v = v;

intern/cycles/kernel/integrator/shade_surface.h

@@ -188,7 +188,7 @@ ccl_device_forceinline void integrate_surface_direct_light(KernelGlobals kg,
   const uint16_t transparent_bounce = INTEGRATOR_STATE(state, path, transparent_bounce);
   uint32_t shadow_flag = INTEGRATOR_STATE(state, path, flag);
   shadow_flag |= (is_light) ? PATH_RAY_SHADOW_FOR_LIGHT : 0;
-  shadow_flag |= (shadow_flag & PATH_RAY_ANY_PASS) ? 0 : PATH_RAY_SURFACE_PASS;
+  shadow_flag |= PATH_RAY_SURFACE_PASS;
   const float3 throughput = INTEGRATOR_STATE(state, path, throughput) * bsdf_eval_sum(&bsdf_eval);
 
   if (kernel_data.kernel_features & KERNEL_FEATURE_LIGHT_PASSES) {

intern/cycles/kernel/integrator/shade_volume.h

@@ -797,7 +797,7 @@ ccl_device_forceinline void integrate_volume_direct_light(
   const uint16_t transparent_bounce = INTEGRATOR_STATE(state, path, transparent_bounce);
   uint32_t shadow_flag = INTEGRATOR_STATE(state, path, flag);
   shadow_flag |= (is_light) ? PATH_RAY_SHADOW_FOR_LIGHT : 0;
-  shadow_flag |= (shadow_flag & PATH_RAY_ANY_PASS) ? 0 : PATH_RAY_VOLUME_PASS;
+  shadow_flag |= PATH_RAY_VOLUME_PASS;
   const float3 throughput_phase = throughput * bsdf_eval_sum(&phase_eval);
 
   if (kernel_data.kernel_features & KERNEL_FEATURE_LIGHT_PASSES) {

intern/cycles/kernel/light/sample.h

@@ -141,23 +141,14 @@ ccl_device_inline float3 shadow_ray_smooth_surface_offset(
     KernelGlobals kg, ccl_private const ShaderData *ccl_restrict sd, float3 Ng)
 {
   float3 V[3], N[3];
-
-  if (sd->type == PRIMITIVE_MOTION_TRIANGLE) {
-    motion_triangle_vertices_and_normals(kg, sd->object, sd->prim, sd->time, V, N);
-  }
-  else {
-    kernel_assert(sd->type == PRIMITIVE_TRIANGLE);
-    triangle_vertices_and_normals(kg, sd->prim, V, N);
-  }
+  triangle_vertices_and_normals(kg, sd->prim, V, N);
 
   const float u = sd->u, v = sd->v;
   const float w = 1 - u - v;
   float3 P = V[0] * u + V[1] * v + V[2] * w; /* Local space */
   float3 n = N[0] * u + N[1] * v + N[2] * w; /* We get away without normalization */
 
-  if (!(sd->object_flag & SD_OBJECT_TRANSFORM_APPLIED)) {
-    object_normal_transform(kg, sd, &n); /* Normal x scale, world space */
-  }
+  object_normal_transform(kg, sd, &n); /* Normal x scale, world space */
 
   /* Parabolic approximation */
   float a = dot(N[2] - N[0], V[0] - V[2]);

intern/cycles/kernel/osl/shaders/node_normal_map.osl

@@ -85,4 +85,6 @@ shader node_normal_map(normal NormalIn = N,
 
   if (Strength != 1.0)
     Normal = normalize(NormalIn + (Normal - NormalIn) * max(Strength, 0.0));
+
+  Normal = ensure_valid_reflection(Ng, I, Normal);
 }

intern/cycles/kernel/svm/attribute.h

@@ -87,9 +87,7 @@ ccl_device_noinline void svm_node_attr(KernelGlobals kg,
   if (node.y == ATTR_STD_GENERATED && desc.element == ATTR_ELEMENT_NONE) {
     /* No generated attribute, fall back to object coordinates. */
     float3 f = sd->P;
-    if (sd->object != OBJECT_NONE) {
-      object_inverse_position_transform(kg, sd, &f);
-    }
+    object_inverse_position_transform(kg, sd, &f);
     if (type == NODE_ATTR_OUTPUT_FLOAT) {
       stack_store_float(stack, out_offset, average(f));
     }
@@ -181,9 +179,7 @@ ccl_device_noinline void svm_node_attr_bump_dx(KernelGlobals kg,
   if (node.y == ATTR_STD_GENERATED && desc.element == ATTR_ELEMENT_NONE) {
     /* No generated attribute, fall back to object coordinates. */
     float3 f = sd->P + sd->dP.dx;
-    if (sd->object != OBJECT_NONE) {
-      object_inverse_position_transform(kg, sd, &f);
-    }
+    object_inverse_position_transform(kg, sd, &f);
     if (type == NODE_ATTR_OUTPUT_FLOAT) {
       stack_store_float(stack, out_offset, average(f));
     }
@@ -279,9 +275,7 @@ ccl_device_noinline void svm_node_attr_bump_dy(KernelGlobals kg,
   if (node.y == ATTR_STD_GENERATED && desc.element == ATTR_ELEMENT_NONE) {
     /* No generated attribute, fall back to object coordinates. */
     float3 f = sd->P + sd->dP.dy;
-    if (sd->object != OBJECT_NONE) {
-      object_inverse_position_transform(kg, sd, &f);
-    }
+    object_inverse_position_transform(kg, sd, &f);
     if (type == NODE_ATTR_OUTPUT_FLOAT) {
       stack_store_float(stack, out_offset, average(f));
     }

intern/cycles/kernel/svm/magic.h

@@ -25,7 +25,7 @@ ccl_device_noinline_cpu float3 svm_magic(float3 p, float scale, int n, float dis
   /*
    * Prevent NaNs due to input p
    * Sin and Cosine are periodic about [0 2*PI) so the following
-   * will yield a more accurate result. As it stops the input values
+   * will yeild a more accurate result. As it stops the input values
    * going out of range for floats which caused a NaN. The
    * calculation of (px + py + pz)*5 can cause an Inf when one or more
    * values are very large the cos or sin of this results in a NaN

intern/cycles/kernel/svm/types.h

@@ -124,7 +124,7 @@ typedef enum ShaderNodeType {
   NODE_AOV_VALUE,
   NODE_FLOAT_CURVE,
   /* NOTE: for best OpenCL performance, item definition in the enum must
-   * match the switch case order in `svm.h`. */
+   * match the switch case order in svm.h. */
 } ShaderNodeType;
 
 typedef enum NodeAttributeOutputType {

intern/cycles/kernel/types.h

@@ -202,7 +202,7 @@ enum SamplingPattern {
 
 /* These flags values correspond to `raytypes` in `osl.cpp`, so keep them in sync! */
 
-enum PathRayFlag : uint32_t {
+enum PathRayFlag {
   /* --------------------------------------------------------------------
    * Ray visibility.
    *
@@ -388,7 +388,6 @@ typedef enum PassType {
   PASS_DENOISING_NORMAL,
   PASS_DENOISING_ALBEDO,
   PASS_DENOISING_DEPTH,
-  PASS_DENOISING_PREVIOUS,
 
   /* PASS_SHADOW_CATCHER accumulates contribution of shadow catcher object which is not affected by
    * any other object. The pass accessor will divide the combined pass by the shadow catcher. The
@@ -1560,7 +1559,7 @@ enum {
 
 /* Kernel Features */
 
-enum KernelFeatureFlag : uint32_t {
+enum KernelFeatureFlag : unsigned int {
   /* Shader nodes. */
   KERNEL_FEATURE_NODE_BSDF = (1U << 0U),
   KERNEL_FEATURE_NODE_EMISSION = (1U << 1U),

intern/cycles/scene/alembic.cpp

@@ -742,7 +742,6 @@ NODE_DEFINE(AlembicProcedural)
   NodeType *type = NodeType::add("alembic", create);
 
   SOCKET_STRING(filepath, "Filename", ustring());
-  SOCKET_STRING_ARRAY(layers, "Layers", array<ustring>());
   SOCKET_FLOAT(frame, "Frame", 1.0f);
   SOCKET_FLOAT(start_frame, "Start Frame", 1.0f);
   SOCKET_FLOAT(end_frame, "End Frame", 1.0f);
@@ -840,26 +839,14 @@ void AlembicProcedural::generate(Scene *scene, Progress &progress)
     return;
   }
 
-  if (!archive.valid() || filepath_is_modified() || layers_is_modified()) {
+  if (!archive.valid()) {
     Alembic::AbcCoreFactory::IFactory factory;
     factory.setPolicy(Alembic::Abc::ErrorHandler::kQuietNoopPolicy);
-
-    std::vector<std::string> filenames;
-    filenames.push_back(filepath.c_str());
-
-    for (const ustring &layer : layers) {
-      filenames.push_back(layer.c_str());
-    }
-
-    /* We need to reverse the order as overriding archives should come first. */
-    std::reverse(filenames.begin(), filenames.end());
-
-    archive = factory.getArchive(filenames);
+    archive = factory.getArchive(filepath.c_str());
 
     if (!archive.valid()) {
       /* avoid potential infinite update loops in viewport synchronization */
       filepath.clear();
-      layers.clear();
       clear_modified();
       return;
     }
@@ -1178,12 +1165,6 @@ void AlembicProcedural::read_subd(AlembicObject *abc_object, Abc::chrono_t frame
   cached_data.subd_creases_weight.copy_to_socket(
       frame_time, mesh, mesh->get_subd_creases_weight_socket());
 
-  cached_data.subd_vertex_crease_indices.copy_to_socket(
-      frame_time, mesh, mesh->get_subd_vert_creases_socket());
-
-  cached_data.subd_vertex_crease_weights.copy_to_socket(
-      frame_time, mesh, mesh->get_subd_vert_creases_weight_socket());
-
   mesh->set_num_subd_faces(mesh->get_subd_shader().size());
 
   /* Update attributes. */

intern/cycles/scene/alembic.h

@@ -320,8 +320,6 @@ struct CachedData {
   DataStore<int> num_ngons;
   DataStore<array<int>> subd_creases_edge;
   DataStore<array<float>> subd_creases_weight;
-  DataStore<array<int>> subd_vertex_crease_indices;
-  DataStore<array<float>> subd_vertex_crease_weights;
 
   /* hair data */
   DataStore<array<float3>> curve_keys;
@@ -481,10 +479,6 @@ class AlembicProcedural : public Procedural {
   /* The file path to the Alembic archive */
   NODE_SOCKET_API(ustring, filepath)
 
-  /* Layers for the Alembic archive. Layers are in the order in which they override data, with the
-   * latter elements overriding the former ones. */
-  NODE_SOCKET_API_ARRAY(array<ustring>, layers)
-
   /* The current frame to render. */
   NODE_SOCKET_API(float, frame)
 

intern/cycles/scene/alembic_read.cpp

@@ -478,9 +478,7 @@ static void add_subd_polygons(CachedData &cached_data, const SubDSchemaData &dat
   cached_data.uv_loops.add_data(uv_loops, time);
 }
 
-static void add_subd_edge_creases(CachedData &cached_data,
-                                  const SubDSchemaData &data,
-                                  chrono_t time)
+static void add_subd_creases(CachedData &cached_data, const SubDSchemaData &data, chrono_t time)
 {
   if (!(data.crease_indices.valid() && data.crease_indices.valid() &&
         data.crease_sharpnesses.valid())) {
@@ -519,37 +517,6 @@ static void add_subd_edge_creases(CachedData &cached_data,
   }
 }
 
-static void add_subd_vertex_creases(CachedData &cached_data,
-                                    const SubDSchemaData &data,
-                                    chrono_t time)
-{
-  if (!(data.corner_indices.valid() && data.crease_sharpnesses.valid())) {
-    return;
-  }
-
-  const ISampleSelector iss = ISampleSelector(time);
-  const Int32ArraySamplePtr creases_indices = data.crease_indices.getValue(iss);
-  const FloatArraySamplePtr creases_sharpnesses = data.crease_sharpnesses.getValue(iss);
-
-  if (!(creases_indices && creases_sharpnesses) ||
-      creases_indices->size() != creases_sharpnesses->size()) {
-    return;
-  }
-
-  array<float> sharpnesses;
-  sharpnesses.reserve(creases_indices->size());
-  array<int> indices;
-  indices.reserve(creases_indices->size());
-
-  for (size_t i = 0; i < creases_indices->size(); i++) {
-    indices.push_back_reserved((*creases_indices)[i]);
-    sharpnesses.push_back_reserved((*creases_sharpnesses)[i]);
-  }
-
-  cached_data.subd_vertex_crease_indices.add_data(indices, time);
-  cached_data.subd_vertex_crease_weights.add_data(sharpnesses, time);
-}
-
 static void read_subd_geometry(CachedData &cached_data, const SubDSchemaData &data, chrono_t time)
 {
   const ISampleSelector iss = ISampleSelector(time);
@@ -558,8 +525,7 @@ static void read_subd_geometry(CachedData &cached_data, const SubDSchemaData &da
 
   if (data.topology_variance != kHomogenousTopology || cached_data.shader.size() == 0) {
     add_subd_polygons(cached_data, data, time);
-    add_subd_edge_creases(cached_data, data, time);
-    add_subd_vertex_creases(cached_data, data, time);
+    add_subd_creases(cached_data, data, time);
   }
 }
 

intern/cycles/scene/alembic_read.h

@@ -76,10 +76,9 @@ struct SubDSchemaData {
 
   vector<FaceSetShaderIndexPair> shader_face_sets;
 
+  // Those are unsupported for now.
   Alembic::AbcGeom::IInt32ArrayProperty corner_indices;
   Alembic::AbcGeom::IFloatArrayProperty corner_sharpnesses;
-
-  // Those are unsupported for now.
   Alembic::AbcGeom::IInt32Property face_varying_interpolate_boundary;
   Alembic::AbcGeom::IInt32Property face_varying_propagate_corners;
   Alembic::AbcGeom::IInt32Property interpolate_boundary;

intern/cycles/scene/colorspace.cpp

@@ -263,9 +263,7 @@ template<typename T> inline void cast_from_float4(T *data, float4 value)
 
 /* Slower versions for other all data types, which needs to convert to float and back. */
 template<typename T, bool compress_as_srgb = false>
-inline void processor_apply_pixels_rgba(const OCIO::Processor *processor,
-                                        T *pixels,
-                                        size_t num_pixels)
+inline void processor_apply_pixels(const OCIO::Processor *processor, T *pixels, size_t num_pixels)
 {
   /* TODO: implement faster version for when we know the conversion
    * is a simple matrix transform between linear spaces. In that case
@@ -312,79 +310,25 @@ inline void processor_apply_pixels_rgba(const OCIO::Processor *processor,
     }
   }
 }
-
-template<typename T, bool compress_as_srgb = false>
-inline void processor_apply_pixels_grayscale(const OCIO::Processor *processor,
-                                             T *pixels,
-                                             size_t num_pixels)
-{
-  OCIO::ConstCPUProcessorRcPtr device_processor = processor->getDefaultCPUProcessor();
-
-  /* Process large images in chunks to keep temporary memory requirement down. */
-  const size_t chunk_size = std::min((size_t)(16 * 1024 * 1024), num_pixels);
-  vector<float> float_pixels(chunk_size * 3);
-
-  for (size_t j = 0; j < num_pixels; j += chunk_size) {
-    size_t width = std::min(chunk_size, num_pixels - j);
-
-    /* Convert to 3 channels, since that's the minimum required by OpenColorIO. */
-    {
-      const T *pixel = pixels + j;
-      float *fpixel = float_pixels.data();
-      for (size_t i = 0; i < width; i++, pixel++, fpixel += 3) {
-        const float f = util_image_cast_to_float<T>(*pixel);
-        fpixel[0] = f;
-        fpixel[1] = f;
-        fpixel[2] = f;
-      }
-    }
-
-    OCIO::PackedImageDesc desc((float *)float_pixels.data(), width, 1, 3);
-    device_processor->apply(desc);
-
-    {
-      T *pixel = pixels + j;
-      const float *fpixel = float_pixels.data();
-      for (size_t i = 0; i < width; i++, pixel++, fpixel += 3) {
-        float f = average(make_float3(fpixel[0], fpixel[1], fpixel[2]));
-        if (compress_as_srgb) {
-          f = color_linear_to_srgb(f);
-        }
-        *pixel = util_image_cast_from_float<T>(f);
-      }
-    }
-  }
-}
-
 #endif
 
 template<typename T>
-void ColorSpaceManager::to_scene_linear(
-    ustring colorspace, T *pixels, size_t num_pixels, bool is_rgba, bool compress_as_srgb)
+void ColorSpaceManager::to_scene_linear(ustring colorspace,
+                                        T *pixels,
+                                        size_t num_pixels,
+                                        bool compress_as_srgb)
 {
 #ifdef WITH_OCIO
   const OCIO::Processor *processor = (const OCIO::Processor *)get_processor(colorspace);
 
   if (processor) {
-    if (is_rgba) {
-      if (compress_as_srgb) {
-        /* Compress output as sRGB. */
-        processor_apply_pixels_rgba<T, true>(processor, pixels, num_pixels);
-      }
-      else {
-        /* Write output as scene linear directly. */
-        processor_apply_pixels_rgba<T>(processor, pixels, num_pixels);
-      }
+    if (compress_as_srgb) {
+      /* Compress output as sRGB. */
+      processor_apply_pixels<T, true>(processor, pixels, num_pixels);
     }
     else {
-      if (compress_as_srgb) {
-        /* Compress output as sRGB. */
-        processor_apply_pixels_grayscale<T, true>(processor, pixels, num_pixels);
-      }
-      else {
-        /* Write output as scene linear directly. */
-        processor_apply_pixels_grayscale<T>(processor, pixels, num_pixels);
-      }
+      /* Write output as scene linear directly. */
+      processor_apply_pixels<T>(processor, pixels, num_pixels);
     }
   }
 #else
@@ -404,11 +348,6 @@ void ColorSpaceManager::to_scene_linear(ColorSpaceProcessor *processor_,
 
   if (processor) {
     OCIO::ConstCPUProcessorRcPtr device_processor = processor->getDefaultCPUProcessor();
-    if (channels == 1) {
-      float3 rgb = make_float3(pixel[0], pixel[0], pixel[0]);
-      device_processor->applyRGB(&rgb.x);
-      pixel[0] = average(rgb);
-    }
     if (channels == 3) {
       device_processor->applyRGB(pixel);
     }
@@ -451,9 +390,9 @@ void ColorSpaceManager::free_memory()
 }
 
 /* Template instantiations so we don't have to inline functions. */
-template void ColorSpaceManager::to_scene_linear(ustring, uchar *, size_t, bool, bool);
-template void ColorSpaceManager::to_scene_linear(ustring, ushort *, size_t, bool, bool);
-template void ColorSpaceManager::to_scene_linear(ustring, half *, size_t, bool, bool);
-template void ColorSpaceManager::to_scene_linear(ustring, float *, size_t, bool, bool);
+template void ColorSpaceManager::to_scene_linear(ustring, uchar *, size_t, bool);
+template void ColorSpaceManager::to_scene_linear(ustring, ushort *, size_t, bool);
+template void ColorSpaceManager::to_scene_linear(ustring, half *, size_t, bool);
+template void ColorSpaceManager::to_scene_linear(ustring, float *, size_t, bool);
 
 CCL_NAMESPACE_END

intern/cycles/scene/colorspace.h

@@ -43,8 +43,10 @@ class ColorSpaceManager {
   /* Convert pixels in the specified colorspace to scene linear color for
    * rendering. Must be a colorspace returned from detect_known_colorspace. */
   template<typename T>
-  static void to_scene_linear(
-      ustring colorspace, T *pixels, size_t num_pixels, bool is_rgba, bool compress_as_srgb);
+  static void to_scene_linear(ustring colorspace,
+                              T *pixels,
+                              size_t num_pixels,
+                              bool compress_as_srgb);
 
   /* Efficiently convert pixels to scene linear colorspace at render time,
    * for OSL where the image texture cache contains original pixels. The

intern/cycles/scene/constant_fold.cpp

@@ -441,13 +441,9 @@ void ConstantFolder::fold_mapping(NodeMappingType type) const
   if (is_zero(scale_in)) {
     make_zero();
   }
-  else if (
-      /* Can't constant fold since we always need to normalize the output. */
-      (type != NODE_MAPPING_TYPE_NORMAL) &&
-      /* Check all use values are zero, note location is not used by vector and normal types. */
-      (is_zero(location_in) || type == NODE_MAPPING_TYPE_VECTOR ||
-       type == NODE_MAPPING_TYPE_NORMAL) &&
-      is_zero(rotation_in) && is_one(scale_in)) {
+  else if ((is_zero(location_in) || type == NODE_MAPPING_TYPE_VECTOR ||
+            type == NODE_MAPPING_TYPE_NORMAL) &&
+           is_zero(rotation_in) && is_one(scale_in)) {
     try_bypass_or_make_constant(vector_in);
   }
 }

intern/cycles/scene/geometry.cpp

@@ -1002,10 +1002,10 @@ void GeometryManager::device_update_attributes(Device *device,
 
   /* After mesh attributes and patch tables have been copied to device memory,
    * we need to update offsets in the objects. */
-  scene->object_manager->device_update_geom_offsets(device, dscene, scene);
+  scene->object_manager->device_update_mesh_offsets(device, dscene, scene);
 }
 
-void GeometryManager::geom_calc_offset(Scene *scene, BVHLayout bvh_layout)
+void GeometryManager::mesh_calc_offset(Scene *scene, BVHLayout bvh_layout)
 {
   size_t vert_size = 0;
   size_t tri_size = 0;
@@ -1922,7 +1922,7 @@ void GeometryManager::device_update(Device *device,
 
   const BVHLayout bvh_layout = BVHParams::best_bvh_layout(scene->params.bvh_layout,
                                                           device->get_bvh_layout_mask());
-  geom_calc_offset(scene, bvh_layout);
+  mesh_calc_offset(scene, bvh_layout);
   if (true_displacement_used || curve_shadow_transparency_used) {
     scoped_callback_timer timer([scene](double time) {
       if (scene->update_stats) {

intern/cycles/scene/geometry.h

@@ -242,7 +242,7 @@ class GeometryManager {
                              vector<AttributeRequestSet> &object_attributes);
 
   /* Compute verts/triangles/curves offsets in global arrays. */
-  void geom_calc_offset(Scene *scene, BVHLayout bvh_layout);
+  void mesh_calc_offset(Scene *scene, BVHLayout bvh_layout);
 
   void device_update_object(Device *device, DeviceScene *dscene, Scene *scene, Progress &progress);
 

intern/cycles/scene/image.cpp

@@ -381,15 +381,8 @@ ImageHandle ImageManager::add_image(const string &filename,
 
   foreach (int tile, tiles) {
     string tile_filename = filename;
-
-    /* Since we don't have information about the exact tile format used in this code location,
-     * just attempt all replacement patterns that Blender supports. */
     if (tile != 0) {
       string_replace(tile_filename, "<UDIM>", string_printf("%04d", tile));
-
-      int u = ((tile - 1001) % 10);
-      int v = ((tile - 1001) / 10);
-      string_replace(tile_filename, "<UVTILE>", string_printf("u%d_v%d", u + 1, v + 1));
     }
     const int slot = add_image_slot(new OIIOImageLoader(tile_filename), params, false);
     handle.tile_slots.push_back(slot);
@@ -576,13 +569,13 @@ bool ImageManager::file_load_image(Image *img, int texture_limit)
         pixels[i * 4 + 3] = one;
       }
     }
-  }
 
-  if (img->metadata.colorspace != u_colorspace_raw &&
-      img->metadata.colorspace != u_colorspace_srgb) {
-    /* Convert to scene linear. */
-    ColorSpaceManager::to_scene_linear(
-        img->metadata.colorspace, pixels, num_pixels, is_rgba, img->metadata.compress_as_srgb);
+    if (img->metadata.colorspace != u_colorspace_raw &&
+        img->metadata.colorspace != u_colorspace_srgb) {
+      /* Convert to scene linear. */
+      ColorSpaceManager::to_scene_linear(
+          img->metadata.colorspace, pixels, num_pixels, img->metadata.compress_as_srgb);
+    }
   }
 
   /* Make sure we don't have buggy values. */
@@ -891,10 +884,6 @@ void ImageManager::device_free(Device *device)
 void ImageManager::collect_statistics(RenderStats *stats)
 {
   foreach (const Image *image, images) {
-    if (!image) {
-      /* Image may have been freed due to lack of users. */
-      continue;
-    }
     stats->image.textures.add_entry(
         NamedSizeEntry(image->loader->name(), image->mem->memory_size()));
   }