Add tooltip hover for 2d vector

The leak happened when the vector set had to grow when it
was empty but it had allocated the keys array already.

CMakeLists.txt

@@ -685,7 +685,7 @@ if(WIN32 OR XCODE)
   option(IDE_GROUP_PROJECTS_IN_FOLDERS "Organize the projects according to source folder structure." ON)
   mark_as_advanced(IDE_GROUP_PROJECTS_IN_FOLDERS)
 
-  if (IDE_GROUP_PROJECTS_IN_FOLDERS)
+  if(IDE_GROUP_PROJECTS_IN_FOLDERS)
     set_property(GLOBAL PROPERTY USE_FOLDERS ON)
   endif()
 endif()

build_files/build_environment/cmake/python_site_packages.cmake

@@ -31,7 +31,7 @@ ExternalProject_Add(external_python_site_packages
   CONFIGURE_COMMAND ${PIP_CONFIGURE_COMMAND}
   BUILD_COMMAND ""
   PREFIX ${BUILD_DIR}/site_packages
-  INSTALL_COMMAND ${PYTHON_BINARY} -m pip install ${SITE_PACKAGES_EXTRA} cython==${CYTHON_VERSION} idna==${IDNA_VERSION} charset-normalizer==${CHARSET_NORMALIZER_VERSION} urllib3==${URLLIB3_VERSION} certifi==${CERTIFI_VERSION} requests==${REQUESTS_VERSION} zstandard==${ZSTANDARD_VERSION} --no-binary :all:
+  INSTALL_COMMAND ${PYTHON_BINARY} -m pip install --no-cache-dir ${SITE_PACKAGES_EXTRA} cython==${CYTHON_VERSION} idna==${IDNA_VERSION} charset-normalizer==${CHARSET_NORMALIZER_VERSION} urllib3==${URLLIB3_VERSION} certifi==${CERTIFI_VERSION} requests==${REQUESTS_VERSION} zstandard==${ZSTANDARD_VERSION} --no-binary :all:
 )
 
 if(USE_PIP_NUMPY)

build_files/build_environment/cmake/versions.cmake

@@ -189,11 +189,11 @@ set(OSL_HASH 1abd7ce40481771a9fa937f19595d2f2)
 set(OSL_HASH_TYPE MD5)
 set(OSL_FILE OpenShadingLanguage-${OSL_VERSION}.tar.gz)
 
-set(PYTHON_VERSION 3.9.7)
-set(PYTHON_SHORT_VERSION 3.9)
-set(PYTHON_SHORT_VERSION_NO_DOTS 39)
+set(PYTHON_VERSION 3.10.2)
+set(PYTHON_SHORT_VERSION 3.10)
+set(PYTHON_SHORT_VERSION_NO_DOTS 310)
 set(PYTHON_URI https://www.python.org/ftp/python/${PYTHON_VERSION}/Python-${PYTHON_VERSION}.tar.xz)
-set(PYTHON_HASH fddb060b483bc01850a3f412eea1d954)
+set(PYTHON_HASH 14e8c22458ed7779a1957b26cde01db9)
 set(PYTHON_HASH_TYPE MD5)
 set(PYTHON_FILE Python-${PYTHON_VERSION}.tar.xz)
 
@@ -215,18 +215,20 @@ set(NANOVDB_HASH e7b9e863ec2f3b04ead171dec2322807)
 set(NANOVDB_HASH_TYPE MD5)
 set(NANOVDB_FILE nano-vdb-${NANOVDB_GIT_UID}.tar.gz)
 
-set(IDNA_VERSION 3.2)
-set(CHARSET_NORMALIZER_VERSION 2.0.6)
-set(URLLIB3_VERSION 1.26.7)
+set(IDNA_VERSION 3.3)
+set(CHARSET_NORMALIZER_VERSION 2.0.10)
+set(URLLIB3_VERSION 1.26.8)
 set(CERTIFI_VERSION 2021.10.8)
-set(REQUESTS_VERSION 2.26.0)
-set(CYTHON_VERSION 0.29.24)
-set(ZSTANDARD_VERSION 0.15.2 )
+set(REQUESTS_VERSION 2.27.1)
+set(CYTHON_VERSION 0.29.26)
+# The version of the zstd library used to build the Python package should match ZSTD_VERSION defined below.
+# At this time of writing, 0.17.0 was already released, but built against zstd 1.5.1, while we use 1.5.0.
+set(ZSTANDARD_VERSION 0.16.0)
 
-set(NUMPY_VERSION 1.21.2)
-set(NUMPY_SHORT_VERSION 1.21)
+set(NUMPY_VERSION 1.22.0)
+set(NUMPY_SHORT_VERSION 1.22)
 set(NUMPY_URI https://github.com/numpy/numpy/releases/download/v${NUMPY_VERSION}/numpy-${NUMPY_VERSION}.zip)
-set(NUMPY_HASH 5638d5dae3ca387be562912312db842e)
+set(NUMPY_HASH 252de134862a27bd66705d29622edbfe)
 set(NUMPY_HASH_TYPE MD5)
 set(NUMPY_FILE numpy-${NUMPY_VERSION}.zip)
 

build_files/build_environment/install_deps.sh

@@ -379,27 +379,27 @@ USE_CXX11=true
 CLANG_FORMAT_VERSION_MIN="6.0"
 CLANG_FORMAT_VERSION_MEX="10.0"
 
-PYTHON_VERSION="3.9.7"
-PYTHON_VERSION_SHORT="3.9"
-PYTHON_VERSION_MIN="3.7"
-PYTHON_VERSION_MEX="3.11"
+PYTHON_VERSION="3.10.2"
+PYTHON_VERSION_SHORT="3.10"
+PYTHON_VERSION_MIN="3.9"
+PYTHON_VERSION_MEX="3.12"
 PYTHON_VERSION_INSTALLED=$PYTHON_VERSION_SHORT
 PYTHON_FORCE_BUILD=false
 PYTHON_FORCE_REBUILD=false
 PYTHON_SKIP=false
 
 # Additional Python modules.
-PYTHON_IDNA_VERSION="3.2"
+PYTHON_IDNA_VERSION="3.3"
 PYTHON_IDNA_VERSION_MIN="2.0"
 PYTHON_IDNA_VERSION_MEX="4.0"
 PYTHON_IDNA_NAME="idna"
 
-PYTHON_CHARSET_NORMALIZER_VERSION="2.0.6"
+PYTHON_CHARSET_NORMALIZER_VERSION="2.0.10"
 PYTHON_CHARSET_NORMALIZER_VERSION_MIN="2.0.6"
 PYTHON_CHARSET_NORMALIZER_VERSION_MEX="2.1.0"  # requests uses `charset_normalizer~=2.0.0`
 PYTHON_CHARSET_NORMALIZER_NAME="charset-normalizer"
 
-PYTHON_URLLIB3_VERSION="1.26.7"
+PYTHON_URLLIB3_VERSION="1.26.8"
 PYTHON_URLLIB3_VERSION_MIN="1.0"
 PYTHON_URLLIB3_VERSION_MEX="2.0"
 PYTHON_URLLIB3_NAME="urllib3"
@@ -409,17 +409,17 @@ PYTHON_CERTIFI_VERSION_MIN="2021.0"
 PYTHON_CERTIFI_VERSION_MEX="2023.0"
 PYTHON_CERTIFI_NAME="certifi"
 
-PYTHON_REQUESTS_VERSION="2.23.0"
+PYTHON_REQUESTS_VERSION="2.27.1"
 PYTHON_REQUESTS_VERSION_MIN="2.0"
 PYTHON_REQUESTS_VERSION_MEX="3.0"
 PYTHON_REQUESTS_NAME="requests"
 
-PYTHON_ZSTANDARD_VERSION="0.15.2"
+PYTHON_ZSTANDARD_VERSION="0.16.0"
 PYTHON_ZSTANDARD_VERSION_MIN="0.15.2"
-PYTHON_ZSTANDARD_VERSION_MEX="0.16.0"
+PYTHON_ZSTANDARD_VERSION_MEX="0.20.0"
 PYTHON_ZSTANDARD_NAME="zstandard"
 
-PYTHON_NUMPY_VERSION="1.21.2"
+PYTHON_NUMPY_VERSION="1.22.0"
 PYTHON_NUMPY_VERSION_MIN="1.14"
 PYTHON_NUMPY_VERSION_MEX="2.0"
 PYTHON_NUMPY_NAME="numpy"
@@ -499,7 +499,7 @@ LLVM_FORCE_REBUILD=false
 LLVM_SKIP=false
 
 # OSL needs to be compiled for now!
-OSL_VERSION="1.11.14.1"
+OSL_VERSION="1.11.17.0"
 OSL_VERSION_SHORT="1.11"
 OSL_VERSION_MIN="1.11"
 OSL_VERSION_MEX="2.0"
@@ -4036,14 +4036,14 @@ install_DEB() {
     INFO "Forced Python building, as requested..."
     _do_compile_python=true
   else
-    check_package_version_ge_lt_DEB python3-dev $PYTHON_VERSION_MIN $PYTHON_VERSION_MEX
+    check_package_version_ge_lt_DEB python${PYTHON_VERSION_SHORT}-dev $PYTHON_VERSION_MIN $PYTHON_VERSION_MEX
     if [ $? -eq 0 ]; then
-      PYTHON_VERSION_INSTALLED=$(echo `get_package_version_DEB python3-dev` | sed -r 's/^([0-9]+\.[0-9]+).*/\1/')
-
-      install_packages_DEB python3-dev
+      install_packages_DEB python${PYTHON_VERSION_SHORT}-dev
       clean_Python
       PRINT ""
 
+      PYTHON_VERSION_INSTALLED=$(echo `get_package_version_DEB python${PYTHON_VERSION_SHORT}-dev` | sed -r 's/^([0-9]+\.[0-9]+).*/\1/')
+
       for module in "${PYTHON_MODULES_PACKAGES[@]}"
       do
         module=($module)
@@ -4681,11 +4681,11 @@ install_RPM() {
   else
     check_package_version_ge_lt_RPM python3-devel $PYTHON_VERSION_MIN $PYTHON_VERSION_MEX
     if [ $? -eq 0 ]; then
-      PYTHON_VERSION_INSTALLED=$(echo `get_package_version_RPM python3-devel` | sed -r 's/^([0-9]+\.[0-9]+).*/\1/')
-
       install_packages_RPM python3-devel
       clean_Python
 
+      PYTHON_VERSION_INSTALLED=$(echo `get_package_version_RPM python3-devel` | sed -r 's/^([0-9]+\.[0-9]+).*/\1/')
+
       for module in "${PYTHON_MODULES_PACKAGES[@]}"
       do
         module=($module)
@@ -5224,12 +5224,12 @@ install_ARCH() {
   else
     check_package_version_ge_lt_ARCH python $PYTHON_VERSION_MIN $PYTHON_VERSION_MEX
     if [ $? -eq 0 ]; then
-      PYTHON_VERSION_INSTALLED=$(echo `get_package_version_ARCH python` | sed -r 's/^([0-9]+\.[0-9]+).*/\1/')
-
       install_packages_ARCH python
       clean_Python
       PRINT ""
 
+      PYTHON_VERSION_INSTALLED=$(echo `get_package_version_ARCH python` | sed -r 's/^([0-9]+\.[0-9]+).*/\1/')
+
       for module in "${PYTHON_MODULES_PACKAGES[@]}"
       do
         module=($module)

build_files/cmake/Modules/FindPythonLibsUnix.cmake

@@ -34,7 +34,7 @@ IF(NOT PYTHON_ROOT_DIR AND NOT $ENV{PYTHON_ROOT_DIR} STREQUAL "")
   SET(PYTHON_ROOT_DIR $ENV{PYTHON_ROOT_DIR})
 ENDIF()
 
-SET(PYTHON_VERSION 3.9 CACHE STRING "Python Version (major and minor only)")
+SET(PYTHON_VERSION 3.10 CACHE STRING "Python Version (major and minor only)")
 MARK_AS_ADVANCED(PYTHON_VERSION)
 
 

build_files/cmake/macros.cmake

@@ -1197,21 +1197,21 @@ endfunction()
 macro(openmp_delayload
   projectname
   )
-    if(MSVC)
-      if(WITH_OPENMP)
-        if(MSVC_CLANG)
-          set(OPENMP_DLL_NAME "libomp")
-        elseif(MSVC_VERSION EQUAL 1800)
-          set(OPENMP_DLL_NAME "vcomp120")
-        else()
-          set(OPENMP_DLL_NAME "vcomp140")
-        endif()
-        set_property(TARGET ${projectname} APPEND_STRING  PROPERTY LINK_FLAGS_RELEASE " /DELAYLOAD:${OPENMP_DLL_NAME}.dll delayimp.lib")
-        set_property(TARGET ${projectname} APPEND_STRING  PROPERTY LINK_FLAGS_DEBUG " /DELAYLOAD:${OPENMP_DLL_NAME}d.dll delayimp.lib")
-        set_property(TARGET ${projectname} APPEND_STRING  PROPERTY LINK_FLAGS_RELWITHDEBINFO " /DELAYLOAD:${OPENMP_DLL_NAME}.dll delayimp.lib")
-        set_property(TARGET ${projectname} APPEND_STRING  PROPERTY LINK_FLAGS_MINSIZEREL " /DELAYLOAD:${OPENMP_DLL_NAME}.dll delayimp.lib")
+  if(MSVC)
+    if(WITH_OPENMP)
+      if(MSVC_CLANG)
+        set(OPENMP_DLL_NAME "libomp")
+      elseif(MSVC_VERSION EQUAL 1800)
+        set(OPENMP_DLL_NAME "vcomp120")
+      else()
+        set(OPENMP_DLL_NAME "vcomp140")
       endif()
+      set_property(TARGET ${projectname} APPEND_STRING  PROPERTY LINK_FLAGS_RELEASE " /DELAYLOAD:${OPENMP_DLL_NAME}.dll delayimp.lib")
+      set_property(TARGET ${projectname} APPEND_STRING  PROPERTY LINK_FLAGS_DEBUG " /DELAYLOAD:${OPENMP_DLL_NAME}d.dll delayimp.lib")
+      set_property(TARGET ${projectname} APPEND_STRING  PROPERTY LINK_FLAGS_RELWITHDEBINFO " /DELAYLOAD:${OPENMP_DLL_NAME}.dll delayimp.lib")
+      set_property(TARGET ${projectname} APPEND_STRING  PROPERTY LINK_FLAGS_MINSIZEREL " /DELAYLOAD:${OPENMP_DLL_NAME}.dll delayimp.lib")
     endif()
+  endif()
 endmacro()
 
 macro(set_and_warn_dependency

build_files/cmake/platform/platform_apple.cmake

@@ -128,25 +128,20 @@ if(WITH_CODEC_SNDFILE)
 endif()
 
 if(WITH_PYTHON)
-  # we use precompiled libraries for py 3.9 and up by default
-  set(PYTHON_VERSION 3.9)
+  # Use precompiled libraries by default.
+  set(PYTHON_VERSION 3.10)
   if(NOT WITH_PYTHON_MODULE AND NOT WITH_PYTHON_FRAMEWORK)
-    # normally cached but not since we include them with blender
+    # Normally cached but not since we include them with blender.
     set(PYTHON_INCLUDE_DIR "${LIBDIR}/python/include/python${PYTHON_VERSION}")
     set(PYTHON_EXECUTABLE "${LIBDIR}/python/bin/python${PYTHON_VERSION}")
     set(PYTHON_LIBRARY ${LIBDIR}/python/lib/libpython${PYTHON_VERSION}.a)
     set(PYTHON_LIBPATH "${LIBDIR}/python/lib/python${PYTHON_VERSION}")
-    # set(PYTHON_LINKFLAGS "-u _PyMac_Error")  # won't  build with this enabled
   else()
-    # module must be compiled against Python framework
+    # Module must be compiled against Python framework.
     set(_py_framework "/Library/Frameworks/Python.framework/Versions/${PYTHON_VERSION}")
-
     set(PYTHON_INCLUDE_DIR "${_py_framework}/include/python${PYTHON_VERSION}")
     set(PYTHON_EXECUTABLE "${_py_framework}/bin/python${PYTHON_VERSION}")
     set(PYTHON_LIBPATH "${_py_framework}/lib/python${PYTHON_VERSION}")
-    # set(PYTHON_LIBRARY python${PYTHON_VERSION})
-    # set(PYTHON_LINKFLAGS "-u _PyMac_Error -framework Python")  # won't  build with this enabled
-
     unset(_py_framework)
   endif()
 

build_files/cmake/platform/platform_unix.cmake

@@ -109,9 +109,14 @@ if(NOT WITH_SYSTEM_FREETYPE)
   find_package_wrapper(Freetype REQUIRED)
   if(EXISTS ${LIBDIR})
     find_package_wrapper(Brotli REQUIRED)
-    list(APPEND FREETYPE_LIBRARIES
-      ${BROTLI_LIBRARIES}
-    )
+
+    # NOTE: This is done on WIN32 & APPLE but fails on some Linux systems.
+    # See: https://devtalk.blender.org/t/22536
+    # So `BROTLI_LIBRARIES` need to be added directly after `FREETYPE_LIBRARIES`.
+    #
+    # list(APPEND FREETYPE_LIBRARIES
+    #   ${BROTLI_LIBRARIES}
+    # )
   endif()
 endif()
 

build_files/cmake/platform/platform_win32.cmake

@@ -55,6 +55,10 @@ if(CMAKE_C_COMPILER_ID MATCHES "Clang")
     message(WARNING "stripped pdb not supported with clang, disabling..")
     set(WITH_WINDOWS_STRIPPED_PDB OFF)
   endif()
+else()
+  if(CMAKE_CXX_COMPILER_VERSION VERSION_LESS 19.28.29921) # MSVC 2019 16.9.16
+    message(FATAL_ERROR "Compiler is unsupported, MSVC 2019 16.9.16 or newer is required for building blender.")
+  endif()
 endif()
 
 if(NOT WITH_PYTHON_MODULE)
@@ -265,12 +269,6 @@ if(NOT DEFINED LIBDIR)
   elseif(MSVC_VERSION GREATER 1919)
     message(STATUS "Visual Studio 2019 detected.")
     set(LIBDIR ${CMAKE_SOURCE_DIR}/../lib/${LIBDIR_BASE}_vc15)
-  elseif(MSVC_VERSION GREATER 1909)
-    message(STATUS "Visual Studio 2017 detected.")
-    set(LIBDIR ${CMAKE_SOURCE_DIR}/../lib/${LIBDIR_BASE}_vc15)
-  elseif(MSVC_VERSION EQUAL 1900)
-    message(STATUS "Visual Studio 2015 detected.")
-    set(LIBDIR ${CMAKE_SOURCE_DIR}/../lib/${LIBDIR_BASE}_vc15)
   endif()
 else()
   message(STATUS "Using pre-compiled LIBDIR: ${LIBDIR}")
@@ -465,7 +463,7 @@ if(WITH_JACK)
 endif()
 
 if(WITH_PYTHON)
-  set(PYTHON_VERSION 3.9) # CACHE STRING)
+  set(PYTHON_VERSION 3.10) # CACHE STRING)
 
   string(REPLACE "." "" _PYTHON_VERSION_NO_DOTS ${PYTHON_VERSION})
   set(PYTHON_LIBRARY ${LIBDIR}/python/${_PYTHON_VERSION_NO_DOTS}/libs/python${_PYTHON_VERSION_NO_DOTS}.lib)

build_files/windows/autodetect_msvc.cmd

@@ -3,9 +3,6 @@ echo No explicit msvc version requested, autodetecting version.
 call "%~dp0\detect_msvc2019.cmd"
 if %ERRORLEVEL% EQU 0 goto DetectionComplete
 
-call "%~dp0\detect_msvc2017.cmd"
-if %ERRORLEVEL% EQU 0 goto DetectionComplete
-
 call "%~dp0\detect_msvc2022.cmd"
 if %ERRORLEVEL% EQU 0 goto DetectionComplete
 

build_files/windows/check_libraries.cmd

@@ -1,4 +1,3 @@
-if "%BUILD_VS_YEAR%"=="2017" set BUILD_VS_LIBDIRPOST=vc15
 if "%BUILD_VS_YEAR%"=="2019" set BUILD_VS_LIBDIRPOST=vc15
 if "%BUILD_VS_YEAR%"=="2022" set BUILD_VS_LIBDIRPOST=vc15
 

build_files/windows/configure_msbuild.cmd

@@ -19,12 +19,6 @@ if "%WITH_PYDEBUG%"=="1" (
 	set PYDEBUG_CMAKE_ARGS=-DWINDOWS_PYTHON_DEBUG=On
 )
 
-if "%BUILD_VS_YEAR%"=="2017" (
-	set BUILD_GENERATOR_POST=%WINDOWS_ARCH%
-) else (
-	set BUILD_PLATFORM_SELECT=-A %MSBUILD_PLATFORM%
-)
-
 set BUILD_CMAKE_ARGS=%BUILD_CMAKE_ARGS% -G "Visual Studio %BUILD_VS_VER% %BUILD_VS_YEAR%%BUILD_GENERATOR_POST%" %BUILD_PLATFORM_SELECT% %TESTS_CMAKE_ARGS% %CLANG_CMAKE_ARGS% %ASAN_CMAKE_ARGS% %PYDEBUG_CMAKE_ARGS%
 
 if NOT EXIST %BUILD_DIR%\nul (

build_files/windows/configure_ninja.cmd

@@ -37,15 +37,9 @@ set LLVM_DIR=
 :DetectionComplete	
 	set CC=%LLVM_DIR%\bin\clang-cl
 	set CXX=%LLVM_DIR%\bin\clang-cl
-	if "%BUILD_VS_YEAR%" == "2019" (
-		rem build and tested against 2019 16.2
-		set CFLAGS=-m64 -fmsc-version=1922
-		set CXXFLAGS=-m64 -fmsc-version=1922
-	) else (
-		rem build and tested against 2017 15.7
-		set CFLAGS=-m64 -fmsc-version=1914
-		set CXXFLAGS=-m64 -fmsc-version=1914
-	)
+	rem build and tested against 2019 16.2
+	set CFLAGS=-m64 -fmsc-version=1922
+	set CXXFLAGS=-m64 -fmsc-version=1922
 )
 
 if "%WITH_ASAN%"=="1" (

build_files/windows/detect_msvc2017.cmd

@@ -1,3 +0,0 @@
-set BUILD_VS_VER=15
-set BUILD_VS_YEAR=2017
-call "%~dp0\detect_msvc_vswhere.cmd"

build_files/windows/parse_arguments.cmd

@@ -50,14 +50,6 @@ if NOT "%1" == "" (
 		goto ERR
 	) else if "%1" == "x64" (
 		set BUILD_ARCH=x64
-	) else if "%1" == "2017" (
-		set BUILD_VS_YEAR=2017
-	) else if "%1" == "2017pre" (
-		set BUILD_VS_YEAR=2017
-		set VSWHERE_ARGS=-prerelease
-	) else if "%1" == "2017b" (
-		set BUILD_VS_YEAR=2017
-		set VSWHERE_ARGS=-products Microsoft.VisualStudio.Product.BuildTools
 	) else if "%1" == "2019" (
 		set BUILD_VS_YEAR=2019
 	) else if "%1" == "2019pre" (

build_files/windows/show_help.cmd

@@ -24,12 +24,12 @@ echo - nobuildinfo ^(disable buildinfo^)
 echo - debug ^(Build an unoptimized debuggable build^)
 echo - packagename [newname] ^(override default cpack package name^)
 echo - builddir [newdir] ^(override default build folder^)
-echo - 2017 ^(build with visual studio 2017^)
-echo - 2017pre ^(build with visual studio 2017 pre-release^)
-echo - 2017b ^(build with visual studio 2017 Build Tools^)
 echo - 2019 ^(build with visual studio 2019^)
 echo - 2019pre ^(build with visual studio 2019 pre-release^)
 echo - 2019b ^(build with visual studio 2019 Build Tools^)
+echo - 2022 ^(build with visual studio 2022^)
+echo - 2022pre ^(build with visual studio 2022 pre-release^)
+echo - 2022b ^(build with visual studio 2022 Build Tools^)
 
 echo.
 echo Documentation Targets ^(Not associated with building^)

build_files/windows/svn_fix.cmd

@@ -1,4 +1,3 @@
-if "%BUILD_VS_YEAR%"=="2017" set BUILD_VS_LIBDIRPOST=vc15
 if "%BUILD_VS_YEAR%"=="2019" set BUILD_VS_LIBDIRPOST=vc15
 if "%BUILD_VS_YEAR%"=="2022" set BUILD_VS_LIBDIRPOST=vc15
 

doc/doxygen/Doxyfile

@@ -38,7 +38,7 @@ PROJECT_NAME           = Blender
 # could be handy for archiving the generated documentation or if some version
 # control system is used.
 
-PROJECT_NUMBER         = V3.1
+PROJECT_NUMBER         = V3.2
 
 # Using the PROJECT_BRIEF tag one can provide an optional one line description
 # for a project that appears at the top of each page and should give viewer a

extern/CMakeLists.txt

@@ -113,6 +113,6 @@ if(WITH_MOD_FLUID)
   add_subdirectory(mantaflow)
 endif()
 
-if (WITH_COMPOSITOR)
+if(WITH_COMPOSITOR)
   add_subdirectory(smaa_areatex)
 endif()

intern/cycles/bvh/embree.cpp

@@ -61,6 +61,26 @@ static_assert(Object::MAX_MOTION_STEPS == Geometry::MAX_MOTION_STEPS,
 
 #  define IS_HAIR(x) (x & 1)
 
+/* This gets called by Embree at every valid ray/object intersection.
+ * Things like recording subsurface or shadow hits for later evaluation
+ * as well as filtering for volume objects happen here.
+ * Cycles' own BVH does that directly inside the traversal calls.
+ */
+static void rtc_filter_intersection_func(const RTCFilterFunctionNArguments *args)
+{
+  /* Current implementation in Cycles assumes only single-ray intersection queries. */
+  assert(args->N == 1);
+
+  RTCHit *hit = (RTCHit *)args->hit;
+  CCLIntersectContext *ctx = ((IntersectContext *)args->context)->userRayExt;
+  const KernelGlobalsCPU *kg = ctx->kg;
+  const Ray *cray = ctx->ray;
+
+  if (kernel_embree_is_self_intersection(kg, hit, cray)) {
+    *args->valid = 0;
+  }
+}
+
 /* This gets called by Embree at every valid ray/object intersection.
  * Things like recording subsurface or shadow hits for later evaluation
  * as well as filtering for volume objects happen here.
@@ -75,12 +95,16 @@ static void rtc_filter_occluded_func(const RTCFilterFunctionNArguments *args)
   RTCHit *hit = (RTCHit *)args->hit;
   CCLIntersectContext *ctx = ((IntersectContext *)args->context)->userRayExt;
   const KernelGlobalsCPU *kg = ctx->kg;
+  const Ray *cray = ctx->ray;
 
   switch (ctx->type) {
     case CCLIntersectContext::RAY_SHADOW_ALL: {
       Intersection current_isect;
       kernel_embree_convert_hit(kg, ray, hit, &current_isect);
-
+      if (intersection_skip_self_shadow(cray->self, current_isect.object, current_isect.prim)) {
+        *args->valid = 0;
+        return;
+      }
       /* If no transparent shadows or max number of hits exceeded, all light is blocked. */
       const int flags = intersection_get_shader_flags(kg, current_isect.prim, current_isect.type);
       if (!(flags & (SD_HAS_TRANSPARENT_SHADOW)) || ctx->num_hits >= ctx->max_hits) {
@@ -160,6 +184,10 @@ static void rtc_filter_occluded_func(const RTCFilterFunctionNArguments *args)
           break;
         }
       }
+      if (intersection_skip_self_local(cray->self, current_isect.prim)) {
+        *args->valid = 0;
+        return;
+      }
 
       /* No intersection information requested, just return a hit. */
       if (ctx->max_hits == 0) {
@@ -225,6 +253,11 @@ static void rtc_filter_occluded_func(const RTCFilterFunctionNArguments *args)
       if (ctx->num_hits < ctx->max_hits) {
         Intersection current_isect;
         kernel_embree_convert_hit(kg, ray, hit, &current_isect);
+        if (intersection_skip_self(cray->self, current_isect.object, current_isect.prim)) {
+          *args->valid = 0;
+          return;
+        }
+
         Intersection *isect = &ctx->isect_s[ctx->num_hits];
         ++ctx->num_hits;
         *isect = current_isect;
@@ -236,12 +269,15 @@ static void rtc_filter_occluded_func(const RTCFilterFunctionNArguments *args)
         }
         /* This tells Embree to continue tracing. */
         *args->valid = 0;
-        break;
       }
+      break;
     }
     case CCLIntersectContext::RAY_REGULAR:
     default:
-      /* Nothing to do here. */
+      if (kernel_embree_is_self_intersection(kg, hit, cray)) {
+        *args->valid = 0;
+        return;
+      }
       break;
   }
 }
@@ -257,6 +293,14 @@ static void rtc_filter_func_backface_cull(const RTCFilterFunctionNArguments *arg
     *args->valid = 0;
     return;
   }
+
+  CCLIntersectContext *ctx = ((IntersectContext *)args->context)->userRayExt;
+  const KernelGlobalsCPU *kg = ctx->kg;
+  const Ray *cray = ctx->ray;
+
+  if (kernel_embree_is_self_intersection(kg, hit, cray)) {
+    *args->valid = 0;
+  }
 }
 
 static void rtc_filter_occluded_func_backface_cull(const RTCFilterFunctionNArguments *args)
@@ -505,6 +549,7 @@ void BVHEmbree::add_triangles(const Object *ob, const Mesh *mesh, int i)
 
   rtcSetGeometryUserData(geom_id, (void *)prim_offset);
   rtcSetGeometryOccludedFilterFunction(geom_id, rtc_filter_occluded_func);
+  rtcSetGeometryIntersectFilterFunction(geom_id, rtc_filter_intersection_func);
   rtcSetGeometryMask(geom_id, ob->visibility_for_tracing());
 
   rtcCommitGeometry(geom_id);
@@ -767,6 +812,7 @@ void BVHEmbree::add_curves(const Object *ob, const Hair *hair, int i)
 
   rtcSetGeometryUserData(geom_id, (void *)prim_offset);
   if (hair->curve_shape == CURVE_RIBBON) {
+    rtcSetGeometryIntersectFilterFunction(geom_id, rtc_filter_intersection_func);
     rtcSetGeometryOccludedFilterFunction(geom_id, rtc_filter_occluded_func);
   }
   else {

intern/cycles/cmake/external_libs.cmake

@@ -559,10 +559,10 @@ if(WITH_CYCLES_DEVICE_METAL)
   find_library(METAL_LIBRARY Metal)
 
   # This file was added in the 12.0 SDK, use it as a way to detect the version.
-  if (METAL_LIBRARY AND NOT EXISTS "${METAL_LIBRARY}/Headers/MTLFunctionStitching.h")
+  if(METAL_LIBRARY AND NOT EXISTS "${METAL_LIBRARY}/Headers/MTLFunctionStitching.h")
     message(STATUS "Metal version too old, must be SDK 12.0 or newer, disabling WITH_CYCLES_DEVICE_METAL")
     set(WITH_CYCLES_DEVICE_METAL OFF)
-  elseif (NOT METAL_LIBRARY)
+  elseif(NOT METAL_LIBRARY)
     message(STATUS "Metal not found, disabling WITH_CYCLES_DEVICE_METAL")
     set(WITH_CYCLES_DEVICE_METAL OFF)
   else()

intern/cycles/device/hip/device_impl.cpp

@@ -905,8 +905,8 @@ void HIPDevice::tex_alloc(device_texture &mem)
       address_mode = hipAddressModeClamp;
       break;
     case EXTENSION_CLIP:
-      // TODO : (Arya) setting this to Mode Clamp instead of Mode Border because it's unsupported
-      // in hip
+      /* TODO(@arya): setting this to Mode Clamp instead of Mode Border
+       * because it's unsupported in HIP. */
       address_mode = hipAddressModeClamp;
       break;
     default:

intern/cycles/device/optix/device_impl.cpp

@@ -226,7 +226,7 @@ bool OptiXDevice::load_kernels(const uint kernel_features)
   pipeline_options.usesMotionBlur = false;
   pipeline_options.traversableGraphFlags =
       OPTIX_TRAVERSABLE_GRAPH_FLAG_ALLOW_SINGLE_LEVEL_INSTANCING;
-  pipeline_options.numPayloadValues = 6;
+  pipeline_options.numPayloadValues = 8;
   pipeline_options.numAttributeValues = 2; /* u, v */
   pipeline_options.exceptionFlags = OPTIX_EXCEPTION_FLAG_NONE;
   pipeline_options.pipelineLaunchParamsVariableName = "__params"; /* See globals.h */

intern/cycles/integrator/denoiser.cpp

@@ -125,20 +125,41 @@ static Device *find_best_device(Device *device, DenoiserType type)
   return best_device;
 }
 
+static DeviceInfo find_best_denoiser_device_info(const vector<DeviceInfo> &device_infos,
+                                                 DenoiserType denoiser_type)
+{
+  for (const DeviceInfo &device_info : device_infos) {
+    if ((device_info.denoisers & denoiser_type) == 0) {
+      continue;
+    }
+
+    /* TODO(sergey): Use one of the already configured devices, so that OptiX denoising can happen
+     * on a physical CUDA device which is already used for rendering. */
+
+    /* TODO(sergey): Choose fastest device for denoising. */
+
+    return device_info;
+  }
+
+  DeviceInfo none_device;
+  none_device.type = DEVICE_NONE;
+  return none_device;
+}
+
 static unique_ptr<Device> create_denoiser_device(Device *path_trace_device,
-                                                 const uint device_type_mask)
+                                                 const uint device_type_mask,
+                                                 DenoiserType denoiser_type)
 {
   const vector<DeviceInfo> device_infos = Device::available_devices(device_type_mask);
   if (device_infos.empty()) {
     return nullptr;
   }
 
-  /* TODO(sergey): Use one of the already configured devices, so that OptiX denoising can happen on
-   * a physical CUDA device which is already used for rendering. */
-
-  /* TODO(sergey): Choose fastest device for denoising. */
-
-  const DeviceInfo denoiser_device_info = device_infos.front();
+  const DeviceInfo denoiser_device_info = find_best_denoiser_device_info(device_infos,
+                                                                         denoiser_type);
+  if (denoiser_device_info.type == DEVICE_NONE) {
+    return nullptr;
+  }
 
   unique_ptr<Device> denoiser_device(
       Device::create(denoiser_device_info, path_trace_device->stats, path_trace_device->profiler));
@@ -186,7 +207,8 @@ Device *Denoiser::ensure_denoiser_device(Progress *progress)
   device_creation_attempted_ = true;
 
   const uint device_type_mask = get_device_type_mask();
-  local_denoiser_device_ = create_denoiser_device(path_trace_device_, device_type_mask);
+  local_denoiser_device_ = create_denoiser_device(
+      path_trace_device_, device_type_mask, params_.type);
   denoiser_device_ = local_denoiser_device_.get();
 
   return denoiser_device_;

intern/cycles/integrator/denoiser_oidn.cpp

@@ -37,8 +37,6 @@ OIDNDenoiser::OIDNDenoiser(Device *path_trace_device, const DenoiseParams &param
     : Denoiser(path_trace_device, params)
 {
   DCHECK_EQ(params.type, DENOISER_OPENIMAGEDENOISE);
-
-  DCHECK(openimagedenoise_supported()) << "OpenImageDenoiser is not supported on this platform.";
 }
 
 #ifdef WITH_OPENIMAGEDENOISE
@@ -585,6 +583,9 @@ bool OIDNDenoiser::denoise_buffer(const BufferParams &buffer_params,
                                   const int num_samples,
                                   bool allow_inplace_modification)
 {
+  DCHECK(openimagedenoise_supported())
+      << "OpenImageDenoiser is not supported on this platform or build.";
+
 #ifdef WITH_OPENIMAGEDENOISE
   thread_scoped_lock lock(mutex_);
 
@@ -635,4 +636,20 @@ uint OIDNDenoiser::get_device_type_mask() const
   return DEVICE_MASK_CPU;
 }
 
+Device *OIDNDenoiser::ensure_denoiser_device(Progress *progress)
+{
+#ifndef WITH_OPENIMAGEDENOISE
+  path_trace_device_->set_error("Build without OpenImageDenoiser");
+  return nullptr;
+#else
+  if (!openimagedenoise_supported()) {
+    path_trace_device_->set_error(
+        "OpenImageDenoiser is not supported on this CPU: missing SSE 4.1 support");
+    return nullptr;
+  }
+
+  return Denoiser::ensure_denoiser_device(progress);
+#endif
+}
+
 CCL_NAMESPACE_END

intern/cycles/integrator/denoiser_oidn.h

@@ -38,6 +38,7 @@ class OIDNDenoiser : public Denoiser {
 
  protected:
   virtual uint get_device_type_mask() const override;
+  virtual Device *ensure_denoiser_device(Progress *progress) override;
 
   /* We only perform one denoising at a time, since OpenImageDenoise itself is multithreaded.
    * Use this mutex whenever images are passed to the OIDN and needs to be denoised. */

intern/cycles/integrator/shader_eval.cpp

@@ -157,7 +157,7 @@ bool ShaderEval::eval_gpu(Device *device,
   queue->init_execution();
 
   /* Execute work on GPU in chunk, so we can cancel.
-   * TODO : query appropriate size from device.*/
+   * TODO: query appropriate size from device. */
   const int32_t chunk_size = 65536;
 
   device_ptr d_input = input.device_pointer;

intern/cycles/kernel/bvh/bvh.h

@@ -173,15 +173,16 @@ ccl_device_intersect bool scene_intersect(KernelGlobals kg,
   uint p3 = 0;
   uint p4 = visibility;
   uint p5 = PRIMITIVE_NONE;
+  uint p6 = ((uint64_t)ray) & 0xFFFFFFFF;
+  uint p7 = (((uint64_t)ray) >> 32) & 0xFFFFFFFF;
 
   uint ray_mask = visibility & 0xFF;
-  uint ray_flags = OPTIX_RAY_FLAG_NONE;
+  uint ray_flags = OPTIX_RAY_FLAG_ENFORCE_ANYHIT;
   if (0 == ray_mask && (visibility & ~0xFF) != 0) {
     ray_mask = 0xFF;
-    ray_flags = OPTIX_RAY_FLAG_ENFORCE_ANYHIT;
   }
   else if (visibility & PATH_RAY_SHADOW_OPAQUE) {
-    ray_flags = OPTIX_RAY_FLAG_TERMINATE_ON_FIRST_HIT;
+    ray_flags |= OPTIX_RAY_FLAG_TERMINATE_ON_FIRST_HIT;
   }
 
   optixTrace(scene_intersect_valid(ray) ? kernel_data.bvh.scene : 0,
@@ -200,7 +201,9 @@ ccl_device_intersect bool scene_intersect(KernelGlobals kg,
              p2,
              p3,
              p4,
-             p5);
+             p5,
+             p6,
+             p7);
 
   isect->t = __uint_as_float(p0);
   isect->u = __uint_as_float(p1);
@@ -242,6 +245,7 @@ ccl_device_intersect bool scene_intersect(KernelGlobals kg,
   }
 
   MetalRTIntersectionPayload payload;
+  payload.self = ray->self;
   payload.u = 0.0f;
   payload.v = 0.0f;
   payload.visibility = visibility;
@@ -309,6 +313,7 @@ ccl_device_intersect bool scene_intersect(KernelGlobals kg,
     CCLIntersectContext ctx(kg, CCLIntersectContext::RAY_REGULAR);
     IntersectContext rtc_ctx(&ctx);
     RTCRayHit ray_hit;
+    ctx.ray = ray;
     kernel_embree_setup_rayhit(*ray, ray_hit, visibility);
     rtcIntersect1(kernel_data.bvh.scene, &rtc_ctx.context, &ray_hit);
     if (ray_hit.hit.geomID != RTC_INVALID_GEOMETRY_ID &&
@@ -356,6 +361,9 @@ ccl_device_intersect bool scene_intersect_local(KernelGlobals kg,
   uint p2 = pointer_pack_to_uint_0(local_isect);
   uint p3 = pointer_pack_to_uint_1(local_isect);
   uint p4 = local_object;
+  uint p6 = ((uint64_t)ray) & 0xFFFFFFFF;
+  uint p7 = (((uint64_t)ray) >> 32) & 0xFFFFFFFF;
+
   /* Is set to zero on miss or if ray is aborted, so can be used as return value. */
   uint p5 = max_hits;
 
@@ -379,7 +387,9 @@ ccl_device_intersect bool scene_intersect_local(KernelGlobals kg,
              p2,
              p3,
              p4,
-             p5);
+             p5,
+             p6,
+             p7);
 
   return p5;
 #  elif defined(__METALRT__)
@@ -417,6 +427,7 @@ ccl_device_intersect bool scene_intersect_local(KernelGlobals kg,
   }
 
   MetalRTIntersectionLocalPayload payload;
+  payload.self = ray->self;
   payload.local_object = local_object;
   payload.max_hits = max_hits;
   payload.local_isect.num_hits = 0;
@@ -460,6 +471,7 @@ ccl_device_intersect bool scene_intersect_local(KernelGlobals kg,
         kg, has_bvh ? CCLIntersectContext::RAY_SSS : CCLIntersectContext::RAY_LOCAL);
     ctx.lcg_state = lcg_state;
     ctx.max_hits = max_hits;
+    ctx.ray = ray;
     ctx.local_isect = local_isect;
     if (local_isect) {
       local_isect->num_hits = 0;
@@ -532,6 +544,8 @@ ccl_device_intersect bool scene_intersect_shadow_all(KernelGlobals kg,
   uint p3 = max_hits;
   uint p4 = visibility;
   uint p5 = false;
+  uint p6 = ((uint64_t)ray) & 0xFFFFFFFF;
+  uint p7 = (((uint64_t)ray) >> 32) & 0xFFFFFFFF;
 
   uint ray_mask = visibility & 0xFF;
   if (0 == ray_mask && (visibility & ~0xFF) != 0) {
@@ -555,7 +569,9 @@ ccl_device_intersect bool scene_intersect_shadow_all(KernelGlobals kg,
              p2,
              p3,
              p4,
-             p5);
+             p5,
+             p6,
+             p7);
 
   *num_recorded_hits = uint16_unpack_from_uint_0(p2);
   *throughput = __uint_as_float(p1);
@@ -588,6 +604,7 @@ ccl_device_intersect bool scene_intersect_shadow_all(KernelGlobals kg,
   }
 
   MetalRTIntersectionShadowPayload payload;
+  payload.self = ray->self;
   payload.visibility = visibility;
   payload.max_hits = max_hits;
   payload.num_hits = 0;
@@ -634,6 +651,7 @@ ccl_device_intersect bool scene_intersect_shadow_all(KernelGlobals kg,
     Intersection *isect_array = (Intersection *)state->shadow_isect;
     ctx.isect_s = isect_array;
     ctx.max_hits = max_hits;
+    ctx.ray = ray;
     IntersectContext rtc_ctx(&ctx);
     RTCRay rtc_ray;
     kernel_embree_setup_ray(*ray, rtc_ray, visibility);
@@ -685,6 +703,8 @@ ccl_device_intersect bool scene_intersect_volume(KernelGlobals kg,
   uint p3 = 0;
   uint p4 = visibility;
   uint p5 = PRIMITIVE_NONE;
+  uint p6 = ((uint64_t)ray) & 0xFFFFFFFF;
+  uint p7 = (((uint64_t)ray) >> 32) & 0xFFFFFFFF;
 
   uint ray_mask = visibility & 0xFF;
   if (0 == ray_mask && (visibility & ~0xFF) != 0) {
@@ -708,7 +728,9 @@ ccl_device_intersect bool scene_intersect_volume(KernelGlobals kg,
              p2,
              p3,
              p4,
-             p5);
+             p5,
+             p6,
+             p7);
 
   isect->t = __uint_as_float(p0);
   isect->u = __uint_as_float(p1);
@@ -744,6 +766,7 @@ ccl_device_intersect bool scene_intersect_volume(KernelGlobals kg,
   }
 
   MetalRTIntersectionPayload payload;
+  payload.self = ray->self;
   payload.visibility = visibility;
 
   typename metalrt_intersector_type::result_type intersection;
@@ -820,6 +843,7 @@ ccl_device_intersect uint scene_intersect_volume_all(KernelGlobals kg,
     ctx.isect_s = isect;
     ctx.max_hits = max_hits;
     ctx.num_hits = 0;
+    ctx.ray = ray;
     IntersectContext rtc_ctx(&ctx);
     RTCRay rtc_ray;
     kernel_embree_setup_ray(*ray, rtc_ray, visibility);

intern/cycles/kernel/bvh/embree.h

@@ -22,6 +22,8 @@
 #include "kernel/device/cpu/compat.h"
 #include "kernel/device/cpu/globals.h"
 
+#include "kernel/bvh/util.h"
+
 #include "util/vector.h"
 
 CCL_NAMESPACE_BEGIN
@@ -38,6 +40,9 @@ struct CCLIntersectContext {
   KernelGlobals kg;
   RayType type;
 
+  /* For avoiding self intersections */
+  const Ray *ray;
+
   /* for shadow rays */
   Intersection *isect_s;
   uint max_hits;
@@ -56,6 +61,7 @@ struct CCLIntersectContext {
   {
     kg = kg_;
     type = type_;
+    ray = NULL;
     max_hits = 1;
     num_hits = 0;
     num_recorded_hits = 0;
@@ -102,7 +108,34 @@ ccl_device_inline void kernel_embree_setup_rayhit(const Ray &ray,
 {
   kernel_embree_setup_ray(ray, rayhit.ray, visibility);
   rayhit.hit.geomID = RTC_INVALID_GEOMETRY_ID;
-  rayhit.hit.primID = RTC_INVALID_GEOMETRY_ID;
+  rayhit.hit.instID[0] = RTC_INVALID_GEOMETRY_ID;
+}
+
+ccl_device_inline bool kernel_embree_is_self_intersection(const KernelGlobals kg,
+                                                          const RTCHit *hit,
+                                                          const Ray *ray)
+{
+  bool status = false;
+  if (hit->instID[0] != RTC_INVALID_GEOMETRY_ID) {
+    const int oID = hit->instID[0] / 2;
+    if ((ray->self.object == oID) || (ray->self.light_object == oID)) {
+      RTCScene inst_scene = (RTCScene)rtcGetGeometryUserData(
+          rtcGetGeometry(kernel_data.bvh.scene, hit->instID[0]));
+      const int pID = hit->primID +
+                      (intptr_t)rtcGetGeometryUserData(rtcGetGeometry(inst_scene, hit->geomID));
+      status = intersection_skip_self_shadow(ray->self, oID, pID);
+    }
+  }
+  else {
+    const int oID = hit->geomID / 2;
+    if ((ray->self.object == oID) || (ray->self.light_object == oID)) {
+      const int pID = hit->primID + (intptr_t)rtcGetGeometryUserData(
+                                        rtcGetGeometry(kernel_data.bvh.scene, hit->geomID));
+      status = intersection_skip_self_shadow(ray->self, oID, pID);
+    }
+  }
+
+  return status;
 }
 
 ccl_device_inline void kernel_embree_convert_hit(KernelGlobals kg,

intern/cycles/kernel/bvh/local.h

@@ -157,7 +157,11 @@ ccl_device_inline
                 }
               }
 
+              /* Skip self intersection. */
               const int prim = kernel_tex_fetch(__prim_index, prim_addr);
+              if (intersection_skip_self_local(ray->self, prim)) {
+                continue;
+              }
 
               if (triangle_intersect_local(kg,
                                            local_isect,
@@ -188,7 +192,11 @@ ccl_device_inline
                 }
               }
 
+              /* Skip self intersection. */
               const int prim = kernel_tex_fetch(__prim_index, prim_addr);
+              if (intersection_skip_self_local(ray->self, prim)) {
+                continue;
+              }
 
               if (motion_triangle_intersect_local(kg,
                                                   local_isect,

intern/cycles/kernel/bvh/metal.h

@@ -15,6 +15,7 @@
  */
 
 struct MetalRTIntersectionPayload {
+  RaySelfPrimitives self;
   uint visibility;
   float u, v;
   int prim;
@@ -25,6 +26,7 @@ struct MetalRTIntersectionPayload {
 };
 
 struct MetalRTIntersectionLocalPayload {
+  RaySelfPrimitives self;
   uint local_object;
   uint lcg_state;
   short max_hits;
@@ -34,6 +36,7 @@ struct MetalRTIntersectionLocalPayload {
 };
 
 struct MetalRTIntersectionShadowPayload {
+  RaySelfPrimitives self;
   uint visibility;
 #if defined(__METALRT_MOTION__)
   float time;

intern/cycles/kernel/bvh/shadow_all.h

@@ -160,6 +160,9 @@ ccl_device_inline
                                         kernel_tex_fetch(__prim_object, prim_addr) :
                                         object;
             const int prim = kernel_tex_fetch(__prim_index, prim_addr);
+            if (intersection_skip_self_shadow(ray->self, prim_object, prim)) {
+              continue;
+            }
 
             switch (type & PRIMITIVE_ALL) {
               case PRIMITIVE_TRIANGLE: {

intern/cycles/kernel/bvh/traversal.h

@@ -133,35 +133,29 @@ ccl_device_noinline bool BVH_FUNCTION_FULL_NAME(BVH)(KernelGlobals kg,
           --stack_ptr;
 
           /* primitive intersection */
-          switch (type & PRIMITIVE_ALL) {
-            case PRIMITIVE_TRIANGLE: {
-              for (; prim_addr < prim_addr2; prim_addr++) {
-                kernel_assert(kernel_tex_fetch(__prim_type, prim_addr) == type);
+          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);
+            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 (intersection_skip_self_shadow(ray->self, prim_object, prim)) {
+              continue;
+            }
 
+            switch (type & PRIMITIVE_ALL) {
+              case PRIMITIVE_TRIANGLE: {
                 if (triangle_intersect(
                         kg, isect, P, dir, isect->t, visibility, prim_object, prim, prim_addr)) {
                   /* shadow ray early termination */
                   if (visibility & PATH_RAY_SHADOW_OPAQUE)
                     return true;
                 }
+                break;
               }
-              break;
-            }
 #if BVH_FEATURE(BVH_MOTION)
-            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);
-
+              case PRIMITIVE_MOTION_TRIANGLE: {
                 if (motion_triangle_intersect(kg,
                                               isect,
                                               P,
@@ -176,28 +170,21 @@ ccl_device_noinline bool BVH_FUNCTION_FULL_NAME(BVH)(KernelGlobals kg,
                   if (visibility & PATH_RAY_SHADOW_OPAQUE)
                     return true;
                 }
+                break;
               }
-              break;
-            }
 #endif /* BVH_FEATURE(BVH_MOTION) */
 #if BVH_FEATURE(BVH_HAIR)
-            case PRIMITIVE_CURVE_THICK:
-            case PRIMITIVE_MOTION_CURVE_THICK:
-            case PRIMITIVE_CURVE_RIBBON:
-            case PRIMITIVE_MOTION_CURVE_RIBBON: {
-              for (; prim_addr < prim_addr2; prim_addr++) {
+              case PRIMITIVE_CURVE_THICK:
+              case PRIMITIVE_MOTION_CURVE_THICK:
+              case PRIMITIVE_CURVE_RIBBON:
+              case PRIMITIVE_MOTION_CURVE_RIBBON: {
                 if ((type & PRIMITIVE_MOTION) && kernel_data.bvh.use_bvh_steps) {
                   const float2 prim_time = kernel_tex_fetch(__prim_time, prim_addr);
                   if (ray->time < prim_time.x || ray->time > prim_time.y) {
-                    continue;
+                    break;
                   }
                 }
 
-                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_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);
@@ -206,26 +193,19 @@ ccl_device_noinline bool BVH_FUNCTION_FULL_NAME(BVH)(KernelGlobals kg,
                   if (visibility & PATH_RAY_SHADOW_OPAQUE)
                     return true;
                 }
+                break;
               }
-              break;
-            }
 #endif /* BVH_FEATURE(BVH_HAIR) */
 #if BVH_FEATURE(BVH_POINTCLOUD)
-            case PRIMITIVE_POINT:
-            case PRIMITIVE_MOTION_POINT: {
-              for (; prim_addr < prim_addr2; prim_addr++) {
+              case PRIMITIVE_POINT:
+              case PRIMITIVE_MOTION_POINT: {
                 if ((type & PRIMITIVE_MOTION) && kernel_data.bvh.use_bvh_steps) {
                   const float2 prim_time = kernel_tex_fetch(__prim_time, prim_addr);
                   if (ray->time < prim_time.x || ray->time > prim_time.y) {
-                    continue;
+                    break;
                   }
                 }
 
-                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_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);
@@ -234,10 +214,10 @@ ccl_device_noinline bool BVH_FUNCTION_FULL_NAME(BVH)(KernelGlobals kg,
                   if (visibility & PATH_RAY_SHADOW_OPAQUE)
                     return true;
                 }
+                break;
               }
-              break;
-            }
 #endif /* BVH_FEATURE(BVH_POINTCLOUD) */
+            }
           }
         }
         else {

intern/cycles/kernel/bvh/util.h

@@ -21,54 +21,22 @@ CCL_NAMESPACE_BEGIN
 /* Ray offset to avoid self intersection.
  *
  * This function should be used to compute a modified ray start position for
- * rays leaving from a surface. */
-
+ * rays leaving from a surface. This is from "A Fast and Robust Method for Avoiding
+ * Self-Intersection" see https://research.nvidia.com/publication/2019-03_A-Fast-and
+ */
 ccl_device_inline float3 ray_offset(float3 P, float3 Ng)
 {
-#ifdef __INTERSECTION_REFINE__
-  const float epsilon_f = 1e-5f;
-  /* ideally this should match epsilon_f, but instancing and motion blur
-   * precision makes it problematic */
-  const float epsilon_test = 1.0f;
-  const int epsilon_i = 32;
+  const float int_scale = 256.0f;
+  int3 of_i = make_int3((int)(int_scale * Ng.x), (int)(int_scale * Ng.y), (int)(int_scale * Ng.z));
 
-  float3 res;
-
-  /* x component */
-  if (fabsf(P.x) < epsilon_test) {
-    res.x = P.x + Ng.x * epsilon_f;
-  }
-  else {
-    uint ix = __float_as_uint(P.x);
-    ix += ((ix ^ __float_as_uint(Ng.x)) >> 31) ? -epsilon_i : epsilon_i;
-    res.x = __uint_as_float(ix);
-  }
-
-  /* y component */
-  if (fabsf(P.y) < epsilon_test) {
-    res.y = P.y + Ng.y * epsilon_f;
-  }
-  else {
-    uint iy = __float_as_uint(P.y);
-    iy += ((iy ^ __float_as_uint(Ng.y)) >> 31) ? -epsilon_i : epsilon_i;
-    res.y = __uint_as_float(iy);
-  }
-
-  /* z component */
-  if (fabsf(P.z) < epsilon_test) {
-    res.z = P.z + Ng.z * epsilon_f;
-  }
-  else {
-    uint iz = __float_as_uint(P.z);
-    iz += ((iz ^ __float_as_uint(Ng.z)) >> 31) ? -epsilon_i : epsilon_i;
-    res.z = __uint_as_float(iz);
-  }
-
-  return res;
-#else
-  const float epsilon_f = 1e-4f;
-  return P + epsilon_f * Ng;
-#endif
+  float3 p_i = make_float3(__int_as_float(__float_as_int(P.x) + ((P.x < 0) ? -of_i.x : of_i.x)),
+                           __int_as_float(__float_as_int(P.y) + ((P.y < 0) ? -of_i.y : of_i.y)),
+                           __int_as_float(__float_as_int(P.z) + ((P.z < 0) ? -of_i.z : of_i.z)));
+  const float origin = 1.0f / 32.0f;
+  const float float_scale = 1.0f / 65536.0f;
+  return make_float3(fabsf(P.x) < origin ? P.x + float_scale * Ng.x : p_i.x,
+                     fabsf(P.y) < origin ? P.y + float_scale * Ng.y : p_i.y,
+                     fabsf(P.z) < origin ? P.z + float_scale * Ng.z : p_i.z);
 }
 
 #if defined(__KERNEL_CPU__)
@@ -227,4 +195,25 @@ ccl_device_inline float intersection_curve_shadow_transparency(KernelGlobals kg,
   return (1.0f - u) * f0 + u * f1;
 }
 
+ccl_device_inline bool intersection_skip_self(ccl_private const RaySelfPrimitives &self,
+                                              const int object,
+                                              const int prim)
+{
+  return (self.prim == prim) && (self.object == object);
+}
+
+ccl_device_inline bool intersection_skip_self_shadow(ccl_private const RaySelfPrimitives &self,
+                                                     const int object,
+                                                     const int prim)
+{
+  return ((self.prim == prim) && (self.object == object)) ||
+         ((self.light_prim == prim) && (self.light_object == object));
+}
+
+ccl_device_inline bool intersection_skip_self_local(ccl_private const RaySelfPrimitives &self,
+                                                    const int prim)
+{
+  return (self.prim == prim);
+}
+
 CCL_NAMESPACE_END

intern/cycles/kernel/bvh/volume.h

@@ -144,6 +144,9 @@ ccl_device_inline
                                             kernel_tex_fetch(__prim_object, prim_addr) :
                                             object;
                 const int prim = kernel_tex_fetch(__prim_index, prim_addr);
+                if (intersection_skip_self(ray->self, prim_object, prim)) {
+                  continue;
+                }
 
                 int object_flag = kernel_tex_fetch(__object_flag, prim_object);
                 if ((object_flag & SD_OBJECT_HAS_VOLUME) == 0) {
@@ -164,6 +167,9 @@ ccl_device_inline
                                             kernel_tex_fetch(__prim_object, prim_addr) :
                                             object;
                 const int prim = kernel_tex_fetch(__prim_index, prim_addr);
+                if (intersection_skip_self(ray->self, prim_object, prim)) {
+                  continue;
+                }
                 int object_flag = kernel_tex_fetch(__object_flag, prim_object);
                 if ((object_flag & SD_OBJECT_HAS_VOLUME) == 0) {
                   continue;

intern/cycles/kernel/bvh/volume_all.h

@@ -147,6 +147,9 @@ ccl_device_inline
                                             kernel_tex_fetch(__prim_object, prim_addr) :
                                             object;
                 const int prim = kernel_tex_fetch(__prim_index, prim_addr);
+                if (intersection_skip_self(ray->self, prim_object, prim)) {
+                  continue;
+                }
                 int object_flag = kernel_tex_fetch(__object_flag, prim_object);
                 if ((object_flag & SD_OBJECT_HAS_VOLUME) == 0) {
                   continue;
@@ -188,6 +191,9 @@ ccl_device_inline
                                             kernel_tex_fetch(__prim_object, prim_addr) :
                                             object;
                 const int prim = kernel_tex_fetch(__prim_index, prim_addr);
+                if (intersection_skip_self(ray->self, prim_object, prim)) {
+                  continue;
+                }
                 int object_flag = kernel_tex_fetch(__object_flag, prim_object);
                 if ((object_flag & SD_OBJECT_HAS_VOLUME) == 0) {
                   continue;

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

@@ -40,6 +40,27 @@ struct TriangleIntersectionResult
 
 enum { METALRT_HIT_TRIANGLE, METALRT_HIT_BOUNDING_BOX };
 
+ccl_device_inline bool intersection_skip_self(ray_data const RaySelfPrimitives& self,
+                                              const int object,
+                                              const int prim)
+{
+  return (self.prim == prim) && (self.object == object);
+}
+
+ccl_device_inline bool intersection_skip_self_shadow(ray_data const RaySelfPrimitives& self,
+                                                     const int object,
+                                                     const int prim)
+{
+  return ((self.prim == prim) && (self.object == object)) ||
+         ((self.light_prim == prim) && (self.light_object == object));
+}
+
+ccl_device_inline bool intersection_skip_self_local(ray_data const RaySelfPrimitives& self,
+                                                    const int prim)
+{
+  return (self.prim == prim);
+}
+
 template<typename TReturn, uint intersection_type>
 TReturn metalrt_local_hit(constant KernelParamsMetal &launch_params_metal,
                           ray_data MetalKernelContext::MetalRTIntersectionLocalPayload &payload,
@@ -53,8 +74,8 @@ TReturn metalrt_local_hit(constant KernelParamsMetal &launch_params_metal,
 #ifdef __BVH_LOCAL__
   uint prim = primitive_id + kernel_tex_fetch(__object_prim_offset, object);
 
-  if (object != payload.local_object) {
-    /* Only intersect with matching object */
+  if ((object != payload.local_object) || intersection_skip_self_local(payload.self, prim)) {
+    /* Only intersect with matching object and skip self-intersecton. */
     result.accept = false;
     result.continue_search = true;
     return result;
@@ -166,6 +187,11 @@ bool metalrt_shadow_all_hit(constant KernelParamsMetal &launch_params_metal,
   }
 #  endif
 
+  if (intersection_skip_self_shadow(payload.self, object, prim)) {
+    /* continue search */
+    return true;
+  }
+
   float u = 0.0f, v = 0.0f;
   int type = 0;
   if (intersection_type == METALRT_HIT_TRIANGLE) {
@@ -322,21 +348,35 @@ inline TReturnType metalrt_visibility_test(constant KernelParamsMetal &launch_pa
   }
 #  endif
 
-#  ifdef __VISIBILITY_FLAG__
   uint visibility = payload.visibility;
+#  ifdef __VISIBILITY_FLAG__
   if ((kernel_tex_fetch(__objects, object).visibility & visibility) == 0) {
     result.accept = false;
     result.continue_search = true;
     return result;
   }
+#  endif
 
   /* Shadow ray early termination. */
   if (visibility & PATH_RAY_SHADOW_OPAQUE) {
-    result.accept = true;
-    result.continue_search = false;
-    return result;
+    if (intersection_skip_self_shadow(payload.self, object, prim)) {
+      result.accept = false;
+      result.continue_search = true;
+      return result;
+    }
+    else {
+      result.accept = true;
+      result.continue_search = false;
+      return result;
+    }
+  }
+  else {
+    if (intersection_skip_self(payload.self, object, prim)) {
+      result.accept = false;
+      result.continue_search = true;
+      return result;
+    }
   }
-#  endif
 
   result.accept = true;
   result.continue_search = true;

intern/cycles/kernel/device/optix/kernel.cu

@@ -45,6 +45,11 @@ template<typename T> ccl_device_forceinline T *get_payload_ptr_2()
   return pointer_unpack_from_uint<T>(optixGetPayload_2(), optixGetPayload_3());
 }
 
+template<typename T> ccl_device_forceinline T *get_payload_ptr_6()
+{
+  return (T *)(((uint64_t)optixGetPayload_7() << 32) | optixGetPayload_6());
+}
+
 ccl_device_forceinline int get_object_id()
 {
 #ifdef __OBJECT_MOTION__
@@ -111,6 +116,12 @@ extern "C" __global__ void __anyhit__kernel_optix_local_hit()
     return optixIgnoreIntersection();
   }
 
+  const int prim = optixGetPrimitiveIndex();
+  ccl_private Ray *const ray = get_payload_ptr_6<Ray>();
+  if (intersection_skip_self_local(ray->self, prim)) {
+    return optixIgnoreIntersection();
+  }
+
   const uint max_hits = optixGetPayload_5();
   if (max_hits == 0) {
     /* Special case for when no hit information is requested, just report that something was hit */
@@ -149,8 +160,6 @@ extern "C" __global__ void __anyhit__kernel_optix_local_hit()
     local_isect->num_hits = 1;
   }
 
-  const int prim = optixGetPrimitiveIndex();
-
   Intersection *isect = &local_isect->hits[hit];
   isect->t = optixGetRayTmax();
   isect->prim = prim;
@@ -185,6 +194,11 @@ extern "C" __global__ void __anyhit__kernel_optix_shadow_all_hit()
   }
 #  endif
 
+  ccl_private Ray *const ray = get_payload_ptr_6<Ray>();
+  if (intersection_skip_self_shadow(ray->self, object, prim)) {
+    return optixIgnoreIntersection();
+  }
+
   float u = 0.0f, v = 0.0f;
   int type = 0;
   if (optixIsTriangleHit()) {
@@ -314,6 +328,12 @@ extern "C" __global__ void __anyhit__kernel_optix_volume_test()
   if ((kernel_tex_fetch(__object_flag, object) & SD_OBJECT_HAS_VOLUME) == 0) {
     return optixIgnoreIntersection();
   }
+
+  const int prim = optixGetPrimitiveIndex();
+  ccl_private Ray *const ray = get_payload_ptr_6<Ray>();
+  if (intersection_skip_self(ray->self, object, prim)) {
+    return optixIgnoreIntersection();
+  }
 }
 
 extern "C" __global__ void __anyhit__kernel_optix_visibility_test()
@@ -330,18 +350,31 @@ extern "C" __global__ void __anyhit__kernel_optix_visibility_test()
 #  endif
 #endif
 
-#ifdef __VISIBILITY_FLAG__
   const uint object = get_object_id();
   const uint visibility = optixGetPayload_4();
+#ifdef __VISIBILITY_FLAG__
   if ((kernel_tex_fetch(__objects, object).visibility & visibility) == 0) {
     return optixIgnoreIntersection();
   }
-
-  /* Shadow ray early termination. */
-  if (visibility & PATH_RAY_SHADOW_OPAQUE) {
-    return optixTerminateRay();
-  }
 #endif
+
+  const int prim = optixGetPrimitiveIndex();
+  ccl_private Ray *const ray = get_payload_ptr_6<Ray>();
+
+  if (visibility & PATH_RAY_SHADOW_OPAQUE) {
+    if (intersection_skip_self_shadow(ray->self, object, prim)) {
+      return optixIgnoreIntersection();
+    }
+    else {
+      /* Shadow ray early termination. */
+      return optixTerminateRay();
+    }
+  }
+  else {
+    if (intersection_skip_self(ray->self, object, prim)) {
+      return optixIgnoreIntersection();
+    }
+  }
 }
 
 extern "C" __global__ void __closesthit__kernel_optix_hit()

intern/cycles/kernel/geom/motion_triangle_intersect.h

@@ -29,46 +29,19 @@
 
 CCL_NAMESPACE_BEGIN
 
-/* Refine triangle intersection to more precise hit point. For rays that travel
- * far the precision is often not so good, this reintersects the primitive from
- * a closer distance.
+/**
+ * Use the barycentric coordinates to get the intersection location
  */
-
-ccl_device_inline float3 motion_triangle_refine(KernelGlobals kg,
-                                                ccl_private ShaderData *sd,
-                                                float3 P,
-                                                float3 D,
-                                                float t,
-                                                const int isect_object,
-                                                const int isect_prim,
-                                                float3 verts[3])
+ccl_device_inline float3 motion_triangle_point_from_uv(KernelGlobals kg,
+                                                       ccl_private ShaderData *sd,
+                                                       const int isect_object,
+                                                       const int isect_prim,
+                                                       const float u,
+                                                       const float v,
+                                                       float3 verts[3])
 {
-#ifdef __INTERSECTION_REFINE__
-  if (!(sd->object_flag & SD_OBJECT_TRANSFORM_APPLIED)) {
-    if (UNLIKELY(t == 0.0f)) {
-      return P;
-    }
-    const Transform tfm = object_get_inverse_transform(kg, sd);
-
-    P = transform_point(&tfm, P);
-    D = transform_direction(&tfm, D * t);
-    D = normalize_len(D, &t);
-  }
-
-  P = P + D * t;
-
-  /* Compute refined intersection distance. */
-  const float3 e1 = verts[0] - verts[2];
-  const float3 e2 = verts[1] - verts[2];
-  const float3 s1 = cross(D, e2);
-
-  const float invdivisor = 1.0f / dot(s1, e1);
-  const float3 d = P - verts[2];
-  const float3 s2 = cross(d, e1);
-  float rt = dot(e2, s2) * invdivisor;
-
-  /* Compute refined position. */
-  P = P + D * rt;
+  float w = 1.0f - u - v;
+  float3 P = u * verts[0] + v * verts[1] + w * verts[2];
 
   if (!(sd->object_flag & SD_OBJECT_TRANSFORM_APPLIED)) {
     const Transform tfm = object_get_transform(kg, sd);
@@ -76,71 +49,8 @@ ccl_device_inline float3 motion_triangle_refine(KernelGlobals kg,
   }
 
   return P;
-#else
-  return P + D * t;
-#endif
 }
 
-/* Same as above, except that t is assumed to be in object space
- * for instancing.
- */
-
-#ifdef __BVH_LOCAL__
-#  if defined(__KERNEL_CUDA__) && (defined(i386) || defined(_M_IX86))
-ccl_device_noinline
-#  else
-ccl_device_inline
-#  endif
-    float3
-    motion_triangle_refine_local(KernelGlobals kg,
-                                 ccl_private ShaderData *sd,
-                                 float3 P,
-                                 float3 D,
-                                 float t,
-                                 const int isect_object,
-                                 const int isect_prim,
-                                 float3 verts[3])
-{
-#  if defined(__KERNEL_GPU_RAYTRACING__)
-  /* t is always in world space with OptiX and MetalRT. */
-  return motion_triangle_refine(kg, sd, P, D, t, isect_object, isect_prim, verts);
-#  else
-#    ifdef __INTERSECTION_REFINE__
-  if (!(sd->object_flag & SD_OBJECT_TRANSFORM_APPLIED)) {
-    const Transform tfm = object_get_inverse_transform(kg, sd);
-
-    P = transform_point(&tfm, P);
-    D = transform_direction(&tfm, D);
-    D = normalize(D);
-  }
-
-  P = P + D * t;
-
-  /* compute refined intersection distance */
-  const float3 e1 = verts[0] - verts[2];
-  const float3 e2 = verts[1] - verts[2];
-  const float3 s1 = cross(D, e2);
-
-  const float invdivisor = 1.0f / dot(s1, e1);
-  const float3 d = P - verts[2];
-  const float3 s2 = cross(d, e1);
-  float rt = dot(e2, s2) * invdivisor;
-
-  P = P + D * rt;
-
-  if (!(sd->object_flag & SD_OBJECT_TRANSFORM_APPLIED)) {
-    const Transform tfm = object_get_transform(kg, sd);
-    P = transform_point(&tfm, P);
-  }
-
-  return P;
-#    else  /* __INTERSECTION_REFINE__ */
-  return P + D * t;
-#    endif /* __INTERSECTION_REFINE__ */
-#  endif
-}
-#endif /* __BVH_LOCAL__ */
-
 /* Ray intersection. We simply compute the vertex positions at the given ray
  * time and do a ray intersection with the resulting triangle.
  */

intern/cycles/kernel/geom/motion_triangle_shader.h

@@ -68,15 +68,7 @@ ccl_device_noinline void motion_triangle_shader_setup(KernelGlobals kg,
   verts[1] = (1.0f - t) * verts[1] + t * next_verts[1];
   verts[2] = (1.0f - t) * verts[2] + t * next_verts[2];
   /* Compute refined position. */
-#ifdef __BVH_LOCAL__
-  if (is_local) {
-    sd->P = motion_triangle_refine_local(kg, sd, P, D, ray_t, isect_object, isect_prim, verts);
-  }
-  else
-#endif /* __BVH_LOCAL__*/
-  {
-    sd->P = motion_triangle_refine(kg, sd, P, D, ray_t, isect_object, isect_prim, verts);
-  }
+  sd->P = motion_triangle_point_from_uv(kg, sd, isect_object, isect_prim, sd->u, sd->v, verts);
   /* Compute face normal. */
   float3 Ng;
   if (sd->object_flag & SD_OBJECT_NEGATIVE_SCALE_APPLIED) {

intern/cycles/kernel/geom/shader_data.h

@@ -89,7 +89,7 @@ ccl_device_inline void shader_setup_from_ray(KernelGlobals kg,
       sd->shader = kernel_tex_fetch(__tri_shader, sd->prim);
 
       /* vectors */
-      sd->P = triangle_refine(kg, sd, ray->P, ray->D, isect->t, isect->object, isect->prim);
+      sd->P = triangle_point_from_uv(kg, sd, isect->object, isect->prim, isect->u, isect->v);
       sd->Ng = Ng;
       sd->N = Ng;
 
@@ -190,40 +190,46 @@ ccl_device_inline void shader_setup_from_sample(KernelGlobals kg,
 #ifdef __OBJECT_MOTION__
     shader_setup_object_transforms(kg, sd, time);
 #endif
-  }
-  else if (lamp != LAMP_NONE) {
-    sd->lamp = lamp;
-  }
 
-  /* transform into world space */
-  if (object_space) {
-    object_position_transform_auto(kg, sd, &sd->P);
-    object_normal_transform_auto(kg, sd, &sd->Ng);
-    sd->N = sd->Ng;
-    object_dir_transform_auto(kg, sd, &sd->I);
-  }
+    /* transform into world space */
+    if (object_space) {
+      object_position_transform_auto(kg, sd, &sd->P);
+      object_normal_transform_auto(kg, sd, &sd->Ng);
+      sd->N = sd->Ng;
+      object_dir_transform_auto(kg, sd, &sd->I);
+    }
 
-  if (sd->type == PRIMITIVE_TRIANGLE) {
-    /* smooth normal */
-    if (sd->shader & SHADER_SMOOTH_NORMAL) {
-      sd->N = triangle_smooth_normal(kg, Ng, sd->prim, sd->u, sd->v);
+    if (sd->type == PRIMITIVE_TRIANGLE) {
+      /* smooth normal */
+      if (sd->shader & SHADER_SMOOTH_NORMAL) {
+        sd->N = triangle_smooth_normal(kg, Ng, sd->prim, sd->u, sd->v);
+
+        if (!(sd->object_flag & SD_OBJECT_TRANSFORM_APPLIED)) {
+          object_normal_transform_auto(kg, sd, &sd->N);
+        }
+      }
+
+      /* dPdu/dPdv */
+#ifdef __DPDU__
+      triangle_dPdudv(kg, sd->prim, &sd->dPdu, &sd->dPdv);
 
       if (!(sd->object_flag & SD_OBJECT_TRANSFORM_APPLIED)) {
-        object_normal_transform_auto(kg, sd, &sd->N);
+        object_dir_transform_auto(kg, sd, &sd->dPdu);
+        object_dir_transform_auto(kg, sd, &sd->dPdv);
       }
-    }
-
-    /* dPdu/dPdv */
-#ifdef __DPDU__
-    triangle_dPdudv(kg, sd->prim, &sd->dPdu, &sd->dPdv);
-
-    if (!(sd->object_flag & SD_OBJECT_TRANSFORM_APPLIED)) {
-      object_dir_transform_auto(kg, sd, &sd->dPdu);
-      object_dir_transform_auto(kg, sd, &sd->dPdv);
-    }
 #endif
+    }
+    else {
+#ifdef __DPDU__
+      sd->dPdu = zero_float3();
+      sd->dPdv = zero_float3();
+#endif
+    }
   }
   else {
+    if (lamp != LAMP_NONE) {
+      sd->lamp = lamp;
+    }
 #ifdef __DPDU__
     sd->dPdu = zero_float3();
     sd->dPdv = zero_float3();

intern/cycles/kernel/geom/triangle_intersect.h

@@ -142,58 +142,23 @@ ccl_device_inline bool triangle_intersect_local(KernelGlobals kg,
 }
 #endif /* __BVH_LOCAL__ */
 
-/* Refine triangle intersection to more precise hit point. For rays that travel
- * far the precision is often not so good, this reintersects the primitive from
- * a closer distance. */
-
-/* Reintersections uses the paper:
- *
- * Tomas Moeller
- * Fast, minimum storage ray/triangle intersection
- * http://www.cs.virginia.edu/~gfx/Courses/2003/ImageSynthesis/papers/Acceleration/Fast%20MinimumStorage%20RayTriangle%20Intersection.pdf
+/**
+ * Use the barycentric coordinates to get the intersection location
  */
-
-ccl_device_inline float3 triangle_refine(KernelGlobals kg,
-                                         ccl_private ShaderData *sd,
-                                         float3 P,
-                                         float3 D,
-                                         float t,
-                                         const int isect_object,
-                                         const int isect_prim)
+ccl_device_inline float3 triangle_point_from_uv(KernelGlobals kg,
+                                                ccl_private ShaderData *sd,
+                                                const int isect_object,
+                                                const int isect_prim,
+                                                const float u,
+                                                const float v)
 {
-#ifdef __INTERSECTION_REFINE__
-  if (!(sd->object_flag & SD_OBJECT_TRANSFORM_APPLIED)) {
-    if (UNLIKELY(t == 0.0f)) {
-      return P;
-    }
-    const Transform tfm = object_get_inverse_transform(kg, sd);
-
-    P = transform_point(&tfm, P);
-    D = transform_direction(&tfm, D * t);
-    D = normalize_len(D, &t);
-  }
-
-  P = P + D * t;
-
   const uint tri_vindex = kernel_tex_fetch(__tri_vindex, isect_prim).w;
   const packed_float3 tri_a = kernel_tex_fetch(__tri_verts, tri_vindex + 0),
                       tri_b = kernel_tex_fetch(__tri_verts, tri_vindex + 1),
                       tri_c = kernel_tex_fetch(__tri_verts, tri_vindex + 2);
-  float3 edge1 = make_float3(tri_a.x - tri_c.x, tri_a.y - tri_c.y, tri_a.z - tri_c.z);
-  float3 edge2 = make_float3(tri_b.x - tri_c.x, tri_b.y - tri_c.y, tri_b.z - tri_c.z);
-  float3 tvec = make_float3(P.x - tri_c.x, P.y - tri_c.y, P.z - tri_c.z);
-  float3 qvec = cross(tvec, edge1);
-  float3 pvec = cross(D, edge2);
-  float det = dot(edge1, pvec);
-  if (det != 0.0f) {
-    /* If determinant is zero it means ray lies in the plane of
-     * the triangle. It is possible in theory due to watertight
-     * nature of triangle intersection. For such cases we simply
-     * don't refine intersection hoping it'll go all fine.
-     */
-    float rt = dot(edge2, qvec) / det;
-    P = P + D * rt;
-  }
+  float w = 1.0f - u - v;
+
+  float3 P = u * tri_a + v * tri_b + w * tri_c;
 
   if (!(sd->object_flag & SD_OBJECT_TRANSFORM_APPLIED)) {
     const Transform tfm = object_get_transform(kg, sd);
@@ -201,65 +166,6 @@ ccl_device_inline float3 triangle_refine(KernelGlobals kg,
   }
 
   return P;
-#else
-  return P + D * t;
-#endif
-}
-
-/* Same as above, except that t is assumed to be in object space for
- * instancing.
- */
-ccl_device_inline float3 triangle_refine_local(KernelGlobals kg,
-                                               ccl_private ShaderData *sd,
-                                               float3 P,
-                                               float3 D,
-                                               float t,
-                                               const int isect_object,
-                                               const int isect_prim)
-{
-#if defined(__KERNEL_GPU_RAYTRACING__)
-  /* t is always in world space with OptiX and MetalRT. */
-  return triangle_refine(kg, sd, P, D, t, isect_object, isect_prim);
-#else
-  if (!(sd->object_flag & SD_OBJECT_TRANSFORM_APPLIED)) {
-    const Transform tfm = object_get_inverse_transform(kg, sd);
-
-    P = transform_point(&tfm, P);
-    D = transform_direction(&tfm, D);
-    D = normalize(D);
-  }
-
-  P = P + D * t;
-
-#  ifdef __INTERSECTION_REFINE__
-  const uint tri_vindex = kernel_tex_fetch(__tri_vindex, isect_prim).w;
-  const packed_float3 tri_a = kernel_tex_fetch(__tri_verts, tri_vindex + 0),
-                      tri_b = kernel_tex_fetch(__tri_verts, tri_vindex + 1),
-                      tri_c = kernel_tex_fetch(__tri_verts, tri_vindex + 2);
-  float3 edge1 = make_float3(tri_a.x - tri_c.x, tri_a.y - tri_c.y, tri_a.z - tri_c.z);
-  float3 edge2 = make_float3(tri_b.x - tri_c.x, tri_b.y - tri_c.y, tri_b.z - tri_c.z);
-  float3 tvec = make_float3(P.x - tri_c.x, P.y - tri_c.y, P.z - tri_c.z);
-  float3 qvec = cross(tvec, edge1);
-  float3 pvec = cross(D, edge2);
-  float det = dot(edge1, pvec);
-  if (det != 0.0f) {
-    /* If determinant is zero it means ray lies in the plane of
-     * the triangle. It is possible in theory due to watertight
-     * nature of triangle intersection. For such cases we simply
-     * don't refine intersection hoping it'll go all fine.
-     */
-    float rt = dot(edge2, qvec) / det;
-    P = P + D * rt;
-  }
-#  endif /* __INTERSECTION_REFINE__ */
-
-  if (!(sd->object_flag & SD_OBJECT_TRANSFORM_APPLIED)) {
-    const Transform tfm = object_get_transform(kg, sd);
-    P = transform_point(&tfm, P);
-  }
-
-  return P;
-#endif
 }
 
 CCL_NAMESPACE_END

intern/cycles/kernel/integrator/intersect_closest.h

@@ -328,6 +328,12 @@ ccl_device void integrator_intersect_closest(KernelGlobals kg,
 
   /* Scene Intersection. */
   Intersection isect ccl_optional_struct_init;
+  isect.object = OBJECT_NONE;
+  isect.prim = PRIM_NONE;
+  ray.self.object = last_isect_object;
+  ray.self.prim = last_isect_prim;
+  ray.self.light_object = OBJECT_NONE;
+  ray.self.light_prim = PRIM_NONE;
   bool hit = scene_intersect(kg, &ray, visibility, &isect);
 
   /* TODO: remove this and do it in the various intersection functions instead. */

intern/cycles/kernel/integrator/intersect_shadow.h

@@ -156,7 +156,10 @@ ccl_device void integrator_intersect_shadow(KernelGlobals kg, IntegratorShadowSt
   /* Read ray from integrator state into local memory. */
   Ray ray ccl_optional_struct_init;
   integrator_state_read_shadow_ray(kg, state, &ray);
-
+  ray.self.object = INTEGRATOR_STATE_ARRAY(state, shadow_isect, 0, object);
+  ray.self.prim = INTEGRATOR_STATE_ARRAY(state, shadow_isect, 0, prim);
+  ray.self.light_object = INTEGRATOR_STATE_ARRAY(state, shadow_isect, 1, object);
+  ray.self.light_prim = INTEGRATOR_STATE_ARRAY(state, shadow_isect, 1, prim);
   /* Compute visibility. */
   const uint visibility = integrate_intersect_shadow_visibility(kg, state);
 

intern/cycles/kernel/integrator/intersect_volume_stack.h

@@ -38,7 +38,10 @@ ccl_device void integrator_volume_stack_update_for_subsurface(KernelGlobals kg,
   Ray volume_ray ccl_optional_struct_init;
   volume_ray.P = from_P;
   volume_ray.D = normalize_len(to_P - from_P, &volume_ray.t);
-
+  volume_ray.self.object = INTEGRATOR_STATE(state, isect, object);
+  volume_ray.self.prim = INTEGRATOR_STATE(state, isect, prim);
+  volume_ray.self.light_object = OBJECT_NONE;
+  volume_ray.self.light_prim = PRIM_NONE;
   /* Store to avoid global fetches on every intersection step. */
   const uint volume_stack_size = kernel_data.volume_stack_size;
 
@@ -68,7 +71,7 @@ ccl_device void integrator_volume_stack_update_for_subsurface(KernelGlobals kg,
     volume_stack_enter_exit(kg, state, stack_sd);
 
     /* Move ray forward. */
-    volume_ray.P = ray_offset(stack_sd->P, -stack_sd->Ng);
+    volume_ray.P = stack_sd->P;
     if (volume_ray.t != FLT_MAX) {
       volume_ray.D = normalize_len(to_P - volume_ray.P, &volume_ray.t);
     }
@@ -91,6 +94,10 @@ ccl_device void integrator_volume_stack_init(KernelGlobals kg, IntegratorState s
    * fewest hits. */
   volume_ray.D = make_float3(0.0f, 0.0f, 1.0f);
   volume_ray.t = FLT_MAX;
+  volume_ray.self.object = OBJECT_NONE;
+  volume_ray.self.prim = PRIM_NONE;
+  volume_ray.self.light_object = OBJECT_NONE;
+  volume_ray.self.light_prim = PRIM_NONE;
 
   int stack_index = 0, enclosed_index = 0;
 
@@ -203,7 +210,7 @@ ccl_device void integrator_volume_stack_init(KernelGlobals kg, IntegratorState s
     }
 
     /* Move ray forward. */
-    volume_ray.P = ray_offset(stack_sd->P, -stack_sd->Ng);
+    volume_ray.P = stack_sd->P;
     ++step;
   }
 #endif

intern/cycles/kernel/integrator/shade_light.h

@@ -37,8 +37,9 @@ ccl_device_inline void integrate_light(KernelGlobals kg,
 
   /* Advance ray beyond light. */
   /* TODO: can we make this more numerically robust to avoid reintersecting the
-   * same light in some cases? */
-  const float3 new_ray_P = ray_offset(ray_P + ray_D * isect.t, ray_D);
+   * same light in some cases? Ray should not intersect surface anymore as the
+   * object and prim ids will prevent self intersection. */
+  const float3 new_ray_P = ray_P + ray_D * isect.t;
   INTEGRATOR_STATE_WRITE(state, ray, P) = new_ray_P;
   INTEGRATOR_STATE_WRITE(state, ray, t) -= isect.t;
 
@@ -46,7 +47,7 @@ ccl_device_inline void integrate_light(KernelGlobals kg,
   const float mis_ray_t = INTEGRATOR_STATE(state, path, mis_ray_t);
   ray_P -= ray_D * mis_ray_t;
   isect.t += mis_ray_t;
-  INTEGRATOR_STATE_WRITE(state, path, mis_ray_t) = mis_ray_t + isect.t;
+  INTEGRATOR_STATE_WRITE(state, path, mis_ray_t) = isect.t;
 
   LightSample ls ccl_optional_struct_init;
   const bool use_light_sample = light_sample_from_intersection(kg, &isect, ray_P, ray_D, &ls);

intern/cycles/kernel/integrator/shade_shadow.h

@@ -83,7 +83,10 @@ ccl_device_inline void integrate_transparent_volume_shadow(KernelGlobals kg,
   /* Setup shader data. */
   Ray ray ccl_optional_struct_init;
   integrator_state_read_shadow_ray(kg, state, &ray);
-
+  ray.self.object = OBJECT_NONE;
+  ray.self.prim = PRIM_NONE;
+  ray.self.light_object = OBJECT_NONE;
+  ray.self.light_prim = PRIM_NONE;
   /* Modify ray position and length to match current segment. */
   const float start_t = (hit == 0) ? 0.0f :
                                      INTEGRATOR_STATE_ARRAY(state, shadow_isect, hit - 1, t);
@@ -149,7 +152,7 @@ ccl_device_inline bool integrate_transparent_shadow(KernelGlobals kg,
     const float last_hit_t = INTEGRATOR_STATE_ARRAY(state, shadow_isect, num_recorded_hits - 1, t);
     const float3 ray_P = INTEGRATOR_STATE(state, shadow_ray, P);
     const float3 ray_D = INTEGRATOR_STATE(state, shadow_ray, D);
-    INTEGRATOR_STATE_WRITE(state, shadow_ray, P) = ray_offset(ray_P + last_hit_t * ray_D, ray_D);
+    INTEGRATOR_STATE_WRITE(state, shadow_ray, P) = ray_P + last_hit_t * ray_D;
     INTEGRATOR_STATE_WRITE(state, shadow_ray, t) -= last_hit_t;
   }
 

intern/cycles/kernel/integrator/shade_surface.h

@@ -182,6 +182,11 @@ ccl_device_forceinline void integrate_surface_direct_light(KernelGlobals kg,
 
   /* Write shadow ray and associated state to global memory. */
   integrator_state_write_shadow_ray(kg, shadow_state, &ray);
+  // Save memory by storing the light and object indices in the shadow_isect
+  INTEGRATOR_STATE_ARRAY_WRITE(shadow_state, shadow_isect, 0, object) = ray.self.object;
+  INTEGRATOR_STATE_ARRAY_WRITE(shadow_state, shadow_isect, 0, prim) = ray.self.prim;
+  INTEGRATOR_STATE_ARRAY_WRITE(shadow_state, shadow_isect, 1, object) = ray.self.light_object;
+  INTEGRATOR_STATE_ARRAY_WRITE(shadow_state, shadow_isect, 1, prim) = ray.self.light_prim;
 
   /* Copy state from main path to shadow path. */
   const uint16_t bounce = INTEGRATOR_STATE(state, path, bounce);
@@ -266,13 +271,11 @@ ccl_device_forceinline int integrate_surface_bsdf_bssrdf_bounce(
   }
 
   /* Setup ray. Note that clipping works through transparent bounces. */
-  INTEGRATOR_STATE_WRITE(state, ray, P) = ray_offset(sd->P,
-                                                     (label & LABEL_TRANSMIT) ? -sd->Ng : sd->Ng);
+  INTEGRATOR_STATE_WRITE(state, ray, P) = sd->P;
   INTEGRATOR_STATE_WRITE(state, ray, D) = normalize(bsdf_omega_in);
   INTEGRATOR_STATE_WRITE(state, ray, t) = (label & LABEL_TRANSPARENT) ?
                                               INTEGRATOR_STATE(state, ray, t) - sd->ray_length :
                                               FLT_MAX;
-
 #ifdef __RAY_DIFFERENTIALS__
   INTEGRATOR_STATE_WRITE(state, ray, dP) = differential_make_compact(sd->dP);
   INTEGRATOR_STATE_WRITE(state, ray, dD) = differential_make_compact(bsdf_domega_in);
@@ -316,7 +319,7 @@ ccl_device_forceinline bool integrate_surface_volume_only_bounce(IntegratorState
   }
 
   /* Setup ray position, direction stays unchanged. */
-  INTEGRATOR_STATE_WRITE(state, ray, P) = ray_offset(sd->P, -sd->Ng);
+  INTEGRATOR_STATE_WRITE(state, ray, P) = sd->P;
 
   /* Clipping works through transparent. */
   INTEGRATOR_STATE_WRITE(state, ray, t) -= sd->ray_length;
@@ -360,10 +363,14 @@ ccl_device_forceinline void integrate_surface_ao(KernelGlobals kg,
   }
 
   Ray ray ccl_optional_struct_init;
-  ray.P = ray_offset(sd->P, sd->Ng);
+  ray.P = sd->P;
   ray.D = ao_D;
   ray.t = kernel_data.integrator.ao_bounces_distance;
   ray.time = sd->time;
+  ray.self.object = sd->object;
+  ray.self.prim = sd->prim;
+  ray.self.light_object = OBJECT_NONE;
+  ray.self.light_prim = PRIM_NONE;
   ray.dP = differential_zero_compact();
   ray.dD = differential_zero_compact();
 
@@ -375,6 +382,10 @@ ccl_device_forceinline void integrate_surface_ao(KernelGlobals kg,
 
   /* Write shadow ray and associated state to global memory. */
   integrator_state_write_shadow_ray(kg, shadow_state, &ray);
+  INTEGRATOR_STATE_ARRAY_WRITE(shadow_state, shadow_isect, 0, object) = ray.self.object;
+  INTEGRATOR_STATE_ARRAY_WRITE(shadow_state, shadow_isect, 0, prim) = ray.self.prim;
+  INTEGRATOR_STATE_ARRAY_WRITE(shadow_state, shadow_isect, 1, object) = ray.self.light_object;
+  INTEGRATOR_STATE_ARRAY_WRITE(shadow_state, shadow_isect, 1, prim) = ray.self.light_prim;
 
   /* Copy state from main path to shadow path. */
   const uint16_t bounce = INTEGRATOR_STATE(state, path, bounce);

intern/cycles/kernel/integrator/shade_volume.h

@@ -791,6 +791,10 @@ ccl_device_forceinline void integrate_volume_direct_light(
 
   /* Write shadow ray and associated state to global memory. */
   integrator_state_write_shadow_ray(kg, shadow_state, &ray);
+  INTEGRATOR_STATE_ARRAY_WRITE(shadow_state, shadow_isect, 0, object) = ray.self.object;
+  INTEGRATOR_STATE_ARRAY_WRITE(shadow_state, shadow_isect, 0, prim) = ray.self.prim;
+  INTEGRATOR_STATE_ARRAY_WRITE(shadow_state, shadow_isect, 1, object) = ray.self.light_object;
+  INTEGRATOR_STATE_ARRAY_WRITE(shadow_state, shadow_isect, 1, prim) = ray.self.light_prim;
 
   /* Copy state from main path to shadow path. */
   const uint16_t bounce = INTEGRATOR_STATE(state, path, bounce);
@@ -873,11 +877,13 @@ ccl_device_forceinline bool integrate_volume_phase_scatter(
   INTEGRATOR_STATE_WRITE(state, ray, P) = sd->P;
   INTEGRATOR_STATE_WRITE(state, ray, D) = normalize(phase_omega_in);
   INTEGRATOR_STATE_WRITE(state, ray, t) = FLT_MAX;
-
 #  ifdef __RAY_DIFFERENTIALS__
   INTEGRATOR_STATE_WRITE(state, ray, dP) = differential_make_compact(sd->dP);
   INTEGRATOR_STATE_WRITE(state, ray, dD) = differential_make_compact(phase_domega_in);
 #  endif
+  // Save memory by storing last hit prim and object in isect
+  INTEGRATOR_STATE_WRITE(state, isect, prim) = sd->prim;
+  INTEGRATOR_STATE_WRITE(state, isect, object) = sd->object;
 
   /* Update throughput. */
   const float3 throughput = INTEGRATOR_STATE(state, path, throughput);

intern/cycles/kernel/integrator/shadow_state_template.h

@@ -61,6 +61,7 @@ KERNEL_STRUCT_MEMBER(shadow_ray, packed_float3, D, KERNEL_FEATURE_PATH_TRACING)
 KERNEL_STRUCT_MEMBER(shadow_ray, float, t, KERNEL_FEATURE_PATH_TRACING)
 KERNEL_STRUCT_MEMBER(shadow_ray, float, time, KERNEL_FEATURE_PATH_TRACING)
 KERNEL_STRUCT_MEMBER(shadow_ray, float, dP, KERNEL_FEATURE_PATH_TRACING)
+KERNEL_STRUCT_MEMBER(shadow_ray, int, object, KERNEL_FEATURE_PATH_TRACING)
 KERNEL_STRUCT_END(shadow_ray)
 
 /*********************** Shadow Intersection result **************************/

intern/cycles/kernel/integrator/subsurface.h

@@ -57,7 +57,6 @@ ccl_device int subsurface_bounce(KernelGlobals kg,
 
   /* Pass along object info, reusing isect to save memory. */
   INTEGRATOR_STATE_WRITE(state, subsurface, Ng) = sd->Ng;
-  INTEGRATOR_STATE_WRITE(state, isect, object) = sd->object;
 
   uint32_t path_flag = (INTEGRATOR_STATE(state, path, flag) & ~PATH_RAY_CAMERA) |
                        ((sc->type == CLOSURE_BSSRDF_BURLEY_ID) ? PATH_RAY_SUBSURFACE_DISK :
@@ -165,10 +164,8 @@ ccl_device_inline bool subsurface_scatter(KernelGlobals kg, IntegratorState stat
 
     if (object_flag & SD_OBJECT_INTERSECTS_VOLUME) {
       float3 P = INTEGRATOR_STATE(state, ray, P);
-      const float3 Ng = INTEGRATOR_STATE(state, subsurface, Ng);
-      const float3 offset_P = ray_offset(P, -Ng);
 
-      integrator_volume_stack_update_for_subsurface(kg, state, offset_P, ray.P);
+      integrator_volume_stack_update_for_subsurface(kg, state, P, ray.P);
     }
   }
 #  endif /* __VOLUME__ */

intern/cycles/kernel/integrator/subsurface_disk.h

@@ -99,6 +99,10 @@ ccl_device_inline bool subsurface_disk(KernelGlobals kg,
   ray.dP = ray_dP;
   ray.dD = differential_zero_compact();
   ray.time = time;
+  ray.self.object = OBJECT_NONE;
+  ray.self.prim = PRIM_NONE;
+  ray.self.light_object = OBJECT_NONE;
+  ray.self.light_prim = OBJECT_NONE;
 
   /* Intersect with the same object. if multiple intersections are found it
    * will use at most BSSRDF_MAX_HITS hits, a random subset of all hits. */

intern/cycles/kernel/integrator/subsurface_random_walk.h

@@ -195,6 +195,7 @@ ccl_device_inline bool subsurface_random_walk(KernelGlobals kg,
   const float time = INTEGRATOR_STATE(state, ray, time);
   const float3 Ng = INTEGRATOR_STATE(state, subsurface, Ng);
   const int object = INTEGRATOR_STATE(state, isect, object);
+  const int prim = INTEGRATOR_STATE(state, isect, prim);
 
   /* Sample diffuse surface scatter into the object. */
   float3 D;
@@ -205,12 +206,16 @@ ccl_device_inline bool subsurface_random_walk(KernelGlobals kg,
   }
 
   /* Setup ray. */
-  ray.P = ray_offset(P, -Ng);
+  ray.P = P;
   ray.D = D;
   ray.t = FLT_MAX;
   ray.time = time;
   ray.dP = ray_dP;
   ray.dD = differential_zero_compact();
+  ray.self.object = object;
+  ray.self.prim = prim;
+  ray.self.light_object = OBJECT_NONE;
+  ray.self.light_prim = PRIM_NONE;
 
 #ifndef __KERNEL_GPU_RAYTRACING__
   /* Compute or fetch object transforms. */
@@ -377,7 +382,15 @@ ccl_device_inline bool subsurface_random_walk(KernelGlobals kg,
      * If yes, we will later use backwards guided sampling in order to have a decent
      * chance of connecting to it.
      * TODO: Maybe use less than 10 times the mean free path? */
-    ray.t = (bounce == 0) ? max(t, 10.0f / (min3(sigma_t))) : t;
+    if (bounce == 0) {
+      ray.t = max(t, 10.0f / (min3(sigma_t)));
+    }
+    else {
+      ray.t = t;
+      /* After the first bounce the object can intersect the same surface again */
+      ray.self.object = OBJECT_NONE;
+      ray.self.prim = PRIM_NONE;
+    }
     scene_intersect_local(kg, &ray, &ss_isect, object, NULL, 1);
     hit = (ss_isect.num_hits > 0);
 
@@ -408,13 +421,6 @@ ccl_device_inline bool subsurface_random_walk(KernelGlobals kg,
     if (hit) {
       t = ray.t;
     }
-    else if (bounce == 0) {
-      /* Restore original position if nothing was hit after the first bounce,
-       * without the ray_offset() that was added to avoid self-intersection.
-       * Otherwise if that offset is relatively large compared to the scattering
-       * radius, we never go back up high enough to exit the surface. */
-      ray.P = P;
-    }
 
     /* Advance to new scatter location. */
     ray.P += t * ray.D;

intern/cycles/kernel/light/light.h

@@ -418,8 +418,8 @@ ccl_device bool light_sample_from_intersection(KernelGlobals kg,
   LightType type = (LightType)klight->type;
   ls->type = type;
   ls->shader = klight->shader_id;
-  ls->object = PRIM_NONE;
-  ls->prim = PRIM_NONE;
+  ls->object = isect->object;
+  ls->prim = isect->prim;
   ls->lamp = lamp;
   /* todo: missing texture coordinates */
   ls->t = isect->t;

intern/cycles/kernel/light/sample.h

@@ -198,7 +198,7 @@ ccl_device_inline float3 shadow_ray_offset(KernelGlobals kg,
   float NL = dot(sd->N, L);
   bool transmit = (NL < 0.0f);
   float3 Ng = (transmit ? -sd->Ng : sd->Ng);
-  float3 P = ray_offset(sd->P, Ng);
+  float3 P = sd->P;
 
   if ((sd->type & PRIMITIVE_TRIANGLE) && (sd->shader & SHADER_SMOOTH_NORMAL)) {
     const float offset_cutoff =
@@ -243,7 +243,7 @@ ccl_device_inline void shadow_ray_setup(ccl_private const ShaderData *ccl_restri
     }
     else {
       /* other lights, avoid self-intersection */
-      ray->D = ray_offset(ls->P, ls->Ng) - P;
+      ray->D = ls->P - P;
       ray->D = normalize_len(ray->D, &ray->t);
     }
   }
@@ -257,6 +257,12 @@ ccl_device_inline void shadow_ray_setup(ccl_private const ShaderData *ccl_restri
   ray->dP = differential_make_compact(sd->dP);
   ray->dD = differential_zero_compact();
   ray->time = sd->time;
+
+  /* Fill in intersection surface and light details. */
+  ray->self.prim = sd->prim;
+  ray->self.object = sd->object;
+  ray->self.light_prim = ls->prim;
+  ray->self.light_object = ls->object;
 }
 
 /* Create shadow ray towards light sample. */

intern/cycles/kernel/svm/ao.h

@@ -70,10 +70,14 @@ ccl_device float svm_ao(
 
     /* Create ray. */
     Ray ray;
-    ray.P = ray_offset(sd->P, N);
+    ray.P = sd->P;
     ray.D = D.x * T + D.y * B + D.z * N;
     ray.t = max_dist;
     ray.time = sd->time;
+    ray.self.object = sd->object;
+    ray.self.prim = sd->prim;
+    ray.self.light_object = OBJECT_NONE;
+    ray.self.light_prim = PRIM_NONE;
     ray.dP = differential_zero_compact();
     ray.dD = differential_zero_compact();
 

intern/cycles/kernel/svm/bevel.h

@@ -196,6 +196,10 @@ ccl_device float3 svm_bevel(
     ray.dP = differential_zero_compact();
     ray.dD = differential_zero_compact();
     ray.time = sd->time;
+    ray.self.object = OBJECT_NONE;
+    ray.self.prim = PRIM_NONE;
+    ray.self.light_object = OBJECT_NONE;
+    ray.self.light_prim = PRIM_NONE;
 
     /* Intersect with the same object. if multiple intersections are found it
      * will use at most LOCAL_MAX_HITS hits, a random subset of all hits. */
@@ -207,15 +211,24 @@ ccl_device float3 svm_bevel(
       /* Quickly retrieve P and Ng without setting up ShaderData. */
       float3 hit_P;
       if (sd->type == PRIMITIVE_TRIANGLE) {
-        hit_P = triangle_refine_local(
-            kg, sd, ray.P, ray.D, ray.t, isect.hits[hit].object, isect.hits[hit].prim);
+        hit_P = triangle_point_from_uv(kg,
+                                       sd,
+                                       isect.hits[hit].object,
+                                       isect.hits[hit].prim,
+                                       isect.hits[hit].u,
+                                       isect.hits[hit].v);
       }
 #  ifdef __OBJECT_MOTION__
       else if (sd->type == PRIMITIVE_MOTION_TRIANGLE) {
         float3 verts[3];
         motion_triangle_vertices(kg, sd->object, isect.hits[hit].prim, sd->time, verts);
-        hit_P = motion_triangle_refine_local(
-            kg, sd, ray.P, ray.D, ray.t, isect.hits[hit].object, isect.hits[hit].prim, verts);
+        hit_P = motion_triangle_point_from_uv(kg,
+                                              sd,
+                                              isect.hits[hit].object,
+                                              isect.hits[hit].prim,
+                                              isect.hits[hit].u,
+                                              isect.hits[hit].v,
+                                              verts);
       }
 #  endif /* __OBJECT_MOTION__ */
 

intern/cycles/kernel/types.h

@@ -512,12 +512,21 @@ typedef struct differential {
 
 /* Ray */
 
+typedef struct RaySelfPrimitives {
+  int prim;         /* Primitive the ray is starting from */
+  int object;       /* Instance prim is a part of */
+  int light_prim;   /* Light primitive */
+  int light_object; /* Light object */
+} RaySelfPrimitives;
+
 typedef struct Ray {
   float3 P;   /* origin */
   float3 D;   /* direction */
   float t;    /* length of the ray */
   float time; /* time (for motion blur) */
 
+  RaySelfPrimitives self;
+
 #ifdef __RAY_DIFFERENTIALS__
   float dP;
   float dD;

intern/ghost/CMakeLists.txt

@@ -291,7 +291,7 @@ elseif(WITH_GHOST_X11 OR WITH_GHOST_WAYLAND)
     include(CheckSymbolExists)
     set(CMAKE_REQUIRED_DEFINITIONS "-D_GNU_SOURCE")
     check_symbol_exists(memfd_create "sys/mman.h" HAVE_MEMFD_CREATE)
-    if (HAVE_MEMFD_CREATE)
+    if(HAVE_MEMFD_CREATE)
       add_definitions(-DHAVE_MEMFD_CREATE)
     endif()
 
@@ -307,7 +307,7 @@ elseif(WITH_GHOST_X11 OR WITH_GHOST_WAYLAND)
     pkg_get_variable(WAYLAND_SCANNER wayland-scanner wayland_scanner)
 
     pkg_check_modules(wayland-protocols wayland-protocols>=1.15)
-    if (${wayland-protocols_FOUND})
+    if(${wayland-protocols_FOUND})
       pkg_get_variable(WAYLAND_PROTOCOLS_DIR wayland-protocols pkgdatadir)
     else()
       find_path(WAYLAND_PROTOCOLS_DIR
@@ -316,7 +316,7 @@ elseif(WITH_GHOST_X11 OR WITH_GHOST_WAYLAND)
       )
     endif()
 
-    if (NOT EXISTS ${WAYLAND_PROTOCOLS_DIR})
+    if(NOT EXISTS ${WAYLAND_PROTOCOLS_DIR})
       message(FATAL_ERROR "path to wayland-protocols not found")
     endif()
 
@@ -518,11 +518,11 @@ if(WITH_XR_OPENXR)
     )
   elseif(UNIX AND NOT APPLE)
     list(APPEND XR_PLATFORM_DEFINES -DXR_OS_LINUX)
-    if (WITH_GHOST_WAYLAND)
+    if(WITH_GHOST_WAYLAND)
       list(APPEND XR_PLATFORM_DEFINES -DXR_USE_PLATFORM_WAYLAND)
     endif()
-    if (WITH_GHOST_X11)
-      if (WITH_GL_EGL)
+    if(WITH_GHOST_X11)
+      if(WITH_GL_EGL)
         list(APPEND XR_PLATFORM_DEFINES -DXR_USE_PLATFORM_EGL)
       else()
         list(APPEND XR_PLATFORM_DEFINES -DXR_USE_PLATFORM_XLIB)

intern/ghost/intern/GHOST_ImeWin32.cpp

@@ -96,7 +96,7 @@ bool GHOST_ImeWin32::IsEnglishMode()
          !(conversion_modes_ & (IME_CMODE_NATIVE | IME_CMODE_FULLSHAPE));
 }
 
-bool GHOST_ImeWin32::IsImeKeyEvent(char ascii)
+bool GHOST_ImeWin32::IsImeKeyEvent(char ascii, GHOST_TKey key)
 {
   if (!(IsEnglishMode())) {
     /* In Chinese, Japanese, Korean, all alpha keys are processed by IME. */
@@ -106,7 +106,8 @@ bool GHOST_ImeWin32::IsImeKeyEvent(char ascii)
     if (IsLanguage(IMELANG_JAPANESE) && (ascii >= ' ' && ascii <= '~')) {
       return true;
     }
-    else if (IsLanguage(IMELANG_CHINESE) && ascii && strchr("!\"$'(),.:;<>?[\\]^_`/", ascii)) {
+    else if (IsLanguage(IMELANG_CHINESE) && ascii && strchr("!\"$'(),.:;<>?[\\]^_`/", ascii) &&
+             !(key == GHOST_kKeyNumpadPeriod)) {
       return true;
     }
   }

intern/ghost/intern/GHOST_ImeWin32.h

@@ -161,7 +161,7 @@ class GHOST_ImeWin32 {
   bool IsEnglishMode();
 
   /* Checks a key whether IME has to do handling. */
-  bool IsImeKeyEvent(char ascii);
+  bool IsImeKeyEvent(char ascii, GHOST_TKey key);
 
   /**
    * Create the IME windows, and allocate required resources for them.

intern/ghost/intern/GHOST_SystemCocoa.mm

@@ -890,7 +890,7 @@ bool GHOST_SystemCocoa::processEvents(bool waitForEvent)
   bool anyProcessed = false;
   NSEvent *event;
 
-  // TODO : implement timer ??
+  /* TODO: implement timer? */
 #if 0
   do {
     GHOST_TimerManager* timerMgr = getTimerManager();

intern/ghost/intern/GHOST_SystemWin32.cpp

@@ -1220,7 +1220,7 @@ GHOST_EventKey *GHOST_SystemWin32::processKeyEvent(GHOST_WindowWin32 *window, RA
     }
 
 #ifdef WITH_INPUT_IME
-    if (window->getImeInput()->IsImeKeyEvent(ascii)) {
+    if (window->getImeInput()->IsImeKeyEvent(ascii, key)) {
       return NULL;
     }
 #endif /* WITH_INPUT_IME */

intern/ghost/intern/GHOST_WindowWin32.cpp

@@ -71,6 +71,8 @@ GHOST_WindowWin32::GHOST_WindowWin32(GHOST_SystemWin32 *system,
       m_mousePresent(false),
       m_inLiveResize(false),
       m_system(system),
+      m_dropTarget(NULL),
+      m_hWnd(0),
       m_hDC(0),
       m_isDialog(dialog),
       m_hasMouseCaptured(false),
@@ -78,6 +80,7 @@ GHOST_WindowWin32::GHOST_WindowWin32(GHOST_SystemWin32 *system,
       m_nPressedButtons(0),
       m_customCursor(0),
       m_wantAlphaBackground(alphaBackground),
+      m_Bar(NULL),
       m_wintab(NULL),
       m_lastPointerTabletData(GHOST_TABLET_DATA_NONE),
       m_normal_state(GHOST_kWindowStateNormal),
@@ -129,8 +132,24 @@ GHOST_WindowWin32::GHOST_WindowWin32(GHOST_SystemWin32 *system,
   m_hDC = ::GetDC(m_hWnd);
 
   if (!setDrawingContextType(type)) {
+    const char *title = "Blender - Unsupported Graphics Card Configuration";
+    const char *text =
+        "A graphics card and driver with support for OpenGL 3.3 or higher is "
+        "required.\n\nInstalling the latest driver for your graphics card might resolve the "
+        "issue.";
+    if (GetSystemMetrics(SM_CMONITORS) > 1) {
+      text =
+          "A graphics card and driver with support for OpenGL 3.3 or higher is "
+          "required.\n\nPlugging all monitors into your primary graphics card might resolve "
+          "this issue. Installing the latest driver for your graphics card could also help.";
+    }
+    MessageBox(m_hWnd, text, title, MB_OK | MB_ICONERROR);
+    ::ReleaseDC(m_hWnd, m_hDC);
     ::DestroyWindow(m_hWnd);
     m_hWnd = NULL;
+    if (!parentwindow) {
+      exit(0);
+    }
     return;
   }
 
@@ -564,20 +583,13 @@ GHOST_Context *GHOST_WindowWin32::newDrawingContext(GHOST_TDrawingContextType ty
                                    (m_debug_context ? WGL_CONTEXT_DEBUG_BIT_ARB : 0),
                                    GHOST_OPENGL_WGL_RESET_NOTIFICATION_STRATEGY);
 
-    if (context->initializeDrawingContext()) {
-      return context;
-    }
-    else {
-      MessageBox(m_hWnd,
-                 "A graphics card and driver with support for OpenGL 3.3 or higher is required.\n"
-                 "Installing the latest driver for your graphics card may resolve the issue.\n\n"
-                 "The program will now close.",
-                 "Blender - Unsupported Graphics Card or Driver",
-                 MB_OK | MB_ICONERROR);
+    if (context && !context->initializeDrawingContext()) {
       delete context;
-      exit(0);
+      context = nullptr;
     }
 
+    return context;
+
 #elif defined(WITH_GL_PROFILE_COMPAT)
     // ask for 2.1 context, driver gives any GL version >= 2.1
     // (hopefully the latest compatibility profile)

intern/ghost/intern/GHOST_XrContext.cpp

@@ -86,6 +86,7 @@ void GHOST_XrContext::initialize(const GHOST_XrContextCreateInfo *create_info)
   initApiLayers();
   initExtensions();
   if (isDebugMode()) {
+    printSDKVersion();
     printAvailableAPILayersAndExtensionsInfo();
   }
 
@@ -156,6 +157,16 @@ void GHOST_XrContext::storeInstanceProperties()
 /** \name Debug Printing
  * \{ */
 
+void GHOST_XrContext::printSDKVersion()
+{
+  const XrVersion sdk_version = XR_CURRENT_API_VERSION;
+
+  printf("OpenXR SDK Version: %u.%u.%u\n",
+         XR_VERSION_MAJOR(sdk_version),
+         XR_VERSION_MINOR(sdk_version),
+         XR_VERSION_PATCH(sdk_version));
+}
+
 void GHOST_XrContext::printInstanceInfo()
 {
   assert(m_oxr->instance != XR_NULL_HANDLE);

intern/ghost/intern/GHOST_XrContext.h

@@ -126,6 +126,7 @@ class GHOST_XrContext : public GHOST_IXrContext {
   void storeInstanceProperties();
   void initDebugMessenger();
 
+  void printSDKVersion();
   void printInstanceInfo();
   void printAvailableAPILayersAndExtensionsInfo();
   void printExtensionsAndAPILayersToEnable();

intern/ghost/test/CMakeLists.txt

@@ -292,7 +292,7 @@ target_link_libraries(multitest_c
   guardedalloc_lib
   wcwidth_lib
   ${OPENGL_gl_LIBRARY}
-  ${FREETYPE_LIBRARIES}
+  ${FREETYPE_LIBRARIES} ${BROTLI_LIBRARIES}
   ${ZLIB_LIBRARIES}
   ${CMAKE_DL_LIBS}
   ${PLATFORM_LINKLIBS}

intern/guardedalloc/MEM_guardedalloc.h

@@ -39,8 +39,8 @@
  * second intern/ module with MEM_ prefix, for use in c++.
  *
  * \subsection memdependencies Dependencies
- * - stdlib
- * - stdio
+ * - `stdlib`
+ * - `stdio`
  *
  * \subsection memdocs API Documentation
  * See \ref MEM_guardedalloc.h

intern/locale/CMakeLists.txt

@@ -56,7 +56,6 @@ if(WITH_INTERNATIONAL)
   list(APPEND LIB
     ${BOOST_LIBRARIES}
   )
-  add_definitions(-DWITH_INTERNATIONAL)
   add_definitions(${BOOST_DEFINITIONS})
 endif()
 

release/scripts/modules/rna_xml.py

@@ -298,7 +298,7 @@ def xml2rna(
                         del value_xml_split
                     tp_name = 'ARRAY'
 
-#                print("  %s.%s (%s) --- %s" % (type(value).__name__, attr, tp_name, subvalue_type))
+                    # print("  %s.%s (%s) --- %s" % (type(value).__name__, attr, tp_name, subvalue_type))
                 try:
                     setattr(value, attr, value_xml_coerce)
                 except ValueError:
@@ -340,7 +340,6 @@ def xml2rna(
 
                     else:
                         # print(elems)
-
                         if len(elems) == 1:
                             # sub node named by its type
                             child_xml_real, = elems
@@ -376,7 +375,6 @@ def _get_context_val(context, path):
 
 
 def xml_file_run(context, filepath, rna_map):
-
     import xml.dom.minidom
 
     xml_nodes = xml.dom.minidom.parse(filepath)
@@ -391,27 +389,25 @@ def xml_file_run(context, filepath, rna_map):
         value = _get_context_val(context, rna_path)
 
         if value is not Ellipsis and value is not None:
-            print("  loading XML: %r -> %r" % (filepath, rna_path))
+            # print("  loading XML: %r -> %r" % (filepath, rna_path))
             xml2rna(xml_node, root_rna=value)
 
 
 def xml_file_write(context, filepath, rna_map, *, skip_typemap=None):
+    with open(filepath, "w", encoding="utf-8") as file:
+        fw = file.write
+        fw("<bpy>\n")
 
-    file = open(filepath, "w", encoding="utf-8")
-    fw = file.write
-
-    fw("<bpy>\n")
-
-    for rna_path, _xml_tag in rna_map:
-        # xml_tag is ignored, we get this from the rna
-        value = _get_context_val(context, rna_path)
-        rna2xml(fw,
+        for rna_path, _xml_tag in rna_map:
+            # xml_tag is ignored, we get this from the rna
+            value = _get_context_val(context, rna_path)
+            rna2xml(
+                fw=fw,
                 root_rna=value,
                 method='ATTR',
                 root_ident="  ",
                 ident_val="  ",
                 skip_typemap=skip_typemap,
-                )
+            )
 
-    fw("</bpy>\n")
-    file.close()
+        fw("</bpy>\n")

release/scripts/startup/bl_operators/wm.py

@@ -2961,93 +2961,75 @@ class WM_MT_splash_quick_setup(Menu):
     bl_label = "Quick Setup"
 
     def draw(self, context):
-        wm = context.window_manager
-        # prefs = context.preferences
-
         layout = self.layout
-
         layout.operator_context = 'EXEC_DEFAULT'
 
         layout.label(text="Quick Setup")
 
-        split = layout.split(factor=0.25)
+        split = layout.split(factor=0.14) # Left margin.
         split.label()
-        split = split.split(factor=2.0 / 3.0)
+        split = split.split(factor=0.73) # Content width.
 
         col = split.column()
 
+        col.use_property_split = True
+        col.use_property_decorate = False
+
+        # Languages.
         if bpy.app.build_options.international:
-            sub = col.split(factor=0.35)
-            row = sub.row()
-            row.alignment = 'RIGHT'
-            row.label(text="Language")
             prefs = context.preferences
-            sub.prop(prefs.view, "language", text="")
+            col.prop(prefs.view, "language")
+            col.separator()
 
-        col.separator()
+        # Shortcuts.
+        wm = context.window_manager
+        kc = wm.keyconfigs.active
+        kc_prefs = kc.preferences
 
-        sub = col.split(factor=0.35)
-        row = sub.row()
-        row.alignment = 'RIGHT'
-        row.label(text="Shortcuts")
-        text = bpy.path.display_name(wm.keyconfigs.active.name)
+        sub = col.column(heading="Shortcuts")
+        text = bpy.path.display_name(kc.name)
         if not text:
             text = "Blender"
         sub.menu("USERPREF_MT_keyconfigs", text=text)
 
-        kc = wm.keyconfigs.active
-        kc_prefs = kc.preferences
         has_select_mouse = hasattr(kc_prefs, "select_mouse")
         if has_select_mouse:
-            sub = col.split(factor=0.35)
-            row = sub.row()
-            row.alignment = 'RIGHT'
-            row.label(text="Select With")
-            sub.row().prop(kc_prefs, "select_mouse", expand=True)
-            has_select_mouse = True
+            col.row().prop(kc_prefs, "select_mouse", text="Select With", expand=True)
 
         has_spacebar_action = hasattr(kc_prefs, "spacebar_action")
         if has_spacebar_action:
-            sub = col.split(factor=0.35)
-            row = sub.row()
-            row.alignment = 'RIGHT'
-            row.label(text="Spacebar")
-            sub.row().prop(kc_prefs, "spacebar_action", expand=True)
-            has_select_mouse = True
+            col.row().prop(kc_prefs, "spacebar_action", text="Spacebar")
 
         col.separator()
 
-        sub = col.split(factor=0.35)
-        row = sub.row()
-        row.alignment = 'RIGHT'
-        row.label(text="Theme")
+        # Themes.
+        sub = col.column(heading="Theme")
         label = bpy.types.USERPREF_MT_interface_theme_presets.bl_label
         if label == "Presets":
             label = "Blender Dark"
         sub.menu("USERPREF_MT_interface_theme_presets", text=label)
 
-        # Keep height constant
+        # Keep height constant.
         if not has_select_mouse:
             col.label()
         if not has_spacebar_action:
             col.label()
 
-        layout.label()
+        layout.separator(factor=2.0)
 
-        row = layout.row()
+        # Save settings buttons.
+        sub = layout.row()
 
-        sub = row.row()
         old_version = bpy.types.PREFERENCES_OT_copy_prev.previous_version()
         if bpy.types.PREFERENCES_OT_copy_prev.poll(context) and old_version:
-            sub.operator("preferences.copy_prev", text=iface_("Load %d.%d Settings", "Operator") % old_version)
+            sub.operator("preferences.copy_prev", text="Load %d.%d Settings" % old_version)
             sub.operator("wm.save_userpref", text="Save New Settings")
         else:
             sub.label()
             sub.label()
             sub.operator("wm.save_userpref", text="Next")
 
-        layout.separator()
-        layout.separator()
+        layout.separator(factor=2.4)
 
 
 class WM_MT_splash(Menu):

release/scripts/startup/bl_ui/space_filebrowser.py

@@ -761,6 +761,15 @@ class ASSETBROWSER_PT_metadata_preview(asset_utils.AssetMetaDataPanel, Panel):
         col.operator("ed.lib_id_load_custom_preview", icon='FILEBROWSER', text="")
         col.separator()
         col.operator("ed.lib_id_generate_preview", icon='FILE_REFRESH', text="")
+        col.menu("ASSETBROWSER_MT_metadata_preview_menu", icon='DOWNARROW_HLT', text="")
+
+
+class ASSETBROWSER_MT_metadata_preview_menu(bpy.types.Menu):
+    bl_label = "Preview"
+
+    def draw(self, context):
+        layout = self.layout
+        layout.operator("ed.lib_id_generate_preview_from_object", text="Render Active Object")
 
 
 class ASSETBROWSER_PT_metadata_tags(asset_utils.AssetMetaDataPanel, Panel):
@@ -840,6 +849,7 @@ classes = (
     ASSETBROWSER_MT_view,
     ASSETBROWSER_MT_select,
     ASSETBROWSER_MT_edit,
+    ASSETBROWSER_MT_metadata_preview_menu,
     ASSETBROWSER_PT_metadata,
     ASSETBROWSER_PT_metadata_preview,
     ASSETBROWSER_PT_metadata_tags,

release/scripts/startup/bl_ui/space_userpref.py

@@ -2316,7 +2316,6 @@ class USERPREF_PT_experimental_debugging(ExperimentalPanel, Panel):
             context, (
                 ({"property": "use_undo_legacy"}, "T60695"),
                 ({"property": "override_auto_resync"}, "T83811"),
-                ({"property": "proxy_to_override_auto_conversion"}, "T91671"),
                 ({"property": "use_cycles_debug"}, None),
                 ({"property": "use_geometry_nodes_legacy"}, "T91274"),
                 ({"property": "show_asset_debug_info"}, None),

release/scripts/startup/nodeitems_builtins.py

@@ -547,6 +547,8 @@ compositor_node_categories = [
         NodeItem("CompositorNodeCombYUVA"),
         NodeItem("CompositorNodeSepYCCA"),
         NodeItem("CompositorNodeCombYCCA"),
+        NodeItem("CompositorNodeSeparateXYZ"),
+        NodeItem("CompositorNodeCombineXYZ"),
         NodeItem("CompositorNodeSwitchView"),
         NodeItem("CompositorNodeConvertColorSpace"),
     ]),

source/CMakeLists.txt

@@ -27,7 +27,7 @@ if(WITH_CLANG_TIDY AND NOT MSVC)
     message(WARNING "Currently Clang-Tidy might fail with GCC toolchain, switch to Clang toolchain if that happens")
     if(COMMAND target_precompile_headers)
       message(STATUS "Clang-Tidy and GCC precompiled headers are incompatible, disabling precompiled headers")
-      set(CMAKE_DISABLE_PRECOMPILE_HEADERS On)
+      set(CMAKE_DISABLE_PRECOMPILE_HEADERS ON)
     endif()
   endif()
 

source/blender/blenfont/CMakeLists.txt

@@ -54,7 +54,7 @@ set(LIB
   bf_gpu
   bf_intern_guardedalloc
 
-  ${FREETYPE_LIBRARIES}
+  ${FREETYPE_LIBRARIES} ${BROTLI_LIBRARIES}
 )
 
 if(WIN32)
@@ -63,10 +63,6 @@ if(WIN32)
   )
 endif()
 
-if(WITH_INTERNATIONAL)
-  add_definitions(-DWITH_INTERNATIONAL)
-endif()
-
 if(WITH_PYTHON)
   add_definitions(-DWITH_PYTHON)
   list(APPEND INC

source/blender/blenkernel/BKE_DerivedMesh.h

@@ -104,14 +104,6 @@ typedef enum DerivedMeshType {
   DM_TYPE_CCGDM,
 } DerivedMeshType;
 
-typedef enum DMDirtyFlag {
-  /* dm has valid tessellated faces, but tessellated CDDATA need to be updated. */
-  DM_DIRTY_TESS_CDLAYERS = 1 << 0,
-
-  /* check this with modifier dependsOnNormals callback to see if normals need recalculation */
-  DM_DIRTY_NORMALS = 1 << 1,
-} DMDirtyFlag;
-
 typedef struct DerivedMesh DerivedMesh;
 struct DerivedMesh {
   /** Private DerivedMesh data, only for internal DerivedMesh use */
@@ -120,7 +112,6 @@ struct DerivedMesh {
   int needsFree;    /* checked on ->release, is set to 0 for cached results */
   int deformedOnly; /* set by modifier stack if only deformed from original */
   DerivedMeshType type;
-  DMDirtyFlag dirty;
 
   /**
    * \warning Typical access is done via #getLoopTriArray, #getNumLoopTri.
@@ -139,9 +130,6 @@ struct DerivedMesh {
 
   short tangent_mask; /* which tangent layers are calculated */
 
-  /** Calculate vert and face normals */
-  void (*calcNormals)(DerivedMesh *dm);
-
   /** Loop tessellation cache (WARNING! Only call inside threading-protected code!) */
   void (*recalcLoopTri)(DerivedMesh *dm);
   /** accessor functions */
@@ -164,7 +152,6 @@ struct DerivedMesh {
    */
   struct MVert *(*getVertArray)(DerivedMesh *dm);
   struct MEdge *(*getEdgeArray)(DerivedMesh *dm);
-  struct MFace *(*getTessFaceArray)(DerivedMesh *dm);
   struct MLoop *(*getLoopArray)(DerivedMesh *dm);
   struct MPoly *(*getPolyArray)(DerivedMesh *dm);
 
@@ -173,7 +160,6 @@ struct DerivedMesh {
    */
   void (*copyVertArray)(DerivedMesh *dm, struct MVert *r_vert);
   void (*copyEdgeArray)(DerivedMesh *dm, struct MEdge *r_edge);
-  void (*copyTessFaceArray)(DerivedMesh *dm, struct MFace *r_face);
   void (*copyLoopArray)(DerivedMesh *dm, struct MLoop *r_loop);
   void (*copyPolyArray)(DerivedMesh *dm, struct MPoly *r_poly);
 
@@ -182,37 +168,18 @@ struct DerivedMesh {
    */
   struct MVert *(*dupVertArray)(DerivedMesh *dm);
   struct MEdge *(*dupEdgeArray)(DerivedMesh *dm);
-  struct MFace *(*dupTessFaceArray)(DerivedMesh *dm);
   struct MLoop *(*dupLoopArray)(DerivedMesh *dm);
   struct MPoly *(*dupPolyArray)(DerivedMesh *dm);
 
-  /** Return a pointer to a single element of vert/edge/face custom data
-   * from the derived mesh (this gives a pointer to the actual data, not
-   * a copy)
-   */
-  void *(*getVertData)(DerivedMesh *dm, int index, int type);
-  void *(*getEdgeData)(DerivedMesh *dm, int index, int type);
-  void *(*getTessFaceData)(DerivedMesh *dm, int index, int type);
-  void *(*getPolyData)(DerivedMesh *dm, int index, int type);
-
   /** Return a pointer to the entire array of vert/edge/face custom data
    * from the derived mesh (this gives a pointer to the actual data, not
    * a copy)
    */
   void *(*getVertDataArray)(DerivedMesh *dm, int type);
   void *(*getEdgeDataArray)(DerivedMesh *dm, int type);
-  void *(*getTessFaceDataArray)(DerivedMesh *dm, int type);
   void *(*getLoopDataArray)(DerivedMesh *dm, int type);
   void *(*getPolyDataArray)(DerivedMesh *dm, int type);
 
-  /** Retrieves the base CustomData structures for
-   * verts/edges/tessfaces/loops/faces. */
-  CustomData *(*getVertDataLayout)(DerivedMesh *dm);
-  CustomData *(*getEdgeDataLayout)(DerivedMesh *dm);
-  CustomData *(*getTessFaceDataLayout)(DerivedMesh *dm);
-  CustomData *(*getLoopDataLayout)(DerivedMesh *dm);
-  CustomData *(*getPolyDataLayout)(DerivedMesh *dm);
-
   /** Optional grid access for subsurf */
   int (*getNumGrids)(DerivedMesh *dm);
   int (*getGridSize)(DerivedMesh *dm);
@@ -231,11 +198,6 @@ struct DerivedMesh {
 
   /** Get smooth vertex normal, undefined if index is not valid */
   void (*getVertNo)(DerivedMesh *dm, int index, float r_no[3]);
-  void (*getPolyNo)(DerivedMesh *dm, int index, float r_no[3]);
-
-  /** Get a map of vertices to faces
-   */
-  const struct MeshElemMap *(*getPolyMap)(struct Object *ob, DerivedMesh *dm);
 
   /** Release reference to the DerivedMesh. This function decides internally
    * if the DerivedMesh will be freed, or cached for later use. */
@@ -265,15 +227,6 @@ void DM_init(DerivedMesh *dm,
  * Utility function to initialize a DerivedMesh for the desired number
  * of vertices, edges and faces, with a layer setup copied from source
  */
-void DM_from_template_ex(DerivedMesh *dm,
-                         DerivedMesh *source,
-                         DerivedMeshType type,
-                         int numVerts,
-                         int numEdges,
-                         int numTessFaces,
-                         int numLoops,
-                         int numPolys,
-                         const struct CustomData_MeshMasks *mask);
 void DM_from_template(DerivedMesh *dm,
                       DerivedMesh *source,
                       DerivedMeshType type,
@@ -303,25 +256,8 @@ void DM_set_only_copy(DerivedMesh *dm, const struct CustomData_MeshMasks *mask);
 
 void DM_add_vert_layer(struct DerivedMesh *dm, int type, eCDAllocType alloctype, void *layer);
 void DM_add_edge_layer(struct DerivedMesh *dm, int type, eCDAllocType alloctype, void *layer);
-void DM_add_tessface_layer(struct DerivedMesh *dm, int type, eCDAllocType alloctype, void *layer);
-void DM_add_loop_layer(DerivedMesh *dm, int type, eCDAllocType alloctype, void *layer);
 void DM_add_poly_layer(struct DerivedMesh *dm, int type, eCDAllocType alloctype, void *layer);
 
-/* -------------------------------------------------------------------- */
-/** \name Custom Data Access Functions
- *
- * \return pointer to data from first layer which matches type
- * if they return NULL for valid indices, data doesn't exist.
- * \note these return pointers - any change modifies the internals of the mesh.
- * \{ */
-
-void *DM_get_vert_data(struct DerivedMesh *dm, int index, int type);
-void *DM_get_edge_data(struct DerivedMesh *dm, int index, int type);
-void *DM_get_tessface_data(struct DerivedMesh *dm, int index, int type);
-void *DM_get_poly_data(struct DerivedMesh *dm, int index, int type);
-
-/** \} */
-
 /* -------------------------------------------------------------------- */
 /** \name Custom Data Layer Access Functions
  *
@@ -332,7 +268,6 @@ void *DM_get_poly_data(struct DerivedMesh *dm, int index, int type);
 
 void *DM_get_vert_data_layer(struct DerivedMesh *dm, int type);
 void *DM_get_edge_data_layer(struct DerivedMesh *dm, int type);
-void *DM_get_tessface_data_layer(struct DerivedMesh *dm, int type);
 void *DM_get_poly_data_layer(struct DerivedMesh *dm, int type);
 void *DM_get_loop_data_layer(struct DerivedMesh *dm, int type);
 
@@ -354,8 +289,6 @@ void DM_copy_vert_data(struct DerivedMesh *source,
  */
 void DM_DupPolys(DerivedMesh *source, DerivedMesh *target);
 
-void DM_ensure_normals(DerivedMesh *dm);
-
 /**
  * Ensure the array is large enough.
  *

source/blender/blenkernel/BKE_blender_version.h

@@ -31,7 +31,7 @@ extern "C" {
  */
 
 /* Blender major and minor version. */
-#define BLENDER_VERSION 301
+#define BLENDER_VERSION 302
 /* Blender patch version for bugfix releases. */
 #define BLENDER_VERSION_PATCH 0
 /** Blender release cycle stage: alpha/beta/rc/release. */
@@ -39,7 +39,7 @@ extern "C" {
 
 /* Blender file format version. */
 #define BLENDER_FILE_VERSION BLENDER_VERSION
-#define BLENDER_FILE_SUBVERSION 5
+#define BLENDER_FILE_SUBVERSION 1
 
 /* Minimum Blender version that supports reading file written with the current
  * version. Older Blender versions will test this and show a warning if the file

source/blender/blenkernel/BKE_camera.h

@@ -46,7 +46,7 @@ void *BKE_camera_add(struct Main *bmain, const char *name);
 /**
  * Get the camera's DOF value, takes the DOF object into account.
  */
-float BKE_camera_object_dof_distance(struct Object *ob);
+float BKE_camera_object_dof_distance(const struct Object *ob);
 
 int BKE_camera_sensor_fit(int sensor_fit, float sizex, float sizey);
 float BKE_camera_sensor_size(int sensor_fit, float sensor_x, float sensor_y);

source/blender/blenkernel/BKE_customdata.h

@@ -436,6 +436,12 @@ int CustomData_get_render_layer(const struct CustomData *data, int type);
 int CustomData_get_clone_layer(const struct CustomData *data, int type);
 int CustomData_get_stencil_layer(const struct CustomData *data, int type);
 
+/**
+ * Returns name of the active layer of the given type or NULL
+ * if no such active layer is defined.
+ */
+const char *CustomData_get_active_layer_name(const struct CustomData *data, int type);
+
 /**
  * Copies the data from source to the data element at index in the first layer of type
  * no effect if there is no layer of type.

source/blender/blenkernel/BKE_image.h

@@ -24,6 +24,8 @@
 
 #include "BLI_utildefines.h"
 
+#include "BLI_rect.h"
+
 #ifdef __cplusplus
 extern "C" {
 #endif
@@ -561,19 +563,27 @@ struct GPUTexture *BKE_image_get_gpu_tilemap(struct Image *image,
  * Is the alpha of the `GPUTexture` for a given image/ibuf premultiplied.
  */
 bool BKE_image_has_gpu_texture_premultiplied_alpha(struct Image *image, struct ImBuf *ibuf);
+
 /**
  * Partial update of texture for texture painting.
  * This is often much quicker than fully updating the texture for high resolution images.
  */
 void BKE_image_update_gputexture(
     struct Image *ima, struct ImageUser *iuser, int x, int y, int w, int h);
+
 /**
  * Mark areas on the #GPUTexture that needs to be updated. The areas are marked in chunks.
  * The next time the #GPUTexture is used these tiles will be refreshes. This saves time
  * when writing to the same place multiple times This happens for during foreground rendering.
  */
-void BKE_image_update_gputexture_delayed(
-    struct Image *ima, struct ImBuf *ibuf, int x, int y, int w, int h);
+void BKE_image_update_gputexture_delayed(struct Image *ima,
+                                         struct ImageTile *image_tile,
+                                         struct ImBuf *ibuf,
+                                         int x,
+                                         int y,
+                                         int w,
+                                         int h);
+
 /**
  * Called on entering and exiting texture paint mode,
  * temporary disabling/enabling mipmapping on all images for quick texture
@@ -591,6 +601,32 @@ bool BKE_image_remove_renderslot(struct Image *ima, struct ImageUser *iuser, int
 struct RenderSlot *BKE_image_get_renderslot(struct Image *ima, int index);
 bool BKE_image_clear_renderslot(struct Image *ima, struct ImageUser *iuser, int slot);
 
+/* --- image_partial_update.cc --- */
+/** Image partial updates. */
+struct PartialUpdateUser;
+
+/**
+ * \brief Create a new PartialUpdateUser. An Object that contains data to use partial updates.
+ */
+struct PartialUpdateUser *BKE_image_partial_update_create(const struct Image *image);
+
+/**
+ * \brief free a partial update user.
+ */
+void BKE_image_partial_update_free(struct PartialUpdateUser *user);
+
+/* --- partial updater (image side) --- */
+struct PartialUpdateRegister;
+
+void BKE_image_partial_update_register_free(struct Image *image);
+/** \brief Mark a region of the image to update. */
+void BKE_image_partial_update_mark_region(struct Image *image,
+                                          const struct ImageTile *image_tile,
+                                          const struct ImBuf *image_buffer,
+                                          const rcti *updated_region);
+/** \brief Mark the whole image to be updated. */
+void BKE_image_partial_update_mark_full_update(struct Image *image);
+
 #ifdef __cplusplus
 }
 #endif

source/blender/blenkernel/BKE_image_partial_update.hh

@@ -0,0 +1,298 @@
+/*
+ * 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.
+ *
+ * Copyright 2021, Blender Foundation.
+ */
+
+/** \file
+ * \ingroup bke
+ *
+ * To reduce the overhead of image processing this file contains a mechanism to detect areas of the
+ * image that are changed. These areas are organized in chunks. Changes that happen over time are
+ * organized in changesets.
+ *
+ * A common use case is to update #GPUTexture for drawing where only that part is uploaded that
+ * only changed.
+ */
+
+#pragma once
+
+#include "BLI_utildefines.h"
+
+#include "BLI_rect.h"
+
+#include "DNA_image_types.h"
+
+extern "C" {
+struct PartialUpdateUser;
+struct PartialUpdateRegister;
+}
+
+namespace blender::bke::image {
+
+using TileNumber = int;
+
+namespace partial_update {
+
+/* --- image_partial_update.cc --- */
+/** Image partial updates. */
+
+/**
+ * \brief Result codes of #BKE_image_partial_update_collect_changes.
+ */
+enum class ePartialUpdateCollectResult {
+  /** \brief Unable to construct partial updates. Caller should perform a full update. */
+  FullUpdateNeeded,
+
+  /** \brief No changes detected since the last time requested. */
+  NoChangesDetected,
+
+  /** \brief Changes detected since the last time requested. */
+  PartialChangesDetected,
+};
+
+/**
+ * \brief A region to update.
+ *
+ * Data is organized in tiles. These tiles are in texel space (1 unit is a single texel). When
+ * tiles are requested they are merged with neighboring tiles.
+ */
+struct PartialUpdateRegion {
+  /** \brief region of the image that has been updated. Region can be bigger than actual changes.
+   */
+  struct rcti region;
+
+  /**
+   * \brief Tile number (UDIM) that this region belongs to.
+   */
+  TileNumber tile_number;
+};
+
+/**
+ * \brief Return codes of #BKE_image_partial_update_get_next_change.
+ */
+enum class ePartialUpdateIterResult {
+  /** \brief no tiles left when iterating over tiles. */
+  Finished = 0,
+
+  /** \brief a chunk was available and has been loaded. */
+  ChangeAvailable = 1,
+};
+
+/**
+ * \brief collect the partial update since the last request.
+ *
+ * Invoke #BKE_image_partial_update_get_next_change to iterate over the collected tiles.
+ *
+ * \returns ePartialUpdateCollectResult::FullUpdateNeeded: called should not use partial updates
+ * but recalculate the full image. This result can be expected when called for the first time for a
+ * user and when it isn't possible to reconstruct the changes as the internal state doesn't have
+ * enough data stored. ePartialUpdateCollectResult::NoChangesDetected: The have been no changes
+ * detected since last invoke for the same user.
+ * ePartialUpdateCollectResult::PartialChangesDetected: Parts of the image has been updated since
+ * last invoke for the same user. The changes can be read by using
+ * #BKE_image_partial_update_get_next_change.
+ */
+ePartialUpdateCollectResult BKE_image_partial_update_collect_changes(
+    struct Image *image, struct PartialUpdateUser *user);
+
+ePartialUpdateIterResult BKE_image_partial_update_get_next_change(
+    struct PartialUpdateUser *user, struct PartialUpdateRegion *r_region);
+
+/** \brief Abstract class to load tile data when using the PartialUpdateChecker. */
+class AbstractTileData {
+ protected:
+  virtual ~AbstractTileData() = default;
+
+ public:
+  /**
+   * \brief Load the data for the given tile_number.
+   *
+   * Invoked when changes are on a different tile compared to the previous tile..
+   */
+  virtual void init_data(TileNumber tile_number) = 0;
+  /**
+   * \brief Unload the data that has been loaded.
+   *
+   * Invoked when changes are on a different tile compared to the previous tile or when finished
+   * iterating over the changes.
+   */
+  virtual void free_data() = 0;
+};
+
+/**
+ * \brief Class to not load any tile specific data when iterating over changes.
+ */
+class NoTileData : AbstractTileData {
+ public:
+  NoTileData(Image *UNUSED(image), ImageUser *UNUSED(image_user))
+  {
+  }
+
+  void init_data(TileNumber UNUSED(new_tile_number)) override
+  {
+  }
+
+  void free_data() override
+  {
+  }
+};
+
+/**
+ * \brief Load the ImageTile and ImBuf associated with the partial change.
+ */
+class ImageTileData : AbstractTileData {
+ public:
+  /**
+   * \brief Not owned Image that is being iterated over.
+   */
+  Image *image;
+
+  /**
+   * \brief Local copy of the image user.
+   *
+   * The local copy is required so we don't change the image user of the caller.
+   * We need to change it in order to request data for a specific tile.
+   */
+  ImageUser image_user = {0};
+
+  /**
+   * \brief ImageTile associated with the loaded tile.
+   * Data is not owned by this instance but by the `image`.
+   */
+  ImageTile *tile = nullptr;
+
+  /**
+   * \brief ImBuf of the loaded tile.
+   *
+   * Can be nullptr when the file doesn't exist or when the tile hasn't been initialized.
+   */
+  ImBuf *tile_buffer = nullptr;
+
+  ImageTileData(Image *image, ImageUser *image_user) : image(image)
+  {
+    if (image_user != nullptr) {
+      this->image_user = *image_user;
+    }
+  }
+
+  void init_data(TileNumber new_tile_number) override
+  {
+    image_user.tile = new_tile_number;
+    tile = BKE_image_get_tile(image, new_tile_number);
+    tile_buffer = BKE_image_acquire_ibuf(image, &image_user, NULL);
+  }
+
+  void free_data() override
+  {
+    BKE_image_release_ibuf(image, tile_buffer, nullptr);
+    tile = nullptr;
+    tile_buffer = nullptr;
+  }
+};
+
+template<typename TileData = NoTileData> struct PartialUpdateChecker {
+
+  /**
+   * \brief Not owned Image that is being iterated over.
+   */
+  Image *image;
+  ImageUser *image_user;
+
+  /**
+   * \brief the collected changes are stored inside the PartialUpdateUser.
+   */
+  PartialUpdateUser *user;
+
+  struct CollectResult {
+    PartialUpdateChecker<TileData> *checker;
+
+    /**
+     * \brief Tile specific data.
+     */
+    TileData tile_data;
+    PartialUpdateRegion changed_region;
+    ePartialUpdateCollectResult result_code;
+
+   private:
+    TileNumber last_tile_number;
+
+   public:
+    CollectResult(PartialUpdateChecker<TileData> *checker, ePartialUpdateCollectResult result_code)
+        : checker(checker),
+          tile_data(checker->image, checker->image_user),
+          result_code(result_code)
+    {
+    }
+
+    const ePartialUpdateCollectResult get_result_code() const
+    {
+      return result_code;
+    }
+
+    /**
+     * \brief Load the next changed region.
+     *
+     * This member function can only be called when partial changes are detected.
+     * (`get_result_code()` returns `ePartialUpdateCollectResult::PartialChangesDetected`).
+     *
+     * When changes for another tile than the previous tile is loaded the #tile_data will be
+     * updated.
+     */
+    ePartialUpdateIterResult get_next_change()
+    {
+      BLI_assert(result_code == ePartialUpdateCollectResult::PartialChangesDetected);
+      ePartialUpdateIterResult result = BKE_image_partial_update_get_next_change(checker->user,
+                                                                                 &changed_region);
+      switch (result) {
+        case ePartialUpdateIterResult::Finished:
+          tile_data.free_data();
+          return result;
+
+        case ePartialUpdateIterResult::ChangeAvailable:
+          if (last_tile_number == changed_region.tile_number) {
+            return result;
+          }
+          tile_data.free_data();
+          tile_data.init_data(changed_region.tile_number);
+          last_tile_number = changed_region.tile_number;
+          return result;
+
+        default:
+          BLI_assert_unreachable();
+          return result;
+      }
+    }
+  };
+
+ public:
+  PartialUpdateChecker(Image *image, ImageUser *image_user, PartialUpdateUser *user)
+      : image(image), image_user(image_user), user(user)
+  {
+  }
+
+  /**
+   * \brief Check for new changes since the last time this method was invoked for this #user.
+   */
+  CollectResult collect_changes()
+  {
+    ePartialUpdateCollectResult collect_result = BKE_image_partial_update_collect_changes(image,
+                                                                                          user);
+    return CollectResult(this, collect_result);
+  }
+};
+
+}  // namespace partial_update
+}  // namespace blender::bke::image

source/blender/blenkernel/BKE_lib_override.h

@@ -100,6 +100,9 @@ struct ID *BKE_lib_override_library_create_from_id(struct Main *bmain,
  * main. You can add more local IDs to be remapped to use new overriding ones by setting their
  * LIB_TAG_DOIT tag.
  *
+ * \param owner_library: the library in which the overrides should be created. Besides versioning
+ * and resync code path, this should always be NULL (i.e. the local .blend file).
+ *
  * \param reference_library: the library from which the linked data being overridden come from
  * (i.e. the library of the linked reference ID).
  *
@@ -109,6 +112,7 @@ struct ID *BKE_lib_override_library_create_from_id(struct Main *bmain,
  * \return \a true on success, \a false otherwise.
  */
 bool BKE_lib_override_library_create_from_tag(struct Main *bmain,
+                                              struct Library *owner_library,
                                               const struct Library *reference_library,
                                               bool do_no_main);
 /**
@@ -122,16 +126,24 @@ bool BKE_lib_override_library_create_from_tag(struct Main *bmain,
  *
  * \param view_layer: the active view layer to search instantiated collections in, can be NULL (in
  *                    which case \a scene's master collection children hierarchy is used instead).
+ *
+ * \param owner_library: the library in which the overrides should be created. Besides versioning
+ * and resync code path, this should always be NULL (i.e. the local .blend file).
+ *
  * \param id_root: The root ID to create an override from.
+ *
  * \param id_reference: Some reference ID used to do some post-processing after overrides have been
  * created, may be NULL. Typically, the Empty object instantiating the linked collection we
  * override, currently.
+ *
  * \param r_id_root_override: if not NULL, the override generated for the given \a id_root.
+ *
  * \return true if override was successfully created.
  */
 bool BKE_lib_override_library_create(struct Main *bmain,
                                      struct Scene *scene,
                                      struct ViewLayer *view_layer,
+                                     struct Library *owner_library,
                                      struct ID *id_root,
                                      struct ID *id_reference,
                                      struct ID **r_id_root_override);

source/blender/blenkernel/BKE_lib_remap.h

@@ -38,6 +38,9 @@
 extern "C" {
 #endif
 
+struct ID;
+struct IDRemapper;
+
 /* BKE_libblock_free, delete are declared in BKE_lib_id.h for convenience. */
 
 /* Also IDRemap->flag. */
@@ -97,6 +100,19 @@ enum {
   ID_REMAP_FORCE_OBDATA_IN_EDITMODE = 1 << 9,
 };
 
+/**
+ * Replace all references in given Main using the given \a mappings
+ *
+ * \note Is preferred over BKE_libblock_remap_locked due to performance.
+ */
+void BKE_libblock_remap_multiple_locked(struct Main *bmain,
+                                        const struct IDRemapper *mappings,
+                                        const short remap_flags);
+
+void BKE_libblock_remap_multiple(struct Main *bmain,
+                                 const struct IDRemapper *mappings,
+                                 const short remap_flags);
+
 /**
  * Replace all references in given Main to \a old_id by \a new_id
  * (if \a new_id is NULL, it unlinks \a old_id).
@@ -146,12 +162,61 @@ void BKE_libblock_relink_to_newid(struct Main *bmain, struct ID *id, int remap_f
     ATTR_NONNULL();
 
 typedef void (*BKE_library_free_notifier_reference_cb)(const void *);
-typedef void (*BKE_library_remap_editor_id_reference_cb)(struct ID *, struct ID *);
+typedef void (*BKE_library_remap_editor_id_reference_cb)(const struct IDRemapper *mappings);
 
 void BKE_library_callback_free_notifier_reference_set(BKE_library_free_notifier_reference_cb func);
 void BKE_library_callback_remap_editor_id_reference_set(
     BKE_library_remap_editor_id_reference_cb func);
 
+/* IDRemapper */
+struct IDRemapper;
+typedef enum IDRemapperApplyResult {
+  /** No remapping rules available for the source. */
+  ID_REMAP_RESULT_SOURCE_UNAVAILABLE,
+  /** Source isn't mappable (e.g. NULL). */
+  ID_REMAP_RESULT_SOURCE_NOT_MAPPABLE,
+  /** Source has been remapped to a new pointer. */
+  ID_REMAP_RESULT_SOURCE_REMAPPED,
+  /** Source has been set to NULL. */
+  ID_REMAP_RESULT_SOURCE_UNASSIGNED,
+} IDRemapperApplyResult;
+
+typedef enum IDRemapperApplyOptions {
+  ID_REMAP_APPLY_UPDATE_REFCOUNT = (1 << 0),
+  ID_REMAP_APPLY_ENSURE_REAL = (1 << 1),
+
+  ID_REMAP_APPLY_DEFAULT = 0,
+} IDRemapperApplyOptions;
+
+typedef void (*IDRemapperIterFunction)(struct ID *old_id, struct ID *new_id, void *user_data);
+
+/**
+ * Create a new ID Remapper.
+ *
+ * An ID remapper stores multiple remapping rules.
+ */
+struct IDRemapper *BKE_id_remapper_create(void);
+
+void BKE_id_remapper_clear(struct IDRemapper *id_remapper);
+bool BKE_id_remapper_is_empty(const struct IDRemapper *id_remapper);
+/** Free the given ID Remapper. */
+void BKE_id_remapper_free(struct IDRemapper *id_remapper);
+/** Add a new remapping. */
+void BKE_id_remapper_add(struct IDRemapper *id_remapper, struct ID *old_id, struct ID *new_id);
+
+/**
+ * Apply a remapping.
+ *
+ * Update the id pointer stored in the given r_id_ptr if a remapping rule exists.
+ */
+IDRemapperApplyResult BKE_id_remapper_apply(const struct IDRemapper *id_remapper,
+                                            struct ID **r_id_ptr,
+                                            IDRemapperApplyOptions options);
+bool BKE_id_remapper_has_mapping_for(const struct IDRemapper *id_remapper, uint64_t type_filter);
+void BKE_id_remapper_iter(const struct IDRemapper *id_remapper,
+                          IDRemapperIterFunction func,
+                          void *user_data);
+
 #ifdef __cplusplus
 }
 #endif

source/blender/blenkernel/BKE_modifier.h

@@ -245,10 +245,6 @@ typedef struct ModifierTypeInfo {
                              const struct ModifierEvalContext *ctx,
                              struct Mesh *mesh);
 
-  struct Hair *(*modifyHair)(struct ModifierData *md,
-                             const struct ModifierEvalContext *ctx,
-                             struct Hair *hair);
-
   /**
    * The modifier has to change the geometry set in-place. The geometry set can contain zero or
    * more geometry components. This callback can be used by modifiers that don't work on any

source/blender/blenkernel/BKE_node.h

@@ -1291,6 +1291,8 @@ void BKE_nodetree_remove_layer_n(struct bNodeTree *ntree, struct Scene *scene, i
 #define CMP_NODE_POSTERIZE 327
 #define CMP_NODE_CONVERT_COLOR_SPACE 328
 #define CMP_NODE_SCENE_TIME 329
+#define CMP_NODE_SEPARATE_XYZ 330
+#define CMP_NODE_COMBINE_XYZ 331
 
 /* channel toggles */
 #define CMP_CHAN_RGB 1

source/blender/blenkernel/BKE_paint.h

@@ -499,7 +499,6 @@ typedef struct SculptSession {
 
   /* These are always assigned to base mesh data when using PBVH_FACES and PBVH_GRIDS. */
   struct MVert *mvert;
-  const float (*vert_normals)[3];
   struct MPoly *mpoly;
   struct MLoop *mloop;
 

source/blender/blenkernel/BKE_screen.h

@@ -38,6 +38,7 @@ struct BlendLibReader;
 struct BlendWriter;
 struct Header;
 struct ID;
+struct IDRemapper;
 struct LibraryForeachIDData;
 struct ListBase;
 struct Menu;
@@ -117,10 +118,7 @@ typedef struct SpaceType {
   bContextDataCallback context;
 
   /* Used when we want to replace an ID by another (or NULL). */
-  void (*id_remap)(struct ScrArea *area,
-                   struct SpaceLink *sl,
-                   struct ID *old_id,
-                   struct ID *new_id);
+  void (*id_remap)(struct ScrArea *area, struct SpaceLink *sl, const struct IDRemapper *mappings);
 
   int (*space_subtype_get)(struct ScrArea *area);
   void (*space_subtype_set)(struct ScrArea *area, int value);

source/blender/blenkernel/CMakeLists.txt

@@ -165,6 +165,7 @@ set(SRC
   intern/idprop_utils.c
   intern/idtype.c
   intern/image.c
+  intern/image_partial_update.cc
   intern/image_gen.c
   intern/image_gpu.cc
   intern/image_save.c
@@ -179,6 +180,7 @@ set(SRC
   intern/lib_id.c
   intern/lib_id_delete.c
   intern/lib_id_eval.c
+  intern/lib_id_remapper.cc
   intern/lib_override.c
   intern/lib_query.c
   intern/lib_remap.c
@@ -522,7 +524,7 @@ set(LIB
   bf_simulation
 
   # For `vfontdata_freetype.c`.
-  ${FREETYPE_LIBRARIES}
+  ${FREETYPE_LIBRARIES} ${BROTLI_LIBRARIES}
 )
 
 if(WITH_BINRELOC)
@@ -717,10 +719,6 @@ if(WITH_FFTW3)
   add_definitions(-DFFTW3=1)
 endif()
 
-if(WITH_INTERNATIONAL)
-  add_definitions(-DWITH_INTERNATIONAL)
-endif()
-
 if(WITH_FREESTYLE)
   add_definitions(-DWITH_FREESTYLE)
 endif()
@@ -821,8 +819,10 @@ if(WITH_GTESTS)
     intern/cryptomatte_test.cc
     intern/fcurve_test.cc
     intern/idprop_serialize_test.cc
+    intern/image_partial_update_test.cc
     intern/lattice_deform_test.cc
     intern/layer_test.cc
+    intern/lib_id_remapper_test.cc
     intern/lib_id_test.cc
     intern/lib_remap_test.cc
     intern/tracking_test.cc

source/blender/blenkernel/intern/DerivedMesh.cc

@@ -127,30 +127,6 @@ static MEdge *dm_getEdgeArray(DerivedMesh *dm)
   return medge;
 }
 
-static MFace *dm_getTessFaceArray(DerivedMesh *dm)
-{
-  MFace *mface = (MFace *)CustomData_get_layer(&dm->faceData, CD_MFACE);
-
-  if (!mface) {
-    int numTessFaces = dm->getNumTessFaces(dm);
-
-    if (!numTessFaces) {
-      /* Do not add layer if there's no elements in it, this leads to issues later when
-       * this layer is needed with non-zero size, but currently CD stuff does not check
-       * for requested layer size on creation and just returns layer which was previously
-       * added (sergey) */
-      return nullptr;
-    }
-
-    mface = (MFace *)CustomData_add_layer(
-        &dm->faceData, CD_MFACE, CD_CALLOC, nullptr, numTessFaces);
-    CustomData_set_layer_flag(&dm->faceData, CD_MFACE, CD_FLAG_TEMPORARY);
-    dm->copyTessFaceArray(dm, mface);
-  }
-
-  return mface;
-}
-
 static MLoop *dm_getLoopArray(DerivedMesh *dm)
 {
   MLoop *mloop = (MLoop *)CustomData_get_layer(&dm->loopData, CD_MLOOP);
@@ -203,18 +179,6 @@ static MEdge *dm_dupEdgeArray(DerivedMesh *dm)
   return tmp;
 }
 
-static MFace *dm_dupFaceArray(DerivedMesh *dm)
-{
-  MFace *tmp = (MFace *)MEM_malloc_arrayN(
-      dm->getNumTessFaces(dm), sizeof(*tmp), "dm_dupFaceArray tmp");
-
-  if (tmp) {
-    dm->copyTessFaceArray(dm, tmp);
-  }
-
-  return tmp;
-}
-
 static MLoop *dm_dupLoopArray(DerivedMesh *dm)
 {
   MLoop *tmp = (MLoop *)MEM_malloc_arrayN(
@@ -270,42 +234,15 @@ static const MLoopTri *dm_getLoopTriArray(DerivedMesh *dm)
   return looptri;
 }
 
-static CustomData *dm_getVertCData(DerivedMesh *dm)
-{
-  return &dm->vertData;
-}
-
-static CustomData *dm_getEdgeCData(DerivedMesh *dm)
-{
-  return &dm->edgeData;
-}
-
-static CustomData *dm_getTessFaceCData(DerivedMesh *dm)
-{
-  return &dm->faceData;
-}
-
-static CustomData *dm_getLoopCData(DerivedMesh *dm)
-{
-  return &dm->loopData;
-}
-
-static CustomData *dm_getPolyCData(DerivedMesh *dm)
-{
-  return &dm->polyData;
-}
-
 void DM_init_funcs(DerivedMesh *dm)
 {
   /* default function implementations */
   dm->getVertArray = dm_getVertArray;
   dm->getEdgeArray = dm_getEdgeArray;
-  dm->getTessFaceArray = dm_getTessFaceArray;
   dm->getLoopArray = dm_getLoopArray;
   dm->getPolyArray = dm_getPolyArray;
   dm->dupVertArray = dm_dupVertArray;
   dm->dupEdgeArray = dm_dupEdgeArray;
-  dm->dupTessFaceArray = dm_dupFaceArray;
   dm->dupLoopArray = dm_dupLoopArray;
   dm->dupPolyArray = dm_dupPolyArray;
 
@@ -314,19 +251,8 @@ void DM_init_funcs(DerivedMesh *dm)
   /* subtypes handle getting actual data */
   dm->getNumLoopTri = dm_getNumLoopTri;
 
-  dm->getVertDataLayout = dm_getVertCData;
-  dm->getEdgeDataLayout = dm_getEdgeCData;
-  dm->getTessFaceDataLayout = dm_getTessFaceCData;
-  dm->getLoopDataLayout = dm_getLoopCData;
-  dm->getPolyDataLayout = dm_getPolyCData;
-
-  dm->getVertData = DM_get_vert_data;
-  dm->getEdgeData = DM_get_edge_data;
-  dm->getTessFaceData = DM_get_tessface_data;
-  dm->getPolyData = DM_get_poly_data;
   dm->getVertDataArray = DM_get_vert_data_layer;
   dm->getEdgeDataArray = DM_get_edge_data_layer;
-  dm->getTessFaceDataArray = DM_get_tessface_data_layer;
   dm->getPolyDataArray = DM_get_poly_data_layer;
   dm->getLoopDataArray = DM_get_loop_data_layer;
 }
@@ -349,7 +275,6 @@ void DM_init(DerivedMesh *dm,
   DM_init_funcs(dm);
 
   dm->needsFree = 1;
-  dm->dirty = (DMDirtyFlag)0;
 
   /* Don't use #CustomData_reset because we don't want to touch custom-data. */
   copy_vn_i(dm->vertData.typemap, CD_NUMTYPES, -1);
@@ -359,16 +284,16 @@ void DM_init(DerivedMesh *dm,
   copy_vn_i(dm->polyData.typemap, CD_NUMTYPES, -1);
 }
 
-void DM_from_template_ex(DerivedMesh *dm,
-                         DerivedMesh *source,
-                         DerivedMeshType type,
-                         int numVerts,
-                         int numEdges,
-                         int numTessFaces,
-                         int numLoops,
-                         int numPolys,
-                         const CustomData_MeshMasks *mask)
+void DM_from_template(DerivedMesh *dm,
+                      DerivedMesh *source,
+                      DerivedMeshType type,
+                      int numVerts,
+                      int numEdges,
+                      int numTessFaces,
+                      int numLoops,
+                      int numPolys)
 {
+  const CustomData_MeshMasks *mask = &CD_MASK_DERIVEDMESH;
   CustomData_copy(&source->vertData, &dm->vertData, mask->vmask, CD_CALLOC, numVerts);
   CustomData_copy(&source->edgeData, &dm->edgeData, mask->emask, CD_CALLOC, numEdges);
   CustomData_copy(&source->faceData, &dm->faceData, mask->fmask, CD_CALLOC, numTessFaces);
@@ -387,26 +312,6 @@ void DM_from_template_ex(DerivedMesh *dm,
   DM_init_funcs(dm);
 
   dm->needsFree = 1;
-  dm->dirty = (DMDirtyFlag)0;
-}
-void DM_from_template(DerivedMesh *dm,
-                      DerivedMesh *source,
-                      DerivedMeshType type,
-                      int numVerts,
-                      int numEdges,
-                      int numTessFaces,
-                      int numLoops,
-                      int numPolys)
-{
-  DM_from_template_ex(dm,
-                      source,
-                      type,
-                      numVerts,
-                      numEdges,
-                      numTessFaces,
-                      numLoops,
-                      numPolys,
-                      &CD_MASK_DERIVEDMESH);
 }
 
 bool DM_release(DerivedMesh *dm)
@@ -464,14 +369,6 @@ void DM_DupPolys(DerivedMesh *source, DerivedMesh *target)
   }
 }
 
-void DM_ensure_normals(DerivedMesh *dm)
-{
-  if (dm->dirty & DM_DIRTY_NORMALS) {
-    dm->calcNormals(dm);
-  }
-  BLI_assert((dm->dirty & DM_DIRTY_NORMALS) == 0);
-}
-
 void DM_ensure_looptri_data(DerivedMesh *dm)
 {
   const unsigned int totpoly = dm->numPolyData;
@@ -524,7 +421,7 @@ void DM_set_only_copy(DerivedMesh *dm, const CustomData_MeshMasks *mask)
    * see replies to r50969, Campbell */
 #if 0
   CustomData_set_only_copy(&dm->loopData, mask->lmask);
-  CustomData_set_only_copy(&dm->polyData, mask->pmask);
+  Custom(&dm->polyData, mask->pmask);
 #endif
 }
 
@@ -552,45 +449,11 @@ void DM_add_edge_layer(DerivedMesh *dm, int type, eCDAllocType alloctype, void *
   CustomData_add_layer(&dm->edgeData, type, alloctype, layer, dm->numEdgeData);
 }
 
-void DM_add_tessface_layer(DerivedMesh *dm, int type, eCDAllocType alloctype, void *layer)
-{
-  CustomData_add_layer(&dm->faceData, type, alloctype, layer, dm->numTessFaceData);
-}
-
-void DM_add_loop_layer(DerivedMesh *dm, int type, eCDAllocType alloctype, void *layer)
-{
-  CustomData_add_layer(&dm->loopData, type, alloctype, layer, dm->numLoopData);
-}
-
 void DM_add_poly_layer(DerivedMesh *dm, int type, eCDAllocType alloctype, void *layer)
 {
   CustomData_add_layer(&dm->polyData, type, alloctype, layer, dm->numPolyData);
 }
 
-void *DM_get_vert_data(DerivedMesh *dm, int index, int type)
-{
-  BLI_assert(index >= 0 && index < dm->getNumVerts(dm));
-  return CustomData_get(&dm->vertData, index, type);
-}
-
-void *DM_get_edge_data(DerivedMesh *dm, int index, int type)
-{
-  BLI_assert(index >= 0 && index < dm->getNumEdges(dm));
-  return CustomData_get(&dm->edgeData, index, type);
-}
-
-void *DM_get_tessface_data(DerivedMesh *dm, int index, int type)
-{
-  BLI_assert(index >= 0 && index < dm->getNumTessFaces(dm));
-  return CustomData_get(&dm->faceData, index, type);
-}
-
-void *DM_get_poly_data(DerivedMesh *dm, int index, int type)
-{
-  BLI_assert(index >= 0 && index < dm->getNumPolys(dm));
-  return CustomData_get(&dm->polyData, index, type);
-}
-
 void *DM_get_vert_data_layer(DerivedMesh *dm, int type)
 {
   if (type == CD_MVERT) {
@@ -609,15 +472,6 @@ void *DM_get_edge_data_layer(DerivedMesh *dm, int type)
   return CustomData_get_layer(&dm->edgeData, type);
 }
 
-void *DM_get_tessface_data_layer(DerivedMesh *dm, int type)
-{
-  if (type == CD_MFACE) {
-    return dm->getTessFaceArray(dm);
-  }
-
-  return CustomData_get_layer(&dm->faceData, type);
-}
-
 void *DM_get_poly_data_layer(DerivedMesh *dm, int type)
 {
   return CustomData_get_layer(&dm->polyData, type);

source/blender/blenkernel/intern/blendfile.c

@@ -78,6 +78,23 @@
 /** \name High Level `.blend` file read/write.
  * \{ */
 
+static bool blendfile_or_libraries_versions_atleast(Main *bmain,
+                                                    const short versionfile,
+                                                    const short subversionfile)
+{
+  if (!MAIN_VERSION_ATLEAST(bmain, versionfile, subversionfile)) {
+    return false;
+  }
+
+  LISTBASE_FOREACH (Library *, library, &bmain->libraries) {
+    if (!MAIN_VERSION_ATLEAST(library, versionfile, subversionfile)) {
+      return false;
+    }
+  }
+
+  return true;
+}
+
 static bool foreach_path_clean_cb(BPathForeachPathData *UNUSED(bpath_data),
                                   char *path_dst,
                                   const char *path_src)
@@ -349,10 +366,11 @@ static void setup_app_data(bContext *C,
     do_versions_ipos_to_animato(bmain);
   }
 
-  /* FIXME: Same as above, readfile's `do_version` do not allow to create new IDs. */
-  /* TODO: Once this is definitively validated for 3.0 and option to not do it is removed, add a
-   * version bump and check here. */
-  if (mode != LOAD_UNDO && !USER_EXPERIMENTAL_TEST(&U, no_proxy_to_override_conversion)) {
+  /* NOTE: readfile's `do_version` does not allow to create new IDs, and only operates on a single
+   * library at a time. This code needs to operate on the whole Main at once. */
+  /* NOTE: Check bmain version (i.e. current blend file version), AND the versions of all the
+   * linked libraries. */
+  if (mode != LOAD_UNDO && !blendfile_or_libraries_versions_atleast(bmain, 302, 1)) {
     BKE_lib_override_library_main_proxy_convert(bmain, reports);
   }
 

source/blender/blenkernel/intern/blendfile_link_append.c

@@ -993,6 +993,27 @@ static int foreach_libblock_link_append_callback(LibraryIDLinkCallbackData *cb_d
 /** \name Library link/append code.
  * \{ */
 
+static void blendfile_link_append_proxies_convert(Main *bmain, ReportList *reports)
+{
+  /* NOTE: Do not bother checking file versions here, if there are no proxies to convert this code
+   * is quite fast anyway. */
+
+  BlendFileReadReport bf_reports = {.reports = reports};
+  BKE_lib_override_library_main_proxy_convert(bmain, &bf_reports);
+
+  if (bf_reports.count.proxies_to_lib_overrides_success != 0 ||
+      bf_reports.count.proxies_to_lib_overrides_failures != 0) {
+    BKE_reportf(
+        bf_reports.reports,
+        RPT_WARNING,
+        "Proxies have been removed from Blender (%d proxies were automatically converted "
+        "to library overrides, %d proxies could not be converted and were cleared). "
+        "Please consider re-saving any library .blend file with the newest Blender version.",
+        bf_reports.count.proxies_to_lib_overrides_success,
+        bf_reports.count.proxies_to_lib_overrides_failures);
+  }
+}
+
 void BKE_blendfile_append(BlendfileLinkAppendContext *lapp_context, ReportList *reports)
 {
   if (lapp_context->num_items == 0) {
@@ -1259,6 +1280,8 @@ void BKE_blendfile_append(BlendfileLinkAppendContext *lapp_context, ReportList *
   }
 
   BKE_main_id_newptr_and_tag_clear(bmain);
+
+  blendfile_link_append_proxies_convert(bmain, reports);
 }
 
 void BKE_blendfile_link(BlendfileLinkAppendContext *lapp_context, ReportList *reports)
@@ -1361,6 +1384,10 @@ void BKE_blendfile_link(BlendfileLinkAppendContext *lapp_context, ReportList *re
                                                        .active_collection = NULL};
     loose_data_instantiate(&instantiate_context);
   }
+
+  if ((lapp_context->params->flag & FILE_LINK) != 0) {
+    blendfile_link_append_proxies_convert(lapp_context->params->bmain, reports);
+  }
 }
 
 /** \} */

source/blender/blenkernel/intern/camera.c

@@ -218,7 +218,7 @@ void *BKE_camera_add(Main *bmain, const char *name)
   return cam;
 }
 
-float BKE_camera_object_dof_distance(Object *ob)
+float BKE_camera_object_dof_distance(const Object *ob)
 {
   Camera *cam = (Camera *)ob->data;
   if (ob->type != OB_CAMERA) {

source/blender/blenkernel/intern/cdderivedmesh.c

@@ -116,12 +116,6 @@ static void cdDM_copyEdgeArray(DerivedMesh *dm, MEdge *r_edge)
   memcpy(r_edge, cddm->medge, sizeof(*r_edge) * dm->numEdgeData);
 }
 
-static void cdDM_copyTessFaceArray(DerivedMesh *dm, MFace *r_face)
-{
-  CDDerivedMesh *cddm = (CDDerivedMesh *)dm;
-  memcpy(r_face, cddm->mface, sizeof(*r_face) * dm->numTessFaceData);
-}
-
 static void cdDM_copyLoopArray(DerivedMesh *dm, MLoop *r_loop)
 {
   CDDerivedMesh *cddm = (CDDerivedMesh *)dm;
@@ -147,20 +141,6 @@ static void cdDM_getVertNo(DerivedMesh *dm, int index, float r_no[3])
   copy_v3_v3(r_no, cddm->vert_normals[index]);
 }
 
-static const MeshElemMap *cdDM_getPolyMap(Object *ob, DerivedMesh *dm)
-{
-  CDDerivedMesh *cddm = (CDDerivedMesh *)dm;
-
-  if (!cddm->pmap && ob->type == OB_MESH) {
-    Mesh *me = ob->data;
-
-    BKE_mesh_vert_poly_map_create(
-        &cddm->pmap, &cddm->pmap_mem, me->mpoly, me->mloop, me->totvert, me->totpoly, me->totloop);
-  }
-
-  return cddm->pmap;
-}
-
 static void cdDM_recalc_looptri(DerivedMesh *dm)
 {
   CDDerivedMesh *cddm = (CDDerivedMesh *)dm;
@@ -216,24 +196,17 @@ static CDDerivedMesh *cdDM_create(const char *desc)
 
   dm->copyVertArray = cdDM_copyVertArray;
   dm->copyEdgeArray = cdDM_copyEdgeArray;
-  dm->copyTessFaceArray = cdDM_copyTessFaceArray;
   dm->copyLoopArray = cdDM_copyLoopArray;
   dm->copyPolyArray = cdDM_copyPolyArray;
 
-  dm->getVertData = DM_get_vert_data;
-  dm->getEdgeData = DM_get_edge_data;
-  dm->getTessFaceData = DM_get_tessface_data;
   dm->getVertDataArray = DM_get_vert_data_layer;
   dm->getEdgeDataArray = DM_get_edge_data_layer;
-  dm->getTessFaceDataArray = DM_get_tessface_data_layer;
 
   dm->recalcLoopTri = cdDM_recalc_looptri;
 
   dm->getVertCo = cdDM_getVertCo;
   dm->getVertNo = cdDM_getVertNo;
 
-  dm->getPolyMap = cdDM_getPolyMap;
-
   dm->release = cdDM_release;
 
   return cddm;
@@ -265,12 +238,6 @@ static DerivedMesh *cdDM_from_mesh_ex(Mesh *mesh,
   dm->deformedOnly = 1;
   dm->cd_flag = mesh->cd_flag;
 
-  if (mesh->runtime.cd_dirty_vert & CD_MASK_NORMAL) {
-    dm->dirty |= DM_DIRTY_NORMALS;
-  }
-  /* TODO: DM_DIRTY_TESS_CDLAYERS ? Maybe not though,
-   * since we probably want to switch to looptris? */
-
   CustomData_merge(&mesh->vdata, &dm->vertData, cddata_masks.vmask, alloctype, mesh->totvert);
   CustomData_merge(&mesh->edata, &dm->edgeData, cddata_masks.emask, alloctype, mesh->totedge);
   CustomData_merge(&mesh->fdata,
@@ -282,7 +249,9 @@ static DerivedMesh *cdDM_from_mesh_ex(Mesh *mesh,
   CustomData_merge(&mesh->pdata, &dm->polyData, cddata_masks.pmask, alloctype, mesh->totpoly);
 
   cddm->mvert = CustomData_get_layer(&dm->vertData, CD_MVERT);
-  cddm->vert_normals = CustomData_get_layer(&dm->vertData, CD_NORMAL);
+  /* Though this may be an unnecessary calculation, simply retrieving the layer may return nothing
+   * or dirty normals. */
+  cddm->vert_normals = BKE_mesh_vertex_normals_ensure(mesh);
   cddm->medge = CustomData_get_layer(&dm->edgeData, CD_MEDGE);
   cddm->mloop = CustomData_get_layer(&dm->loopData, CD_MLOOP);
   cddm->mpoly = CustomData_get_layer(&dm->polyData, CD_MPOLY);
@@ -327,12 +296,6 @@ DerivedMesh *CDDM_copy(DerivedMesh *source)
   DM_from_template(dm, source, DM_TYPE_CDDM, numVerts, numEdges, numTessFaces, numLoops, numPolys);
   dm->deformedOnly = source->deformedOnly;
   dm->cd_flag = source->cd_flag;
-  dm->dirty = source->dirty;
-
-  /* Tessellation data is never copied, so tag it here.
-   * Only tag dirty layers if we really ignored tessellation faces.
-   */
-  dm->dirty |= DM_DIRTY_TESS_CDLAYERS;
 
   CustomData_copy_data(&source->vertData, &dm->vertData, 0, 0, numVerts);
   CustomData_copy_data(&source->edgeData, &dm->edgeData, 0, 0, numEdges);

source/blender/blenkernel/intern/crazyspace.c

@@ -194,13 +194,10 @@ void BKE_crazyspace_set_quats_mesh(Mesh *me,
                                    float (*mappedcos)[3],
                                    float (*quats)[4])
 {
-  MVert *mvert = me->mvert;
-  for (int i = 0; i < me->totvert; i++, mvert++) {
-    mvert->flag &= ~ME_VERT_TMP_TAG;
-  }
+  BLI_bitmap *vert_tag = BLI_BITMAP_NEW(me->totvert, __func__);
 
   /* first store two sets of tangent vectors in vertices, we derive it just from the face-edges */
-  mvert = me->mvert;
+  MVert *mvert = me->mvert;
   MPoly *mp = me->mpoly;
   MLoop *mloop = me->mloop;
 
@@ -210,7 +207,7 @@ void BKE_crazyspace_set_quats_mesh(Mesh *me,
     MLoop *ml_prev = &ml_next[mp->totloop - 2];
 
     for (int j = 0; j < mp->totloop; j++) {
-      if ((mvert[ml_curr->v].flag & ME_VERT_TMP_TAG) == 0) {
+      if (!BLI_BITMAP_TEST(vert_tag, ml_curr->v)) {
         const float *co_prev, *co_curr, *co_next; /* orig */
         const float *vd_prev, *vd_curr, *vd_next; /* deform */
 
@@ -233,7 +230,7 @@ void BKE_crazyspace_set_quats_mesh(Mesh *me,
         set_crazy_vertex_quat(
             quats[ml_curr->v], co_curr, co_next, co_prev, vd_curr, vd_next, vd_prev);
 
-        mvert[ml_curr->v].flag |= ME_VERT_TMP_TAG;
+        BLI_BITMAP_ENABLE(vert_tag, ml_curr->v);
       }
 
       ml_prev = ml_curr;
@@ -241,6 +238,8 @@ void BKE_crazyspace_set_quats_mesh(Mesh *me,
       ml_next++;
     }
   }
+
+  MEM_freeN(vert_tag);
 }
 
 int BKE_crazyspace_get_first_deform_matrices_editbmesh(struct Depsgraph *depsgraph,

source/blender/blenkernel/intern/curve_to_mesh_convert.cc

@@ -66,8 +66,8 @@ static void vert_extrude_to_mesh_data(const Spline &spline,
 
   if (spline.is_cyclic() && spline.evaluated_edges_size() > 1) {
     MEdge &edge = r_edges[edge_offset + spline.evaluated_edges_size() - 1];
-    edge.v1 = vert_offset;
-    edge.v2 = vert_offset + eval_size - 1;
+    edge.v1 = vert_offset + eval_size - 1;
+    edge.v2 = vert_offset;
     edge.flag = ME_LOOSEEDGE;
   }
 

source/blender/blenkernel/intern/customdata.cc

@@ -2427,6 +2427,13 @@ int CustomData_get_stencil_layer(const CustomData *data, int type)
   return (layer_index != -1) ? data->layers[layer_index].active_mask : -1;
 }
 
+const char *CustomData_get_active_layer_name(const struct CustomData *data, const int type)
+{
+  /* Get the layer index of the active layer of this type. */
+  const int layer_index = CustomData_get_active_layer_index(data, type);
+  return layer_index < 0 ? NULL : data->layers[layer_index].name;
+}
+
 void CustomData_set_layer_active(CustomData *data, int type, int n)
 {
   for (int i = 0; i < data->totlayer; i++) {

source/blender/blenkernel/intern/dynamicpaint.c

@@ -1518,7 +1518,7 @@ static void dynamic_paint_set_init_color_tex_to_vcol_cb(
     multitex_ext_safe(tex, uv, &texres, pool, scene_color_manage, false);
 
     if (texres.tin > pPoint[vert].color[3]) {
-      copy_v3_v3(pPoint[vert].color, &texres.tr);
+      copy_v3_v3(pPoint[vert].color, texres.trgba);
       pPoint[vert].color[3] = texres.tin;
     }
   }
@@ -1559,7 +1559,7 @@ static void dynamic_paint_set_init_color_tex_to_imseq_cb(
   multitex_ext_safe(tex, uv_final, &texres, NULL, scene_color_manage, false);
 
   /* apply color */
-  copy_v3_v3(pPoint[i].color, &texres.tr);
+  copy_v3_v3(pPoint[i].color, texres.trgba);
   pPoint[i].color[3] = texres.tin;
 }
 
@@ -1985,9 +1985,6 @@ static Mesh *dynamicPaint_Modifier_apply(DynamicPaintModifierData *pmd, Object *
             }
 
             MEM_freeN(fcolor);
-
-            /* Mark tessellated CD layers as dirty. */
-            // result->dirty |= DM_DIRTY_TESS_CDLAYERS;
           }
           /* vertex group paint */
           else if (surface->type == MOD_DPAINT_SURFACE_T_WEIGHT) {

source/blender/blenkernel/intern/effect.c

@@ -908,9 +908,9 @@ static void do_texture_effector(EffectorCache *eff,
       eff->pd->tex, tex_co, NULL, NULL, 0, result, 0, NULL, scene_color_manage, false);
 
   if (hasrgb && mode == PFIELD_TEX_RGB) {
-    force[0] = (0.5f - result->tr) * strength;
-    force[1] = (0.5f - result->tg) * strength;
-    force[2] = (0.5f - result->tb) * strength;
+    force[0] = (0.5f - result->trgba[0]) * strength;
+    force[1] = (0.5f - result->trgba[1]) * strength;
+    force[2] = (0.5f - result->trgba[2]) * strength;
   }
   else if (nabla != 0) {
     strength /= nabla;
@@ -933,7 +933,8 @@ static void do_texture_effector(EffectorCache *eff,
       /* generate intensity if texture only has rgb value */
       if (hasrgb & TEX_RGB) {
         for (int i = 0; i < 4; i++) {
-          result[i].tin = (1.0f / 3.0f) * (result[i].tr + result[i].tg + result[i].tb);
+          result[i].tin = (1.0f / 3.0f) *
+                          (result[i].trgba[0] + result[i].trgba[1] + result[i].trgba[2]);
         }
       }
       force[0] = (result[0].tin - result[1].tin) * strength;
@@ -943,12 +944,12 @@ static void do_texture_effector(EffectorCache *eff,
     else { /*PFIELD_TEX_CURL*/
       float dbdy, dgdz, drdz, dbdx, dgdx, drdy;
 
-      dbdy = result[2].tb - result[0].tb;
-      dgdz = result[3].tg - result[0].tg;
-      drdz = result[3].tr - result[0].tr;
-      dbdx = result[1].tb - result[0].tb;
-      dgdx = result[1].tg - result[0].tg;
-      drdy = result[2].tr - result[0].tr;
+      dbdy = result[2].trgba[2] - result[0].trgba[2];
+      dgdz = result[3].trgba[1] - result[0].trgba[1];
+      drdz = result[3].trgba[0] - result[0].trgba[0];
+      dbdx = result[1].trgba[2] - result[0].trgba[2];
+      dgdx = result[1].trgba[1] - result[0].trgba[1];
+      drdy = result[2].trgba[0] - result[0].trgba[0];
 
       force[0] = (dbdy - dgdz) * strength;
       force[1] = (drdz - dbdx) * strength;

source/blender/blenkernel/intern/hair.cc

@@ -31,7 +31,7 @@
 #include "BLI_listbase.h"
 #include "BLI_math_base.h"
 #include "BLI_math_vec_types.hh"
-#include "BLI_rand.h"
+#include "BLI_rand.hh"
 #include "BLI_string.h"
 #include "BLI_utildefines.h"
 
@@ -54,6 +54,7 @@
 #include "BLO_read_write.h"
 
 using blender::float3;
+using blender::RandomNumberGenerator;
 
 static const char *HAIR_ATTR_POSITION = "position";
 static const char *HAIR_ATTR_RADIUS = "radius";
@@ -220,38 +221,32 @@ static void hair_random(Hair *hair)
   CustomData_realloc(&hair->cdata, hair->totcurve);
   BKE_hair_update_customdata_pointers(hair);
 
-  RNG *rng = BLI_rng_new(0);
+  RandomNumberGenerator rng;
 
   for (int i = 0; i < hair->totcurve; i++) {
     HairCurve *curve = &hair->curves[i];
     curve->firstpoint = i * numpoints;
     curve->numpoints = numpoints;
 
-    float theta = 2.0f * M_PI * BLI_rng_get_float(rng);
-    float phi = saacosf(2.0f * BLI_rng_get_float(rng) - 1.0f);
+    const float theta = 2.0f * M_PI * rng.get_float();
+    const float phi = saacosf(2.0f * rng.get_float() - 1.0f);
 
-    float no[3] = {sinf(theta) * sinf(phi), cosf(theta) * sinf(phi), cosf(phi)};
-    normalize_v3(no);
+    float3 no = {std::sin(theta) * std::sin(phi), std::cos(theta) * std::sin(phi), std::cos(phi)};
+    blender::math::normalize(no);
 
-    float co[3];
-    copy_v3_v3(co, no);
-
-    float(*curve_co)[3] = hair->co + curve->firstpoint;
-    float *curve_radius = hair->radius + curve->firstpoint;
+    float(*curve_positions)[3] = hair->co + curve->firstpoint;
+    float *curve_radii = hair->radius + curve->firstpoint;
+    float3 co = no;
     for (int key = 0; key < numpoints; key++) {
       float t = key / (float)(numpoints - 1);
-      copy_v3_v3(curve_co[key], co);
-      curve_radius[key] = 0.02f * (1.0f - t);
+      copy_v3_v3(curve_positions[key], co);
+      curve_radii[key] = 0.02f * (1.0f - t);
 
-      float offset[3] = {2.0f * BLI_rng_get_float(rng) - 1.0f,
-                         2.0f * BLI_rng_get_float(rng) - 1.0f,
-                         2.0f * BLI_rng_get_float(rng) - 1.0f};
-      add_v3_v3(offset, no);
-      madd_v3_v3fl(co, offset, 1.0f / numpoints);
+      float3 offset = float3(rng.get_float(), rng.get_float(), rng.get_float()) * 2.0f - 1.0f;
+
+      co += (offset + no) / numpoints;
     }
   }
-
-  BLI_rng_free(rng);
 }
 
 void *BKE_hair_add(Main *bmain, const char *name)
@@ -380,22 +375,6 @@ static Hair *hair_evaluate_modifiers(struct Depsgraph *depsgraph,
       /* Created deformed coordinates array on demand. */
       mti->deformVerts(md, &mectx, nullptr, hair->co, hair->totpoint);
     }
-    else if (mti->modifyHair) {
-      /* Ensure we are not modifying the input. */
-      if (hair == hair_input) {
-        hair = BKE_hair_copy_for_eval(hair, true);
-      }
-
-      Hair *hair_next = mti->modifyHair(md, &mectx, hair);
-
-      if (hair_next && hair_next != hair) {
-        /* If the modifier returned a new hair, release the old one. */
-        if (hair != hair_input) {
-          BKE_id_free(nullptr, hair);
-        }
-        hair = hair_next;
-      }
-    }
   }
 
   return hair;