continue with derived node tree

rename
DerivedNodeTree data structure
2020-05-05 15:13:02 +02:00 · 2020-05-05 14:35:27 +02:00 · 2020-05-05 14:31:43 +02:00 · 2020-05-04 17:41:29 +02:00 · 2020-05-04 17:35:26 +02:00 · 2020-05-04 17:15:20 +02:00
1093 changed files with 16611 additions and 16446 deletions
--- a/.clang-format
+++ b/.clang-format
@@ -256,7 +256,6 @@ ForEachMacros:
  - SEQ_BEGIN
  - SURFACE_QUAD_ITER_BEGIN
  - foreach
-  - ED_screen_areas_iter

 # Use once we bump the minimum version to version 8.
 # # Without this string literals that in-line 'STRINGIFY' behave strangely (a bug?).
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -281,7 +281,7 @@ option(WITH_ALEMBIC             "Enable Alembic Support" ON)
 option(WITH_ALEMBIC_HDF5        "Enable Legacy Alembic Support (not officially supported)" OFF)

 # Universal Scene Description support
-option(WITH_USD                 "Enable Universal Scene Description (USD) Support" ON)
+option(WITH_USD                 "Enable Universal Scene Description (USD) Support" OFF)

 # 3D format support
 # Disable opencollada when we don't have precompiled libs
@@ -628,6 +628,12 @@ if(WITH_PYTHON_MODULE AND WITH_PYTHON_INSTALL)
  message(FATAL_ERROR "WITH_PYTHON_MODULE requires WITH_PYTHON_INSTALL to be OFF")
 endif()

+
+# may as well build python module without a UI
+if(WITH_PYTHON_MODULE)
+  set(WITH_HEADLESS ON)
+endif()
+
 if(NOT WITH_PYTHON)
  set(WITH_CYCLES OFF)
  set(WITH_DRACO OFF)
@@ -1733,7 +1739,6 @@ if(FIRST_RUN)
  info_cfg_option(WITH_ALEMBIC)
  info_cfg_option(WITH_QUADRIFLOW)
  info_cfg_option(WITH_USD)
-  info_cfg_option(WITH_TBB)

  info_cfg_text("Compiler Options:")
  info_cfg_option(WITH_BUILDINFO)
--- a/build_files/build_environment/CMakeLists.txt
+++ b/build_files/build_environment/CMakeLists.txt
@@ -130,6 +130,7 @@ if(NOT WIN32 OR ENABLE_MINGW64)
    include(cmake/vpx.cmake)
    include(cmake/x264.cmake)
    include(cmake/xvidcore.cmake)
+    include(cmake/faad.cmake)
    include(cmake/ffmpeg.cmake)
    include(cmake/fftw.cmake)
    include(cmake/sndfile.cmake)
--- a/build_files/build_environment/cmake/faad.cmake
+++ b/build_files/build_environment/cmake/faad.cmake
@@ -0,0 +1,40 @@
+# ***** BEGIN GPL LICENSE BLOCK *****
+#
+# This program is free software; you can redistribute it and/or
+# modify it under the terms of the GNU General Public License
+# as published by the Free Software Foundation; either version 2
+# of the License, or (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation,
+# Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+#
+# ***** END GPL LICENSE BLOCK *****
+
+set(FAAD_EXTRA_ARGS)
+
+if(WIN32)
+  set(FAAD_EXTRA_CONFIGURE "utils\\win32\\ac2ver.exe" "faad2" "configure.ac" > libfaad\\win32_ver.h)
+else()
+  set(FAAD_EXTRA_CONFIGURE echo .)
+endif()
+
+ExternalProject_Add(external_faad
+  URL ${FAAD_URI}
+  DOWNLOAD_DIR ${DOWNLOAD_DIR}
+  URL_HASH MD5=${FAAD_HASH}
+  PREFIX ${BUILD_DIR}/faad
+  CONFIGURE_COMMAND ${CONFIGURE_ENV} && cd ${BUILD_DIR}/faad/src/external_faad/ && ${FAAD_EXTRA_CONFIGURE} && ${CONFIGURE_COMMAND} --disable-shared --enable-static --prefix=${LIBDIR}/faad
+  BUILD_COMMAND ${CONFIGURE_ENV} && cd ${BUILD_DIR}/faad/src/external_faad/ && make -j${MAKE_THREADS}
+  INSTALL_COMMAND ${CONFIGURE_ENV} && cd ${BUILD_DIR}/faad/src/external_faad/ && make install
+  INSTALL_DIR ${LIBDIR}/faad
+)
+
+if(MSVC)
+  set_target_properties(external_faad PROPERTIES FOLDER Mingw)
+endif()
--- a/build_files/build_environment/cmake/ffmpeg.cmake
+++ b/build_files/build_environment/cmake/ffmpeg.cmake
@@ -127,6 +127,7 @@ endif()
 add_dependencies(
  external_ffmpeg
  external_zlib
+  external_faad
  external_openjpeg
  external_xvidcore
  external_x264
--- a/build_files/build_environment/cmake/fftw.cmake
+++ b/build_files/build_environment/cmake/fftw.cmake
@@ -19,12 +19,8 @@
 set(FFTW_EXTRA_ARGS)

 if(WIN32)
+  set(FFTW3_ENV set CFLAGS=-fno-stack-check -fno-stack-protector -mno-stack-arg-probe -fno-lto &&)
  set(FFTW3_PATCH_COMMAND ${PATCH_CMD} --verbose -p 0 -N -d ${BUILD_DIR}/fftw3/src/external_fftw3 < ${PATCH_DIR}/fftw3.diff)
-  set(FFTW_EXTRA_ARGS --disable-static --enable-shared)
-  set(FFTW_INSTALL install-strip)
-else()
-  set(FFTW_EXTRA_ARGS --enable-static)
-  set(FFTW_INSTALL install)
 endif()

 ExternalProject_Add(external_fftw3
@@ -32,10 +28,10 @@ ExternalProject_Add(external_fftw3
  DOWNLOAD_DIR ${DOWNLOAD_DIR}
  URL_HASH MD5=${FFTW_HASH}
  PREFIX ${BUILD_DIR}/fftw3
-  CONFIGURE_COMMAND ${CONFIGURE_ENV} && cd ${BUILD_DIR}/fftw3/src/external_fftw3/ && ${CONFIGURE_COMMAND} ${FFTW_EXTRA_ARGS} --prefix=${mingw_LIBDIR}/fftw3
+  CONFIGURE_COMMAND ${CONFIGURE_ENV} && ${FFTW3_ENV} cd ${BUILD_DIR}/fftw3/src/external_fftw3/ && ${CONFIGURE_COMMAND} --enable-static --prefix=${mingw_LIBDIR}/fftw3
  PATCH_COMMAND ${FFTW3_PATCH_COMMAND}
  BUILD_COMMAND ${CONFIGURE_ENV} && cd ${BUILD_DIR}/fftw3/src/external_fftw3/ && make -j${MAKE_THREADS}
-  INSTALL_COMMAND ${CONFIGURE_ENV} && cd ${BUILD_DIR}/fftw3/src/external_fftw3/ && make ${FFTW_INSTALL}
+  INSTALL_COMMAND ${CONFIGURE_ENV} && cd ${BUILD_DIR}/fftw3/src/external_fftw3/ && make install
  INSTALL_DIR ${LIBDIR}/fftw3
 )

@@ -43,8 +39,7 @@ if(MSVC)
  set_target_properties(external_fftw3 PROPERTIES FOLDER Mingw)
  if(BUILD_MODE STREQUAL Release)
    ExternalProject_Add_Step(external_fftw3 after_install
-      COMMAND ${CMAKE_COMMAND} -E copy ${LIBDIR}/fftw3/lib/libfftw3.dll.a ${HARVEST_TARGET}/fftw3/lib/libfftw.lib
-      COMMAND ${CMAKE_COMMAND} -E copy ${LIBDIR}/fftw3/bin/libfftw3-3.dll ${HARVEST_TARGET}/fftw3/lib/libfftw3-3.dll
+      COMMAND ${CMAKE_COMMAND} -E copy ${LIBDIR}/fftw3/lib/libfftw3.a ${HARVEST_TARGET}/fftw3/lib/libfftw.lib
      COMMAND ${CMAKE_COMMAND} -E copy ${LIBDIR}/fftw3/include/fftw3.h ${HARVEST_TARGET}/fftw3/include/fftw3.h
      DEPENDEES install
    )
--- a/build_files/build_environment/cmake/harvest.cmake
+++ b/build_files/build_environment/cmake/harvest.cmake
@@ -44,6 +44,10 @@ if(BUILD_MODE STREQUAL Release)
        # glew-> opengl
        ${CMAKE_COMMAND} -E copy ${LIBDIR}/glew/lib/libglew32.lib ${HARVEST_TARGET}/opengl/lib/glew.lib &&
        ${CMAKE_COMMAND} -E copy_directory ${LIBDIR}/glew/include/ ${HARVEST_TARGET}/opengl/include/ &&
+        # sndfile
+        ${CMAKE_COMMAND} -E copy ${LIBDIR}/sndfile/lib/libsndfile.dll.a ${HARVEST_TARGET}/sndfile/lib/libsndfile-1.lib &&
+        ${CMAKE_COMMAND} -E copy ${LIBDIR}/sndfile/bin/libsndfile-1.dll ${HARVEST_TARGET}/sndfile/lib/libsndfile-1.dll &&
+        ${CMAKE_COMMAND} -E copy ${LIBDIR}/sndfile/include/sndfile.h ${HARVEST_TARGET}/sndfile/include/sndfile.h &&
        # tiff
        ${CMAKE_COMMAND} -E copy ${LIBDIR}/tiff/lib/tiff.lib ${HARVEST_TARGET}/tiff/lib/libtiff.lib &&
        ${CMAKE_COMMAND} -E copy_directory ${LIBDIR}/tiff/include/ ${HARVEST_TARGET}/tiff/include/ &&
--- a/build_files/build_environment/cmake/jpeg.cmake
+++ b/build_files/build_environment/cmake/jpeg.cmake
@@ -18,7 +18,7 @@

 if(WIN32)
  # cmake for windows
-  set(JPEG_EXTRA_ARGS -DNASM=${NASM_PATH} -DWITH_JPEG8=ON  -DCMAKE_DEBUG_POSTFIX=d -DWITH_CRT_DLL=On)
+  set(JPEG_EXTRA_ARGS -DNASM=${NASM_PATH} -DWITH_JPEG8=ON  -DCMAKE_DEBUG_POSTFIX=d)

  ExternalProject_Add(external_jpeg
    URL ${JPEG_URI}
--- a/build_files/build_environment/cmake/openal.cmake
+++ b/build_files/build_environment/cmake/openal.cmake
@@ -52,6 +52,7 @@ if(BUILD_MODE STREQUAL Release)
    PREFIX ${BUILD_DIR}/openal
    CMAKE_ARGS -DCMAKE_INSTALL_PREFIX=${LIBDIR}/openal ${DEFAULT_CMAKE_FLAGS} ${OPENAL_EXTRA_ARGS}
    INSTALL_DIR ${LIBDIR}/openal
+    PATCH_COMMAND ${PATCH_CMD} -p 1 -d ${BUILD_DIR}/openal/src/external_openal < ${PATCH_DIR}/openal.diff
  )

  if(WIN32)
--- a/build_files/build_environment/cmake/opensubdiv.cmake
+++ b/build_files/build_environment/cmake/opensubdiv.cmake
@@ -67,7 +67,7 @@ endif()
 ExternalProject_Add(external_opensubdiv
  URL ${OPENSUBDIV_URI}
  DOWNLOAD_DIR ${DOWNLOAD_DIR}
-  URL_HASH MD5=${OPENSUBDIV_HASH}
+  URL_HASH MD5=${OPENSUBDIV_Hash}
  PREFIX ${BUILD_DIR}/opensubdiv
  CMAKE_ARGS -DCMAKE_INSTALL_PREFIX=${LIBDIR}/opensubdiv -Wno-dev ${DEFAULT_CMAKE_FLAGS} ${OPENSUBDIV_EXTRA_ARGS}
  INSTALL_DIR ${LIBDIR}/opensubdiv
--- a/build_files/build_environment/cmake/pthreads.cmake
+++ b/build_files/build_environment/cmake/pthreads.cmake
@@ -24,7 +24,7 @@ if(WIN32)
      set(PTHREAD_CPPFLAGS "/I. /DHAVE_CONFIG_H ")
    endif()

-    set(PTHREADS_BUILD cd ${BUILD_DIR}/pthreads/src/external_pthreads/ && cd && nmake VC-static /e CPPFLAGS=${PTHREAD_CPPFLAGS})
+    set(PTHREADS_BUILD cd ${BUILD_DIR}/pthreads/src/external_pthreads/ && cd && nmake VC-static /e CPPFLAGS=${PTHREAD_CPPFLAGS} /e XLIBS=/NODEFAULTLIB:msvcr)

    ExternalProject_Add(external_pthreads
      URL ${PTHREADS_URI}
@@ -32,7 +32,6 @@ if(WIN32)
      URL_HASH MD5=${PTHREADS_HASH}
      PREFIX ${BUILD_DIR}/pthreads
      CONFIGURE_COMMAND echo .
-      PATCH_COMMAND COMMAND ${PATCH_CMD} -p 1 -d ${BUILD_DIR}/pthreads/src/external_pthreads < ${PATCH_DIR}/pthreads.diff
      BUILD_COMMAND ${PTHREADS_BUILD}
      INSTALL_COMMAND COMMAND
        ${CMAKE_COMMAND} -E copy ${BUILD_DIR}/pthreads/src/external_pthreads/libpthreadVC3${LIBEXT} ${LIBDIR}/pthreads/lib/pthreadVC3${LIBEXT} &&
--- a/build_files/build_environment/cmake/sndfile.cmake
+++ b/build_files/build_environment/cmake/sndfile.cmake
@@ -60,14 +60,3 @@ if(UNIX)
    external_flac
  )
 endif()
-
-if(BUILD_MODE STREQUAL Release AND WIN32)
-  ExternalProject_Add_Step(external_sndfile after_install
-      COMMAND  lib /def:${BUILD_DIR}/sndfile/src/external_sndfile/src/libsndfile-1.def /machine:x64 /out:${BUILD_DIR}/sndfile/src/external_sndfile/src/libsndfile-1.lib
-      COMMAND  ${CMAKE_COMMAND} -E copy ${LIBDIR}/sndfile/bin/libsndfile-1.dll ${HARVEST_TARGET}/sndfile/lib/libsndfile-1.dll
-      COMMAND  ${CMAKE_COMMAND} -E copy ${BUILD_DIR}/sndfile/src/external_sndfile/src/libsndfile-1.lib ${HARVEST_TARGET}/sndfile/lib/libsndfile-1.lib
-      COMMAND  ${CMAKE_COMMAND} -E copy ${LIBDIR}/sndfile/include/sndfile.h ${HARVEST_TARGET}/sndfile/include/sndfile.h
-
-    DEPENDEES install
-  )
-endif()
--- a/build_files/build_environment/cmake/versions.cmake
+++ b/build_files/build_environment/cmake/versions.cmake
@@ -20,9 +20,9 @@ set(ZLIB_VERSION 1.2.11)
 set(ZLIB_URI https://zlib.net/zlib-${ZLIB_VERSION}.tar.gz)
 set(ZLIB_HASH 1c9f62f0778697a09d36121ead88e08e)

-set(OPENAL_VERSION 1.20.1)
+set(OPENAL_VERSION 1.18.2)
 set(OPENAL_URI http://openal-soft.org/openal-releases/openal-soft-${OPENAL_VERSION}.tar.bz2)
-set(OPENAL_HASH 556695068ce8375b89986083d810fd35)
+set(OPENAL_HASH d4eeb0889812e2fdeaa1843523d76190)

 set(PNG_VERSION 1.6.35)
 set(PNG_URI http://prdownloads.sourceforge.net/libpng/libpng-${PNG_VERSION}.tar.xz)
@@ -66,9 +66,9 @@ else()
  set(OPENEXR_VERSION_POSTFIX)
 endif()

-set(FREETYPE_VERSION 2.10.2)
+set(FREETYPE_VERSION 2.10.1)
 set(FREETYPE_URI http://prdownloads.sourceforge.net/freetype/freetype-${FREETYPE_VERSION}.tar.gz)
-set(FREETYPE_HASH b1cb620e4c875cd4d1bfa04945400945)
+set(FREETYPE_HASH c50a3c9e5e62bdc938a6e1598a782947)

 set(GLEW_VERSION 1.13.0)
 set(GLEW_URI http://prdownloads.sourceforge.net/glew/glew/${GLEW_VERSION}/glew-${GLEW_VERSION}.tgz)
@@ -101,13 +101,13 @@ set(CUEW_GIT_UID 1744972026de9cf27c8a7dc39cf39cd83d5f922f)
 set(CUEW_URI https://github.com/CudaWrangler/cuew/archive/${CUEW_GIT_UID}.zip)
 set(CUEW_HASH 86760d62978ebfd96cd93f5aa1abaf4a)

-set(OPENSUBDIV_VERSION v3_4_3)
+set(OPENSUBDIV_VERSION v3_4_0_RC2)
+set(OPENSUBDIV_Hash f6a10ba9efaa82fde86fe65aad346319)
 set(OPENSUBDIV_URI https://github.com/PixarAnimationStudios/OpenSubdiv/archive/${OPENSUBDIV_VERSION}.tar.gz)
-set(OPENSUBDIV_HASH 7bbfa275d021fb829e521df749160edb)

-set(SDL_VERSION 2.0.12)
+set(SDL_VERSION 2.0.8)
 set(SDL_URI https://www.libsdl.org/release/SDL2-${SDL_VERSION}.tar.gz)
-set(SDL_HASH 783b6f2df8ff02b19bb5ce492b99c8ff)
+set(SDL_HASH 3800d705cef742c6a634f202c37f263f)

 set(OPENCOLLADA_VERSION v1.6.68)
 set(OPENCOLLADA_URI https://github.com/KhronosGroup/OpenCOLLADA/archive/${OPENCOLLADA_VERSION}.tar.gz)
@@ -168,9 +168,9 @@ set(LAME_VERSION 3.100)
 set(LAME_URI http://downloads.sourceforge.net/project/lame/lame/3.100/lame-${LAME_VERSION}.tar.gz)
 set(LAME_HASH 83e260acbe4389b54fe08e0bdbf7cddb)

-set(OGG_VERSION 1.3.4)
+set(OGG_VERSION 1.3.3)
 set(OGG_URI http://downloads.xiph.org/releases/ogg/libogg-${OGG_VERSION}.tar.gz)
-set(OGG_HASH fe5670640bd49e828d64d2879c31cb4dde9758681bb664f9bdbf159a01b0c76e)
+set(OGG_HASH c2e8a485110b97550f453226ec644ebac6cb29d1caef2902c007edab4308d985)

 set(VORBIS_VERSION 1.3.6)
 set(VORBIS_URI http://downloads.xiph.org/releases/vorbis/libvorbis-${VORBIS_VERSION}.tar.gz)
@@ -180,24 +180,24 @@ set(THEORA_VERSION 1.1.1)
 set(THEORA_URI http://downloads.xiph.org/releases/theora/libtheora-${THEORA_VERSION}.tar.bz2)
 set(THEORA_HASH b6ae1ee2fa3d42ac489287d3ec34c5885730b1296f0801ae577a35193d3affbc)

-set(FLAC_VERSION 1.3.3)
+set(FLAC_VERSION 1.3.2)
 set(FLAC_URI http://downloads.xiph.org/releases/flac/flac-${FLAC_VERSION}.tar.xz)
-set(FLAC_HASH 213e82bd716c9de6db2f98bcadbc4c24c7e2efe8c75939a1a84e28539c4e1748)
+set(FLAC_HASH 91cfc3ed61dc40f47f050a109b08610667d73477af6ef36dcad31c31a4a8d53f)

-set(VPX_VERSION 1.8.2)
+set(VPX_VERSION 1.7.0)
 set(VPX_URI https://github.com/webmproject/libvpx/archive/v${VPX_VERSION}/libvpx-v${VPX_VERSION}.tar.gz)
-set(VPX_HASH 8735d9fcd1a781ae6917f28f239a8aa358ce4864ba113ea18af4bb2dc8b474ac)
+set(VPX_HASH 1fec931eb5c94279ad219a5b6e0202358e94a93a90cfb1603578c326abfc1238)

 set(OPUS_VERSION 1.3.1)
 set(OPUS_URI https://archive.mozilla.org/pub/opus/opus-${OPUS_VERSION}.tar.gz)
 set(OPUS_HASH 65b58e1e25b2a114157014736a3d9dfeaad8d41be1c8179866f144a2fb44ff9d)

-set(X264_URI https://code.videolan.org/videolan/x264/-/archive/master/x264-33f9e1474613f59392be5ab6a7e7abf60fa63622.tar.gz)
-set(X264_HASH 300dfb5b6c35722516f168868ce9419252a9e9eb77a05d82c9cede925b691bd6)
+set(X264_URI http://download.videolan.org/pub/videolan/x264/snapshots/x264-snapshot-20180811-2245-stable.tar.bz2)
+set(X264_HASH ae8a868a0e236a348b35d79f3ee80294b169d1195408b689f9851383661ed7aa)

-set(XVIDCORE_VERSION 1.3.7)
-set(XVIDCORE_URI https://downloads.xvid.com/downloads/xvidcore-${XVIDCORE_VERSION}.tar.gz)
-set(XVIDCORE_HASH abbdcbd39555691dd1c9b4d08f0a031376a3b211652c0d8b3b8aa9be1303ce2d)
+set(XVIDCORE_VERSION 1.3.5)
+set(XVIDCORE_URI http://downloads.xvid.org/downloads/xvidcore-${XVIDCORE_VERSION}.tar.gz)
+set(XVIDCORE_HASH 165ba6a2a447a8375f7b06db5a3c91810181f2898166e7c8137401d7fc894cf0)

 # This has to be in sync with the version in blenders /extern folder.
 set(OPENJPEG_VERSION 2.3.0)
@@ -206,9 +206,13 @@ set(OPENJPEG_SHORT_VERSION 2.3)
 set(OPENJPEG_URI https://github.com/uclouvain/openjpeg/archive/66297f07a43.zip)
 set(OPENJPEG_HASH 8242b18d908c7c42174e4231a741cfa7ce7c26b6ed5c9644feb9df7b3054310b)

-set(FFMPEG_VERSION 4.2.3)
+set(FAAD_VERSION 2-2.8.8)
+set(FAAD_URI http://downloads.sourceforge.net/faac/faad${FAAD_VERSION}.tar.gz)
+set(FAAD_HASH 28f6116efdbe9378269f8a6221767d1f)
+
+set(FFMPEG_VERSION 4.0.2)
 set(FFMPEG_URI http://ffmpeg.org/releases/ffmpeg-${FFMPEG_VERSION}.tar.bz2)
-set(FFMPEG_HASH 695fad11f3baf27784e24cb0e977b65a)
+set(FFMPEG_HASH 5576e8a22f80b6a336db39808f427cfb)

 set(FFTW_VERSION 3.3.8)
 set(FFTW_URI http://www.fftw.org/fftw-${FFTW_VERSION}.tar.gz)
@@ -315,6 +319,6 @@ set(MESA_VERSION 18.3.1)
 set(MESA_URI ftp://ftp.freedesktop.org/pub/mesa//mesa-${MESA_VERSION}.tar.xz)
 set(MESA_HASH d60828056d77bfdbae0970f9b15fb1be)

-set(XR_OPENXR_SDK_VERSION 1.0.8)
+set(XR_OPENXR_SDK_VERSION 1.0.6)
 set(XR_OPENXR_SDK_URI https://github.com/KhronosGroup/OpenXR-SDK/archive/release-${XR_OPENXR_SDK_VERSION}.tar.gz)
-set(XR_OPENXR_SDK_HASH c6de63d2e0f9029aa58dfa97cad8ce07)
+set(XR_OPENXR_SDK_HASH 21daea7c3bfec365298d779a0e19caa1)
--- a/build_files/build_environment/cmake/x264.cmake
+++ b/build_files/build_environment/cmake/x264.cmake
@@ -18,6 +18,9 @@

 if(WIN32)
  set(X264_EXTRA_ARGS --enable-win32thread --cross-prefix=${MINGW_HOST}- --host=${MINGW_HOST})
+  set(X264_PATCH_CMD ${PATCH_CMD} --verbose -p 1 -N -d ${BUILD_DIR}/x264/src/external_x264 < ${PATCH_DIR}/x264.diff)
+else()
+  set(X264_PATCH_CMD echo .)
 endif()


@@ -26,6 +29,7 @@ ExternalProject_Add(external_x264
  DOWNLOAD_DIR ${DOWNLOAD_DIR}
  URL_HASH SHA256=${X264_HASH}
  PREFIX ${BUILD_DIR}/x264
+  PATCH_COMMAND ${X264_PATCH_CMD}
  CONFIGURE_COMMAND ${CONFIGURE_ENV} && cd ${BUILD_DIR}/x264/src/external_x264/ && ${CONFIGURE_COMMAND} --prefix=${LIBDIR}/x264
    --enable-static
    --enable-pic
--- a/build_files/build_environment/install_deps.sh
+++ b/build_files/build_environment/install_deps.sh
@@ -53,15 +53,15 @@ getopt \
 --long source:,install:,tmp:,info:,threads:,help,show-deps,no-sudo,no-build,no-confirm,\
 with-all,with-opencollada,with-jack,with-embree,with-oidn,\
 ver-ocio:,ver-oiio:,ver-llvm:,ver-osl:,ver-osd:,ver-openvdb:,ver-xr-openxr:,\
-force-all,force-python,force-numpy,force-boost,force-tbb,\
+force-all,force-python,force-numpy,force-boost,\
 force-ocio,force-openexr,force-oiio,force-llvm,force-osl,force-osd,force-openvdb,\
 force-ffmpeg,force-opencollada,force-alembic,force-embree,force-oidn,force-usd,\
 force-xr-openxr,\
-build-all,build-python,build-numpy,build-boost,build-tbb,\
+build-all,build-python,build-numpy,build-boost,\
 build-ocio,build-openexr,build-oiio,build-llvm,build-osl,build-osd,build-openvdb,\
 build-ffmpeg,build-opencollada,build-alembic,build-embree,build-oidn,build-usd,\
 build-xr-openxr,\
-skip-python,skip-numpy,skip-boost,skip-tbb,\
+skip-python,skip-numpy,skip-boost,\
 skip-ocio,skip-openexr,skip-oiio,skip-llvm,skip-osl,skip-osd,skip-openvdb,\
 skip-ffmpeg,skip-opencollada,skip-alembic,skip-embree,skip-oidn,skip-usd,\
 skip-xr-openxr \
@@ -191,9 +191,6 @@ ARGUMENTS_INFO="\"COMMAND LINE ARGUMENTS:
    --build-boost
        Force the build of Boost.

-    --build-tbb
-        Force the build of TBB.
-
    --build-ocio
        Force the build of OpenColorIO.

@@ -258,9 +255,6 @@ ARGUMENTS_INFO="\"COMMAND LINE ARGUMENTS:
    --force-boost
        Force the rebuild of Boost.

-    --force-tbb
-        Force the rebuild of TBB.
-
    --force-ocio
        Force the rebuild of OpenColorIO.

@@ -318,9 +312,6 @@ ARGUMENTS_INFO="\"COMMAND LINE ARGUMENTS:
    --skip-boost
        Unconditionally skip Boost installation/building.

-    --skip-tbb
-        Unconditionally skip TBB installation/building.
-
    --skip-ocio
        Unconditionally skip OpenColorIO installation/building.

@@ -394,13 +385,6 @@ BOOST_FORCE_BUILD=false
 BOOST_FORCE_REBUILD=false
 BOOST_SKIP=false

-TBB_VERSION="2019"
-TBB_VERSION_UPDATE="_U9"  # Used for source packages...
-TBB_VERSION_MIN="2018"
-TBB_FORCE_BUILD=false
-TBB_FORCE_REBUILD=false
-TBB_SKIP=false
-
 OCIO_VERSION="1.1.0"
 OCIO_VERSION_MIN="1.0"
 OCIO_FORCE_BUILD=false
@@ -669,7 +653,6 @@ while true; do
      PYTHON_FORCE_BUILD=true
      NUMPY_FORCE_BUILD=true
      BOOST_FORCE_BUILD=true
-      TBB_FORCE_BUILD=true
      OCIO_FORCE_BUILD=true
      OPENEXR_FORCE_BUILD=true
      OIIO_FORCE_BUILD=true
@@ -699,9 +682,6 @@ while true; do
    --build-boost)
      BOOST_FORCE_BUILD=true; shift; continue
    ;;
-    --build-tbb)
-      TBB_FORCE_BUILD=true; shift; continue
-    ;;
    --build-ocio)
      OCIO_FORCE_BUILD=true; shift; continue
    ;;
@@ -748,7 +728,6 @@ while true; do
      PYTHON_FORCE_REBUILD=true
      NUMPY_FORCE_REBUILD=true
      BOOST_FORCE_REBUILD=true
-      TBB_FORCE_REBUILD=true
      OCIO_FORCE_REBUILD=true
      OPENEXR_FORCE_REBUILD=true
      OIIO_FORCE_REBUILD=true
@@ -776,9 +755,6 @@ while true; do
    --force-boost)
      BOOST_FORCE_REBUILD=true; shift; continue
    ;;
-    --force-tbb)
-      TBB_FORCE_REBUILD=true; shift; continue
-    ;;
    --force-ocio)
      OCIO_FORCE_REBUILD=true; shift; continue
    ;;
@@ -830,9 +806,6 @@ while true; do
    --skip-boost)
      BOOST_SKIP=true; shift; continue
    ;;
-    --skip-tbb)
-      TBB_SKIP=true; shift; continue
-    ;;
    --skip-ocio)
      OCIO_SKIP=true; shift; continue
    ;;
@@ -925,12 +898,9 @@ PYTHON_SOURCE=( "https://www.python.org/ftp/python/$PYTHON_VERSION/Python-$PYTHO
 NUMPY_SOURCE=( "https://github.com/numpy/numpy/releases/download/v$NUMPY_VERSION/numpy-$NUMPY_VERSION.tar.gz" )

 _boost_version_nodots=`echo "$BOOST_VERSION" | sed -r 's/\./_/g'`
-BOOST_SOURCE=( "https://dl.bintray.com/boostorg/release/$BOOST_VERSION/source/boost_$_boost_version_nodots.tar.bz2" )
+BOOST_SOURCE=( "http://sourceforge.net/projects/boost/files/boost/$BOOST_VERSION/boost_$_boost_version_nodots.tar.bz2/download" )
 BOOST_BUILD_MODULES="--with-system --with-filesystem --with-thread --with-regex --with-locale --with-date_time --with-wave --with-iostreams --with-python --with-program_options"

-TBB_SOURCE=( "https://github.com/oneapi-src/oneTBB/archive/$TBB_VERSION$TBB_VERSION_UPDATE.tar.gz" )
-TBB_SOURCE_CMAKE=( "https://raw.githubusercontent.com/wjakob/tbb/master/CMakeLists.txt" )
-
 OCIO_USE_REPO=false
 OCIO_SOURCE=( "https://github.com/AcademySoftwareFoundation/OpenColorIO/archive/v$OCIO_VERSION.tar.gz")
 #~ OCIO_SOURCE_REPO=( "https://github.com/imageworks/OpenColorIO.git" )
@@ -1041,7 +1011,6 @@ You may also want to build them yourself (optional ones are [between brackets]):
    * Python $PYTHON_VERSION_MIN (from $PYTHON_SOURCE).
    * [NumPy $NUMPY_VERSION_MIN] (from $NUMPY_SOURCE).
    * Boost $BOOST_VERSION_MIN (from $BOOST_SOURCE, modules: $BOOST_BUILD_MODULES).
-    * TBB $TBB_VERSION_MIN (from $TBB_SOURCE).
    * [FFMpeg $FFMPEG_VERSION_MIN (needs libvorbis, libogg, libtheora, libx264, libmp3lame, libxvidcore, libvpx, ...)] (from $FFMPEG_SOURCE).
    * [OpenColorIO $OCIO_VERSION_MIN] (from $OCIO_SOURCE).
    * ILMBase $OPENEXR_VERSION_MIN (from $OPENEXR_SOURCE).
@@ -1279,10 +1248,8 @@ _update_deps_python() {
 clean_Python() {
  clean_Numpy
  _init_python
-  if [ -d $_inst ]; then
-    _update_deps_python
-  fi
  _clean
+  _update_deps_python
 }

 compile_Python() {
@@ -1365,10 +1332,8 @@ _update_deps_numpy() {

 clean_Numpy() {
  _init_numpy
-  if [ -d $_inst ]; then
-    _update_deps_numpy
-  fi
  _clean
+  _update_deps_numpy
 }

 compile_Numpy() {
@@ -1453,10 +1418,8 @@ _update_deps_boost() {

 clean_Boost() {
  _init_boost
-  if [ -d $_inst ]; then
-    _update_deps_boost
-  fi
  _clean
+  _update_deps_boost
 }

 compile_Boost() {
@@ -1521,128 +1484,6 @@ compile_Boost() {
  run_ldconfig "boost"
 }

-# ----------------------------------------------------------------------------
-# Build TBB
-
-_init_tbb() {
-  _src=$SRC/TBB-$TBB_VERSION
-  _git=false
-  _inst=$INST/tbb-$TBB_VERSION
-  _inst_shortcut=$INST/tbb
-}
-
-_update_deps_tbb() {
-  OSD_FORCE_REBUILD=true
-  OPENVDB_FORCE_REBUILD=true
-  USD_FORCE_REBUILD=true
-  OIDN_FORCE_REBUILD=true
-  if [ "$_is_building" = true ]; then
-    OSD_FORCE_BUILD=true
-    OPENVDB_FORCE_BUILD=true
-    USD_FORCE_BUILD=true
-    OIDN_FORCE_BUILD=true
-  fi
-}
-
-clean_TBB() {
-  _init_tbb
-  if [ -d $_inst ]; then
-    _update_deps_tbb
-  fi
-  _clean
-}
-
-compile_TBB() {
-  if [ "$NO_BUILD" = true ]; then
-    WARNING "--no-build enabled, TBB will not be compiled!"
-    return
-  fi
-
-  # To be changed each time we make edits that would modify the compiled result!
-  tbb_magic=0
-  _init_tbb
-
-  # Clean install if needed!
-  magic_compile_check tbb-$TBB_VERSION $tbb_magic
-  if [ $? -eq 1 -o "$TBB_FORCE_REBUILD" = true ]; then
-    clean_TBB
-  fi
-
-  if [ ! -d $_inst ]; then
-    INFO "Building TBB-$TBB_VERSION$TBB_VERSION_UPDATE"
-    _is_building=true
-    
-    # Rebuild dependencies as well!
-    _update_deps_tbb
-
-    prepare_opt
-
-    if [ ! -d $_src ]; then
-      INFO "Downloading TBB-$TBB_VERSION$TBB_VERSION_UPDATE"
-      mkdir -p $SRC
-
-      download TBB_SOURCE[@] $_src.tar.gz
-      INFO "Unpacking TBB-$TBB_VERSION$TBB_VERSION_UPDATE"
-      tar -C $SRC --transform "s,(.*/?)oneTBB[^/]*(.*),\1TBB-$TBB_VERSION\2,x" \
-          -xf $_src.tar.gz
-
-      INFO
-
-      # Super-hack: Add some cmake builder to tbb... since they don't even have an install target by default, sic.
-      download TBB_SOURCE_CMAKE[@] $_src/CMakeLists.txt
-      cp $_src/build/vs2013/version_string.ver $_src/build/version_string.ver.in
-    fi
-
-    cd $_src
-
-    # Always refresh the whole build!
-    if [ -d cmake_build ]; then
-      rm -rf cmake_build
-    fi
-    mkdir cmake_build
-    cd cmake_build
-
-    cmake_d="-D CMAKE_BUILD_TYPE=Release"
-    cmake_d="$cmake_d -D CMAKE_PREFIX_PATH=$_inst"
-    cmake_d="$cmake_d -D CMAKE_INSTALL_PREFIX=$_inst"
-    cmake_d="$cmake_d -D TBB_BUILD_SHARED=ON"
-    cmake_d="$cmake_d -D TBB_BUILD_STATIC=OFF"
-    cmake_d="$cmake_d -D TBB_BUILD_TBBMALLOC=ON"
-    cmake_d="$cmake_d -D TBB_BUILD_TBBMALLOC_PROXY=OFF"
-    cmake_d="$cmake_d -D TBB_BUILD_TESTS=OFF"
-
-    if file /bin/cp | grep -q '32-bit'; then
-      cflags="-fPIC -m32 -march=i686"
-    else
-      cflags="-fPIC"
-    fi
-
-    cmake $cmake_d -D CMAKE_CXX_FLAGS="$cflags" -D CMAKE_EXE_LINKER_FLAGS="-lgcc_s -lgcc" ..
-
-    make -j$THREADS && make install
-
-    make clean
-
-    if [ -d $_inst ]; then
-      _create_inst_shortcut
-    else
-      ERROR "TBB-$TBB_VERSION$TBB_VERSION_UPDATE failed to compile, exiting"
-      exit 1
-    fi
-
-    magic_compile_set tbb-$TBB_VERSION $tbb_magic
-
-    cd $CWD
-    INFO "Done compiling TBB-$TBB_VERSION$TBB_VERSION_UPDATE!"
-    _is_building=false
-  else
-    INFO "Own TBB-$TBB_VERSION$TBB_VERSION_UPDATE is up to date, nothing to do!"
-    INFO "If you want to force rebuild of this lib, use the --force-tbb option."
-  fi
-
-  run_ldconfig "tbb"
-}
-
 # ----------------------------------------------------------------------------
 # Build OCIO

@@ -1663,10 +1504,8 @@ _update_deps_ocio() {

 clean_OCIO() {
  _init_ocio
-  if [ -d $_inst ]; then
-    _update_deps_ocio
-  fi
  _clean
+  _update_deps_ocio
 }

 compile_OCIO() {
@@ -1778,7 +1617,7 @@ compile_OCIO() {
 _init_openexr() {
  _src=$SRC/OpenEXR-$OPENEXR_VERSION
  _git=false
-  _inst=$INST/openexr-$OPENEXR_VERSION
+  _inst=$_openexr_inst
  _inst_shortcut=$INST/openexr
 }

@@ -1793,10 +1632,8 @@ _update_deps_openexr() {

 clean_OPENEXR() {
  _init_openexr
-  if [ -d $_inst ]; then
-    _update_deps_openexr
-  fi
  _clean
+  _update_deps_openexr
 }

 compile_OPENEXR() {
@@ -1814,6 +1651,7 @@ compile_OPENEXR() {
    clean_OPENEXR
  fi

+  _openexr_inst=$INST/openexr-$OPENEXR_VERSION
  PRINT ""
  _init_openexr

@@ -1859,7 +1697,7 @@ compile_OPENEXR() {
    mkdir build
    cd build

-    cmake_d="$cmake_d -D CMAKE_INSTALL_PREFIX=$_inst"
+    cmake_d="$cmake_d -D CMAKE_INSTALL_PREFIX=$_openexr_inst"
    cmake_d="$cmake_d -D CMAKE_INSTALL_DOCDIR=/dev/null"  # Hack, there is no option to disable that currently...
    cmake_d="$cmake_d -D BUILD_SHARED_LIBS=ON"
    cmake_d="$cmake_d -D BUILD_TESTING=OFF"
@@ -1921,10 +1759,8 @@ _update_deps_oiio() {

 clean_OIIO() {
  _init_oiio
-  if [ -d $_inst ]; then
-    _update_deps_oiio
-  fi
  _clean
+  _update_deps_oiio
 }

 compile_OIIO() {
@@ -2076,10 +1912,8 @@ _update_deps_llvm() {

 clean_LLVM() {
  _init_llvm
-  if [ -d $_inst ]; then
-    _update_deps_llvm
-  fi
  _clean
+  _update_deps_llvm
 }

 compile_LLVM() {
@@ -2185,10 +2019,8 @@ _update_deps_osl() {

 clean_OSL() {
  _init_osl
-  if [ -d $_inst ]; then
-    _update_deps_osl
-  fi
  _clean
+  _update_deps_osl
 }

 compile_OSL() {
@@ -2330,10 +2162,8 @@ _update_deps_osd() {

 clean_OSD() {
  _init_osd
-  if [ -d $_inst ]; then
-    _update_deps_osd
-  fi
  _clean
+  _update_deps_osd
 }

 compile_OSD() {
@@ -2392,9 +2222,6 @@ compile_OSD() {
    mkdir build
    cd build

-    if [ -d $INST/tbb ]; then
-      cmake_d="$cmake_d $cmake_d -D TBB_LOCATION=$INST/tbb"
-    fi
    cmake_d="-D CMAKE_BUILD_TYPE=Release"
    cmake_d="$cmake_d -D CMAKE_INSTALL_PREFIX=$_inst"
    # ptex is only needed when nicholas bishop is ready
@@ -2447,10 +2274,8 @@ _update_deps_blosc() {

 clean_BLOSC() {
  _init_blosc
-  if [ -d $_inst ]; then
-    _update_deps_blosc
-  fi
  _clean
+  _update_deps_blosc
 }

 compile_BLOSC() {
@@ -2544,10 +2369,8 @@ _update_deps_openvdb() {

 clean_OPENVDB() {
  _init_openvdb
-  if [ -d $_inst ]; then
-    _update_deps_openvdb
-  fi
  _clean
+  _update_deps_openvdb
 }

 compile_OPENVDB() {
@@ -2606,9 +2429,6 @@ compile_OPENVDB() {
    if [ -d $INST/boost ]; then
      make_d="$make_d BOOST_INCL_DIR=$INST/boost/include BOOST_LIB_DIR=$INST/boost/lib"
    fi
-    if [ -d $INST/tbb ]; then
-      make_d="$make_d TBB_ROOT=$INST/tbb TBB_USE_STATIC_LIBS=OFF"
-    fi

    if [ "$_with_built_openexr" = true ]; then
      make_d="$make_d ILMBASE_INCL_DIR=$INST/openexr/include ILMBASE_LIB_DIR=$INST/openexr/lib"
@@ -2662,10 +2482,8 @@ _update_deps_alembic() {

 clean_ALEMBIC() {
  _init_alembic
-  if [ -d $_inst ]; then
-    _update_deps_alembic
-  fi
  _clean
+  _update_deps_alembic
 }

 compile_ALEMBIC() {
@@ -2767,10 +2585,8 @@ _update_deps_usd() {

 clean_USD() {
  _init_usd
-  if [ -d $_inst ]; then
-    _update_deps_usd
-  fi
  _clean
+  _update_deps_usd
 }

 compile_USD() {
@@ -2814,10 +2630,6 @@ compile_USD() {
    if [ -d $INST/boost ]; then
      cmake_d="$cmake_d $cmake_d -D BOOST_ROOT=$INST/boost"
    fi
-
-    if [ -d $INST/tbb ]; then
-      cmake_d="$cmake_d $cmake_d -D TBB_ROOT_DIR=$INST/tbb"
-    fi
    cmake_d="$cmake_d -DPXR_SET_INTERNAL_NAMESPACE=usdBlender"
    cmake_d="$cmake_d -DPXR_ENABLE_PYTHON_SUPPORT=OFF"
    cmake_d="$cmake_d -DPXR_BUILD_IMAGING=OFF"
@@ -2866,10 +2678,8 @@ _update_deps_collada() {

 clean_OpenCOLLADA() {
  _init_opencollada
-  if [ -d $_inst ]; then
-    _update_deps_collada
-  fi
  _clean
+  _update_deps_collada
 }

 compile_OpenCOLLADA() {
@@ -2972,10 +2782,8 @@ _update_deps_embree() {

 clean_Embree() {
  _init_embree
-  if [ -d $_inst ]; then
-    _update_deps_embree
-  fi
  _clean
+  _update_deps_embree
 }

 compile_Embree() {
@@ -3081,10 +2889,8 @@ _update_deps_oidn() {

 clean_oidn() {
  _init_oidn
-  if [ -d $_inst ]; then
-    _update_deps_oidn
-  fi
  _clean
+  _update_deps_oidn
 }

 compile_OIDN() {
@@ -3146,10 +2952,6 @@ compile_OIDN() {
    cmake_d="$cmake_d -D WITH_TEST=OFF"
    cmake_d="$cmake_d -D OIDN_STATIC_LIB=ON"

-    if [ -d $INST/tbb ]; then
-      make_d="$make_d TBB_ROOT=$INST/tbb"
-    fi
-
    cmake $cmake_d ../

    make -j$THREADS && make install
@@ -3190,10 +2992,8 @@ _update_deps_ffmpeg() {

 clean_FFmpeg() {
  _init_ffmpeg
-  if [ -d $_inst ]; then
-    _update_deps_ffmpeg
-  fi
  _clean
+  _update_deps_ffmpeg
 }

 compile_FFmpeg() {
@@ -3319,10 +3119,8 @@ _update_deps_xr_openxr_sdk() {

 clean_XR_OpenXR_SDK() {
  _init_xr_openxr_sdk
-  if [ -d $_inst ]; then
-    _update_deps_xr_openxr_sdk
-  fi
  _clean
+  _update_deps_xr_openxr_sdk
 }

 compile_XR_OpenXR_SDK() {
@@ -3336,7 +3134,7 @@ compile_XR_OpenXR_SDK() {
  _init_xr_openxr_sdk

  # Clean install if needed!
-  magic_compile_check xr-openxr-$XR_OPENXR_VERSION $xr_openxr_magic
+  magic_compile_check xr-openxr-$OPENXR_VERSION $xr_openxr_magic
  if [ $? -eq 1 -o "$XR_OPENXR_FORCE_REBUILD" = true ]; then
    clean_XR_OpenXR_SDK
  fi
@@ -3513,7 +3311,7 @@ install_DEB() {
  THEORA_DEV="libtheora-dev"

  _packages="gawk cmake cmake-curses-gui build-essential libjpeg-dev libpng-dev libtiff-dev \
-             git libfreetype6-dev libx11-dev flex bison libxxf86vm-dev \
+             git libfreetype6-dev libx11-dev flex bison libtbb-dev libxxf86vm-dev \
             libxcursor-dev libxi-dev wget libsqlite3-dev libxrandr-dev libxinerama-dev \
             libbz2-dev libncurses5-dev libssl-dev liblzma-dev libreadline-dev \
             libopenal-dev libglew-dev yasm $THEORA_DEV $VORBIS_DEV $OGG_DEV \
@@ -3722,23 +3520,6 @@ install_DEB() {
  fi


-  PRINT ""
-  if [ "$TBB_SKIP" = true ]; then
-    WARNING "Skipping TBB installation, as requested..."
-  elif [ "$TBB_FORCE_BUILD" = true ]; then
-    INFO "Forced TBB building, as requested..."
-    compile_TBB
-  else
-    check_package_version_ge_DEB libtbb-dev $TBB_VERSION_MIN
-    if [ $? -eq 0 ]; then
-      install_packages_DEB libtbb-dev
-      clean_TBB
-    else
-      compile_TBB
-    fi
-  fi
-
-
  PRINT ""
  if [ "$OCIO_SKIP" = true ]; then
    WARNING "Skipping OpenColorIO installation, as requested..."
@@ -4175,7 +3956,7 @@ install_RPM() {
  THEORA_USE=true

  if [ "$RPM" = "FEDORA" -o "$RPM" = "RHEL" ]; then
-    _packages="$_packages freetype-devel"
+    _packages="$_packages freetype-devel tbb-devel"

    if [ "$WITH_JACK" = true ]; then
      _packages="$_packages jack-audio-connection-kit-devel"
@@ -4216,6 +3997,17 @@ install_RPM() {
    PRINT ""
    install_packages_RPM $_packages

+    PRINT ""
+    # Install TBB on openSUSE, from temporary repo
+    check_package_RPM tbb-devel
+    if [ $? -eq 0 ]; then
+      install_packages_RPM tbb-devel
+    else
+      $SUDO zypper ar -f http://download.opensuse.org/repositories/devel:/libraries:/c_c++/openSUSE_$_suse_rel/devel:libraries:c_c++.repo
+      $SUDO zypper -n --gpg-auto-import-keys install tbb-devel
+      $SUDO zypper rr devel_libraries_c_c++
+    fi
+
    PRINT ""
    X264_DEV="libx264-devel"
    check_package_version_ge_RPM $X264_DEV $X264_VERSION_MIN
@@ -4351,23 +4143,6 @@ install_RPM() {
  fi


-  PRINT ""
-  if [ "$TBB_SKIP" = true ]; then
-    WARNING "Skipping TBB installation, as requested..."
-  elif [ "$TBB_FORCE_BUILD" = true ]; then
-    INFO "Forced TBB building, as requested..."
-    compile_TBB
-  else
-    check_package_version_ge_RPM tbb-devel $TBB_VERSION_MIN
-    if [ $? -eq 0 ]; then
-      install_packages_RPM tbb-devel
-      clean_TBB
-    else
-      compile_TBB
-    fi
-  fi
-
-
  PRINT ""
  if [ "$OCIO_SKIP" = true ]; then
    WARNING "Skipping OpenColorIO installation, as requested..."
@@ -4724,7 +4499,7 @@ install_ARCH() {

  _packages="$BASE_DEVEL git cmake \
             libxi libxcursor libxrandr libxinerama glew libpng libtiff wget openal \
-             $OPENJPEG_DEV $VORBIS_DEV $OGG_DEV $THEORA_DEV yasm sdl fftw \
+             $OPENJPEG_DEV $VORBIS_DEV $OGG_DEV $THEORA_DEV yasm sdl fftw intel-tbb \
             libxml2 yaml-cpp tinyxml python-requests jemalloc"

  OPENJPEG_USE=true
@@ -4863,23 +4638,6 @@ install_ARCH() {
  fi


-  PRINT ""
-  if [ "$TBB_SKIP" = true ]; then
-    WARNING "Skipping TBB installation, as requested..."
-  elif [ "$TBB_FORCE_BUILD" = true ]; then
-    INFO "Forced TBB building, as requested..."
-    compile_TBB
-  else
-    check_package_version_ge_ARCH intel-tbb $TBB_VERSION_MIN
-    if [ $? -eq 0 ]; then
-      install_packages_ARCH intel-tbb
-      clean_TBB
-    else
-      compile_TBB
-    fi
-  fi
-
-
  PRINT ""
  if [ "$OCIO_SKIP" = true ]; then
    WARNING "Skipping OpenColorIO installation, as requested..."
@@ -5174,10 +4932,15 @@ install_OTHER() {
  fi

  PRINT ""
+  _do_compile_python=false
  if [ "$PYTHON_SKIP" = true ]; then
    WARNING "Skipping Python/NumPy installation, as requested..."
  elif [ "$PYTHON_FORCE_BUILD" = true ]; then
    INFO "Forced Python/NumPy building, as requested..."
+    _do_compile_python=true
+  fi
+
+  if [ "$_do_compile_python" = true ]; then
    compile_Python
    PRINT ""
    if [ "$NUMPY_SKIP" = true ]; then
@@ -5197,15 +4960,6 @@ install_OTHER() {
  fi


-  PRINT ""
-  if [ "$TBB_SKIP" = true ]; then
-    WARNING "Skipping TBB installation, as requested..."
-  elif [ "$TBB_FORCE_BUILD" = true ]; then
-    INFO "Forced TBB building, as requested..."
-    compile_TBB
-  fi
-
-
  PRINT ""
  if [ "$OCIO_SKIP" = true ]; then
    WARNING "Skipping OpenColorIO installation, as requested..."
@@ -5235,10 +4989,16 @@ install_OTHER() {

  PRINT ""
  have_llvm=false
+  _do_compile_llvm=false
  if [ "$LLVM_SKIP" = true ]; then
    WARNING "Skipping LLVM installation, as requested (this also implies skipping OSL!)..."
  elif [ "$LLVM_FORCE_BUILD" = true ]; then
    INFO "Forced LLVM building, as requested..."
+    _do_compile_llvm=true
+  fi
+
+  if [ "$_do_compile_llvm" = true ]; then
+    PRINT ""
    compile_LLVM
    have_llvm=true
    LLVM_VERSION_FOUND=$LLVM_VERSION
@@ -5246,10 +5006,15 @@ install_OTHER() {


  PRINT ""
+  _do_compile_osl=false
  if [ "$OSL_SKIP" = true ]; then
    WARNING "Skipping OpenShadingLanguage installation, as requested..."
  elif [ "$OSL_FORCE_BUILD" = true ]; then
    INFO "Forced OpenShadingLanguage building, as requested..."
+    _do_compile_osl=true
+  fi
+
+  if [ "$_do_compile_osl" = true ]; then
    if [ "$have_llvm" = true ]; then
      PRINT ""
      compile_OSL
@@ -5260,40 +5025,66 @@ install_OTHER() {


  PRINT ""
+  _do_compile_osd=false
  if [ "$OSD_SKIP" = true ]; then
    WARNING "Skipping OpenSubdiv installation, as requested..."
  elif [ "$OSD_FORCE_BUILD" = true ]; then
    INFO "Forced OpenSubdiv building, as requested..."
+    _do_compile_osd=true
+  fi
+
+  if [ "$_do_compile_osd" = true ]; then
+    PRINT ""
    compile_OSD
  fi


  if [ "$WITH_OPENCOLLADA" = true ]; then
+    _do_compile_collada=false
    PRINT ""
    if [ "$OPENCOLLADA_SKIP" = true ]; then
      WARNING "Skipping OpenCOLLADA installation, as requested..."
    elif [ "$OPENCOLLADA_FORCE_BUILD" = true ]; then
      INFO "Forced OpenCollada building, as requested..."
+      _do_compile_collada=true
+    fi
+
+    if [ "$_do_compile_collada" = true ]; then
+      PRINT ""
      compile_OpenCOLLADA
    fi
  fi

  if [ "$WITH_EMBREE" = true ]; then
+    _do_compile_embree=false
    PRINT ""
    if [ "$EMBREE_SKIP" = true ]; then
      WARNING "Skipping Embree installation, as requested..."
    elif [ "$EMBREE_FORCE_BUILD" = true ]; then
      INFO "Forced Embree building, as requested..."
+      _do_compile_embree=true
+    fi
+
+    if [ "$_do_compile_embree" = true ]; then
+      PRINT ""
      compile_Embree
    fi
  fi

  if [ "$WITH_OIDN" = true ]; then
+    _do_compile_oidn=false
    PRINT ""
    if [ "$OIDN_SKIP" = true ]; then
      WARNING "Skipping OpenImgeDenoise installation, as requested..."
    elif [ "$OIDN_FORCE_BUILD" = true ]; then
      INFO "Forced OpenImageDenoise building, as requested..."
+      _do_compile_oidn=true
+    else
+      # No package currently!
+      _do_compile_oidn=true
+    fi
+
+    if [ "$_do_compile_oidn" = true ]; then
      compile_OIDN
    fi
  fi
@@ -5312,6 +5103,9 @@ install_OTHER() {
  elif [ "$XR_OPENXR_FORCE_BUILD" = true ]; then
    INFO "Forced OpenXR-SDK building, as requested..."
    compile_XR_OpenXR_SDK
+  else
+    # No package currently!
+    compile_XR_OpenXR_SDK
  fi
 }

@@ -5397,7 +5191,7 @@ print_info() {
  PRINT ""
  PRINT "If you're using CMake add this to your configuration flags:"

-  _buildargs="-U *SNDFILE* -U PYTHON* -U *BOOST* -U *Boost* -U *TBB*"
+  _buildargs="-U *SNDFILE* -U PYTHON* -U *BOOST* -U *Boost*"
  _buildargs="$_buildargs -U *OPENCOLORIO* -U *OPENEXR* -U *OPENIMAGEIO* -U *LLVM* -U *CYCLES*"
  _buildargs="$_buildargs -U *OPENSUBDIV* -U *OPENVDB* -U *COLLADA* -U *FFMPEG* -U *ALEMBIC* -U *USD*"

@@ -5422,12 +5216,6 @@ print_info() {
    _buildargs="$_buildargs $_1 $_2"
  fi

-  if [ -d $INST/tbb ]; then
-    _1="-D TBB_ROOT_DIR=$INST/tbb"
-    PRINT "  $_1"
-    _buildargs="$_buildargs $_1"
-  fi
-
  if [ "$OCIO_SKIP" = false ]; then
    _1="-D WITH_OPENCOLORIO=ON"
    PRINT "  $_1"
--- a/build_files/build_environment/patches/openal.diff
+++ b/build_files/build_environment/patches/openal.diff
@@ -0,0 +1,13 @@
+diff -Naur external_openal_original/CMakeLists.txt external_openal/CMakeLists.txt
+--- external_openal_original/CMakeLists.txt	2016-01-24 20:12:39 -0700
+++ external_openal/CMakeLists.txt	2018-06-02 12:16:52 -0600
+@@ -885,7 +885,8 @@
+ OPTION(ALSOFT_REQUIRE_MMDEVAPI "Require MMDevApi backend" OFF)
+ IF(HAVE_WINDOWS_H)
+     # Check MMSystem backend
+-    CHECK_INCLUDE_FILES("windows.h;mmsystem.h" HAVE_MMSYSTEM_H -D_WIN32_WINNT=0x0502)
+    set(CMAKE_REQUIRED_FLAGS "-D_WIN32_WINNT=0x0502")
+    CHECK_INCLUDE_FILES("windows.h;mmsystem.h" HAVE_MMSYSTEM_H)
+     IF(HAVE_MMSYSTEM_H)
+         CHECK_SHARED_FUNCTION_EXISTS(waveOutOpen "windows.h;mmsystem.h" winmm "" HAVE_LIBWINMM)
+         IF(HAVE_LIBWINMM)
--- a/build_files/build_environment/patches/pthreads.diff
+++ b/build_files/build_environment/patches/pthreads.diff
@@ -1,12 +1,13 @@
-diff -Naur orig/Makefile external_pthreads/Makefile
--- orig/Makefile	2018-08-08 04:47:40 -0600
-+++ external_pthreads/Makefile	2020-05-09 11:20:28 -0600
-@@ -185,7 +185,7 @@
- 	@ $(MAKE) /E /nologo XCFLAGS="/MTd" EHFLAGS="$(VSEFLAGSD) /D__PTW32_STATIC_LIB /D__PTW32_BUILD_INLINED" CLEANUP=__PTW32_CLEANUP_SEH pthreadVSE$(PTW32_VER_DEBUG).inlined_static_stamp
+--- pthread.h.orig	2012-05-26 22:16:45 -0600
+++ pthread.h	2016-04-01 09:20:36 -0600
+@@ -109,6 +109,10 @@
+ /* Include everything */
+ #endif
 
- VC-static:
-	@ $(MAKE) /E /nologo XCFLAGS="/MT" EHFLAGS="$(VCFLAGS) /D__PTW32_STATIC_LIB /D__PTW32_BUILD_INLINED" CLEANUP=__PTW32_CLEANUP_C pthreadVC$(PTW32_VER).inlined_static_stamp
-+	@ $(MAKE) /E /nologo XCFLAGS="/MD" EHFLAGS="$(VCFLAGS) /D__PTW32_STATIC_LIB /D__PTW32_BUILD_INLINED" CLEANUP=__PTW32_CLEANUP_C pthreadVC$(PTW32_VER).inlined_static_stamp
- 
- VC-static-debug:
- 	@ $(MAKE) /E /nologo XCFLAGS="/MTd" EHFLAGS="$(VCFLAGSD) /D__PTW32_STATIC_LIB /D__PTW32_BUILD_INLINED" CLEANUP=__PTW32_CLEANUP_C pthreadVC$(PTW32_VER_DEBUG).inlined_static_stamp
+#if _MSC_VER >= 1900
+#   define HAVE_STRUCT_TIMESPEC 1
+#endif
+
+ #if defined(_UWIN)
+ #   define HAVE_STRUCT_TIMESPEC 1
+ #   define HAVE_SIGNAL_H        1
--- a/build_files/build_environment/patches/x264.diff
+++ b/build_files/build_environment/patches/x264.diff
@@ -0,0 +1,22 @@
+--- x264-snapshot-20180811-2245-stable\configure	2018-08-11 14:45:05 -0600
+++ external_x264\configure	2018-08-11 23:51:35 -0600
+@@ -396,7 +396,7 @@
+ # list of all preprocessor HAVE values we can define
+ CONFIG_HAVE="MALLOC_H ALTIVEC ALTIVEC_H MMX ARMV6 ARMV6T2 NEON BEOSTHREAD POSIXTHREAD WIN32THREAD THREAD LOG2F SWSCALE \
+              LAVF FFMS GPAC AVS GPL VECTOREXT INTERLACED CPU_COUNT OPENCL THP LSMASH X86_INLINE_ASM AS_FUNC INTEL_DISPATCHER \
+-             MSA MMAP WINRT VSX ARM_INLINE_ASM STRTOK_R BITDEPTH8 BITDEPTH10"
+             MSA MMAP WINRT VSX ARM_INLINE_ASM BITDEPTH8 BITDEPTH10"
+ 
+ # parse options
+ 
+@@ -1071,10 +1071,6 @@
+     define HAVE_LOG2F
+ fi
+ 
+-if cc_check 'string.h' '' 'strtok_r(0, 0, 0);' ; then
+-    define HAVE_STRTOK_R
+-fi
+-
+ if [ "$SYS" != "WINDOWS" ] && cpp_check "sys/mman.h unistd.h" "" "defined(MAP_PRIVATE)"; then
+     define HAVE_MMAP
+ fi
--- a/build_files/cmake/Modules/FindOpenEXR.cmake
+++ b/build_files/cmake/Modules/FindOpenEXR.cmake
@@ -104,7 +104,6 @@ FOREACH(COMPONENT ${_openexr_FIND_COMPONENTS})
  FIND_LIBRARY(OPENEXR_${UPPERCOMPONENT}_LIBRARY
    NAMES
      ${COMPONENT}-${_openexr_libs_ver} ${COMPONENT}
-    NAMES_PER_DIR
    HINTS
      ${_openexr_SEARCH_DIRS}
    PATH_SUFFIXES
--- a/build_files/cmake/Modules/FindUSD.cmake
+++ b/build_files/cmake/Modules/FindUSD.cmake
@@ -43,7 +43,6 @@ FIND_PATH(USD_INCLUDE_DIR
 FIND_LIBRARY(USD_LIBRARY
  NAMES
    usd_m usd_ms
-  NAMES_PER_DIR
  HINTS
    ${_usd_SEARCH_DIRS}
  PATH_SUFFIXES
--- a/build_files/cmake/config/blender_lite.cmake
+++ b/build_files/cmake/config/blender_lite.cmake
@@ -52,7 +52,6 @@ set(WITH_OPENVDB             OFF CACHE BOOL "" FORCE)
 set(WITH_QUADRIFLOW          OFF CACHE BOOL "" FORCE)
 set(WITH_SDL                 OFF CACHE BOOL "" FORCE)
 set(WITH_TBB                 OFF CACHE BOOL "" FORCE)
-set(WITH_USD                 OFF CACHE BOOL "" FORCE)

 if(UNIX AND NOT APPLE)
  set(WITH_GHOST_XDND          OFF CACHE BOOL "" FORCE)
--- a/build_files/cmake/platform/platform_unix.cmake
+++ b/build_files/cmake/platform/platform_unix.cmake
@@ -57,7 +57,6 @@ if(EXISTS ${LIBDIR})
  set(BOOST_ROOT ${LIBDIR}/boost)
  set(BOOST_LIBRARYDIR ${LIBDIR}/boost/lib)
  set(Boost_NO_SYSTEM_PATHS ON)
-  set(OPENEXR_ROOT_DIR ${LIBDIR}/openexr)
 endif()

 if(WITH_STATIC_LIBS)
--- a/build_files/cmake/platform/platform_win32.cmake
+++ b/build_files/cmake/platform/platform_win32.cmake
@@ -190,7 +190,7 @@ if(MSVC_VERSION GREATER 1914 AND NOT MSVC_CLANG)
  set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} /JMC")
 endif()

-set(PLATFORM_LINKFLAGS "${PLATFORM_LINKFLAGS} /SUBSYSTEM:CONSOLE /STACK:2097152")
+set(PLATFORM_LINKFLAGS "${PLATFORM_LINKFLAGS} /SUBSYSTEM:CONSOLE /STACK:2097152 /INCREMENTAL:NO ")
 set(PLATFORM_LINKFLAGS_RELEASE "/NODEFAULTLIB:libcmt.lib /NODEFAULTLIB:libcmtd.lib /NODEFAULTLIB:msvcrtd.lib")
 set(PLATFORM_LINKFLAGS_DEBUG "${PLATFORM_LINKFLAGS_DEBUG} /IGNORE:4099 /NODEFAULTLIB:libcmt.lib /NODEFAULTLIB:msvcrt.lib /NODEFAULTLIB:libcmtd.lib")

--- a/doc/doxygen/Doxyfile
+++ b/doc/doxygen/Doxyfile
@@ -51,7 +51,7 @@ PROJECT_BRIEF          =
 # pixels and the maximum width should not exceed 200 pixels. Doxygen will copy
 # the logo to the output directory.

-PROJECT_LOGO           = ../../release/freedesktop/icons/scalable/apps/blender.svg
+PROJECT_LOGO           = ../../release/freedesktop/icons/48x48/apps/blender.png

 # The OUTPUT_DIRECTORY tag is used to specify the (relative or absolute) path
 # into which the generated documentation will be written. If a relative path is
@@ -1720,7 +1720,7 @@ COMPACT_LATEX          = NO
 # The default value is: a4.
 # This tag requires that the tag GENERATE_LATEX is set to YES.

-PAPER_TYPE             = a4
+PAPER_TYPE             = a4wide

 # The EXTRA_PACKAGES tag can be used to specify one or more LaTeX package names
 # that should be included in the LaTeX output. The package can be specified just
--- a/doc/python_api/requirements.txt
+++ b/doc/python_api/requirements.txt
@@ -1,2 +1,2 @@
-Sphinx==3.0.3
-sphinx_rtd_theme==0.5.0rc1
+Sphinx==1.8.5
+sphinx_rtd_theme==0.4.3
--- a/doc/python_api/rst/info_overview.rst
+++ b/doc/python_api/rst/info_overview.rst
@@ -253,13 +253,7 @@ Registering a class with Blender results in the class definition being loaded in
 where it becomes available alongside existing functionality.

 Once this class is loaded you can access it from :mod:`bpy.types`,
-using the ``bl_idname`` rather than the classes original name.
-
-.. note::
-
-   There are some exceptions to this for class names which aren't guarantee to be unique.
-   In this case use: :func:`bpy.types.Struct.bl_rna_get_subclass`.
-
+using the bl_idname rather than the classes original name.

 When loading a class, Blender performs sanity checks making sure all required properties and functions are found,
 that properties have the correct type, and that functions have the right number of arguments.
--- a/doc/python_api/sphinx_doc_gen.py
+++ b/doc/python_api/sphinx_doc_gen.py
@@ -492,11 +492,6 @@ if _BPY_PROP_COLLECTION_FAKE:
 else:
    _BPY_PROP_COLLECTION_ID = "collection"

-if _BPY_STRUCT_FAKE:
-    bpy_struct = bpy.types.bpy_struct
-else:
-    bpy_struct = None
-

 def escape_rst(text):
    """ Escape plain text which may contain characters used by RST.
@@ -517,7 +512,7 @@ def is_struct_seq(value):


 def undocumented_message(module_name, type_name, identifier):
-    return "Undocumented, consider `contributing <https://developer.blender.org/T51061>`__."
+    return "Undocumented `contribute <https://developer.blender.org/T51061>`"


 def range_str(val):
@@ -1448,7 +1443,7 @@ def pyrna2sphinx(basepath):

            if _BPY_STRUCT_FAKE:
                descr_items = [
-                    (key, descr) for key, descr in sorted(bpy_struct.__dict__.items())
+                    (key, descr) for key, descr in sorted(bpy.types.Struct.__bases__[0].__dict__.items())
                    if not key.startswith("__")
                ]

@@ -1464,6 +1459,9 @@ def pyrna2sphinx(basepath):
                for identifier, py_prop in base.get_py_properties():
                    lines.append("   * :class:`%s.%s`\n" % (base.identifier, identifier))

+                for identifier, py_prop in base.get_py_properties():
+                    lines.append("   * :class:`%s.%s`\n" % (base.identifier, identifier))
+
            if lines:
                fw(".. rubric:: Inherited Properties\n\n")

@@ -1487,8 +1485,6 @@ def pyrna2sphinx(basepath):
                    lines.append("   * :class:`%s.%s`\n" % (base.identifier, func.identifier))
                for identifier, py_func in base.get_py_functions():
                    lines.append("   * :class:`%s.%s`\n" % (base.identifier, identifier))
-                for identifier, py_func in base.get_py_c_functions():
-                    lines.append("   * :class:`%s.%s`\n" % (base.identifier, identifier))

            if lines:
                fw(".. rubric:: Inherited Functions\n\n")
@@ -1576,7 +1572,7 @@ def pyrna2sphinx(basepath):

        # write fake classes
        if _BPY_STRUCT_FAKE:
-            class_value = bpy_struct
+            class_value = bpy.types.Struct.__bases__[0]
            fake_bpy_type(
                "bpy.types", class_value, _BPY_STRUCT_FAKE,
                "built-in base class for all classes in bpy.types.", use_subclasses=True,
@@ -1716,7 +1712,7 @@ class PatchedPythonDomain(PythonDomain):

    fw("def setup(app):\n")
    fw("    app.add_stylesheet('css/theme_overrides.css')\n")
-    fw("    app.add_domain(PatchedPythonDomain, override=True)\n\n")
+    fw("    app.override_domain(PatchedPythonDomain)\n\n")

    file.close()

--- a/extern/mantaflow/preprocessed/gitinfo.h
+++ b/extern/mantaflow/preprocessed/gitinfo.h
@@ -1,3 +1,3 @@


-#define MANTA_GIT_VERSION "commit b61bf9efa7a1d8ca98635076a7e9f2c4dacb2914"
+#define MANTA_GIT_VERSION "commit b4a2742bd743e2913fba94dd35846042e2650212"
--- a/extern/mantaflow/preprocessed/particle.h
+++ b/extern/mantaflow/preprocessed/particle.h
@@ -469,7 +469,6 @@ template<class S> class ParticleSystem : public ParticleBase {
                    const int integrationMode,
                    const bool deleteInObstacle = true,
                    const bool stopInObstacle = true,
-                    const bool skipNew = false,
                    const ParticleDataImpl<int> *ptype = NULL,
                    const int exclude = 0);
  static PyObject *_W_9(PyObject *_self, PyObject *_linargs, PyObject *_kwds)
@@ -487,20 +486,13 @@ template<class S> class ParticleSystem : public ParticleBase {
        const int integrationMode = _args.get<int>("integrationMode", 2, &_lock);
        const bool deleteInObstacle = _args.getOpt<bool>("deleteInObstacle", 3, true, &_lock);
        const bool stopInObstacle = _args.getOpt<bool>("stopInObstacle", 4, true, &_lock);
-        const bool skipNew = _args.getOpt<bool>("skipNew", 5, false, &_lock);
        const ParticleDataImpl<int> *ptype = _args.getPtrOpt<ParticleDataImpl<int>>(
-            "ptype", 6, NULL, &_lock);
-        const int exclude = _args.getOpt<int>("exclude", 7, 0, &_lock);
+            "ptype", 5, NULL, &_lock);
+        const int exclude = _args.getOpt<int>("exclude", 6, 0, &_lock);
        pbo->_args.copy(_args);
        _retval = getPyNone();
-        pbo->advectInGrid(flags,
-                          vel,
-                          integrationMode,
-                          deleteInObstacle,
-                          stopInObstacle,
-                          skipNew,
-                          ptype,
-                          exclude);
+        pbo->advectInGrid(
+            flags, vel, integrationMode, deleteInObstacle, stopInObstacle, ptype, exclude);
        pbo->_args.check();
      }
      pbFinalizePlugin(pbo->getParent(), "ParticleSystem::advectInGrid", !noTiming);
@@ -1871,7 +1863,6 @@ template<class S> struct _GridAdvectKernel : public KernelBase {
                    const Real dt,
                    const bool deleteInObstacle,
                    const bool stopInObstacle,
-                    const bool skipNew,
                    const ParticleDataImpl<int> *ptype,
                    const int exclude,
                    std::vector<Vec3> &u)
@@ -1882,7 +1873,6 @@ template<class S> struct _GridAdvectKernel : public KernelBase {
        dt(dt),
        deleteInObstacle(deleteInObstacle),
        stopInObstacle(stopInObstacle),
-        skipNew(skipNew),
        ptype(ptype),
        exclude(exclude),
        u(u)
@@ -1895,13 +1885,11 @@ template<class S> struct _GridAdvectKernel : public KernelBase {
                 const Real dt,
                 const bool deleteInObstacle,
                 const bool stopInObstacle,
-                 const bool skipNew,
                 const ParticleDataImpl<int> *ptype,
                 const int exclude,
                 std::vector<Vec3> &u) const
  {
-    if ((p[idx].flag & ParticleBase::PDELETE) || (ptype && ((*ptype)[idx] & exclude)) ||
-        (skipNew && (p[idx].flag & ParticleBase::PNEW))) {
+    if ((p[idx].flag & ParticleBase::PDELETE) || (ptype && ((*ptype)[idx] & exclude))) {
      u[idx] = 0.;
      return;
    }
@@ -1922,7 +1910,7 @@ template<class S> struct _GridAdvectKernel : public KernelBase {
  void operator()(const tbb::blocked_range<IndexInt> &__r) const
  {
    for (IndexInt idx = __r.begin(); idx != (IndexInt)__r.end(); idx++)
-      op(idx, p, vel, flags, dt, deleteInObstacle, stopInObstacle, skipNew, ptype, exclude, u);
+      op(idx, p, vel, flags, dt, deleteInObstacle, stopInObstacle, ptype, exclude, u);
  }
  void run()
  {
@@ -1934,7 +1922,6 @@ template<class S> struct _GridAdvectKernel : public KernelBase {
  const Real dt;
  const bool deleteInObstacle;
  const bool stopInObstacle;
-  const bool skipNew;
  const ParticleDataImpl<int> *ptype;
  const int exclude;
  std::vector<Vec3> &u;
@@ -1946,7 +1933,6 @@ template<class S> struct GridAdvectKernel : public KernelBase {
                   const Real dt,
                   const bool deleteInObstacle,
                   const bool stopInObstacle,
-                   const bool skipNew,
                   const ParticleDataImpl<int> *ptype,
                   const int exclude)
      : KernelBase(p.size()),
@@ -1957,7 +1943,6 @@ template<class S> struct GridAdvectKernel : public KernelBase {
               dt,
               deleteInObstacle,
               stopInObstacle,
-               skipNew,
               ptype,
               exclude,
               u),
@@ -1967,7 +1952,6 @@ template<class S> struct GridAdvectKernel : public KernelBase {
        dt(dt),
        deleteInObstacle(deleteInObstacle),
        stopInObstacle(stopInObstacle),
-        skipNew(skipNew),
        ptype(ptype),
        exclude(exclude),
        u((size))
@@ -2017,21 +2001,16 @@ template<class S> struct GridAdvectKernel : public KernelBase {
    return stopInObstacle;
  }
  typedef bool type5;
-  inline const bool &getArg6()
-  {
-    return skipNew;
-  }
-  typedef bool type6;
-  inline const ParticleDataImpl<int> *getArg7()
+  inline const ParticleDataImpl<int> *getArg6()
  {
    return ptype;
  }
-  typedef ParticleDataImpl<int> type7;
-  inline const int &getArg8()
+  typedef ParticleDataImpl<int> type6;
+  inline const int &getArg7()
  {
    return exclude;
  }
-  typedef int type8;
+  typedef int type7;
  void runMessage()
  {
    debMsg("Executing kernel GridAdvectKernel ", 3);
@@ -2046,7 +2025,6 @@ template<class S> struct GridAdvectKernel : public KernelBase {
  const Real dt;
  const bool deleteInObstacle;
  const bool stopInObstacle;
-  const bool skipNew;
  const ParticleDataImpl<int> *ptype;
  const int exclude;
  std::vector<Vec3> u;
@@ -2217,7 +2195,6 @@ void ParticleSystem<S>::advectInGrid(const FlagGrid &flags,
                                     const int integrationMode,
                                     const bool deleteInObstacle,
                                     const bool stopInObstacle,
-                                     const bool skipNew,
                                     const ParticleDataImpl<int> *ptype,
                                     const int exclude)
 {
@@ -2231,15 +2208,8 @@ void ParticleSystem<S>::advectInGrid(const FlagGrid &flags,
  }

  // update positions
-  GridAdvectKernel<S> kernel(mData,
-                             vel,
-                             flags,
-                             getParent()->getDt(),
-                             deleteInObstacle,
-                             stopInObstacle,
-                             skipNew,
-                             ptype,
-                             exclude);
+  GridAdvectKernel<S> kernel(
+      mData, vel, flags, getParent()->getDt(), deleteInObstacle, stopInObstacle, ptype, exclude);
  integratePointSet(kernel, integrationMode);

  if (!deleteInObstacle) {
@@ -2468,15 +2438,15 @@ template<class S> void ParticleSystem<S>::compress()
 //! insert buffered positions as new particles, update additional particle data
 template<class S> void ParticleSystem<S>::insertBufferedParticles()
 {
-  // clear new flag everywhere
-  for (IndexInt i = 0; i < (IndexInt)mData.size(); ++i)
-    mData[i].flag &= ~PNEW;
-
  if (mNewBufferPos.size() == 0)
    return;
  IndexInt newCnt = mData.size();
  resizeAll(newCnt + mNewBufferPos.size());

+  // clear new flag everywhere
+  for (IndexInt i = 0; i < (IndexInt)mData.size(); ++i)
+    mData[i].flag &= ~PNEW;
+
  for (IndexInt i = 0; i < (IndexInt)mNewBufferPos.size(); ++i) {
    int flag = (mNewBufferFlag.size() > 0) ? mNewBufferFlag[i] : 0;
    // note, other fields are not initialized here...
--- a/extern/mantaflow/preprocessed/plugin/secondaryparticles.cpp
+++ b/extern/mantaflow/preprocessed/plugin/secondaryparticles.cpp
@@ -982,14 +982,9 @@ void flipSampleSecondaryParticles(const std::string mode,
                                  const Real c_b,
                                  const Real k_ta,
                                  const Real k_wc,
-                                  const Real dt = 0,
+                                  const Real dt,
                                  const int itype = FlagGrid::TypeFluid)
 {
-
-  float timestep = dt;
-  if (dt <= 0)
-    timestep = flags.getParent()->getDt();
-
  if (mode == "single") {
    knFlipSampleSecondaryParticles(flags,
                                   v,
@@ -1006,7 +1001,7 @@ void flipSampleSecondaryParticles(const std::string mode,
                                   c_b,
                                   k_ta,
                                   k_wc,
-                                   timestep,
+                                   dt,
                                   itype);
  }
  else if (mode == "multiple") {
@@ -1025,7 +1020,7 @@ void flipSampleSecondaryParticles(const std::string mode,
                                                c_b,
                                                k_ta,
                                                k_wc,
-                                                timestep,
+                                                dt,
                                                itype);
  }
  else {
@@ -1058,7 +1053,7 @@ static PyObject *_W_1(PyObject *_self, PyObject *_linargs, PyObject *_kwds)
      const Real c_b = _args.get<Real>("c_b", 13, &_lock);
      const Real k_ta = _args.get<Real>("k_ta", 14, &_lock);
      const Real k_wc = _args.get<Real>("k_wc", 15, &_lock);
-      const Real dt = _args.getOpt<Real>("dt", 16, 0, &_lock);
+      const Real dt = _args.get<Real>("dt", 16, &_lock);
      const int itype = _args.getOpt<int>("itype", 17, FlagGrid::TypeFluid, &_lock);
      _retval = getPyNone();
      flipSampleSecondaryParticles(mode,
@@ -1696,20 +1691,13 @@ void flipUpdateSecondaryParticles(const std::string mode,
                                  const Real k_d,
                                  const Real c_s,
                                  const Real c_b,
-                                  const Real dt = 0,
-                                  bool scale = true,
+                                  const Real dt,
                                  const int exclude = ParticleBase::PTRACER,
                                  const int antitunneling = 0,
                                  const int itype = FlagGrid::TypeFluid)
 {

-  float gridScale = (scale) ? flags.getParent()->getDx() : 1;
-  Vec3 g = gravity / gridScale;
-
-  float timestep = dt;
-  if (dt <= 0)
-    timestep = flags.getParent()->getDt();
-
+  Vec3 g = gravity / flags.getDx();
  if (mode == "linear") {
    knFlipUpdateSecondaryParticlesLinear(pts_sec,
                                         v_sec,
@@ -1723,7 +1711,7 @@ void flipUpdateSecondaryParticles(const std::string mode,
                                         k_d,
                                         c_s,
                                         c_b,
-                                         timestep,
+                                         dt,
                                         exclude,
                                         antitunneling);
  }
@@ -1741,7 +1729,7 @@ void flipUpdateSecondaryParticles(const std::string mode,
                                        k_d,
                                        c_s,
                                        c_b,
-                                        timestep,
+                                        dt,
                                        exclude,
                                        antitunneling,
                                        itype);
@@ -1776,11 +1764,10 @@ static PyObject *_W_2(PyObject *_self, PyObject *_linargs, PyObject *_kwds)
      const Real k_d = _args.get<Real>("k_d", 11, &_lock);
      const Real c_s = _args.get<Real>("c_s", 12, &_lock);
      const Real c_b = _args.get<Real>("c_b", 13, &_lock);
-      const Real dt = _args.getOpt<Real>("dt", 14, 0, &_lock);
-      bool scale = _args.getOpt<bool>("scale", 15, true, &_lock);
-      const int exclude = _args.getOpt<int>("exclude", 16, ParticleBase::PTRACER, &_lock);
-      const int antitunneling = _args.getOpt<int>("antitunneling", 17, 0, &_lock);
-      const int itype = _args.getOpt<int>("itype", 18, FlagGrid::TypeFluid, &_lock);
+      const Real dt = _args.get<Real>("dt", 14, &_lock);
+      const int exclude = _args.getOpt<int>("exclude", 15, ParticleBase::PTRACER, &_lock);
+      const int antitunneling = _args.getOpt<int>("antitunneling", 16, 0, &_lock);
+      const int itype = _args.getOpt<int>("itype", 17, FlagGrid::TypeFluid, &_lock);
      _retval = getPyNone();
      flipUpdateSecondaryParticles(mode,
                                   pts_sec,
@@ -1797,7 +1784,6 @@ static PyObject *_W_2(PyObject *_self, PyObject *_linargs, PyObject *_kwds)
                                   c_s,
                                   c_b,
                                   dt,
-                                   scale,
                                   exclude,
                                   antitunneling,
                                   itype);
--- a/extern/wcwidth/wcwidth.h
+++ b/extern/wcwidth/wcwidth.h
@@ -21,7 +21,7 @@
 #define __WCWIDTH_H__

 #ifndef __cplusplus
-#  if defined(__APPLE__) || defined(__NetBSD__)
+#  if defined(__APPLE__)
 /* The <uchar.h> standard header is missing on macOS. */
 #include <stddef.h>
 typedef unsigned int char32_t;
--- a/intern/cycles/blender/addon/init.py
+++ b/intern/cycles/blender/addon/init.py
@@ -82,8 +82,8 @@ class CyclesRender(bpy.types.RenderEngine):
    def render(self, depsgraph):
        engine.render(self, depsgraph)

-    def bake(self, depsgraph, obj, pass_type, pass_filter, width, height):
-        engine.bake(self, depsgraph, obj, pass_type, pass_filter, width, height)
+    def bake(self, depsgraph, obj, pass_type, pass_filter, object_id, pixel_array, num_pixels, depth, result):
+        engine.bake(self, depsgraph, obj, pass_type, pass_filter, object_id, pixel_array, num_pixels, depth, result)

    # viewport render
    def view_update(self, context, depsgraph):
--- a/intern/cycles/blender/addon/engine.py
+++ b/intern/cycles/blender/addon/engine.py
@@ -168,11 +168,11 @@ def render(engine, depsgraph):
        _cycles.render(engine.session, depsgraph.as_pointer())


-def bake(engine, depsgraph, obj, pass_type, pass_filter, width, height):
+def bake(engine, depsgraph, obj, pass_type, pass_filter, object_id, pixel_array, num_pixels, depth, result):
    import _cycles
    session = getattr(engine, "session", None)
    if session is not None:
-        _cycles.bake(engine.session, depsgraph.as_pointer(), obj.as_pointer(), pass_type, pass_filter, width, height)
+        _cycles.bake(engine.session, depsgraph.as_pointer(), obj.as_pointer(), pass_type, pass_filter, object_id, pixel_array.as_pointer(), num_pixels, depth, result.as_pointer())


 def reset(engine, data, depsgraph):
--- a/intern/cycles/blender/addon/ui.py
+++ b/intern/cycles/blender/addon/ui.py
@@ -1907,15 +1907,10 @@ class CYCLES_RENDER_PT_bake_selected_to_active(CyclesButtonsPanel, Panel):

        col.prop(cbk, "use_cage", text="Cage")
        if cbk.use_cage:
-            col.prop(cbk, "cage_object")
-            col = layout.column()
-            col.prop(cbk, "cage_extrusion")
-            col.active = cbk.cage_object is None
-        else:
            col.prop(cbk, "cage_extrusion", text="Extrusion")
-
-        col = layout.column()
-        col.prop(cbk, "max_ray_distance")
+            col.prop(cbk, "cage_object", text="Cage Object")
+        else:
+            col.prop(cbk, "cage_extrusion", text="Ray Distance")


 class CYCLES_RENDER_PT_bake_output(CyclesButtonsPanel, Panel):
--- a/intern/cycles/blender/blender_object.cpp
+++ b/intern/cycles/blender/blender_object.cpp
@@ -460,12 +460,9 @@ void BlenderSync::sync_motion(BL::RenderSettings &b_render,
    sync_objects(b_depsgraph, b_v3d, 0.0f);
  }

-  /* Insert motion times from camera. Motion times from other objects
-   * have already been added in a sync_objects call. */
-  uint camera_motion_steps = object_motion_steps(b_cam, b_cam);
-  for (size_t step = 0; step < camera_motion_steps; step++) {
-    motion_times.insert(scene->camera->motion_time(step));
-  }
+  /* always sample these times for camera motion */
+  motion_times.insert(-1.0f);
+  motion_times.insert(1.0f);

  /* note iteration over motion_times set happens in sorted order */
  foreach (float relative_time, motion_times) {
@@ -490,8 +487,10 @@ void BlenderSync::sync_motion(BL::RenderSettings &b_render,
    b_engine.frame_set(frame, subframe);
    python_thread_state_save(python_thread_state);

-    /* Syncs camera motion if relative_time is one of the camera's motion times. */
-    sync_camera_motion(b_render, b_cam, width, height, relative_time);
+    /* sync camera, only supports two times at the moment */
+    if (relative_time == -1.0f || relative_time == 1.0f) {
+      sync_camera_motion(b_render, b_cam, width, height, relative_time);
+    }

    /* sync object */
    sync_objects(b_depsgraph, b_v3d, relative_time);
--- a/intern/cycles/blender/blender_python.cpp
+++ b/intern/cycles/blender/blender_python.cpp
@@ -60,12 +60,6 @@ void *pylong_as_voidptr_typesafe(PyObject *object)
  return PyLong_AsVoidPtr(object);
 }

-PyObject *pyunicode_from_string(const char *str)
-{
-  /* Ignore errors if device API returns invalid UTF-8 strings. */
-  return PyUnicode_DecodeUTF8(str, strlen(str), "ignore");
-}
-
 /* Synchronize debug flags from a given Blender scene.
 * Return truth when device list needs invalidation.
 */
@@ -298,18 +292,22 @@ static PyObject *render_func(PyObject * /*self*/, PyObject *args)
 static PyObject *bake_func(PyObject * /*self*/, PyObject *args)
 {
  PyObject *pysession, *pydepsgraph, *pyobject;
+  PyObject *pypixel_array, *pyresult;
  const char *pass_type;
-  int pass_filter, width, height;
+  int num_pixels, depth, object_id, pass_filter;

  if (!PyArg_ParseTuple(args,
-                        "OOOsiii",
+                        "OOOsiiOiiO",
                        &pysession,
                        &pydepsgraph,
                        &pyobject,
                        &pass_type,
                        &pass_filter,
-                        &width,
-                        &height))
+                        &object_id,
+                        &pypixel_array,
+                        &num_pixels,
+                        &depth,
+                        &pyresult))
    return NULL;

  BlenderSession *session = (BlenderSession *)PyLong_AsVoidPtr(pysession);
@@ -322,9 +320,23 @@ static PyObject *bake_func(PyObject * /*self*/, PyObject *args)
  RNA_id_pointer_create((ID *)PyLong_AsVoidPtr(pyobject), &objectptr);
  BL::Object b_object(objectptr);

+  void *b_result = PyLong_AsVoidPtr(pyresult);
+
+  PointerRNA bakepixelptr;
+  RNA_pointer_create(NULL, &RNA_BakePixel, PyLong_AsVoidPtr(pypixel_array), &bakepixelptr);
+  BL::BakePixel b_bake_pixel(bakepixelptr);
+
  python_thread_state_save(&session->python_thread_state);

-  session->bake(b_depsgraph, b_object, pass_type, pass_filter, width, height);
+  session->bake(b_depsgraph,
+                b_object,
+                pass_type,
+                pass_filter,
+                object_id,
+                b_bake_pixel,
+                (size_t)num_pixels,
+                depth,
+                (float *)b_result);

  python_thread_state_restore(&session->python_thread_state);

@@ -417,9 +429,9 @@ static PyObject *available_devices_func(PyObject * /*self*/, PyObject *args)
    DeviceInfo &device = devices[i];
    string type_name = Device::string_from_type(device.type);
    PyObject *device_tuple = PyTuple_New(3);
-    PyTuple_SET_ITEM(device_tuple, 0, pyunicode_from_string(device.description.c_str()));
-    PyTuple_SET_ITEM(device_tuple, 1, pyunicode_from_string(type_name.c_str()));
-    PyTuple_SET_ITEM(device_tuple, 2, pyunicode_from_string(device.id.c_str()));
+    PyTuple_SET_ITEM(device_tuple, 0, PyUnicode_FromString(device.description.c_str()));
+    PyTuple_SET_ITEM(device_tuple, 1, PyUnicode_FromString(type_name.c_str()));
+    PyTuple_SET_ITEM(device_tuple, 2, PyUnicode_FromString(device.id.c_str()));
    PyTuple_SET_ITEM(ret, i, device_tuple);
  }

@@ -630,7 +642,7 @@ static PyObject *osl_compile_func(PyObject * /*self*/, PyObject *args)
 static PyObject *system_info_func(PyObject * /*self*/, PyObject * /*value*/)
 {
  string system_info = Device::device_capabilities();
-  return pyunicode_from_string(system_info.c_str());
+  return PyUnicode_FromString(system_info.c_str());
 }

 #ifdef WITH_OPENCL
--- a/intern/cycles/blender/blender_session.cpp
+++ b/intern/cycles/blender/blender_session.cpp
@@ -247,7 +247,9 @@ void BlenderSession::reset_session(BL::BlendData &b_data, BL::Depsgraph &b_depsg

 void BlenderSession::free_session()
 {
-  delete sync;
+  if (sync)
+    delete sync;
+
  delete session;
 }

@@ -315,7 +317,6 @@ static void end_render_result(BL::RenderEngine &b_engine,

 void BlenderSession::do_write_update_render_tile(RenderTile &rtile,
                                                 bool do_update_only,
-                                                 bool do_read_only,
                                                 bool highlight)
 {
  int x = rtile.x - session->tile_manager.params.full_x;
@@ -341,23 +342,7 @@ void BlenderSession::do_write_update_render_tile(RenderTile &rtile,

  BL::RenderLayer b_rlay = *b_single_rlay;

-  if (do_read_only) {
-    /* copy each pass */
-    BL::RenderLayer::passes_iterator b_iter;
-
-    for (b_rlay.passes.begin(b_iter); b_iter != b_rlay.passes.end(); ++b_iter) {
-      BL::RenderPass b_pass(*b_iter);
-
-      /* find matching pass type */
-      PassType pass_type = BlenderSync::get_pass_type(b_pass);
-      int components = b_pass.channels();
-
-      rtile.buffers->set_pass_rect(pass_type, components, (float *)b_pass.rect());
-    }
-
-    end_render_result(b_engine, b_rr, false, false, false);
-  }
-  else if (do_update_only) {
+  if (do_update_only) {
    /* Sample would be zero at initial tile update, which is only needed
     * to tag tile form blender side as IN PROGRESS for proper highlight
     * no buffers should be sent to blender yet. For denoise we also
@@ -377,14 +362,9 @@ void BlenderSession::do_write_update_render_tile(RenderTile &rtile,
  }
 }

-void BlenderSession::read_render_tile(RenderTile &rtile)
-{
-  do_write_update_render_tile(rtile, false, true, false);
-}
-
 void BlenderSession::write_render_tile(RenderTile &rtile)
 {
-  do_write_update_render_tile(rtile, false, false, false);
+  do_write_update_render_tile(rtile, false, false);
 }

 void BlenderSession::update_render_tile(RenderTile &rtile, bool highlight)
@@ -394,9 +374,9 @@ void BlenderSession::update_render_tile(RenderTile &rtile, bool highlight)
   * would need to be investigated a bit further, but for now shall be fine
   */
  if (!b_engine.is_preview())
-    do_write_update_render_tile(rtile, true, false, highlight);
+    do_write_update_render_tile(rtile, true, highlight);
  else
-    do_write_update_render_tile(rtile, false, false, false);
+    do_write_update_render_tile(rtile, false, false);
 }

 static void add_cryptomatte_layer(BL::RenderResult &b_rr, string name, string manifest)
@@ -613,6 +593,25 @@ void BlenderSession::render(BL::Depsgraph &b_depsgraph_)
 #endif
 }

+static void populate_bake_data(BakeData *data,
+                               const int object_id,
+                               BL::BakePixel &pixel_array,
+                               const int num_pixels)
+{
+  BL::BakePixel bp = pixel_array;
+
+  int i;
+  for (i = 0; i < num_pixels; i++) {
+    if (bp.object_id() == object_id) {
+      data->set(i, bp.primitive_id(), bp.uv(), bp.du_dx(), bp.du_dy(), bp.dv_dx(), bp.dv_dy());
+    }
+    else {
+      data->set_null(i);
+    }
+    bp = bp.next();
+  }
+}
+
 static int bake_pass_filter_get(const int pass_filter)
 {
  int flag = BAKE_FILTER_NONE;
@@ -643,26 +642,43 @@ void BlenderSession::bake(BL::Depsgraph &b_depsgraph_,
                          BL::Object &b_object,
                          const string &pass_type,
                          const int pass_filter,
-                          const int bake_width,
-                          const int bake_height)
+                          const int object_id,
+                          BL::BakePixel &pixel_array,
+                          const size_t num_pixels,
+                          const int /*depth*/,
+                          float result[])
 {
  b_depsgraph = b_depsgraph_;

  ShaderEvalType shader_type = get_shader_type(pass_type);
+
+  /* Set baking flag in advance, so kernel loading can check if we need
+   * any baking capabilities.
+   */
+  scene->bake_manager->set_baking(true);
+
+  /* ensure kernels are loaded before we do any scene updates */
+  session->load_kernels();
+
+  if (shader_type == SHADER_EVAL_UV) {
+    /* force UV to be available */
+    Pass::add(PASS_UV, scene->film->passes);
+  }
+
  int bake_pass_filter = bake_pass_filter_get(pass_filter);
+  bake_pass_filter = BakeManager::shader_type_to_pass_filter(shader_type, bake_pass_filter);

-  /* Initialize bake manager, before we load the baking kernels. */
-  scene->bake_manager->set(scene, b_object.name(), shader_type, bake_pass_filter);
+  /* force use_light_pass to be true if we bake more than just colors */
+  if (bake_pass_filter & ~BAKE_FILTER_COLOR) {
+    Pass::add(PASS_LIGHT, scene->film->passes);
+  }

-  /* Passes are identified by name, so in order to return the combined pass we need to set the
-   * name. */
-  Pass::add(PASS_COMBINED, scene->film->passes, "Combined");
-
-  session->read_bake_tile_cb = function_bind(&BlenderSession::read_render_tile, this, _1);
-  session->write_render_tile_cb = function_bind(&BlenderSession::write_render_tile, this, _1);
+  /* create device and update scene */
+  scene->film->tag_update(scene);
+  scene->integrator->tag_update(scene);

  if (!session->progress.get_cancel()) {
-    /* Sync scene. */
+    /* update scene */
    BL::Object b_camera_override(b_engine.camera_override());
    sync->sync_camera(b_render, b_camera_override, width, height, "");
    sync->sync_data(
@@ -670,43 +686,75 @@ void BlenderSession::bake(BL::Depsgraph &b_depsgraph_,
    builtin_images_load();
  }

-  /* Object might have been disabled for rendering or excluded in some
-   * other way, in that case Blender will report a warning afterwards. */
-  bool object_found = false;
-  foreach (Object *ob, scene->objects) {
-    if (ob->name == b_object.name()) {
-      object_found = true;
-      break;
-    }
-  }
+  BakeData *bake_data = NULL;

-  if (object_found && !session->progress.get_cancel()) {
-    /* Get session and buffer parameters. */
+  if (!session->progress.get_cancel()) {
+    /* get buffer parameters */
    SessionParams session_params = BlenderSync::get_session_params(
        b_engine, b_userpref, b_scene, background);
-    session_params.progressive_refine = false;
+    BufferParams buffer_params = BlenderSync::get_buffer_params(
+        b_scene, b_render, b_v3d, b_rv3d, scene->camera, width, height);

-    BufferParams buffer_params;
-    buffer_params.width = bake_width;
-    buffer_params.height = bake_height;
-    buffer_params.passes = scene->film->passes;
+    scene->bake_manager->set_shader_limit((size_t)b_engine.tile_x(), (size_t)b_engine.tile_y());

-    /* Update session. */
+    /* set number of samples */
    session->tile_manager.set_samples(session_params.samples);
    session->reset(buffer_params, session_params.samples);
+    session->update_scene();
+
+    /* find object index. todo: is arbitrary - copied from mesh_displace.cpp */
+    size_t object_index = OBJECT_NONE;
+    int tri_offset = 0;
+
+    for (size_t i = 0; i < scene->objects.size(); i++) {
+      const Object *object = scene->objects[i];
+      const Geometry *geom = object->geometry;
+      if (object->name == b_object.name() && geom->type == Geometry::MESH) {
+        const Mesh *mesh = static_cast<const Mesh *>(geom);
+        object_index = i;
+        tri_offset = mesh->prim_offset;
+        break;
+      }
+    }
+
+    /* Object might have been disabled for rendering or excluded in some
+     * other way, in that case Blender will report a warning afterwards. */
+    if (object_index != OBJECT_NONE) {
+      int object = object_index;
+
+      bake_data = scene->bake_manager->init(object, tri_offset, num_pixels);
+      populate_bake_data(bake_data, object_id, pixel_array, num_pixels);
+    }
+
+    /* set number of samples */
+    session->tile_manager.set_samples(session_params.samples);
+    session->reset(buffer_params, session_params.samples);
+    session->update_scene();

    session->progress.set_update_callback(
        function_bind(&BlenderSession::update_bake_progress, this));
  }

  /* Perform bake. Check cancel to avoid crash with incomplete scene data. */
-  if (object_found && !session->progress.get_cancel()) {
-    session->start();
-    session->wait();
+  if (!session->progress.get_cancel() && bake_data) {
+    scene->bake_manager->bake(scene->device,
+                              &scene->dscene,
+                              scene,
+                              session->progress,
+                              shader_type,
+                              bake_pass_filter,
+                              bake_data,
+                              result);
  }

-  session->read_bake_tile_cb = function_null;
-  session->write_render_tile_cb = function_null;
+  /* free all memory used (host and device), so we wouldn't leave render
+   * engine with extra memory allocated
+   */
+
+  session->device_free();
+
+  delete sync;
+  sync = NULL;
 }

 void BlenderSession::do_write_update_render_result(BL::RenderLayer &b_rlay,
--- a/intern/cycles/blender/blender_session.h
+++ b/intern/cycles/blender/blender_session.h
@@ -66,12 +66,14 @@ class BlenderSession {
            BL::Object &b_object,
            const string &pass_type,
            const int custom_flag,
-            const int bake_width,
-            const int bake_height);
+            const int object_id,
+            BL::BakePixel &pixel_array,
+            const size_t num_pixels,
+            const int depth,
+            float pixels[]);

  void write_render_result(BL::RenderLayer &b_rlay, RenderTile &rtile);
  void write_render_tile(RenderTile &rtile);
-  void read_render_tile(RenderTile &rtile);

  /* update functions are used to update display buffer only after sample was rendered
   * only needed for better visual feedback */
@@ -153,10 +155,7 @@ class BlenderSession {
  void do_write_update_render_result(BL::RenderLayer &b_rlay,
                                     RenderTile &rtile,
                                     bool do_update_only);
-  void do_write_update_render_tile(RenderTile &rtile,
-                                   bool do_update_only,
-                                   bool do_read_only,
-                                   bool highlight);
+  void do_write_update_render_tile(RenderTile &rtile, bool do_update_only, bool highlight);

  void builtin_images_load();

--- a/intern/cycles/blender/blender_shader.cpp
+++ b/intern/cycles/blender/blender_shader.cpp
@@ -384,16 +384,16 @@ static ShaderNode *add_node(Scene *scene,

    switch (b_aniso_node.distribution()) {
      case BL::ShaderNodeBsdfAnisotropic::distribution_BECKMANN:
-        aniso->distribution = CLOSURE_BSDF_MICROFACET_BECKMANN_ID;
+        aniso->distribution = CLOSURE_BSDF_MICROFACET_BECKMANN_ANISO_ID;
        break;
      case BL::ShaderNodeBsdfAnisotropic::distribution_GGX:
-        aniso->distribution = CLOSURE_BSDF_MICROFACET_GGX_ID;
+        aniso->distribution = CLOSURE_BSDF_MICROFACET_GGX_ANISO_ID;
        break;
      case BL::ShaderNodeBsdfAnisotropic::distribution_MULTI_GGX:
-        aniso->distribution = CLOSURE_BSDF_MICROFACET_MULTI_GGX_ID;
+        aniso->distribution = CLOSURE_BSDF_MICROFACET_MULTI_GGX_ANISO_ID;
        break;
      case BL::ShaderNodeBsdfAnisotropic::distribution_ASHIKHMIN_SHIRLEY:
-        aniso->distribution = CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ID;
+        aniso->distribution = CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ANISO_ID;
        break;
    }

@@ -1231,11 +1231,12 @@ void BlenderSync::sync_materials(BL::Depsgraph &b_depsgraph, bool update_all)
    Shader *shader;

    /* test if we need to sync */
-    if (shader_map.add_or_update(&shader, b_mat) || update_all) {
+    if (shader_map.add_or_update(&shader, b_mat) || shader->need_sync_object || update_all) {
      ShaderGraph *graph = new ShaderGraph();

      shader->name = b_mat.name().c_str();
      shader->pass_id = b_mat.pass_index();
+      shader->need_sync_object = false;

      /* create nodes */
      if (b_mat.use_nodes() && b_mat.node_tree()) {
--- a/intern/cycles/blender/blender_sync.cpp
+++ b/intern/cycles/blender/blender_sync.cpp
@@ -85,6 +85,8 @@ void BlenderSync::sync_recalc(BL::Depsgraph &b_depsgraph, BL::SpaceView3D &b_v3d
  /* Sync recalc flags from blender to cycles. Actual update is done separate,
   * so we can do it later on if doing it immediate is not suitable. */

+  bool has_updated_objects = b_depsgraph.id_type_updated(BL::DriverTarget::id_type_OBJECT);
+
  if (experimental) {
    /* Mark all meshes as needing to be exported again if dicing changed. */
    PointerRNA cscene = RNA_pointer_get(&b_scene.ptr, "cycles");
@@ -187,6 +189,19 @@ void BlenderSync::sync_recalc(BL::Depsgraph &b_depsgraph, BL::SpaceView3D &b_v3d
  if (viewport_parameters.modified(new_viewport_parameters)) {
    world_recalc = true;
  }
+
+  /* Updates shader with object dependency if objects changed. */
+  if (has_updated_objects) {
+    if (scene->default_background->has_object_dependency) {
+      world_recalc = true;
+    }
+
+    foreach (Shader *shader, scene->shaders) {
+      if (shader->has_object_dependency) {
+        shader->need_sync_object = true;
+      }
+    }
+  }
 }

 void BlenderSync::sync_data(BL::RenderSettings &b_render,
@@ -481,9 +496,6 @@ PassType BlenderSync::get_pass_type(BL::RenderPass &b_pass)
  MAP_PASS("AO", PASS_AO);
  MAP_PASS("Shadow", PASS_SHADOW);

-  MAP_PASS("BakePrimitive", PASS_BAKE_PRIMITIVE);
-  MAP_PASS("BakeDifferential", PASS_BAKE_DIFFERENTIAL);
-
 #ifdef __KERNEL_DEBUG__
  MAP_PASS("Debug BVH Traversed Nodes", PASS_BVH_TRAVERSED_NODES);
  MAP_PASS("Debug BVH Traversed Instances", PASS_BVH_TRAVERSED_INSTANCES);
--- a/intern/cycles/blender/blender_volume.cpp
+++ b/intern/cycles/blender/blender_volume.cpp
@@ -34,13 +34,15 @@ CCL_NAMESPACE_BEGIN
 /* TODO: verify this is not loading unnecessary attributes. */
 class BlenderSmokeLoader : public ImageLoader {
 public:
-  BlenderSmokeLoader(BL::Object &b_ob, AttributeStandard attribute)
-      : b_domain(object_fluid_gas_domain_find(b_ob)), b_mesh(b_ob.data()), attribute(attribute)
+  BlenderSmokeLoader(const BL::Object &b_ob, AttributeStandard attribute)
+      : b_ob(b_ob), attribute(attribute)
  {
  }

  bool load_metadata(ImageMetaData &metadata) override
  {
+    BL::FluidDomainSettings b_domain = object_fluid_gas_domain_find(b_ob);
+
    if (!b_domain) {
      return false;
    }
@@ -77,6 +79,7 @@ class BlenderSmokeLoader : public ImageLoader {
    /* Create a matrix to transform from object space to mesh texture space.
     * This does not work with deformations but that can probably only be done
     * well with a volume grid mapping of coordinates. */
+    BL::Mesh b_mesh(b_ob.data());
    float3 loc, size;
    mesh_texture_space(b_mesh, loc, size);
    metadata.transform_3d = transform_translate(-loc) * transform_scale(size);
@@ -87,6 +90,9 @@ class BlenderSmokeLoader : public ImageLoader {

  bool load_pixels(const ImageMetaData &, void *pixels, const size_t, const bool) override
  {
+    /* smoke volume data */
+    BL::FluidDomainSettings b_domain = object_fluid_gas_domain_find(b_ob);
+
    if (!b_domain) {
      return false;
    }
@@ -173,11 +179,10 @@ class BlenderSmokeLoader : public ImageLoader {
  bool equals(const ImageLoader &other) const override
  {
    const BlenderSmokeLoader &other_loader = (const BlenderSmokeLoader &)other;
-    return b_domain == other_loader.b_domain && attribute == other_loader.attribute;
+    return b_ob == other_loader.b_ob && attribute == other_loader.attribute;
  }

-  BL::FluidDomainSettings b_domain;
-  BL::Mesh b_mesh;
+  BL::Object b_ob;
  AttributeStandard attribute;
 };

--- a/intern/cycles/device/cuda/device_cuda.h
+++ b/intern/cycles/device/cuda/device_cuda.h
@@ -223,7 +223,7 @@ class CUDADevice : public Device {
                              CUdeviceptr d_wtile,
                              CUstream stream = 0);

-  void render(DeviceTask &task, RenderTile &rtile, device_vector<WorkTile> &work_tiles);
+  void path_trace(DeviceTask &task, RenderTile &rtile, device_vector<WorkTile> &work_tiles);

  void film_convert(DeviceTask &task,
                    device_ptr buffer,
--- a/intern/cycles/device/cuda/device_cuda_impl.cpp
+++ b/intern/cycles/device/cuda/device_cuda_impl.cpp
@@ -421,10 +421,10 @@ string CUDADevice::compile_kernel(const DeviceRequestedFeatures &requested_featu
        nvcc_cuda_version % 10);
    return string();
  }
-  else if (!(nvcc_cuda_version == 101 || nvcc_cuda_version == 102)) {
+  else if (nvcc_cuda_version != 101) {
    printf(
        "CUDA version %d.%d detected, build may succeed but only "
-        "CUDA 10.1 and 10.2 are officially supported.\n",
+        "CUDA 10.1 is officially supported.\n",
        nvcc_cuda_version / 10,
        nvcc_cuda_version % 10);
  }
@@ -586,23 +586,20 @@ void CUDADevice::reserve_local_memory(const DeviceRequestedFeatures &requested_f
  cuMemGetInfo(&free_before, &total);

  /* Get kernel function. */
-  CUfunction cuRender;
+  CUfunction cuPathTrace;

-  if (requested_features.use_baking) {
-    cuda_assert(cuModuleGetFunction(&cuRender, cuModule, "kernel_cuda_bake"));
-  }
-  else if (requested_features.use_integrator_branched) {
-    cuda_assert(cuModuleGetFunction(&cuRender, cuModule, "kernel_cuda_branched_path_trace"));
+  if (requested_features.use_integrator_branched) {
+    cuda_assert(cuModuleGetFunction(&cuPathTrace, cuModule, "kernel_cuda_branched_path_trace"));
  }
  else {
-    cuda_assert(cuModuleGetFunction(&cuRender, cuModule, "kernel_cuda_path_trace"));
+    cuda_assert(cuModuleGetFunction(&cuPathTrace, cuModule, "kernel_cuda_path_trace"));
  }

-  cuda_assert(cuFuncSetCacheConfig(cuRender, CU_FUNC_CACHE_PREFER_L1));
+  cuda_assert(cuFuncSetCacheConfig(cuPathTrace, CU_FUNC_CACHE_PREFER_L1));

  int min_blocks, num_threads_per_block;
-  cuda_assert(
-      cuOccupancyMaxPotentialBlockSize(&min_blocks, &num_threads_per_block, cuRender, NULL, 0, 0));
+  cuda_assert(cuOccupancyMaxPotentialBlockSize(
+      &min_blocks, &num_threads_per_block, cuPathTrace, NULL, 0, 0));

  /* Launch kernel, using just 1 block appears sufficient to reserve
   * memory for all multiprocessors. It would be good to do this in
@@ -612,7 +609,7 @@ void CUDADevice::reserve_local_memory(const DeviceRequestedFeatures &requested_f

  void *args[] = {&d_work_tiles, &total_work_size};

-  cuda_assert(cuLaunchKernel(cuRender, 1, 1, 1, num_threads_per_block, 1, 1, 0, 0, args, 0));
+  cuda_assert(cuLaunchKernel(cuPathTrace, 1, 1, 1, num_threads_per_block, 1, 1, 0, 0, args, 0));

  cuda_assert(cuCtxSynchronize());

@@ -1783,7 +1780,9 @@ void CUDADevice::adaptive_sampling_post(RenderTile &rtile,
                             0));
 }

-void CUDADevice::render(DeviceTask &task, RenderTile &rtile, device_vector<WorkTile> &work_tiles)
+void CUDADevice::path_trace(DeviceTask &task,
+                            RenderTile &rtile,
+                            device_vector<WorkTile> &work_tiles)
 {
  scoped_timer timer(&rtile.buffers->render_time);

@@ -1791,24 +1790,21 @@ void CUDADevice::render(DeviceTask &task, RenderTile &rtile, device_vector<WorkT
    return;

  CUDAContextScope scope(this);
-  CUfunction cuRender;
+  CUfunction cuPathTrace;

  /* Get kernel function. */
-  if (rtile.task == RenderTile::BAKE) {
-    cuda_assert(cuModuleGetFunction(&cuRender, cuModule, "kernel_cuda_bake"));
-  }
-  else if (task.integrator_branched) {
-    cuda_assert(cuModuleGetFunction(&cuRender, cuModule, "kernel_cuda_branched_path_trace"));
+  if (task.integrator_branched) {
+    cuda_assert(cuModuleGetFunction(&cuPathTrace, cuModule, "kernel_cuda_branched_path_trace"));
  }
  else {
-    cuda_assert(cuModuleGetFunction(&cuRender, cuModule, "kernel_cuda_path_trace"));
+    cuda_assert(cuModuleGetFunction(&cuPathTrace, cuModule, "kernel_cuda_path_trace"));
  }

  if (have_error()) {
    return;
  }

-  cuda_assert(cuFuncSetCacheConfig(cuRender, CU_FUNC_CACHE_PREFER_L1));
+  cuda_assert(cuFuncSetCacheConfig(cuPathTrace, CU_FUNC_CACHE_PREFER_L1));

  /* Allocate work tile. */
  work_tiles.alloc(1);
@@ -1826,8 +1822,8 @@ void CUDADevice::render(DeviceTask &task, RenderTile &rtile, device_vector<WorkT
   * remain conservative for GPUs connected to a display to avoid driver
   * timeouts and display freezing. */
  int min_blocks, num_threads_per_block;
-  cuda_assert(
-      cuOccupancyMaxPotentialBlockSize(&min_blocks, &num_threads_per_block, cuRender, NULL, 0, 0));
+  cuda_assert(cuOccupancyMaxPotentialBlockSize(
+      &min_blocks, &num_threads_per_block, cuPathTrace, NULL, 0, 0));
  if (!info.display_device) {
    min_blocks *= 8;
  }
@@ -1855,7 +1851,7 @@ void CUDADevice::render(DeviceTask &task, RenderTile &rtile, device_vector<WorkT
    void *args[] = {&d_work_tiles, &total_work_size};

    cuda_assert(
-        cuLaunchKernel(cuRender, num_blocks, 1, 1, num_threads_per_block, 1, 1, 0, 0, args, 0));
+        cuLaunchKernel(cuPathTrace, num_blocks, 1, 1, num_threads_per_block, 1, 1, 0, 0, args, 0));

    /* Run the adaptive sampling kernels at selected samples aligned to step samples. */
    uint filter_sample = sample + wtile->num_samples - 1;
@@ -1961,7 +1957,10 @@ void CUDADevice::shader(DeviceTask &task)
  CUdeviceptr d_output = (CUdeviceptr)task.shader_output;

  /* get kernel function */
-  if (task.shader_eval_type == SHADER_EVAL_DISPLACE) {
+  if (task.shader_eval_type >= SHADER_EVAL_BAKE) {
+    cuda_assert(cuModuleGetFunction(&cuShader, cuModule, "kernel_cuda_bake"));
+  }
+  else if (task.shader_eval_type == SHADER_EVAL_DISPLACE) {
    cuda_assert(cuModuleGetFunction(&cuShader, cuModule, "kernel_cuda_displace"));
  }
  else {
@@ -2298,12 +2297,9 @@ void CUDADevice::thread_run(DeviceTask *task)
          split_kernel->path_trace(task, tile, void_buffer, void_buffer);
        }
        else {
-          render(*task, tile, work_tiles);
+          path_trace(*task, tile, work_tiles);
        }
      }
-      else if (tile.task == RenderTile::BAKE) {
-        render(*task, tile, work_tiles);
-      }
      else if (tile.task == RenderTile::DENOISE) {
        tile.sample = tile.start_sample + tile.num_samples;

--- a/intern/cycles/device/device.cpp
+++ b/intern/cycles/device/device.cpp
@@ -494,18 +494,16 @@ vector<DeviceInfo> Device::available_devices(uint mask)
  }
 #endif

-#if defined(WITH_CUDA) || defined(WITH_OPTIX)
-  if (mask & (DEVICE_MASK_CUDA | DEVICE_MASK_OPTIX)) {
+#ifdef WITH_CUDA
+  if (mask & DEVICE_MASK_CUDA) {
    if (!(devices_initialized_mask & DEVICE_MASK_CUDA)) {
      if (device_cuda_init()) {
        device_cuda_info(cuda_devices);
      }
      devices_initialized_mask |= DEVICE_MASK_CUDA;
    }
-    if (mask & DEVICE_MASK_CUDA) {
-      foreach (DeviceInfo &info, cuda_devices) {
-        devices.push_back(info);
-      }
+    foreach (DeviceInfo &info, cuda_devices) {
+      devices.push_back(info);
    }
  }
 #endif
@@ -514,7 +512,7 @@ vector<DeviceInfo> Device::available_devices(uint mask)
  if (mask & DEVICE_MASK_OPTIX) {
    if (!(devices_initialized_mask & DEVICE_MASK_OPTIX)) {
      if (device_optix_init()) {
-        device_optix_info(cuda_devices, optix_devices);
+        device_optix_info(optix_devices);
      }
      devices_initialized_mask |= DEVICE_MASK_OPTIX;
    }
--- a/intern/cycles/device/device_cpu.cpp
+++ b/intern/cycles/device/device_cpu.cpp
@@ -188,7 +188,6 @@ class CPUDevice : public Device {
      convert_to_byte_kernel;
  KernelFunctions<void (*)(KernelGlobals *, uint4 *, float4 *, int, int, int, int, int)>
      shader_kernel;
-  KernelFunctions<void (*)(KernelGlobals *, float *, int, int, int, int, int)> bake_kernel;

  KernelFunctions<void (*)(
      int, TileInfo *, int, int, float *, float *, float *, float *, float *, int *, int, int)>
@@ -271,7 +270,6 @@ class CPUDevice : public Device {
        REGISTER_KERNEL(convert_to_half_float),
        REGISTER_KERNEL(convert_to_byte),
        REGISTER_KERNEL(shader),
-        REGISTER_KERNEL(bake),
        REGISTER_KERNEL(filter_divide_shadow),
        REGISTER_KERNEL(filter_get_feature),
        REGISTER_KERNEL(filter_write_feature),
@@ -897,7 +895,7 @@ class CPUDevice : public Device {
    }
  }

-  void render(DeviceTask &task, RenderTile &tile, KernelGlobals *kg)
+  void path_trace(DeviceTask &task, RenderTile &tile, KernelGlobals *kg)
  {
    const bool use_coverage = kernel_data.film.cryptomatte_passes & CRYPT_ACCURATE;

@@ -921,21 +919,12 @@ class CPUDevice : public Device {
          break;
      }

-      if (tile.task == RenderTile::PATH_TRACE) {
-        for (int y = tile.y; y < tile.y + tile.h; y++) {
-          for (int x = tile.x; x < tile.x + tile.w; x++) {
-            if (use_coverage) {
-              coverage.init_pixel(x, y);
-            }
-            path_trace_kernel()(kg, render_buffer, sample, x, y, tile.offset, tile.stride);
-          }
-        }
-      }
-      else {
-        for (int y = tile.y; y < tile.y + tile.h; y++) {
-          for (int x = tile.x; x < tile.x + tile.w; x++) {
-            bake_kernel()(kg, render_buffer, sample, x, y, tile.offset, tile.stride);
+      for (int y = tile.y; y < tile.y + tile.h; y++) {
+        for (int x = tile.x; x < tile.x + tile.w; x++) {
+          if (use_coverage) {
+            coverage.init_pixel(x, y);
          }
+          path_trace_kernel()(kg, render_buffer, sample, x, y, tile.offset, tile.stride);
        }
      }
      tile.sample = sample + 1;
@@ -1030,12 +1019,9 @@ class CPUDevice : public Device {
          split_kernel->path_trace(&task, tile, kgbuffer, void_buffer);
        }
        else {
-          render(task, tile, kg);
+          path_trace(task, tile, kg);
        }
      }
-      else if (tile.task == RenderTile::BAKE) {
-        render(task, tile, kg);
-      }
      else if (tile.task == RenderTile::DENOISE) {
        denoise(denoising, tile);
        task.update_progress(&tile, tile.w * tile.h);
--- a/intern/cycles/device/device_intern.h
+++ b/intern/cycles/device/device_intern.h
@@ -45,7 +45,7 @@ Device *device_multi_create(DeviceInfo &info, Stats &stats, Profiler &profiler,
 void device_cpu_info(vector<DeviceInfo> &devices);
 void device_opencl_info(vector<DeviceInfo> &devices);
 void device_cuda_info(vector<DeviceInfo> &devices);
-void device_optix_info(const vector<DeviceInfo> &cuda_devices, vector<DeviceInfo> &devices);
+void device_optix_info(vector<DeviceInfo> &devices);
 void device_network_info(vector<DeviceInfo> &devices);

 string device_cpu_capabilities();
--- a/intern/cycles/device/device_optix.cpp
+++ b/intern/cycles/device/device_optix.cpp
@@ -1534,11 +1534,14 @@ bool device_optix_init()
  return true;
 }

-void device_optix_info(const vector<DeviceInfo> &cuda_devices, vector<DeviceInfo> &devices)
+void device_optix_info(vector<DeviceInfo> &devices)
 {
  // Simply add all supported CUDA devices as OptiX devices again
-  for (const DeviceInfo &cuda_info : cuda_devices) {
-    DeviceInfo info = cuda_info;
+  vector<DeviceInfo> cuda_devices;
+  device_cuda_info(cuda_devices);
+
+  for (auto it = cuda_devices.begin(); it != cuda_devices.end();) {
+    DeviceInfo &info = *it;
    assert(info.type == DEVICE_CUDA);
    info.type = DEVICE_OPTIX;
    info.id += "_OptiX";
@@ -1561,10 +1564,13 @@ void device_optix_info(const vector<DeviceInfo> &cuda_devices, vector<DeviceInfo
    }

    // Only add devices with RTX support
-    if (rtcore_version != 0 || getenv("CYCLES_OPTIX_TEST")) {
-      devices.push_back(info);
-    }
+    if (rtcore_version == 0 && !getenv("CYCLES_OPTIX_TEST"))
+      it = cuda_devices.erase(it);
+    else
+      ++it;
  }
+
+  devices.insert(devices.end(), cuda_devices.begin(), cuda_devices.end());
 }

 Device *device_optix_create(DeviceInfo &info, Stats &stats, Profiler &profiler, bool background)
--- a/intern/cycles/device/opencl/device_opencl.h
+++ b/intern/cycles/device/opencl/device_opencl.h
@@ -451,7 +451,6 @@ class OpenCLDevice : public Device {
                    device_ptr rgba_half);
  void shader(DeviceTask &task);
  void update_adaptive(DeviceTask &task, RenderTile &tile, int sample);
-  void bake(DeviceTask &task, RenderTile &tile);

  void denoise(RenderTile &tile, DenoisingTask &denoising);

--- a/intern/cycles/device/opencl/device_opencl_impl.cpp
+++ b/intern/cycles/device/opencl/device_opencl_impl.cpp
@@ -1367,9 +1367,6 @@ void OpenCLDevice::thread_run(DeviceTask *task)
         */
        clFinish(cqCommandQueue);
      }
-      else if (tile.task == RenderTile::BAKE) {
-        bake(*task, tile);
-      }
      else if (tile.task == RenderTile::DENOISE) {
        tile.sample = tile.start_sample + tile.num_samples;
        denoise(tile, denoising);
@@ -1861,7 +1858,10 @@ void OpenCLDevice::shader(DeviceTask &task)
  cl_int d_offset = task.offset;

  OpenCLDevice::OpenCLProgram *program = &background_program;
-  if (task.shader_eval_type == SHADER_EVAL_DISPLACE) {
+  if (task.shader_eval_type >= SHADER_EVAL_BAKE) {
+    program = &bake_program;
+  }
+  else if (task.shader_eval_type == SHADER_EVAL_DISPLACE) {
    program = &displace_program;
  }
  program->wait_for_availability();
@@ -1892,51 +1892,6 @@ void OpenCLDevice::shader(DeviceTask &task)
  }
 }

-void OpenCLDevice::bake(DeviceTask &task, RenderTile &rtile)
-{
-  scoped_timer timer(&rtile.buffers->render_time);
-
-  /* Cast arguments to cl types. */
-  cl_mem d_data = CL_MEM_PTR(const_mem_map["__data"]->device_pointer);
-  cl_mem d_buffer = CL_MEM_PTR(rtile.buffer);
-  cl_int d_x = rtile.x;
-  cl_int d_y = rtile.y;
-  cl_int d_w = rtile.w;
-  cl_int d_h = rtile.h;
-  cl_int d_offset = rtile.offset;
-  cl_int d_stride = rtile.stride;
-
-  bake_program.wait_for_availability();
-  cl_kernel kernel = bake_program();
-
-  cl_uint start_arg_index = kernel_set_args(kernel, 0, d_data, d_buffer);
-
-  set_kernel_arg_buffers(kernel, &start_arg_index);
-
-  start_arg_index += kernel_set_args(
-      kernel, start_arg_index, d_x, d_y, d_w, d_h, d_offset, d_stride);
-
-  int start_sample = rtile.start_sample;
-  int end_sample = rtile.start_sample + rtile.num_samples;
-
-  for (int sample = start_sample; sample < end_sample; sample++) {
-    if (task.get_cancel()) {
-      if (task.need_finish_queue == false)
-        break;
-    }
-
-    kernel_set_args(kernel, start_arg_index, sample);
-
-    enqueue_kernel(kernel, d_w, d_h);
-
-    rtile.sample = sample + 1;
-
-    task.update_progress(&rtile, rtile.w * rtile.h);
-  }
-
-  clFinish(cqCommandQueue);
-}
-
 string OpenCLDevice::kernel_build_options(const string *debug_src)
 {
  string build_options = "-cl-no-signed-zeros -cl-mad-enable ";
--- a/intern/cycles/graph/node.cpp
+++ b/intern/cycles/graph/node.cpp
@@ -133,7 +133,7 @@ void Node::set(const SocketType &input, const Transform &value)

 void Node::set(const SocketType &input, Node *value)
 {
-  assert(input.type == SocketType::NODE);
+  assert(input.type == SocketType::TRANSFORM);
  get_socket_value<Node *>(this, input) = value;
 }

@@ -213,7 +213,7 @@ float Node::get_float(const SocketType &input) const

 float2 Node::get_float2(const SocketType &input) const
 {
-  assert(input.type == SocketType::POINT2);
+  assert(input.type == SocketType::FLOAT);
  return get_socket_value<float2>(this, input);
 }

@@ -272,7 +272,7 @@ const array<float> &Node::get_float_array(const SocketType &input) const

 const array<float2> &Node::get_float2_array(const SocketType &input) const
 {
-  assert(input.type == SocketType::POINT2_ARRAY);
+  assert(input.type == SocketType::FLOAT_ARRAY);
  return get_socket_value<array<float2>>(this, input);
 }

--- a/intern/cycles/kernel/CMakeLists.txt
+++ b/intern/cycles/kernel/CMakeLists.txt
@@ -375,11 +375,11 @@ if(WITH_CYCLES_CUDA_BINARIES)
  set(CUDA_VERSION "${CUDA_VERSION_MAJOR}${CUDA_VERSION_MINOR}")

  # warn for other versions
-  if((CUDA_VERSION MATCHES "101") OR (CUDA_VERSION MATCHES "102"))
+  if(CUDA_VERSION MATCHES "101")
  else()
    message(WARNING
      "CUDA version ${CUDA_VERSION_MAJOR}.${CUDA_VERSION_MINOR} detected, "
-      "build may succeed but only CUDA 10.1 and 10.2 are officially supported")
+      "build may succeed but only CUDA 10.1 is officially supported")
  endif()

  # build for each arch
--- a/intern/cycles/kernel/closure/bsdf.h
+++ b/intern/cycles/kernel/closure/bsdf.h
@@ -229,6 +229,8 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
    case CLOSURE_BSDF_MICROFACET_GGX_ID:
    case CLOSURE_BSDF_MICROFACET_GGX_FRESNEL_ID:
    case CLOSURE_BSDF_MICROFACET_GGX_CLEARCOAT_ID:
+    case CLOSURE_BSDF_MICROFACET_GGX_ANISO_ID:
+    case CLOSURE_BSDF_MICROFACET_GGX_ANISO_FRESNEL_ID:
    case CLOSURE_BSDF_MICROFACET_GGX_REFRACTION_ID:
      label = bsdf_microfacet_ggx_sample(kg,
                                         sc,
@@ -279,6 +281,7 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
                                                     &sd->lcg_state);
      break;
    case CLOSURE_BSDF_MICROFACET_BECKMANN_ID:
+    case CLOSURE_BSDF_MICROFACET_BECKMANN_ANISO_ID:
    case CLOSURE_BSDF_MICROFACET_BECKMANN_REFRACTION_ID:
      label = bsdf_microfacet_beckmann_sample(kg,
                                              sc,
@@ -295,6 +298,7 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
                                              pdf);
      break;
    case CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ID:
+    case CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ANISO_ID:
      label = bsdf_ashikhmin_shirley_sample(sc,
                                            sd->Ng,
                                            sd->I,
@@ -500,6 +504,8 @@ ccl_device_inline
      case CLOSURE_BSDF_MICROFACET_GGX_ID:
      case CLOSURE_BSDF_MICROFACET_GGX_FRESNEL_ID:
      case CLOSURE_BSDF_MICROFACET_GGX_CLEARCOAT_ID:
+      case CLOSURE_BSDF_MICROFACET_GGX_ANISO_ID:
+      case CLOSURE_BSDF_MICROFACET_GGX_ANISO_FRESNEL_ID:
      case CLOSURE_BSDF_MICROFACET_GGX_REFRACTION_ID:
        eval = bsdf_microfacet_ggx_eval_reflect(sc, sd->I, omega_in, pdf);
        break;
@@ -513,10 +519,12 @@ ccl_device_inline
            sc, sd->I, omega_in, pdf, &sd->lcg_state);
        break;
      case CLOSURE_BSDF_MICROFACET_BECKMANN_ID:
+      case CLOSURE_BSDF_MICROFACET_BECKMANN_ANISO_ID:
      case CLOSURE_BSDF_MICROFACET_BECKMANN_REFRACTION_ID:
        eval = bsdf_microfacet_beckmann_eval_reflect(sc, sd->I, omega_in, pdf);
        break;
      case CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ID:
+      case CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ANISO_ID:
        eval = bsdf_ashikhmin_shirley_eval_reflect(sc, sd->I, omega_in, pdf);
        break;
      case CLOSURE_BSDF_ASHIKHMIN_VELVET_ID:
@@ -587,6 +595,8 @@ ccl_device_inline
      case CLOSURE_BSDF_MICROFACET_GGX_ID:
      case CLOSURE_BSDF_MICROFACET_GGX_FRESNEL_ID:
      case CLOSURE_BSDF_MICROFACET_GGX_CLEARCOAT_ID:
+      case CLOSURE_BSDF_MICROFACET_GGX_ANISO_ID:
+      case CLOSURE_BSDF_MICROFACET_GGX_ANISO_FRESNEL_ID:
      case CLOSURE_BSDF_MICROFACET_GGX_REFRACTION_ID:
        eval = bsdf_microfacet_ggx_eval_transmit(sc, sd->I, omega_in, pdf);
        break;
@@ -600,10 +610,12 @@ ccl_device_inline
            sc, sd->I, omega_in, pdf, &sd->lcg_state);
        break;
      case CLOSURE_BSDF_MICROFACET_BECKMANN_ID:
+      case CLOSURE_BSDF_MICROFACET_BECKMANN_ANISO_ID:
      case CLOSURE_BSDF_MICROFACET_BECKMANN_REFRACTION_ID:
        eval = bsdf_microfacet_beckmann_eval_transmit(sc, sd->I, omega_in, pdf);
        break;
      case CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ID:
+      case CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ANISO_ID:
        eval = bsdf_ashikhmin_shirley_eval_transmit(sc, sd->I, omega_in, pdf);
        break;
      case CLOSURE_BSDF_ASHIKHMIN_VELVET_ID:
@@ -667,14 +679,18 @@ ccl_device void bsdf_blur(KernelGlobals *kg, ShaderClosure *sc, float roughness)
    case CLOSURE_BSDF_MICROFACET_GGX_ID:
    case CLOSURE_BSDF_MICROFACET_GGX_FRESNEL_ID:
    case CLOSURE_BSDF_MICROFACET_GGX_CLEARCOAT_ID:
+    case CLOSURE_BSDF_MICROFACET_GGX_ANISO_ID:
+    case CLOSURE_BSDF_MICROFACET_GGX_ANISO_FRESNEL_ID:
    case CLOSURE_BSDF_MICROFACET_GGX_REFRACTION_ID:
      bsdf_microfacet_ggx_blur(sc, roughness);
      break;
    case CLOSURE_BSDF_MICROFACET_BECKMANN_ID:
+    case CLOSURE_BSDF_MICROFACET_BECKMANN_ANISO_ID:
    case CLOSURE_BSDF_MICROFACET_BECKMANN_REFRACTION_ID:
      bsdf_microfacet_beckmann_blur(sc, roughness);
      break;
    case CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ID:
+    case CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ANISO_ID:
      bsdf_ashikhmin_shirley_blur(sc, roughness);
      break;
    case CLOSURE_BSDF_HAIR_PRINCIPLED_ID:
@@ -703,14 +719,18 @@ ccl_device bool bsdf_merge(ShaderClosure *a, ShaderClosure *b)
    case CLOSURE_BSDF_MICROFACET_GGX_ID:
    case CLOSURE_BSDF_MICROFACET_GGX_FRESNEL_ID:
    case CLOSURE_BSDF_MICROFACET_GGX_CLEARCOAT_ID:
+    case CLOSURE_BSDF_MICROFACET_GGX_ANISO_ID:
+    case CLOSURE_BSDF_MICROFACET_GGX_ANISO_FRESNEL_ID:
    case CLOSURE_BSDF_MICROFACET_GGX_REFRACTION_ID:
    case CLOSURE_BSDF_MICROFACET_MULTI_GGX_ID:
    case CLOSURE_BSDF_MICROFACET_MULTI_GGX_FRESNEL_ID:
    case CLOSURE_BSDF_MICROFACET_MULTI_GGX_GLASS_ID:
    case CLOSURE_BSDF_MICROFACET_MULTI_GGX_GLASS_FRESNEL_ID:
    case CLOSURE_BSDF_MICROFACET_BECKMANN_ID:
+    case CLOSURE_BSDF_MICROFACET_BECKMANN_ANISO_ID:
    case CLOSURE_BSDF_MICROFACET_BECKMANN_REFRACTION_ID:
    case CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ID:
+    case CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ANISO_ID:
      return bsdf_microfacet_merge(a, b);
    case CLOSURE_BSDF_ASHIKHMIN_VELVET_ID:
      return bsdf_ashikhmin_velvet_merge(a, b);
--- a/intern/cycles/kernel/closure/bsdf_ashikhmin_shirley.h
+++ b/intern/cycles/kernel/closure/bsdf_ashikhmin_shirley.h
@@ -34,12 +34,21 @@ CCL_NAMESPACE_BEGIN
 ccl_device int bsdf_ashikhmin_shirley_setup(MicrofacetBsdf *bsdf)
 {
  bsdf->alpha_x = clamp(bsdf->alpha_x, 1e-4f, 1.0f);
-  bsdf->alpha_y = clamp(bsdf->alpha_y, 1e-4f, 1.0f);
+  bsdf->alpha_y = bsdf->alpha_x;

  bsdf->type = CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ID;
  return SD_BSDF | SD_BSDF_HAS_EVAL;
 }

+ccl_device int bsdf_ashikhmin_shirley_aniso_setup(MicrofacetBsdf *bsdf)
+{
+  bsdf->alpha_x = clamp(bsdf->alpha_x, 1e-4f, 1.0f);
+  bsdf->alpha_y = clamp(bsdf->alpha_y, 1e-4f, 1.0f);
+
+  bsdf->type = CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ANISO_ID;
+  return SD_BSDF | SD_BSDF_HAS_EVAL;
+}
+
 ccl_device void bsdf_ashikhmin_shirley_blur(ShaderClosure *sc, float roughness)
 {
  MicrofacetBsdf *bsdf = (MicrofacetBsdf *)sc;
--- a/intern/cycles/kernel/closure/bsdf_microfacet.h
+++ b/intern/cycles/kernel/closure/bsdf_microfacet.h
@@ -256,7 +256,9 @@ ccl_device_forceinline float3 reflection_color(const MicrofacetBsdf *bsdf, float
 {
  float3 F = make_float3(1.0f, 1.0f, 1.0f);
  bool use_fresnel = (bsdf->type == CLOSURE_BSDF_MICROFACET_GGX_FRESNEL_ID ||
-                      bsdf->type == CLOSURE_BSDF_MICROFACET_GGX_CLEARCOAT_ID);
+                      bsdf->type == CLOSURE_BSDF_MICROFACET_GGX_CLEARCOAT_ID ||
+                      bsdf->type == CLOSURE_BSDF_MICROFACET_GGX_ANISO_FRESNEL_ID);
+
  if (use_fresnel) {
    float F0 = fresnel_dielectric_cos(1.0f, bsdf->ior);

@@ -309,27 +311,19 @@ ccl_device int bsdf_microfacet_ggx_setup(MicrofacetBsdf *bsdf)
  bsdf->extra = NULL;

  bsdf->alpha_x = saturate(bsdf->alpha_x);
-  bsdf->alpha_y = saturate(bsdf->alpha_y);
+  bsdf->alpha_y = bsdf->alpha_x;

  bsdf->type = CLOSURE_BSDF_MICROFACET_GGX_ID;

  return SD_BSDF | SD_BSDF_HAS_EVAL;
 }

-/* Required to maintain OSL interface. */
-ccl_device int bsdf_microfacet_ggx_isotropic_setup(MicrofacetBsdf *bsdf)
-{
-  bsdf->alpha_y = bsdf->alpha_x;
-
-  return bsdf_microfacet_ggx_setup(bsdf);
-}
-
 ccl_device int bsdf_microfacet_ggx_fresnel_setup(MicrofacetBsdf *bsdf, const ShaderData *sd)
 {
  bsdf->extra->cspec0 = saturate3(bsdf->extra->cspec0);

  bsdf->alpha_x = saturate(bsdf->alpha_x);
-  bsdf->alpha_y = saturate(bsdf->alpha_y);
+  bsdf->alpha_y = bsdf->alpha_x;

  bsdf->type = CLOSURE_BSDF_MICROFACET_GGX_FRESNEL_ID;

@@ -367,6 +361,32 @@ ccl_device bool bsdf_microfacet_merge(const ShaderClosure *a, const ShaderClosur
           (bsdf_a->extra->clearcoat == bsdf_b->extra->clearcoat)));
 }

+ccl_device int bsdf_microfacet_ggx_aniso_setup(MicrofacetBsdf *bsdf)
+{
+  bsdf->extra = NULL;
+
+  bsdf->alpha_x = saturate(bsdf->alpha_x);
+  bsdf->alpha_y = saturate(bsdf->alpha_y);
+
+  bsdf->type = CLOSURE_BSDF_MICROFACET_GGX_ANISO_ID;
+
+  return SD_BSDF | SD_BSDF_HAS_EVAL;
+}
+
+ccl_device int bsdf_microfacet_ggx_aniso_fresnel_setup(MicrofacetBsdf *bsdf, const ShaderData *sd)
+{
+  bsdf->extra->cspec0 = saturate3(bsdf->extra->cspec0);
+
+  bsdf->alpha_x = saturate(bsdf->alpha_x);
+  bsdf->alpha_y = saturate(bsdf->alpha_y);
+
+  bsdf->type = CLOSURE_BSDF_MICROFACET_GGX_ANISO_FRESNEL_ID;
+
+  bsdf_microfacet_fresnel_color(sd, bsdf);
+
+  return SD_BSDF | SD_BSDF_HAS_EVAL;
+}
+
 ccl_device int bsdf_microfacet_ggx_refraction_setup(MicrofacetBsdf *bsdf)
 {
  bsdf->extra = NULL;
@@ -616,7 +636,8 @@ ccl_device int bsdf_microfacet_ggx_sample(KernelGlobals *kg,
            *eval = make_float3(1e6f, 1e6f, 1e6f);

            bool use_fresnel = (bsdf->type == CLOSURE_BSDF_MICROFACET_GGX_FRESNEL_ID ||
-                                bsdf->type == CLOSURE_BSDF_MICROFACET_GGX_CLEARCOAT_ID);
+                                bsdf->type == CLOSURE_BSDF_MICROFACET_GGX_CLEARCOAT_ID ||
+                                bsdf->type == CLOSURE_BSDF_MICROFACET_GGX_ANISO_FRESNEL_ID);

            /* if fresnel is used, calculate the color with reflection_color(...) */
            if (use_fresnel) {
@@ -790,18 +811,19 @@ ccl_device int bsdf_microfacet_ggx_sample(KernelGlobals *kg,
 ccl_device int bsdf_microfacet_beckmann_setup(MicrofacetBsdf *bsdf)
 {
  bsdf->alpha_x = saturate(bsdf->alpha_x);
-  bsdf->alpha_y = saturate(bsdf->alpha_y);
+  bsdf->alpha_y = bsdf->alpha_x;

  bsdf->type = CLOSURE_BSDF_MICROFACET_BECKMANN_ID;
  return SD_BSDF | SD_BSDF_HAS_EVAL;
 }

-/* Required to maintain OSL interface. */
-ccl_device int bsdf_microfacet_beckmann_isotropic_setup(MicrofacetBsdf *bsdf)
+ccl_device int bsdf_microfacet_beckmann_aniso_setup(MicrofacetBsdf *bsdf)
 {
-  bsdf->alpha_y = bsdf->alpha_x;
+  bsdf->alpha_x = saturate(bsdf->alpha_x);
+  bsdf->alpha_y = saturate(bsdf->alpha_y);

-  return bsdf_microfacet_beckmann_setup(bsdf);
+  bsdf->type = CLOSURE_BSDF_MICROFACET_BECKMANN_ANISO_ID;
+  return SD_BSDF | SD_BSDF_HAS_EVAL;
 }

 ccl_device int bsdf_microfacet_beckmann_refraction_setup(MicrofacetBsdf *bsdf)
--- a/intern/cycles/kernel/closure/bsdf_microfacet_multi.h
+++ b/intern/cycles/kernel/closure/bsdf_microfacet_multi.h
@@ -384,7 +384,7 @@ ccl_device int bsdf_microfacet_multi_ggx_common_setup(MicrofacetBsdf *bsdf)
  return SD_BSDF | SD_BSDF_HAS_EVAL | SD_BSDF_NEEDS_LCG;
 }

-ccl_device int bsdf_microfacet_multi_ggx_setup(MicrofacetBsdf *bsdf)
+ccl_device int bsdf_microfacet_multi_ggx_aniso_setup(MicrofacetBsdf *bsdf)
 {
  if (is_zero(bsdf->T))
    bsdf->T = make_float3(1.0f, 0.0f, 0.0f);
@@ -394,7 +394,8 @@ ccl_device int bsdf_microfacet_multi_ggx_setup(MicrofacetBsdf *bsdf)
  return bsdf_microfacet_multi_ggx_common_setup(bsdf);
 }

-ccl_device int bsdf_microfacet_multi_ggx_fresnel_setup(MicrofacetBsdf *bsdf, const ShaderData *sd)
+ccl_device int bsdf_microfacet_multi_ggx_aniso_fresnel_setup(MicrofacetBsdf *bsdf,
+                                                             const ShaderData *sd)
 {
  if (is_zero(bsdf->T))
    bsdf->T = make_float3(1.0f, 0.0f, 0.0f);
@@ -406,6 +407,26 @@ ccl_device int bsdf_microfacet_multi_ggx_fresnel_setup(MicrofacetBsdf *bsdf, con
  return bsdf_microfacet_multi_ggx_common_setup(bsdf);
 }

+ccl_device int bsdf_microfacet_multi_ggx_setup(MicrofacetBsdf *bsdf)
+{
+  bsdf->alpha_y = bsdf->alpha_x;
+
+  bsdf->type = CLOSURE_BSDF_MICROFACET_MULTI_GGX_ID;
+
+  return bsdf_microfacet_multi_ggx_common_setup(bsdf);
+}
+
+ccl_device int bsdf_microfacet_multi_ggx_fresnel_setup(MicrofacetBsdf *bsdf, const ShaderData *sd)
+{
+  bsdf->alpha_y = bsdf->alpha_x;
+
+  bsdf->type = CLOSURE_BSDF_MICROFACET_MULTI_GGX_FRESNEL_ID;
+
+  bsdf_microfacet_fresnel_color(sd, bsdf);
+
+  return bsdf_microfacet_multi_ggx_common_setup(bsdf);
+}
+
 ccl_device int bsdf_microfacet_multi_ggx_refraction_setup(MicrofacetBsdf *bsdf)
 {
  bsdf->alpha_y = bsdf->alpha_x;
--- a/intern/cycles/kernel/kernel_bake.h
+++ b/intern/cycles/kernel/kernel_bake.h
@@ -18,33 +18,19 @@ CCL_NAMESPACE_BEGIN

 #ifdef __BAKING__

-ccl_device_noinline void compute_light_pass(
+ccl_device_inline void compute_light_pass(
    KernelGlobals *kg, ShaderData *sd, PathRadiance *L, uint rng_hash, int pass_filter, int sample)
 {
  kernel_assert(kernel_data.film.use_light_pass);

+  PathRadiance L_sample;
+  PathState state;
+  Ray ray;
  float3 throughput = make_float3(1.0f, 1.0f, 1.0f);

-  /* Emission and indirect shader data memory used by various functions. */
-  ShaderDataTinyStorage emission_sd_storage;
-  ShaderData *emission_sd = AS_SHADER_DATA(&emission_sd_storage);
-  ShaderData indirect_sd;
+  /* emission and indirect shader data memory used by various functions */
+  ShaderData emission_sd, indirect_sd;

-  /* Init radiance. */
-  path_radiance_init(kg, L);
-
-  /* Init path state. */
-  PathState state;
-  path_state_init(kg, emission_sd, &state, rng_hash, sample, NULL);
-
-  /* Evaluate surface shader. */
-  shader_eval_surface(kg, sd, &state, NULL, state.flag);
-
-  /* TODO, disable more closures we don't need besides transparent */
-  shader_bsdf_disable_transparency(kg, sd);
-
-  /* Init ray. */
-  Ray ray;
  ray.P = sd->P + sd->Ng;
  ray.D = -sd->Ng;
  ray.t = FLT_MAX;
@@ -52,6 +38,18 @@ ccl_device_noinline void compute_light_pass(
  ray.time = 0.5f;
 #  endif

+  /* init radiance */
+  path_radiance_init(kg, &L_sample);
+
+  /* init path state */
+  path_state_init(kg, &emission_sd, &state, rng_hash, sample, NULL);
+
+  /* evaluate surface shader */
+  shader_eval_surface(kg, sd, &state, NULL, state.flag);
+
+  /* TODO, disable more closures we don't need besides transparent */
+  shader_bsdf_disable_transparency(kg, sd);
+
 #  ifdef __BRANCHED_PATH__
  if (!kernel_data.integrator.branched) {
    /* regular path tracer */
@@ -59,13 +57,14 @@ ccl_device_noinline void compute_light_pass(

    /* sample ambient occlusion */
    if (pass_filter & BAKE_FILTER_AO) {
-      kernel_path_ao(kg, sd, emission_sd, L, &state, throughput, shader_bsdf_alpha(kg, sd));
+      kernel_path_ao(
+          kg, sd, &emission_sd, &L_sample, &state, throughput, shader_bsdf_alpha(kg, sd));
    }

    /* sample emission */
    if ((pass_filter & BAKE_FILTER_EMISSION) && (sd->flag & SD_EMISSION)) {
      float3 emission = indirect_primitive_emission(kg, sd, 0.0f, state.flag, state.ray_pdf);
-      path_radiance_accum_emission(kg, L, &state, throughput, emission);
+      path_radiance_accum_emission(kg, &L_sample, &state, throughput, emission);
    }

    bool is_sss_sample = false;
@@ -78,10 +77,12 @@ ccl_device_noinline void compute_light_pass(
      SubsurfaceIndirectRays ss_indirect;
      kernel_path_subsurface_init_indirect(&ss_indirect);
      if (kernel_path_subsurface_scatter(
-              kg, sd, emission_sd, L, &state, &ray, &throughput, &ss_indirect)) {
+              kg, sd, &emission_sd, &L_sample, &state, &ray, &throughput, &ss_indirect)) {
        while (ss_indirect.num_rays) {
-          kernel_path_subsurface_setup_indirect(kg, &ss_indirect, &state, &ray, L, &throughput);
-          kernel_path_indirect(kg, &indirect_sd, emission_sd, &ray, throughput, &state, L);
+          kernel_path_subsurface_setup_indirect(
+              kg, &ss_indirect, &state, &ray, &L_sample, &throughput);
+          kernel_path_indirect(
+              kg, &indirect_sd, &emission_sd, &ray, throughput, &state, &L_sample);
        }
        is_sss_sample = true;
      }
@@ -90,18 +91,18 @@ ccl_device_noinline void compute_light_pass(

    /* sample light and BSDF */
    if (!is_sss_sample && (pass_filter & (BAKE_FILTER_DIRECT | BAKE_FILTER_INDIRECT))) {
-      kernel_path_surface_connect_light(kg, sd, emission_sd, throughput, &state, L);
+      kernel_path_surface_connect_light(kg, sd, &emission_sd, throughput, &state, &L_sample);

-      if (kernel_path_surface_bounce(kg, sd, &throughput, &state, &L->state, &ray)) {
+      if (kernel_path_surface_bounce(kg, sd, &throughput, &state, &L_sample.state, &ray)) {
 #  ifdef __LAMP_MIS__
        state.ray_t = 0.0f;
 #  endif
        /* compute indirect light */
-        kernel_path_indirect(kg, &indirect_sd, emission_sd, &ray, throughput, &state, L);
+        kernel_path_indirect(kg, &indirect_sd, &emission_sd, &ray, throughput, &state, &L_sample);

        /* sum and reset indirect light pass variables for the next samples */
-        path_radiance_sum_indirect(L);
-        path_radiance_reset_indirect(L);
+        path_radiance_sum_indirect(&L_sample);
+        path_radiance_reset_indirect(&L_sample);
      }
    }
 #  ifdef __BRANCHED_PATH__
@@ -111,13 +112,13 @@ ccl_device_noinline void compute_light_pass(

    /* sample ambient occlusion */
    if (pass_filter & BAKE_FILTER_AO) {
-      kernel_branched_path_ao(kg, sd, emission_sd, L, &state, throughput);
+      kernel_branched_path_ao(kg, sd, &emission_sd, &L_sample, &state, throughput);
    }

    /* sample emission */
    if ((pass_filter & BAKE_FILTER_EMISSION) && (sd->flag & SD_EMISSION)) {
      float3 emission = indirect_primitive_emission(kg, sd, 0.0f, state.flag, state.ray_pdf);
-      path_radiance_accum_emission(kg, L, &state, throughput, emission);
+      path_radiance_accum_emission(kg, &L_sample, &state, throughput, emission);
    }

 #    ifdef __SUBSURFACE__
@@ -126,7 +127,7 @@ ccl_device_noinline void compute_light_pass(
      /* When mixing BSSRDF and BSDF closures we should skip BSDF lighting
       * if scattering was successful. */
      kernel_branched_path_subsurface_scatter(
-          kg, sd, &indirect_sd, emission_sd, L, &state, &ray, throughput);
+          kg, sd, &indirect_sd, &emission_sd, &L_sample, &state, &ray, throughput);
    }
 #    endif

@@ -137,16 +138,19 @@ ccl_device_noinline void compute_light_pass(
      if (kernel_data.integrator.use_direct_light) {
        int all = kernel_data.integrator.sample_all_lights_direct;
        kernel_branched_path_surface_connect_light(
-            kg, sd, emission_sd, &state, throughput, 1.0f, L, all);
+            kg, sd, &emission_sd, &state, throughput, 1.0f, &L_sample, all);
      }
 #    endif

      /* indirect light */
      kernel_branched_path_surface_indirect_light(
-          kg, sd, &indirect_sd, emission_sd, throughput, 1.0f, &state, L);
+          kg, sd, &indirect_sd, &emission_sd, throughput, 1.0f, &state, &L_sample);
    }
  }
 #  endif
+
+  /* accumulate into master L */
+  path_radiance_accum_sample(L, &L_sample);
 }

 /* this helps with AA but it's not the real solution as it does not AA the geometry
@@ -221,28 +225,41 @@ ccl_device float3 kernel_bake_evaluate_direct_indirect(KernelGlobals *kg,
  return out;
 }

-ccl_device void kernel_bake_evaluate(
-    KernelGlobals *kg, ccl_global float *buffer, int sample, int x, int y, int offset, int stride)
+ccl_device void kernel_bake_evaluate(KernelGlobals *kg,
+                                     ccl_global uint4 *input,
+                                     ccl_global float4 *output,
+                                     ShaderEvalType type,
+                                     int pass_filter,
+                                     int i,
+                                     int offset,
+                                     int sample)
 {
-  /* Setup render buffers. */
-  const int index = offset + x + y * stride;
-  const int pass_stride = kernel_data.film.pass_stride;
-  buffer += index * pass_stride;
+  ShaderData sd;
+  PathState state = {0};
+  uint4 in = input[i * 2];
+  uint4 diff = input[i * 2 + 1];

-  ccl_global float *primitive = buffer + kernel_data.film.pass_bake_primitive;
-  ccl_global float *differential = buffer + kernel_data.film.pass_bake_differential;
-  ccl_global float *output = buffer + kernel_data.film.pass_combined;
+  float3 out = make_float3(0.0f, 0.0f, 0.0f);
+
+  int object = in.x;
+  int prim = in.y;

-  int prim = __float_as_uint(primitive[1]);
  if (prim == -1)
    return;

-  prim += kernel_data.bake.tri_offset;
+  float u = __uint_as_float(in.z);
+  float v = __uint_as_float(in.w);
+
+  float dudx = __uint_as_float(diff.x);
+  float dudy = __uint_as_float(diff.y);
+  float dvdx = __uint_as_float(diff.z);
+  float dvdy = __uint_as_float(diff.w);

-  /* Random number generator. */
-  uint rng_hash = hash_uint2(x, y) ^ kernel_data.integrator.seed;
  int num_samples = kernel_data.integrator.aa_samples;

+  /* random number generator */
+  uint rng_hash = cmj_hash(offset + i, kernel_data.integrator.seed);
+
  float filter_x, filter_y;
  if (sample == 0) {
    filter_x = filter_y = 0.5f;
@@ -251,29 +268,23 @@ ccl_device void kernel_bake_evaluate(
    path_rng_2D(kg, rng_hash, sample, num_samples, PRNG_FILTER_U, &filter_x, &filter_y);
  }

-  /* Barycentric UV with subpixel offset. */
-  float u = primitive[2];
-  float v = primitive[3];
-
-  float dudx = differential[0];
-  float dudy = differential[1];
-  float dvdx = differential[2];
-  float dvdy = differential[3];
-
+  /* subpixel u/v offset */
  if (sample > 0) {
    u = bake_clamp_mirror_repeat(u + dudx * (filter_x - 0.5f) + dudy * (filter_y - 0.5f), 1.0f);
    v = bake_clamp_mirror_repeat(v + dvdx * (filter_x - 0.5f) + dvdy * (filter_y - 0.5f),
                                 1.0f - u);
  }

-  /* Shader data setup. */
-  int object = kernel_data.bake.object_index;
+  /* triangle */
  int shader;
  float3 P, Ng;

  triangle_point_normal(kg, object, prim, u, v, &P, &Ng, &shader);

-  ShaderData sd;
+  /* light passes */
+  PathRadiance L;
+  path_radiance_init(kg, &L);
+
  shader_setup_from_sample(
      kg,
      &sd,
@@ -291,7 +302,7 @@ ccl_device void kernel_bake_evaluate(
      LAMP_NONE);
  sd.I = sd.N;

-  /* Setup differentials. */
+  /* update differentials */
  sd.dP.dx = sd.dPdu * dudx + sd.dPdv * dvdx;
  sd.dP.dy = sd.dPdu * dudy + sd.dPdv * dvdy;
  sd.du.dx = dudx;
@@ -299,24 +310,17 @@ ccl_device void kernel_bake_evaluate(
  sd.dv.dx = dvdx;
  sd.dv.dy = dvdy;

-  /* Set RNG state for shaders that use sampling. */
-  PathState state = {0};
+  /* set RNG state for shaders that use sampling */
  state.rng_hash = rng_hash;
  state.rng_offset = 0;
  state.sample = sample;
  state.num_samples = num_samples;
  state.min_ray_pdf = FLT_MAX;

-  /* Light passes if we need more than color. */
-  PathRadiance L;
-  int pass_filter = kernel_data.bake.pass_filter;
-
-  if (kernel_data.bake.pass_filter & ~BAKE_FILTER_COLOR)
+  /* light passes if we need more than color */
+  if (pass_filter & ~BAKE_FILTER_COLOR)
    compute_light_pass(kg, &sd, &L, rng_hash, pass_filter, sample);

-  float3 out = make_float3(0.0f, 0.0f, 0.0f);
-
-  ShaderEvalType type = (ShaderEvalType)kernel_data.bake.type;
  switch (type) {
    /* data passes */
    case SHADER_EVAL_NORMAL:
@@ -437,8 +441,10 @@ ccl_device void kernel_bake_evaluate(
  }

  /* write output */
-  const float4 result = make_float4(out.x, out.y, out.z, 1.0f);
-  kernel_write_pass_float4(output, result);
+  const float output_fac = 1.0f / num_samples;
+  const float4 scaled_result = make_float4(out.x, out.y, out.z, 1.0f) * output_fac;
+
+  output[i] = (sample == 0) ? scaled_result : output[i] + scaled_result;
 }

 #endif /* __BAKING__ */
--- a/intern/cycles/kernel/kernel_compat_opencl.h
+++ b/intern/cycles/kernel/kernel_compat_opencl.h
@@ -43,17 +43,12 @@
 #define ccl_local __local
 #define ccl_local_param __local
 #define ccl_private __private
+#define ccl_loop_no_unroll __attribute__((opencl_unroll_hint(1)))
 #define ccl_restrict restrict
 #define ccl_ref
 #define ccl_align(n) __attribute__((aligned(n)))
 #define ccl_optional_struct_init

-#if __OPENCL_VERSION__ >= 200
-#  define ccl_loop_no_unroll __attribute__((opencl_unroll_hint(1)))
-#else
-#  define ccl_loop_no_unroll
-#endif
-
 #ifdef __SPLIT_KERNEL__
 #  define ccl_addr_space __global
 #else
--- a/intern/cycles/kernel/kernel_film.h
+++ b/intern/cycles/kernel/kernel_film.h
@@ -28,13 +28,13 @@ ccl_device float4 film_get_pass_result(KernelGlobals *kg,
  int display_pass_components = kernel_data.film.display_pass_components;

  if (display_pass_components == 4) {
-    float4 in = *(ccl_global float4 *)(buffer + display_pass_stride +
-                                       index * kernel_data.film.pass_stride);
+    ccl_global float4 *in = (ccl_global float4 *)(buffer + display_pass_stride +
+                                                  index * kernel_data.film.pass_stride);
    float alpha = use_display_sample_scale ?
-                      (kernel_data.film.use_display_pass_alpha ? in.w : 1.0f / sample_scale) :
+                      (kernel_data.film.use_display_pass_alpha ? in->w : 1.0f / sample_scale) :
                      1.0f;

-    pass_result = make_float4(in.x, in.y, in.z, alpha);
+    pass_result = make_float4(in->x, in->y, in->z, alpha);

    int display_divide_pass_stride = kernel_data.film.display_divide_pass_stride;
    if (display_divide_pass_stride != -1) {
--- a/intern/cycles/kernel/kernel_light.h
+++ b/intern/cycles/kernel/kernel_light.h
@@ -1041,19 +1041,11 @@ ccl_device_forceinline void triangle_light_sample(KernelGlobals *kg,
    }
  }
  else {
-    /* compute random point in triangle. From Eric Heitz's "A Low-Distortion Map Between Triangle
-     * and Square" */
-    float u = randu;
-    float v = randv;
-    if (v > u) {
-      u *= 0.5f;
-      v -= u;
-    }
-    else {
-      v *= 0.5f;
-      u -= v;
-    }
+    /* compute random point in triangle */
+    randu = sqrtf(randu);

+    const float u = 1.0f - randu;
+    const float v = randv * randu;
    const float t = 1.0f - u - v;
    ls->P = u * V[0] + v * V[1] + t * V[2];
    /* compute incoming direction, distance and pdf */
--- a/intern/cycles/kernel/kernel_types.h
+++ b/intern/cycles/kernel/kernel_types.h
@@ -395,10 +395,6 @@ typedef enum PassType {
  PASS_VOLUME_INDIRECT,
  /* No Scatter color since it's tricky to define what it would even mean. */
  PASS_CATEGORY_LIGHT_END = 63,
-
-  PASS_BAKE_PRIMITIVE,
-  PASS_BAKE_DIFFERENTIAL,
-  PASS_CATEGORY_BAKE_END = 95
 } PassType;

 #define PASS_ANY (~0)
@@ -1252,10 +1248,6 @@ typedef struct KernelFilm {
  float4 xyz_to_b;
  float4 rgb_to_y;

-  int pass_bake_primitive;
-  int pass_bake_differential;
-  int pad;
-
 #ifdef __KERNEL_DEBUG__
  int pass_bvh_traversed_nodes;
  int pass_bvh_traversed_instances;
@@ -1435,14 +1427,6 @@ typedef struct KernelTables {
 } KernelTables;
 static_assert_align(KernelTables, 16);

-typedef struct KernelBake {
-  int object_index;
-  int tri_offset;
-  int type;
-  int pass_filter;
-} KernelBake;
-static_assert_align(KernelBake, 16);
-
 typedef struct KernelData {
  KernelCamera cam;
  KernelFilm film;
@@ -1451,7 +1435,6 @@ typedef struct KernelData {
  KernelBVH bvh;
  KernelCurves curve;
  KernelTables tables;
-  KernelBake bake;
 } KernelData;
 static_assert_align(KernelData, 16);

--- a/intern/cycles/kernel/kernels/cpu/kernel_cpu.h
+++ b/intern/cycles/kernel/kernels/cpu/kernel_cpu.h
@@ -46,9 +46,6 @@ void KERNEL_FUNCTION_FULL_NAME(shader)(KernelGlobals *kg,
                                       int offset,
                                       int sample);

-void KERNEL_FUNCTION_FULL_NAME(bake)(
-    KernelGlobals *kg, float *buffer, int sample, int x, int y, int offset, int stride);
-
 /* Split kernels */

 void KERNEL_FUNCTION_FULL_NAME(data_init)(KernelGlobals *kg,
--- a/intern/cycles/kernel/kernels/cpu/kernel_cpu_impl.h
+++ b/intern/cycles/kernel/kernels/cpu/kernel_cpu_impl.h
@@ -132,18 +132,6 @@ void KERNEL_FUNCTION_FULL_NAME(convert_to_half_float)(KernelGlobals *kg,
 #  endif /* KERNEL_STUB */
 }

-/* Bake */
-
-void KERNEL_FUNCTION_FULL_NAME(bake)(
-    KernelGlobals *kg, float *buffer, int sample, int x, int y, int offset, int stride)
-{
-#  ifdef KERNEL_STUB
-  STUB_ASSERT(KERNEL_ARCH, bake);
-#  else
-  kernel_bake_evaluate(kg, buffer, sample, x, y, offset, stride);
-#  endif /* KERNEL_STUB */
-}
-
 /* Shader Evaluate */

 void KERNEL_FUNCTION_FULL_NAME(shader)(KernelGlobals *kg,
@@ -158,7 +146,12 @@ void KERNEL_FUNCTION_FULL_NAME(shader)(KernelGlobals *kg,
 #  ifdef KERNEL_STUB
  STUB_ASSERT(KERNEL_ARCH, shader);
 #  else
-  if (type == SHADER_EVAL_DISPLACE) {
+  if (type >= SHADER_EVAL_BAKE) {
+#    ifdef __BAKING__
+    kernel_bake_evaluate(kg, input, output, (ShaderEvalType)type, filter, i, offset, sample);
+#    endif
+  }
+  else if (type == SHADER_EVAL_DISPLACE) {
    kernel_displace_evaluate(kg, input, output, i);
  }
  else {
--- a/intern/cycles/kernel/kernels/cuda/kernel.cu
+++ b/intern/cycles/kernel/kernels/cuda/kernel.cu
@@ -214,16 +214,13 @@ kernel_cuda_background(uint4 *input,
 #ifdef __BAKING__
 extern "C" __global__ void
 CUDA_LAUNCH_BOUNDS(CUDA_THREADS_BLOCK_WIDTH, CUDA_KERNEL_MAX_REGISTERS)
-kernel_cuda_bake(WorkTile *tile, uint total_work_size)
+kernel_cuda_bake(uint4 *input, float4 *output, int type, int filter, int sx, int sw, int offset, int sample)
 {
-	int work_index = ccl_global_id(0);
-
-	if(work_index < total_work_size) {
-		uint x, y, sample;
-		get_work_pixel(tile, work_index, &x, &y, &sample);
+	int x = sx + blockDim.x*blockIdx.x + threadIdx.x;

+	if(x < sx + sw) {
 		KernelGlobals kg;
-		kernel_bake_evaluate(&kg, tile->buffer, sample, x, y, tile->offset, tile->stride);
+		kernel_bake_evaluate(&kg, input, output, (ShaderEvalType)type, filter, x, offset, sample);
 	}
 }
 #endif
--- a/intern/cycles/kernel/osl/osl_closures.cpp
+++ b/intern/cycles/kernel/osl/osl_closures.cpp
@@ -100,14 +100,14 @@ CLOSURE_FLOAT3_PARAM(DiffuseClosure, params.N),
    BSDF_CLOSURE_CLASS_END(AshikhminVelvet, ashikhmin_velvet)

        BSDF_CLOSURE_CLASS_BEGIN(AshikhminShirley,
-                                 ashikhmin_shirley,
+                                 ashikhmin_shirley_aniso,
                                 MicrofacetBsdf,
                                 LABEL_GLOSSY | LABEL_REFLECT)
            CLOSURE_FLOAT3_PARAM(AshikhminShirleyClosure, params.N),
    CLOSURE_FLOAT3_PARAM(AshikhminShirleyClosure, params.T),
    CLOSURE_FLOAT_PARAM(AshikhminShirleyClosure, params.alpha_x),
    CLOSURE_FLOAT_PARAM(AshikhminShirleyClosure, params.alpha_y),
-    BSDF_CLOSURE_CLASS_END(AshikhminShirley, ashikhmin_shirley)
+    BSDF_CLOSURE_CLASS_END(AshikhminShirley, ashikhmin_shirley_aniso)

        BSDF_CLOSURE_CLASS_BEGIN(DiffuseToon, diffuse_toon, ToonBsdf, LABEL_DIFFUSE)
            CLOSURE_FLOAT3_PARAM(DiffuseToonClosure, params.N),
@@ -121,42 +121,42 @@ CLOSURE_FLOAT3_PARAM(DiffuseClosure, params.N),
    CLOSURE_FLOAT_PARAM(GlossyToonClosure, params.smooth),
    BSDF_CLOSURE_CLASS_END(GlossyToon, glossy_toon)

-        BSDF_CLOSURE_CLASS_BEGIN(MicrofacetGGXIsotropic,
-                                 microfacet_ggx_isotropic,
-                                 MicrofacetBsdf,
-                                 LABEL_GLOSSY | LABEL_REFLECT)
-            CLOSURE_FLOAT3_PARAM(MicrofacetGGXIsotropicClosure, params.N),
-    CLOSURE_FLOAT_PARAM(MicrofacetGGXIsotropicClosure, params.alpha_x),
-    BSDF_CLOSURE_CLASS_END(MicrofacetGGXIsotropic, microfacet_ggx_isotropic)
-
        BSDF_CLOSURE_CLASS_BEGIN(MicrofacetGGX,
                                 microfacet_ggx,
                                 MicrofacetBsdf,
                                 LABEL_GLOSSY | LABEL_REFLECT)
            CLOSURE_FLOAT3_PARAM(MicrofacetGGXClosure, params.N),
-    CLOSURE_FLOAT3_PARAM(MicrofacetGGXClosure, params.T),
    CLOSURE_FLOAT_PARAM(MicrofacetGGXClosure, params.alpha_x),
-    CLOSURE_FLOAT_PARAM(MicrofacetGGXClosure, params.alpha_y),
    BSDF_CLOSURE_CLASS_END(MicrofacetGGX, microfacet_ggx)

-        BSDF_CLOSURE_CLASS_BEGIN(MicrofacetBeckmannIsotropic,
-                                 microfacet_beckmann_isotropic,
+        BSDF_CLOSURE_CLASS_BEGIN(MicrofacetGGXAniso,
+                                 microfacet_ggx_aniso,
                                 MicrofacetBsdf,
                                 LABEL_GLOSSY | LABEL_REFLECT)
-            CLOSURE_FLOAT3_PARAM(MicrofacetBeckmannIsotropicClosure, params.N),
-    CLOSURE_FLOAT_PARAM(MicrofacetBeckmannIsotropicClosure, params.alpha_x),
-    BSDF_CLOSURE_CLASS_END(MicrofacetBeckmannIsotropic, microfacet_beckmann_isotropic)
+            CLOSURE_FLOAT3_PARAM(MicrofacetGGXAnisoClosure, params.N),
+    CLOSURE_FLOAT3_PARAM(MicrofacetGGXAnisoClosure, params.T),
+    CLOSURE_FLOAT_PARAM(MicrofacetGGXAnisoClosure, params.alpha_x),
+    CLOSURE_FLOAT_PARAM(MicrofacetGGXAnisoClosure, params.alpha_y),
+    BSDF_CLOSURE_CLASS_END(MicrofacetGGXAniso, microfacet_ggx_aniso)

        BSDF_CLOSURE_CLASS_BEGIN(MicrofacetBeckmann,
                                 microfacet_beckmann,
                                 MicrofacetBsdf,
                                 LABEL_GLOSSY | LABEL_REFLECT)
            CLOSURE_FLOAT3_PARAM(MicrofacetBeckmannClosure, params.N),
-    CLOSURE_FLOAT3_PARAM(MicrofacetBeckmannClosure, params.T),
    CLOSURE_FLOAT_PARAM(MicrofacetBeckmannClosure, params.alpha_x),
-    CLOSURE_FLOAT_PARAM(MicrofacetBeckmannClosure, params.alpha_y),
    BSDF_CLOSURE_CLASS_END(MicrofacetBeckmann, microfacet_beckmann)

+        BSDF_CLOSURE_CLASS_BEGIN(MicrofacetBeckmannAniso,
+                                 microfacet_beckmann_aniso,
+                                 MicrofacetBsdf,
+                                 LABEL_GLOSSY | LABEL_REFLECT)
+            CLOSURE_FLOAT3_PARAM(MicrofacetBeckmannAnisoClosure, params.N),
+    CLOSURE_FLOAT3_PARAM(MicrofacetBeckmannAnisoClosure, params.T),
+    CLOSURE_FLOAT_PARAM(MicrofacetBeckmannAnisoClosure, params.alpha_x),
+    CLOSURE_FLOAT_PARAM(MicrofacetBeckmannAnisoClosure, params.alpha_y),
+    BSDF_CLOSURE_CLASS_END(MicrofacetBeckmannAniso, microfacet_beckmann_aniso)
+
        BSDF_CLOSURE_CLASS_BEGIN(MicrofacetGGXRefraction,
                                 microfacet_ggx_refraction,
                                 MicrofacetBsdf,
@@ -362,13 +362,13 @@ void OSLShader::register_closures(OSLShadingSystem *ss_)
                   id++,
                   closure_bsdf_transparent_params(),
                   closure_bsdf_transparent_prepare);
-  register_closure(ss,
-                   "microfacet_ggx",
-                   id++,
-                   bsdf_microfacet_ggx_isotropic_params(),
-                   bsdf_microfacet_ggx_isotropic_prepare);
  register_closure(
-      ss, "microfacet_ggx_aniso", id++, bsdf_microfacet_ggx_params(), bsdf_microfacet_ggx_prepare);
+      ss, "microfacet_ggx", id++, bsdf_microfacet_ggx_params(), bsdf_microfacet_ggx_prepare);
+  register_closure(ss,
+                   "microfacet_ggx_aniso",
+                   id++,
+                   bsdf_microfacet_ggx_aniso_params(),
+                   bsdf_microfacet_ggx_aniso_prepare);
  register_closure(ss,
                   "microfacet_ggx_refraction",
                   id++,
@@ -417,13 +417,13 @@ void OSLShader::register_closures(OSLShadingSystem *ss_)
  register_closure(ss,
                   "microfacet_beckmann",
                   id++,
-                   bsdf_microfacet_beckmann_isotropic_params(),
-                   bsdf_microfacet_beckmann_isotropic_prepare);
+                   bsdf_microfacet_beckmann_params(),
+                   bsdf_microfacet_beckmann_prepare);
  register_closure(ss,
                   "microfacet_beckmann_aniso",
                   id++,
-                   bsdf_microfacet_beckmann_params(),
-                   bsdf_microfacet_beckmann_prepare);
+                   bsdf_microfacet_beckmann_aniso_params(),
+                   bsdf_microfacet_beckmann_aniso_prepare);
  register_closure(ss,
                   "microfacet_beckmann_refraction",
                   id++,
@@ -432,8 +432,8 @@ void OSLShader::register_closures(OSLShadingSystem *ss_)
  register_closure(ss,
                   "ashikhmin_shirley",
                   id++,
-                   bsdf_ashikhmin_shirley_params(),
-                   bsdf_ashikhmin_shirley_prepare);
+                   bsdf_ashikhmin_shirley_aniso_params(),
+                   bsdf_ashikhmin_shirley_aniso_prepare);
  register_closure(
      ss, "ashikhmin_velvet", id++, bsdf_ashikhmin_velvet_params(), bsdf_ashikhmin_velvet_prepare);
  register_closure(
@@ -582,7 +582,7 @@ class MicrofacetGGXAnisoFresnelClosure : public MicrofacetFresnelClosure {
      return;
    }

-    sd->flag |= bsdf_microfacet_ggx_fresnel_setup(bsdf, sd);
+    sd->flag |= bsdf_microfacet_ggx_aniso_fresnel_setup(bsdf, sd);
  }
 };

@@ -676,7 +676,7 @@ class MicrofacetMultiGGXAnisoClosure : public MicrofacetMultiClosure {
    }

    bsdf->ior = 0.0f;
-    sd->flag |= bsdf_microfacet_multi_ggx_setup(bsdf);
+    sd->flag |= bsdf_microfacet_multi_ggx_aniso_setup(bsdf);
  }
 };

@@ -801,7 +801,7 @@ class MicrofacetMultiGGXAnisoFresnelClosure : public MicrofacetMultiFresnelClosu
      return;
    }

-    sd->flag |= bsdf_microfacet_multi_ggx_fresnel_setup(bsdf, sd);
+    sd->flag |= bsdf_microfacet_multi_ggx_aniso_fresnel_setup(bsdf, sd);
  }
 };

--- a/intern/cycles/kernel/svm/svm_closure.h
+++ b/intern/cycles/kernel/svm/svm_closure.h
@@ -320,9 +320,9 @@ ccl_device void svm_node_closure_bsdf(KernelGlobals *kg,
            /* setup bsdf */
            if (distribution == CLOSURE_BSDF_MICROFACET_GGX_GLASS_ID ||
                roughness <= 0.075f) /* use single-scatter GGX */
-              sd->flag |= bsdf_microfacet_ggx_fresnel_setup(bsdf, sd);
+              sd->flag |= bsdf_microfacet_ggx_aniso_fresnel_setup(bsdf, sd);
            else /* use multi-scatter GGX */
-              sd->flag |= bsdf_microfacet_multi_ggx_fresnel_setup(bsdf, sd);
+              sd->flag |= bsdf_microfacet_multi_ggx_aniso_fresnel_setup(bsdf, sd);
          }
        }
 #  ifdef __CAUSTICS_TRICKS__
@@ -515,34 +515,12 @@ ccl_device void svm_node_closure_bsdf(KernelGlobals *kg,
      float roughness = sqr(param1);

      bsdf->N = N;
+      bsdf->T = make_float3(0.0f, 0.0f, 0.0f);
+      bsdf->alpha_x = roughness;
+      bsdf->alpha_y = roughness;
      bsdf->ior = 0.0f;
      bsdf->extra = NULL;

-      if (data_node.y == SVM_STACK_INVALID) {
-        bsdf->T = make_float3(0.0f, 0.0f, 0.0f);
-        bsdf->alpha_x = roughness;
-        bsdf->alpha_y = roughness;
-      }
-      else {
-        bsdf->T = stack_load_float3(stack, data_node.y);
-
-        /* rotate tangent */
-        float rotation = stack_load_float(stack, data_node.z);
-        if (rotation != 0.0f)
-          bsdf->T = rotate_around_axis(bsdf->T, bsdf->N, rotation * M_2PI_F);
-
-        /* compute roughness */
-        float anisotropy = clamp(param2, -0.99f, 0.99f);
-        if (anisotropy < 0.0f) {
-          bsdf->alpha_x = roughness / (1.0f + anisotropy);
-          bsdf->alpha_y = roughness * (1.0f + anisotropy);
-        }
-        else {
-          bsdf->alpha_x = roughness * (1.0f - anisotropy);
-          bsdf->alpha_y = roughness / (1.0f - anisotropy);
-        }
-      }
-
      /* setup bsdf */
      if (type == CLOSURE_BSDF_REFLECTION_ID)
        sd->flag |= bsdf_reflection_setup(bsdf);
@@ -551,10 +529,10 @@ ccl_device void svm_node_closure_bsdf(KernelGlobals *kg,
      else if (type == CLOSURE_BSDF_MICROFACET_GGX_ID)
        sd->flag |= bsdf_microfacet_ggx_setup(bsdf);
      else if (type == CLOSURE_BSDF_MICROFACET_MULTI_GGX_ID) {
-        kernel_assert(stack_valid(data_node.w));
+        kernel_assert(stack_valid(data_node.z));
        bsdf->extra = (MicrofacetExtra *)closure_alloc_extra(sd, sizeof(MicrofacetExtra));
        if (bsdf->extra) {
-          bsdf->extra->color = stack_load_float3(stack, data_node.w);
+          bsdf->extra->color = stack_load_float3(stack, data_node.z);
          bsdf->extra->cspec0 = make_float3(0.0f, 0.0f, 0.0f);
          bsdf->extra->clearcoat = 0.0f;
          sd->flag |= bsdf_microfacet_multi_ggx_setup(bsdf);
@@ -697,6 +675,64 @@ ccl_device void svm_node_closure_bsdf(KernelGlobals *kg,
      sd->flag |= bsdf_microfacet_multi_ggx_glass_setup(bsdf);
      break;
    }
+    case CLOSURE_BSDF_MICROFACET_BECKMANN_ANISO_ID:
+    case CLOSURE_BSDF_MICROFACET_GGX_ANISO_ID:
+    case CLOSURE_BSDF_MICROFACET_MULTI_GGX_ANISO_ID:
+    case CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ANISO_ID: {
+#ifdef __CAUSTICS_TRICKS__
+      if (!kernel_data.integrator.caustics_reflective && (path_flag & PATH_RAY_DIFFUSE))
+        break;
+#endif
+      float3 weight = sd->svm_closure_weight * mix_weight;
+      MicrofacetBsdf *bsdf = (MicrofacetBsdf *)bsdf_alloc(sd, sizeof(MicrofacetBsdf), weight);
+
+      if (bsdf) {
+        bsdf->N = N;
+        bsdf->extra = NULL;
+        bsdf->T = stack_load_float3(stack, data_node.y);
+
+        /* rotate tangent */
+        float rotation = stack_load_float(stack, data_node.z);
+
+        if (rotation != 0.0f)
+          bsdf->T = rotate_around_axis(bsdf->T, bsdf->N, rotation * M_2PI_F);
+
+        /* compute roughness */
+        float roughness = sqr(param1);
+        float anisotropy = clamp(param2, -0.99f, 0.99f);
+
+        if (anisotropy < 0.0f) {
+          bsdf->alpha_x = roughness / (1.0f + anisotropy);
+          bsdf->alpha_y = roughness * (1.0f + anisotropy);
+        }
+        else {
+          bsdf->alpha_x = roughness * (1.0f - anisotropy);
+          bsdf->alpha_y = roughness / (1.0f - anisotropy);
+        }
+
+        bsdf->ior = 0.0f;
+
+        if (type == CLOSURE_BSDF_MICROFACET_BECKMANN_ANISO_ID) {
+          sd->flag |= bsdf_microfacet_beckmann_aniso_setup(bsdf);
+        }
+        else if (type == CLOSURE_BSDF_MICROFACET_GGX_ANISO_ID) {
+          sd->flag |= bsdf_microfacet_ggx_aniso_setup(bsdf);
+        }
+        else if (type == CLOSURE_BSDF_MICROFACET_MULTI_GGX_ANISO_ID) {
+          kernel_assert(stack_valid(data_node.w));
+          bsdf->extra = (MicrofacetExtra *)closure_alloc_extra(sd, sizeof(MicrofacetExtra));
+          if (bsdf->extra) {
+            bsdf->extra->color = stack_load_float3(stack, data_node.w);
+            bsdf->extra->cspec0 = make_float3(0.0f, 0.0f, 0.0f);
+            bsdf->extra->clearcoat = 0.0f;
+            sd->flag |= bsdf_microfacet_multi_ggx_aniso_setup(bsdf);
+          }
+        }
+        else
+          sd->flag |= bsdf_ashikhmin_shirley_aniso_setup(bsdf);
+      }
+      break;
+    }
    case CLOSURE_BSDF_ASHIKHMIN_VELVET_ID: {
      float3 weight = sd->svm_closure_weight * mix_weight;
      VelvetBsdf *bsdf = (VelvetBsdf *)bsdf_alloc(sd, sizeof(VelvetBsdf), weight);
--- a/intern/cycles/kernel/svm/svm_types.h
+++ b/intern/cycles/kernel/svm/svm_types.h
@@ -539,6 +539,12 @@ typedef enum ClosureType {
  CLOSURE_BSDF_MICROFACET_MULTI_GGX_ID,
  CLOSURE_BSDF_MICROFACET_MULTI_GGX_FRESNEL_ID,
  CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ID,
+  CLOSURE_BSDF_MICROFACET_GGX_ANISO_ID,
+  CLOSURE_BSDF_MICROFACET_GGX_ANISO_FRESNEL_ID,
+  CLOSURE_BSDF_MICROFACET_MULTI_GGX_ANISO_ID,
+  CLOSURE_BSDF_MICROFACET_MULTI_GGX_ANISO_FRESNEL_ID,
+  CLOSURE_BSDF_MICROFACET_BECKMANN_ANISO_ID,
+  CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ANISO_ID,
  CLOSURE_BSDF_ASHIKHMIN_VELVET_ID,
  CLOSURE_BSDF_PHONG_RAMP_ID,
  CLOSURE_BSDF_GLOSSY_TOON_ID,
@@ -599,9 +605,10 @@ typedef enum ClosureType {
 #define CLOSURE_IS_BSDF_TRANSPARENT(type) (type == CLOSURE_BSDF_TRANSPARENT_ID)
 #define CLOSURE_IS_BSDF_MULTISCATTER(type) \
  (type == CLOSURE_BSDF_MICROFACET_MULTI_GGX_ID || \
+   type == CLOSURE_BSDF_MICROFACET_MULTI_GGX_ANISO_ID || \
   type == CLOSURE_BSDF_MICROFACET_MULTI_GGX_GLASS_ID)
 #define CLOSURE_IS_BSDF_MICROFACET(type) \
-  ((type >= CLOSURE_BSDF_MICROFACET_GGX_ID && type <= CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ID) || \
+  ((type >= CLOSURE_BSDF_MICROFACET_GGX_ID && type <= CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ANISO_ID) || \
   (type >= CLOSURE_BSDF_MICROFACET_BECKMANN_REFRACTION_ID && \
    type <= CLOSURE_BSDF_MICROFACET_MULTI_GGX_GLASS_ID) || \
   (type == CLOSURE_BSDF_MICROFACET_MULTI_GGX_GLASS_FRESNEL_ID))
@@ -609,7 +616,8 @@ typedef enum ClosureType {
  (type == CLOSURE_BSDF_MICROFACET_MULTI_GGX_FRESNEL_ID || \
   type == CLOSURE_BSDF_MICROFACET_MULTI_GGX_GLASS_FRESNEL_ID || \
   type == CLOSURE_BSDF_MICROFACET_GGX_FRESNEL_ID || \
-   type == CLOSURE_BSDF_MICROFACET_GGX_CLEARCOAT_ID)
+   type == CLOSURE_BSDF_MICROFACET_GGX_CLEARCOAT_ID || \
+   type == CLOSURE_BSDF_MICROFACET_GGX_ANISO_FRESNEL_ID)
 #define CLOSURE_IS_BSDF_OR_BSSRDF(type) (type <= CLOSURE_BSSRDF_PRINCIPLED_RANDOM_WALK_ID)
 #define CLOSURE_IS_BSSRDF(type) \
  (type >= CLOSURE_BSSRDF_CUBIC_ID && type <= CLOSURE_BSSRDF_PRINCIPLED_RANDOM_WALK_ID)
--- a/intern/cycles/render/bake.cpp
+++ b/intern/cycles/render/bake.cpp
@@ -15,7 +15,6 @@
 */

 #include "render/bake.h"
-#include "render/buffers.h"
 #include "render/integrator.h"
 #include "render/mesh.h"
 #include "render/object.h"
@@ -25,13 +24,235 @@

 CCL_NAMESPACE_BEGIN

-static int aa_samples(Scene *scene, Object *object, ShaderEvalType type)
+BakeData::BakeData(const int object, const size_t tri_offset, const size_t num_pixels)
+    : m_object(object), m_tri_offset(tri_offset), m_num_pixels(num_pixels)
+{
+  m_primitive.resize(num_pixels);
+  m_u.resize(num_pixels);
+  m_v.resize(num_pixels);
+  m_dudx.resize(num_pixels);
+  m_dudy.resize(num_pixels);
+  m_dvdx.resize(num_pixels);
+  m_dvdy.resize(num_pixels);
+}
+
+BakeData::~BakeData()
+{
+  m_primitive.clear();
+  m_u.clear();
+  m_v.clear();
+  m_dudx.clear();
+  m_dudy.clear();
+  m_dvdx.clear();
+  m_dvdy.clear();
+}
+
+void BakeData::set(int i, int prim, float uv[2], float dudx, float dudy, float dvdx, float dvdy)
+{
+  m_primitive[i] = (prim == -1 ? -1 : m_tri_offset + prim);
+  m_u[i] = uv[0];
+  m_v[i] = uv[1];
+  m_dudx[i] = dudx;
+  m_dudy[i] = dudy;
+  m_dvdx[i] = dvdx;
+  m_dvdy[i] = dvdy;
+}
+
+void BakeData::set_null(int i)
+{
+  m_primitive[i] = -1;
+}
+
+int BakeData::object()
+{
+  return m_object;
+}
+
+size_t BakeData::size()
+{
+  return m_num_pixels;
+}
+
+bool BakeData::is_valid(int i)
+{
+  return m_primitive[i] != -1;
+}
+
+uint4 BakeData::data(int i)
+{
+  return make_uint4(m_object, m_primitive[i], __float_as_int(m_u[i]), __float_as_int(m_v[i]));
+}
+
+uint4 BakeData::differentials(int i)
+{
+  return make_uint4(__float_as_int(m_dudx[i]),
+                    __float_as_int(m_dudy[i]),
+                    __float_as_int(m_dvdx[i]),
+                    __float_as_int(m_dvdy[i]));
+}
+
+BakeManager::BakeManager()
+{
+  m_bake_data = NULL;
+  m_is_baking = false;
+  need_update = true;
+  m_shader_limit = 512 * 512;
+}
+
+BakeManager::~BakeManager()
+{
+  if (m_bake_data)
+    delete m_bake_data;
+}
+
+bool BakeManager::get_baking()
+{
+  return m_is_baking;
+}
+
+void BakeManager::set_baking(const bool value)
+{
+  m_is_baking = value;
+}
+
+BakeData *BakeManager::init(const int object, const size_t tri_offset, const size_t num_pixels)
+{
+  m_bake_data = new BakeData(object, tri_offset, num_pixels);
+  return m_bake_data;
+}
+
+void BakeManager::set_shader_limit(const size_t x, const size_t y)
+{
+  m_shader_limit = x * y;
+  m_shader_limit = (size_t)pow(2, std::ceil(log(m_shader_limit) / log(2)));
+}
+
+bool BakeManager::bake(Device *device,
+                       DeviceScene *dscene,
+                       Scene *scene,
+                       Progress &progress,
+                       ShaderEvalType shader_type,
+                       const int pass_filter,
+                       BakeData *bake_data,
+                       float result[])
+{
+  size_t num_pixels = bake_data->size();
+
+  int num_samples = aa_samples(scene, bake_data, shader_type);
+
+  /* calculate the total pixel samples for the progress bar */
+  total_pixel_samples = 0;
+  for (size_t shader_offset = 0; shader_offset < num_pixels; shader_offset += m_shader_limit) {
+    size_t shader_size = (size_t)fminf(num_pixels - shader_offset, m_shader_limit);
+    total_pixel_samples += shader_size * num_samples;
+  }
+  progress.reset_sample();
+  progress.set_total_pixel_samples(total_pixel_samples);
+
+  /* needs to be up to date for baking specific AA samples */
+  dscene->data.integrator.aa_samples = num_samples;
+  device->const_copy_to("__data", &dscene->data, sizeof(dscene->data));
+
+  for (size_t shader_offset = 0; shader_offset < num_pixels; shader_offset += m_shader_limit) {
+    size_t shader_size = (size_t)fminf(num_pixels - shader_offset, m_shader_limit);
+
+    /* setup input for device task */
+    device_vector<uint4> d_input(device, "bake_input", MEM_READ_ONLY);
+    uint4 *d_input_data = d_input.alloc(shader_size * 2);
+    size_t d_input_size = 0;
+
+    for (size_t i = shader_offset; i < (shader_offset + shader_size); i++) {
+      d_input_data[d_input_size++] = bake_data->data(i);
+      d_input_data[d_input_size++] = bake_data->differentials(i);
+    }
+
+    if (d_input_size == 0) {
+      m_is_baking = false;
+      return false;
+    }
+
+    /* run device task */
+    device_vector<float4> d_output(device, "bake_output", MEM_READ_WRITE);
+    d_output.alloc(shader_size);
+    d_output.zero_to_device();
+    d_input.copy_to_device();
+
+    DeviceTask task(DeviceTask::SHADER);
+    task.shader_input = d_input.device_pointer;
+    task.shader_output = d_output.device_pointer;
+    task.shader_eval_type = shader_type;
+    task.shader_filter = pass_filter;
+    task.shader_x = 0;
+    task.offset = shader_offset;
+    task.shader_w = d_output.size();
+    task.num_samples = num_samples;
+    task.get_cancel = function_bind(&Progress::get_cancel, &progress);
+    task.update_progress_sample = function_bind(&Progress::add_samples_update, &progress, _1, _2);
+
+    device->task_add(task);
+    device->task_wait();
+
+    if (progress.get_cancel()) {
+      d_input.free();
+      d_output.free();
+      m_is_baking = false;
+      return false;
+    }
+
+    d_output.copy_from_device(0, 1, d_output.size());
+    d_input.free();
+
+    /* read result */
+    int k = 0;
+
+    float4 *offset = d_output.data();
+
+    size_t depth = 4;
+    for (size_t i = shader_offset; i < (shader_offset + shader_size); i++) {
+      size_t index = i * depth;
+      float4 out = offset[k++];
+
+      if (bake_data->is_valid(i)) {
+        for (size_t j = 0; j < 4; j++) {
+          result[index + j] = out[j];
+        }
+      }
+    }
+
+    d_output.free();
+  }
+
+  m_is_baking = false;
+  return true;
+}
+
+void BakeManager::device_update(Device * /*device*/,
+                                DeviceScene * /*dscene*/,
+                                Scene * /*scene*/,
+                                Progress &progress)
+{
+  if (!need_update)
+    return;
+
+  if (progress.get_cancel())
+    return;
+
+  need_update = false;
+}
+
+void BakeManager::device_free(Device * /*device*/, DeviceScene * /*dscene*/)
+{
+}
+
+int BakeManager::aa_samples(Scene *scene, BakeData *bake_data, ShaderEvalType type)
 {
  if (type == SHADER_EVAL_UV || type == SHADER_EVAL_ROUGHNESS) {
    return 1;
  }
  else if (type == SHADER_EVAL_NORMAL) {
    /* Only antialias normal if mesh has bump mapping. */
+    Object *object = scene->objects[bake_data->object()];
+
    if (object->geometry) {
      foreach (Shader *shader, object->geometry->used_shaders) {
        if (shader->has_bump) {
@@ -48,7 +269,7 @@ static int aa_samples(Scene *scene, Object *object, ShaderEvalType type)
 }

 /* Keep it synced with kernel_bake.h logic */
-static int shader_type_to_pass_filter(ShaderEvalType type, int pass_filter)
+int BakeManager::shader_type_to_pass_filter(ShaderEvalType type, const int pass_filter)
 {
  const int component_flags = pass_filter &
                              (BAKE_FILTER_DIRECT | BAKE_FILTER_INDIRECT | BAKE_FILTER_COLOR);
@@ -71,84 +292,4 @@ static int shader_type_to_pass_filter(ShaderEvalType type, int pass_filter)
  }
 }

-BakeManager::BakeManager()
-{
-  type = SHADER_EVAL_BAKE;
-  pass_filter = 0;
-
-  need_update = true;
-}
-
-BakeManager::~BakeManager()
-{
-}
-
-bool BakeManager::get_baking()
-{
-  return !object_name.empty();
-}
-
-void BakeManager::set(Scene *scene,
-                      const std::string &object_name_,
-                      ShaderEvalType type_,
-                      int pass_filter_)
-{
-  object_name = object_name_;
-  type = type_;
-  pass_filter = shader_type_to_pass_filter(type_, pass_filter_);
-
-  Pass::add(PASS_BAKE_PRIMITIVE, scene->film->passes);
-  Pass::add(PASS_BAKE_DIFFERENTIAL, scene->film->passes);
-
-  if (type == SHADER_EVAL_UV) {
-    /* force UV to be available */
-    Pass::add(PASS_UV, scene->film->passes);
-  }
-
-  /* force use_light_pass to be true if we bake more than just colors */
-  if (pass_filter & ~BAKE_FILTER_COLOR) {
-    Pass::add(PASS_LIGHT, scene->film->passes);
-  }
-
-  /* create device and update scene */
-  scene->film->tag_update(scene);
-  scene->integrator->tag_update(scene);
-
-  need_update = true;
-}
-
-void BakeManager::device_update(Device * /*device*/,
-                                DeviceScene *dscene,
-                                Scene *scene,
-                                Progress & /* progress */)
-{
-  if (!need_update)
-    return;
-
-  KernelIntegrator *kintegrator = &dscene->data.integrator;
-  KernelBake *kbake = &dscene->data.bake;
-
-  kbake->type = type;
-  kbake->pass_filter = pass_filter;
-
-  int object_index = 0;
-  foreach (Object *object, scene->objects) {
-    const Geometry *geom = object->geometry;
-    if (object->name == object_name && geom->type == Geometry::MESH) {
-      kbake->object_index = object_index;
-      kbake->tri_offset = geom->prim_offset;
-      kintegrator->aa_samples = aa_samples(scene, object, type);
-      break;
-    }
-
-    object_index++;
-  }
-
-  need_update = false;
-}
-
-void BakeManager::device_free(Device * /*device*/, DeviceScene * /*dscene*/)
-{
-}
-
 CCL_NAMESPACE_END
--- a/intern/cycles/render/bake.h
+++ b/intern/cycles/render/bake.h
@@ -25,23 +25,67 @@

 CCL_NAMESPACE_BEGIN

+class BakeData {
+ public:
+  BakeData(const int object, const size_t tri_offset, const size_t num_pixels);
+  ~BakeData();
+
+  void set(int i, int prim, float uv[2], float dudx, float dudy, float dvdx, float dvdy);
+  void set_null(int i);
+  int object();
+  size_t size();
+  uint4 data(int i);
+  uint4 differentials(int i);
+  bool is_valid(int i);
+
+ private:
+  int m_object;
+  size_t m_tri_offset;
+  size_t m_num_pixels;
+  vector<int> m_primitive;
+  vector<float> m_u;
+  vector<float> m_v;
+  vector<float> m_dudx;
+  vector<float> m_dudy;
+  vector<float> m_dvdx;
+  vector<float> m_dvdy;
+};
+
 class BakeManager {
 public:
  BakeManager();
  ~BakeManager();

-  void set(Scene *scene, const std::string &object_name, ShaderEvalType type, int pass_filter);
  bool get_baking();
+  void set_baking(const bool value);
+
+  BakeData *init(const int object, const size_t tri_offset, const size_t num_pixels);
+
+  void set_shader_limit(const size_t x, const size_t y);
+
+  bool bake(Device *device,
+            DeviceScene *dscene,
+            Scene *scene,
+            Progress &progress,
+            ShaderEvalType shader_type,
+            const int pass_filter,
+            BakeData *bake_data,
+            float result[]);

  void device_update(Device *device, DeviceScene *dscene, Scene *scene, Progress &progress);
  void device_free(Device *device, DeviceScene *dscene);

+  static int shader_type_to_pass_filter(ShaderEvalType type, const int pass_filter);
+  static int aa_samples(Scene *scene, BakeData *bake_data, ShaderEvalType type);
+
  bool need_update;

+  size_t total_pixel_samples;
+
 private:
-  ShaderEvalType type;
-  int pass_filter;
-  std::string object_name;
+  BakeData *m_bake_data;
+  bool m_is_baking;
+  size_t m_shader_limit;
 };

 CCL_NAMESPACE_END
--- a/intern/cycles/render/buffers.cpp
+++ b/intern/cycles/render/buffers.cpp
@@ -459,40 +459,6 @@ bool RenderBuffers::get_pass_rect(
  return false;
 }

-bool RenderBuffers::set_pass_rect(PassType type, int components, float *pixels)
-{
-  if (buffer.data() == NULL) {
-    return false;
-  }
-
-  int pass_offset = 0;
-
-  for (size_t j = 0; j < params.passes.size(); j++) {
-    Pass &pass = params.passes[j];
-
-    if (pass.type != type) {
-      pass_offset += pass.components;
-      continue;
-    }
-
-    float *out = buffer.data() + pass_offset;
-    int pass_stride = params.get_passes_size();
-    int size = params.width * params.height;
-
-    assert(pass.components == components);
-
-    for (int i = 0; i < size; i++, out += pass_stride, pixels += components) {
-      for (int j = 0; j < components; j++) {
-        out[j] = pixels[j];
-      }
-    }
-
-    return true;
-  }
-
-  return false;
-}
-
 /* Display Buffer */

 DisplayBuffer::DisplayBuffer(Device *device, bool linear)
--- a/intern/cycles/render/buffers.h
+++ b/intern/cycles/render/buffers.h
@@ -92,7 +92,6 @@ class RenderBuffers {
      const string &name, float exposure, int sample, int components, float *pixels);
  bool get_denoising_pass_rect(
      int offset, float exposure, int sample, int components, float *pixels);
-  bool set_pass_rect(PassType type, int components, float *pixels);
 };

 /* Display Buffer
@@ -131,7 +130,7 @@ class DisplayBuffer {

 class RenderTile {
 public:
-  typedef enum { PATH_TRACE = (1 << 0), BAKE = (1 << 1), DENOISE = (1 << 2) } Task;
+  typedef enum { PATH_TRACE = (1 << 0), DENOISE = (1 << 1) } Task;

  Task task;
  int x, y, w, h;
--- a/intern/cycles/render/film.cpp
+++ b/intern/cycles/render/film.cpp
@@ -196,10 +196,6 @@ void Pass::add(PassType type, vector<Pass> &passes, const char *name)
    case PASS_AOV_VALUE:
      pass.components = 1;
      break;
-    case PASS_BAKE_PRIMITIVE:
-    case PASS_BAKE_DIFFERENTIAL:
-      pass.components = 4;
-      break;
    default:
      assert(false);
      break;
@@ -390,13 +386,11 @@ void Film::device_update(Device *device, DeviceScene *dscene, Scene *scene)
    if (pass.type <= PASS_CATEGORY_MAIN_END) {
      kfilm->pass_flag |= pass_flag;
    }
-    else if (pass.type <= PASS_CATEGORY_LIGHT_END) {
+    else {
+      assert(pass.type <= PASS_CATEGORY_LIGHT_END);
      kfilm->use_light_pass = 1;
      kfilm->light_pass_flag |= pass_flag;
    }
-    else {
-      assert(pass.type <= PASS_CATEGORY_BAKE_END);
-    }

    switch (pass.type) {
      case PASS_COMBINED:
@@ -477,13 +471,6 @@ void Film::device_update(Device *device, DeviceScene *dscene, Scene *scene)
        kfilm->pass_volume_direct = kfilm->pass_stride;
        break;

-      case PASS_BAKE_PRIMITIVE:
-        kfilm->pass_bake_primitive = kfilm->pass_stride;
-        break;
-      case PASS_BAKE_DIFFERENTIAL:
-        kfilm->pass_bake_differential = kfilm->pass_stride;
-        break;
-
 #ifdef WITH_CYCLES_DEBUG
      case PASS_BVH_TRAVERSED_NODES:
        kfilm->pass_bvh_traversed_nodes = kfilm->pass_stride;
--- a/intern/cycles/render/graph.h
+++ b/intern/cycles/render/graph.h
@@ -204,6 +204,10 @@ class ShaderNode : public Node {
  {
    return false;
  }
+  virtual bool has_object_dependency()
+  {
+    return false;
+  }
  virtual bool has_attribute_dependency()
  {
    return false;
--- a/intern/cycles/render/light.cpp
+++ b/intern/cycles/render/light.cpp
@@ -181,7 +181,6 @@ bool Light::has_contribution(Scene *scene)
 LightManager::LightManager()
 {
  need_update = true;
-  need_update_background = true;
  use_light_visibility = false;
 }

@@ -902,7 +901,7 @@ void LightManager::device_update(Device *device,

  VLOG(1) << "Total " << scene->lights.size() << " lights.";

-  device_free(device, dscene, need_update_background);
+  device_free(device, dscene);

  use_light_visibility = false;

@@ -916,11 +915,9 @@ void LightManager::device_update(Device *device,
  if (progress.get_cancel())
    return;

-  if (need_update_background) {
-    device_update_background(device, dscene, scene, progress);
-    if (progress.get_cancel())
-      return;
-  }
+  device_update_background(device, dscene, scene, progress);
+  if (progress.get_cancel())
+    return;

  device_update_ies(dscene);
  if (progress.get_cancel())
@@ -932,17 +929,14 @@ void LightManager::device_update(Device *device,
  }

  need_update = false;
-  need_update_background = false;
 }

-void LightManager::device_free(Device *, DeviceScene *dscene, const bool free_background)
+void LightManager::device_free(Device *, DeviceScene *dscene)
 {
  dscene->light_distribution.free();
  dscene->lights.free();
-  if (free_background) {
-    dscene->light_background_marginal_cdf.free();
-    dscene->light_background_conditional_cdf.free();
-  }
+  dscene->light_background_marginal_cdf.free();
+  dscene->light_background_conditional_cdf.free();
  dscene->ies_lights.free();
 }

@@ -995,7 +989,6 @@ int LightManager::add_ies(const string &content)
  ies_slots[slot]->hash = hash;

  need_update = true;
-  need_update_background = true;

  return slot;
 }
@@ -1014,7 +1007,6 @@ void LightManager::remove_ies(int slot)

  /* If the slot has no more users, update the device to remove it. */
  need_update |= (ies_slots[slot]->users == 0);
-  need_update_background |= need_update;
 }

 void LightManager::device_update_ies(DeviceScene *dscene)
--- a/intern/cycles/render/light.h
+++ b/intern/cycles/render/light.h
@@ -88,9 +88,6 @@ class LightManager {
  bool use_light_visibility;
  bool need_update;

-  /* Need to update background (including multiple importance map) */
-  bool need_update_background;
-
  LightManager();
  ~LightManager();

@@ -100,7 +97,7 @@ class LightManager {
  void remove_ies(int slot);

  void device_update(Device *device, DeviceScene *dscene, Scene *scene, Progress &progress);
-  void device_free(Device *device, DeviceScene *dscene, const bool free_background = true);
+  void device_free(Device *device, DeviceScene *dscene);

  void tag_update(Scene *scene);

--- a/intern/cycles/render/nodes.cpp
+++ b/intern/cycles/render/nodes.cpp
@@ -2166,11 +2166,12 @@ NODE_DEFINE(AnisotropicBsdfNode)
  SOCKET_IN_FLOAT(surface_mix_weight, "SurfaceMixWeight", 0.0f, SocketType::SVM_INTERNAL);

  static NodeEnum distribution_enum;
-  distribution_enum.insert("beckmann", CLOSURE_BSDF_MICROFACET_BECKMANN_ID);
-  distribution_enum.insert("GGX", CLOSURE_BSDF_MICROFACET_GGX_ID);
-  distribution_enum.insert("Multiscatter GGX", CLOSURE_BSDF_MICROFACET_MULTI_GGX_ID);
-  distribution_enum.insert("ashikhmin_shirley", CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ID);
-  SOCKET_ENUM(distribution, "Distribution", distribution_enum, CLOSURE_BSDF_MICROFACET_GGX_ID);
+  distribution_enum.insert("beckmann", CLOSURE_BSDF_MICROFACET_BECKMANN_ANISO_ID);
+  distribution_enum.insert("GGX", CLOSURE_BSDF_MICROFACET_GGX_ANISO_ID);
+  distribution_enum.insert("Multiscatter GGX", CLOSURE_BSDF_MICROFACET_MULTI_GGX_ANISO_ID);
+  distribution_enum.insert("ashikhmin_shirley", CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ANISO_ID);
+  SOCKET_ENUM(
+      distribution, "Distribution", distribution_enum, CLOSURE_BSDF_MICROFACET_GGX_ANISO_ID);

  SOCKET_IN_VECTOR(tangent, "Tangent", make_float3(0.0f, 0.0f, 0.0f), SocketType::LINK_TANGENT);

@@ -2185,7 +2186,7 @@ NODE_DEFINE(AnisotropicBsdfNode)

 AnisotropicBsdfNode::AnisotropicBsdfNode() : BsdfNode(node_type)
 {
-  closure = CLOSURE_BSDF_MICROFACET_GGX_ID;
+  closure = CLOSURE_BSDF_MICROFACET_GGX_ANISO_ID;
 }

 void AnisotropicBsdfNode::attributes(Shader *shader, AttributeRequestSet *attributes)
@@ -2204,7 +2205,7 @@ void AnisotropicBsdfNode::compile(SVMCompiler &compiler)
 {
  closure = distribution;

-  if (closure == CLOSURE_BSDF_MICROFACET_MULTI_GGX_ID)
+  if (closure == CLOSURE_BSDF_MICROFACET_MULTI_GGX_ANISO_ID)
    BsdfNode::compile(
        compiler, input("Roughness"), input("Anisotropy"), input("Rotation"), input("Color"));
  else
@@ -2298,7 +2299,7 @@ void GlossyBsdfNode::compile(SVMCompiler &compiler)
  if (closure == CLOSURE_BSDF_REFLECTION_ID)
    BsdfNode::compile(compiler, NULL, NULL);
  else if (closure == CLOSURE_BSDF_MICROFACET_MULTI_GGX_ID)
-    BsdfNode::compile(compiler, input("Roughness"), NULL, NULL, input("Color"));
+    BsdfNode::compile(compiler, input("Roughness"), NULL, input("Color"));
  else
    BsdfNode::compile(compiler, input("Roughness"), NULL);
 }
--- a/intern/cycles/render/nodes.h
+++ b/intern/cycles/render/nodes.h
@@ -359,6 +359,10 @@ class PointDensityTextureNode : public ShaderNode {
  {
    return true;
  }
+  bool has_object_dependency()
+  {
+    return true;
+  }

  /* Parameters. */
  ustring filename;
@@ -892,6 +896,10 @@ class TextureCoordinateNode : public ShaderNode {
  {
    return true;
  }
+  bool has_object_dependency()
+  {
+    return use_transform;
+  }

  float3 normal_osl;
  bool from_dupli;
--- a/intern/cycles/render/osl.cpp
+++ b/intern/cycles/render/osl.cpp
@@ -764,6 +764,10 @@ void OSLCompiler::add(ShaderNode *node, const char *name, bool isfilepath)
      current_shader->has_volume_attribute_dependency = true;
  }

+  if (node->has_object_dependency()) {
+    current_shader->has_object_dependency = true;
+  }
+
  if (node->has_integrator_dependency()) {
    current_shader->has_integrator_dependency = true;
  }
@@ -1138,6 +1142,7 @@ void OSLCompiler::compile(OSLGlobals *og, Shader *shader)
    shader->has_surface_spatial_varying = false;
    shader->has_volume_spatial_varying = false;
    shader->has_volume_attribute_dependency = false;
+    shader->has_object_dependency = false;
    shader->has_integrator_dependency = false;

    /* generate surface shader */
--- a/intern/cycles/render/session.cpp
+++ b/intern/cycles/render/session.cpp
@@ -410,16 +410,7 @@ bool Session::acquire_tile(RenderTile &rtile, Device *tile_device, uint tile_typ
  rtile.num_samples = tile_manager.state.num_samples;
  rtile.resolution = tile_manager.state.resolution_divider;
  rtile.tile_index = tile->index;
-
-  if (tile->state == Tile::DENOISE) {
-    rtile.task = RenderTile::DENOISE;
-  }
-  else if (read_bake_tile_cb) {
-    rtile.task = RenderTile::BAKE;
-  }
-  else {
-    rtile.task = RenderTile::PATH_TRACE;
-  }
+  rtile.task = tile->state == Tile::DENOISE ? RenderTile::DENOISE : RenderTile::PATH_TRACE;

  tile_lock.unlock();

@@ -460,20 +451,11 @@ bool Session::acquire_tile(RenderTile &rtile, Device *tile_device, uint tile_typ
  rtile.buffers = tile->buffers;
  rtile.sample = tile_manager.state.sample;

-  if (read_bake_tile_cb) {
-    /* This will read any passes needed as input for baking. */
-    {
-      thread_scoped_lock tile_lock(tile_mutex);
-      read_bake_tile_cb(rtile);
-    }
-    rtile.buffers->buffer.copy_to_device();
-  }
-  else {
-    /* This will tag tile as IN PROGRESS in blender-side render pipeline,
-     * which is needed to highlight currently rendering tile before first
-     * sample was processed for it. */
-    update_tile_sample(rtile);
-  }
+  /* this will tag tile as IN PROGRESS in blender-side render pipeline,
+   * which is needed to highlight currently rendering tile before first
+   * sample was processed for it
+   */
+  update_tile_sample(rtile);

  return true;
 }
@@ -502,7 +484,6 @@ void Session::release_tile(RenderTile &rtile, const bool need_denoise)
  bool delete_tile;

  if (tile_manager.finish_tile(rtile.tile_index, need_denoise, delete_tile)) {
-    /* Finished tile pixels write. */
    if (write_render_tile_cb && params.progressive_refine == false) {
      write_render_tile_cb(rtile);
    }
@@ -513,7 +494,6 @@ void Session::release_tile(RenderTile &rtile, const bool need_denoise)
    }
  }
  else {
-    /* In progress tile pixels update. */
    if (update_render_tile_cb && params.progressive_refine == false) {
      update_render_tile_cb(rtile, false);
    }
--- a/intern/cycles/render/session.h
+++ b/intern/cycles/render/session.h
@@ -148,7 +148,6 @@ class Session {

  function<void(RenderTile &)> write_render_tile_cb;
  function<void(RenderTile &, bool)> update_render_tile_cb;
-  function<void(RenderTile &)> read_bake_tile_cb;

  explicit Session(const SessionParams &params);
  ~Session();
--- a/intern/cycles/render/shader.cpp
+++ b/intern/cycles/render/shader.cpp
@@ -206,6 +206,7 @@ Shader::Shader() : Node(node_type)
  has_surface_spatial_varying = false;
  has_volume_spatial_varying = false;
  has_volume_attribute_dependency = false;
+  has_object_dependency = false;
  has_integrator_dependency = false;
  has_volume_connected = false;
  prev_volume_step_rate = 0.0f;
@@ -217,6 +218,7 @@ Shader::Shader() : Node(node_type)

  need_update = true;
  need_update_geometry = true;
+  need_sync_object = false;
 }

 Shader::~Shader()
@@ -318,11 +320,9 @@ void Shader::tag_update(Scene *scene)
   * has use_mis set to false. We are quite close to release now, so
   * better to be safe.
   */
-  if (this == scene->background->get_shader(scene)) {
-    scene->light_manager->need_update_background = true;
-    if (scene->light_manager->has_background_light(scene)) {
-      scene->light_manager->need_update = true;
-    }
+  if (this == scene->background->get_shader(scene) &&
+      scene->light_manager->has_background_light(scene)) {
+    scene->light_manager->need_update = true;
  }

  /* quick detection of which kind of shaders we have to avoid loading
--- a/intern/cycles/render/shader.h
+++ b/intern/cycles/render/shader.h
@@ -98,6 +98,7 @@ class Shader : public Node {
  /* synchronization */
  bool need_update;
  bool need_update_geometry;
+  bool need_sync_object;

  /* If the shader has only volume components, the surface is assumed to
   * be transparent.
@@ -120,6 +121,7 @@ class Shader : public Node {
  bool has_surface_spatial_varying;
  bool has_volume_spatial_varying;
  bool has_volume_attribute_dependency;
+  bool has_object_dependency;
  bool has_integrator_dependency;

  /* displacement */
--- a/intern/cycles/render/svm.cpp
+++ b/intern/cycles/render/svm.cpp
@@ -448,6 +448,10 @@ void SVMCompiler::generate_node(ShaderNode *node, ShaderNodeSet &done)
      current_shader->has_volume_attribute_dependency = true;
  }

+  if (node->has_object_dependency()) {
+    current_shader->has_object_dependency = true;
+  }
+
  if (node->has_integrator_dependency()) {
    current_shader->has_integrator_dependency = true;
  }
@@ -859,6 +863,7 @@ void SVMCompiler::compile(Shader *shader, array<int4> &svm_nodes, int index, Sum
  shader->has_surface_spatial_varying = false;
  shader->has_volume_spatial_varying = false;
  shader->has_volume_attribute_dependency = false;
+  shader->has_object_dependency = false;
  shader->has_integrator_dependency = false;

  /* generate bump shader */
--- a/intern/cycles/util/util_avxb.h
+++ b/intern/cycles/util/util_avxb.h
@@ -57,7 +57,7 @@ struct avxb {
      : m256(_mm256_insertf128_ps(_mm256_castps128_ps256(a), b, 1))
  {
  }
-  __forceinline operator const __m256 &(void)const
+  __forceinline operator const __m256 &(void) const
  {
    return m256;
  }
--- a/intern/cycles/util/util_avxi.h
+++ b/intern/cycles/util/util_avxi.h
@@ -54,7 +54,7 @@ struct avxi {
  __forceinline avxi(const __m256i a) : m256(a)
  {
  }
-  __forceinline operator const __m256i &(void)const
+  __forceinline operator const __m256i &(void) const
  {
    return m256;
  }
--- a/intern/cycles/util/util_sseb.h
+++ b/intern/cycles/util/util_sseb.h
@@ -57,7 +57,7 @@ struct sseb {
  __forceinline sseb(const __m128 input) : m128(input)
  {
  }
-  __forceinline operator const __m128 &(void)const
+  __forceinline operator const __m128 &(void) const
  {
    return m128;
  }
--- a/intern/cycles/util/util_ssei.h
+++ b/intern/cycles/util/util_ssei.h
@@ -57,7 +57,7 @@ struct ssei {
  __forceinline ssei(const __m128i a) : m128(a)
  {
  }
-  __forceinline operator const __m128i &(void)const
+  __forceinline operator const __m128i &(void) const
  {
    return m128;
  }
--- a/intern/ghost/intern/GHOST_System.cpp
+++ b/intern/ghost/intern/GHOST_System.cpp
@@ -23,8 +23,8 @@

 #include "GHOST_System.h"

-#include <chrono>
 #include <stdio.h> /* just for printf */
+#include <time.h>

 #include "GHOST_DisplayManager.h"
 #include "GHOST_EventManager.h"
@@ -58,9 +58,12 @@ GHOST_System::~GHOST_System()

 GHOST_TUns64 GHOST_System::getMilliSeconds() const
 {
-  return std::chrono::duration_cast<std::chrono::milliseconds>(
-             std::chrono::steady_clock::now().time_since_epoch())
-      .count();
+  GHOST_TUns64 millis = ::clock();
+  if (CLOCKS_PER_SEC != 1000) {
+    millis *= 1000;
+    millis /= CLOCKS_PER_SEC;
+  }
+  return millis;
 }

 GHOST_ITimerTask *GHOST_System::installTimer(GHOST_TUns64 delay,
--- a/intern/ghost/intern/GHOST_SystemCocoa.mm
+++ b/intern/ghost/intern/GHOST_SystemCocoa.mm
@@ -1063,7 +1063,14 @@ GHOST_TSuccess GHOST_SystemCocoa::handleWindowEvent(GHOST_TEventType eventType,
  }
  switch (eventType) {
    case GHOST_kEventWindowClose:
-      pushEvent(new GHOST_Event(getMilliSeconds(), GHOST_kEventWindowClose, window));
+      // check for index of mainwindow as it would quit blender without dialog and discard
+      if ([windowsList count] > 1 &&
+          window->getCocoaWindow() != [windowsList objectAtIndex:[windowsList count] - 1]) {
+        pushEvent(new GHOST_Event(getMilliSeconds(), GHOST_kEventWindowClose, window));
+      }
+      else {
+        handleQuitRequest();  // -> quit dialog
+      }
      break;
    case GHOST_kEventWindowActivate:
      m_windowManager->setActiveWindow(window);
--- a/intern/ghost/intern/GHOST_SystemWayland.cpp
+++ b/intern/ghost/intern/GHOST_SystemWayland.cpp
@@ -25,7 +25,6 @@
 #include "GHOST_EventDragnDrop.h"
 #include "GHOST_EventKey.h"
 #include "GHOST_EventWheel.h"
-#include "GHOST_TimerManager.h"
 #include "GHOST_WindowManager.h"

 #include "GHOST_ContextEGL.h"
@@ -35,7 +34,7 @@

 #include <algorithm>
 #include <atomic>
-#include <stdexcept>
+#include <exception>
 #include <thread>
 #include <unordered_map>
 #include <unordered_set>
@@ -92,13 +91,6 @@ struct data_source_t {
  char *buffer_out;
 };

-struct key_repeat_payload_t {
-  GHOST_SystemWayland *system;
-  GHOST_IWindow *window;
-  GHOST_TKey key;
-  GHOST_TEventKeyData key_data;
-};
-
 struct input_t {
  GHOST_SystemWayland *system;

@@ -117,17 +109,6 @@ struct input_t {

  struct xkb_context *xkb_context;
  struct xkb_state *xkb_state;
-  struct {
-    /* Key repetition in character per second. */
-    int32_t rate;
-    /* Time (milliseconds) after which to start repeating keys. */
-    int32_t delay;
-    /* Timer for key repeats. */
-    GHOST_ITimerTask *timer = nullptr;
-  } key_repeat;
-
-  struct wl_surface *focus_pointer = nullptr;
-  struct wl_surface *focus_keyboard = nullptr;

  struct wl_data_device *data_device = nullptr;
  struct data_offer_t *data_offer_dnd;        /* Drag & Drop. */
@@ -140,6 +121,7 @@ struct display_t {
  GHOST_SystemWayland *system;

  struct wl_display *display;
+  struct wl_registry *registry;
  struct wl_compositor *compositor = nullptr;
  struct xdg_wm_base *xdg_shell = nullptr;
  struct wl_shm *shm = nullptr;
@@ -193,11 +175,6 @@ static void display_destroy(display_t *d)
      }
    }
    if (input->keyboard) {
-      if (input->key_repeat.timer) {
-        delete static_cast<key_repeat_payload_t *>(input->key_repeat.timer->getUserData());
-        input->system->removeTimer(input->key_repeat.timer);
-        input->key_repeat.timer = nullptr;
-      }
      wl_keyboard_destroy(input->keyboard);
    }
    if (input->xkb_state) {
@@ -288,7 +265,6 @@ static GHOST_TKey xkb_map_gkey(const xkb_keysym_t &sym)

      GXMAP(gkey, XKB_KEY_Escape, GHOST_kKeyEsc);
      GXMAP(gkey, XKB_KEY_space, GHOST_kKeySpace);
-      GXMAP(gkey, XKB_KEY_apostrophe, GHOST_kKeyQuote);
      GXMAP(gkey, XKB_KEY_comma, GHOST_kKeyComma);
      GXMAP(gkey, XKB_KEY_minus, GHOST_kKeyMinus);
      GXMAP(gkey, XKB_KEY_plus, GHOST_kKeyPlus);
@@ -318,19 +294,29 @@ static GHOST_TKey xkb_map_gkey(const xkb_keysym_t &sym)
      GXMAP(gkey, XKB_KEY_Scroll_Lock, GHOST_kKeyScrollLock);

      GXMAP(gkey, XKB_KEY_Left, GHOST_kKeyLeftArrow);
+      GXMAP(gkey, XKB_KEY_KP_Left, GHOST_kKeyLeftArrow);
      GXMAP(gkey, XKB_KEY_Right, GHOST_kKeyRightArrow);
+      GXMAP(gkey, XKB_KEY_KP_Right, GHOST_kKeyRightArrow);
      GXMAP(gkey, XKB_KEY_Up, GHOST_kKeyUpArrow);
+      GXMAP(gkey, XKB_KEY_KP_Up, GHOST_kKeyUpArrow);
      GXMAP(gkey, XKB_KEY_Down, GHOST_kKeyDownArrow);
+      GXMAP(gkey, XKB_KEY_KP_Down, GHOST_kKeyDownArrow);

      GXMAP(gkey, XKB_KEY_Print, GHOST_kKeyPrintScreen);
      GXMAP(gkey, XKB_KEY_Pause, GHOST_kKeyPause);

      GXMAP(gkey, XKB_KEY_Insert, GHOST_kKeyInsert);
+      GXMAP(gkey, XKB_KEY_KP_Insert, GHOST_kKeyInsert);
      GXMAP(gkey, XKB_KEY_Delete, GHOST_kKeyDelete);
+      GXMAP(gkey, XKB_KEY_KP_Delete, GHOST_kKeyDelete);
      GXMAP(gkey, XKB_KEY_Home, GHOST_kKeyHome);
+      GXMAP(gkey, XKB_KEY_KP_Home, GHOST_kKeyHome);
      GXMAP(gkey, XKB_KEY_End, GHOST_kKeyEnd);
+      GXMAP(gkey, XKB_KEY_KP_End, GHOST_kKeyEnd);
      GXMAP(gkey, XKB_KEY_Page_Up, GHOST_kKeyUpPage);
+      GXMAP(gkey, XKB_KEY_KP_Page_Up, GHOST_kKeyUpPage);
      GXMAP(gkey, XKB_KEY_Page_Down, GHOST_kKeyDownPage);
+      GXMAP(gkey, XKB_KEY_KP_Page_Down, GHOST_kKeyDownPage);

      GXMAP(gkey, XKB_KEY_KP_Decimal, GHOST_kKeyNumpadPeriod);
      GXMAP(gkey, XKB_KEY_KP_Enter, GHOST_kKeyNumpadEnter);
@@ -344,8 +330,7 @@ static GHOST_TKey xkb_map_gkey(const xkb_keysym_t &sym)
      GXMAP(gkey, XKB_KEY_XF86AudioPrev, GHOST_kKeyMediaFirst);
      GXMAP(gkey, XKB_KEY_XF86AudioNext, GHOST_kKeyMediaLast);
      default:
-        GHOST_PRINT("unhandled key: " << std::hex << std::showbase << sym << std::dec << " ("
-                                      << sym << ")" << std::endl);
+        GHOST_PRINT("unhandled key: " << sym << std::endl);
        gkey = GHOST_kKeyUnknown;
    }
 #undef GXMAP
@@ -444,26 +429,22 @@ static void relative_pointer_relative_motion(
  input->x += wl_fixed_to_int(dx);
  input->y += wl_fixed_to_int(dy);

-  GHOST_IWindow *win = static_cast<GHOST_WindowWayland *>(
-      wl_surface_get_user_data(input->focus_pointer));
-
-  input->system->pushEvent(new GHOST_EventCursor(input->system->getMilliSeconds(),
-                                                 GHOST_kEventCursorMove,
-                                                 win,
-                                                 input->x,
-                                                 input->y,
-                                                 GHOST_TABLET_DATA_NONE));
+  input->system->pushEvent(
+      new GHOST_EventCursor(input->system->getMilliSeconds(),
+                            GHOST_kEventCursorMove,
+                            input->system->getWindowManager()->getActiveWindow(),
+                            input->x,
+                            input->y,
+                            GHOST_TABLET_DATA_NONE));
 }

 static const zwp_relative_pointer_v1_listener relative_pointer_listener = {
-    relative_pointer_relative_motion,
-};
+    relative_pointer_relative_motion};

 static void dnd_events(const input_t *const input, const GHOST_TEventType event)
 {
  const GHOST_TUns64 time = input->system->getMilliSeconds();
-  GHOST_IWindow *const window = static_cast<GHOST_WindowWayland *>(
-      wl_surface_get_user_data(input->focus_pointer));
+  GHOST_IWindow *const window = input->system->getWindowManager()->getActiveWindow();
  for (const std::string &type : mime_preference_order) {
    input->system->pushEvent(new GHOST_EventDragnDrop(time,
                                                      event,
@@ -669,7 +650,7 @@ static void data_device_drop(void *data, struct wl_data_device * /*wl_data_devic
                                                       data_offer->types.begin(),
                                                       data_offer->types.end());

-  auto read_uris = [](input_t *const input,
+  auto read_uris = [](GHOST_SystemWayland *const system,
                      data_offer_t *data_offer,
                      const std::string mime_receive) {
    const int x = data_offer->dnd.x;
@@ -683,8 +664,6 @@ static void data_device_drop(void *data, struct wl_data_device * /*wl_data_devic
    delete data_offer;
    data_offer = nullptr;

-    GHOST_SystemWayland *const system = input->system;
-
    if (mime_receive == mime_text_uri) {
      static constexpr const char *file_proto = "file://";
      static constexpr const char *crlf = "\r\n";
@@ -713,12 +692,10 @@ static void data_device_drop(void *data, struct wl_data_device * /*wl_data_devic
            malloc((uris[i].size() + 1) * sizeof(GHOST_TUns8)));
        memcpy(flist->strings[i], uris[i].data(), uris[i].size() + 1);
      }
-      GHOST_IWindow *win = static_cast<GHOST_WindowWayland *>(
-          wl_surface_get_user_data(input->focus_pointer));
      system->pushEvent(new GHOST_EventDragnDrop(system->getMilliSeconds(),
                                                 GHOST_kEventDraggingDropDone,
                                                 GHOST_kDragnDropTypeFilenames,
-                                                 win,
+                                                 system->getWindowManager()->getActiveWindow(),
                                                 x,
                                                 y,
                                                 flist));
@@ -730,7 +707,7 @@ static void data_device_drop(void *data, struct wl_data_device * /*wl_data_devic
    wl_display_roundtrip(system->display());
  };

-  std::thread read_thread(read_uris, input, data_offer, mime_receive);
+  std::thread read_thread(read_uris, input->system, data_offer, mime_receive);
  read_thread.detach();
 }

@@ -784,16 +761,13 @@ static const struct wl_data_device_listener data_device_listener = {
    data_device_selection,
 };

-static void cursor_buffer_release(void *data, struct wl_buffer *wl_buffer)
+static void buffer_release(void * /*data*/, struct wl_buffer *wl_buffer)
 {
-  cursor_t *cursor = static_cast<cursor_t *>(data);
-
  wl_buffer_destroy(wl_buffer);
-  cursor->buffer = nullptr;
 }

-const struct wl_buffer_listener cursor_buffer_listener = {
-    cursor_buffer_release,
+const struct wl_buffer_listener buffer_listener = {
+    buffer_release,
 };

 static void pointer_enter(void *data,
@@ -810,24 +784,17 @@ static void pointer_enter(void *data,
  input->pointer_serial = serial;
  input->x = wl_fixed_to_int(surface_x);
  input->y = wl_fixed_to_int(surface_y);
-  input->focus_pointer = surface;

-  input->system->pushEvent(
-      new GHOST_EventCursor(input->system->getMilliSeconds(),
-                            GHOST_kEventCursorMove,
-                            static_cast<GHOST_WindowWayland *>(wl_surface_get_user_data(surface)),
-                            input->x,
-                            input->y,
-                            GHOST_TABLET_DATA_NONE));
+  static_cast<GHOST_WindowWayland *>(wl_surface_get_user_data(surface))->activate();
 }

-static void pointer_leave(void *data,
+static void pointer_leave(void * /*data*/,
                          struct wl_pointer * /*wl_pointer*/,
                          uint32_t /*serial*/,
                          struct wl_surface *surface)
 {
  if (surface != nullptr) {
-    static_cast<input_t *>(data)->focus_pointer = nullptr;
+    static_cast<GHOST_WindowWayland *>(wl_surface_get_user_data(surface))->deactivate();
  }
 }

@@ -839,22 +806,16 @@ static void pointer_motion(void *data,
 {
  input_t *input = static_cast<input_t *>(data);

-  GHOST_IWindow *win = static_cast<GHOST_WindowWayland *>(
-      wl_surface_get_user_data(input->focus_pointer));
-
-  if (!win) {
-    return;
-  }
-
  input->x = wl_fixed_to_int(surface_x);
  input->y = wl_fixed_to_int(surface_y);

-  input->system->pushEvent(new GHOST_EventCursor(input->system->getMilliSeconds(),
-                                                 GHOST_kEventCursorMove,
-                                                 win,
-                                                 wl_fixed_to_int(surface_x),
-                                                 wl_fixed_to_int(surface_y),
-                                                 GHOST_TABLET_DATA_NONE));
+  input->system->pushEvent(
+      new GHOST_EventCursor(input->system->getMilliSeconds(),
+                            GHOST_kEventCursorMove,
+                            input->system->getWindowManager()->getActiveWindow(),
+                            wl_fixed_to_int(surface_x),
+                            wl_fixed_to_int(surface_y),
+                            GHOST_TABLET_DATA_NONE));
 }

 static void pointer_button(void *data,
@@ -888,12 +849,14 @@ static void pointer_button(void *data,
  }

  input_t *input = static_cast<input_t *>(data);
-  GHOST_IWindow *win = static_cast<GHOST_WindowWayland *>(
-      wl_surface_get_user_data(input->focus_pointer));
  input->data_source->source_serial = serial;
  input->buttons.set(ebutton, state == WL_POINTER_BUTTON_STATE_PRESSED);
-  input->system->pushEvent(new GHOST_EventButton(
-      input->system->getMilliSeconds(), etype, win, ebutton, GHOST_TABLET_DATA_NONE));
+  input->system->pushEvent(
+      new GHOST_EventButton(input->system->getMilliSeconds(),
+                            etype,
+                            input->system->getWindowManager()->getActiveWindow(),
+                            ebutton,
+                            GHOST_TABLET_DATA_NONE));
 }

 static void pointer_axis(void *data,
@@ -906,10 +869,10 @@ static void pointer_axis(void *data,
    return;
  }
  input_t *input = static_cast<input_t *>(data);
-  GHOST_IWindow *win = static_cast<GHOST_WindowWayland *>(
-      wl_surface_get_user_data(input->focus_pointer));
  input->system->pushEvent(
-      new GHOST_EventWheel(input->system->getMilliSeconds(), win, std::signbit(value) ? +1 : -1));
+      new GHOST_EventWheel(input->system->getMilliSeconds(),
+                           input->system->getWindowManager()->getActiveWindow(),
+                           std::signbit(value) ? +1 : -1));
 }

 static const struct wl_pointer_listener pointer_listener = {
@@ -956,15 +919,13 @@ static void keyboard_keymap(
 * Notification that this seat's keyboard focus is on a certain
 * surface.
 */
-static void keyboard_enter(void *data,
+static void keyboard_enter(void * /*data*/,
                           struct wl_keyboard * /*wl_keyboard*/,
                           uint32_t /*serial*/,
-                           struct wl_surface *surface,
+                           struct wl_surface * /*surface*/,
                           struct wl_array * /*keys*/)
 {
-  if (surface != nullptr) {
-    static_cast<input_t *>(data)->focus_keyboard = surface;
-  }
+  /* pass */
 }

 /**
@@ -973,40 +934,12 @@ static void keyboard_enter(void *data,
 * Notification that this seat's keyboard focus is no longer on a
 * certain surface.
 */
-static void keyboard_leave(void *data,
+static void keyboard_leave(void * /*data*/,
                           struct wl_keyboard * /*wl_keyboard*/,
                           uint32_t /*serial*/,
-                           struct wl_surface *surface)
+                           struct wl_surface * /*surface*/)
 {
-  if (surface != nullptr) {
-    static_cast<input_t *>(data)->focus_keyboard = nullptr;
-  }
-}
-
-/**
- * A version of #xkb_state_key_get_one_sym which returns the key without any modifiers pressed.
- * Needed because #GHOST_TKey uses these values as key-codes.
- */
-static xkb_keysym_t xkb_state_key_get_one_sym_without_modifiers(struct xkb_state *xkb_state,
-                                                                xkb_keycode_t key)
-{
-  /* Use an empty keyboard state to access key symbol without modifiers. */
-  xkb_state_get_keymap(xkb_state);
-  struct xkb_keymap *keymap = xkb_state_get_keymap(xkb_state);
-  struct xkb_state *xkb_state_empty = xkb_state_new(keymap);
-
-  /* Enable number-lock. */
-  {
-    const xkb_mod_index_t mod2 = xkb_keymap_mod_get_index(keymap, XKB_MOD_NAME_NUM);
-    const xkb_mod_index_t num = xkb_keymap_mod_get_index(keymap, "NumLock");
-    if (num != XKB_MOD_INVALID && mod2 != XKB_MOD_INVALID) {
-      xkb_state_update_mask(xkb_state_empty, (1 << mod2), 0, (1 << num), 0, 0, 0);
-    }
-  }
-
-  const xkb_keysym_t sym = xkb_state_key_get_one_sym(xkb_state_empty, key);
-  xkb_state_unref(xkb_state_empty);
-  return sym;
+  /* pass */
 }

 static void keyboard_key(void *data,
@@ -1028,21 +961,12 @@ static void keyboard_key(void *data,
      break;
  }

-  const xkb_keysym_t sym = xkb_state_key_get_one_sym_without_modifiers(input->xkb_state, key + 8);
-
+  const xkb_keysym_t sym = xkb_state_key_get_one_sym(input->xkb_state, key + 8);
  if (sym == XKB_KEY_NoSymbol) {
    return;
  }
  const GHOST_TKey gkey = xkb_map_gkey(sym);

-  /* Delete previous timer. */
-  if (xkb_keymap_key_repeats(xkb_state_get_keymap(input->xkb_state), key + 8) &&
-      input->key_repeat.timer) {
-    delete static_cast<key_repeat_payload_t *>(input->key_repeat.timer->getUserData());
-    input->system->removeTimer(input->key_repeat.timer);
-    input->key_repeat.timer = nullptr;
-  }
-
  GHOST_TEventKeyData key_data;

  if (etype == GHOST_kEventKeyDown) {
@@ -1054,38 +978,13 @@ static void keyboard_key(void *data,
  }

  input->data_source->source_serial = serial;
-
-  GHOST_IWindow *win = static_cast<GHOST_WindowWayland *>(
-      wl_surface_get_user_data(input->focus_keyboard));
-  input->system->pushEvent(new GHOST_EventKey(
-      input->system->getMilliSeconds(), etype, win, gkey, '\0', key_data.utf8_buf, false));
-
-  /* Start timer for repeating key, if applicable. */
-  if (input->key_repeat.rate > 0 &&
-      xkb_keymap_key_repeats(xkb_state_get_keymap(input->xkb_state), key + 8) &&
-      etype == GHOST_kEventKeyDown) {
-
-    key_repeat_payload_t *payload = new key_repeat_payload_t({
-        .system = input->system,
-        .window = win,
-        .key = gkey,
-        .key_data = key_data,
-    });
-
-    auto cb = [](GHOST_ITimerTask *task, GHOST_TUns64 /*time*/) {
-      struct key_repeat_payload_t *payload = static_cast<key_repeat_payload_t *>(
-          task->getUserData());
-      payload->system->pushEvent(new GHOST_EventKey(payload->system->getMilliSeconds(),
-                                                    GHOST_kEventKeyDown,
-                                                    payload->window,
-                                                    payload->key,
-                                                    '\0',
-                                                    payload->key_data.utf8_buf,
-                                                    true));
-    };
-    input->key_repeat.timer = input->system->installTimer(
-        input->key_repeat.delay, 1000 / input->key_repeat.rate, cb, payload);
-  }
+  input->system->pushEvent(new GHOST_EventKey(input->system->getMilliSeconds(),
+                                              etype,
+                                              input->system->getWindowManager()->getActiveWindow(),
+                                              gkey,
+                                              '\0',
+                                              key_data.utf8_buf,
+                                              false));
 }

 static void keyboard_modifiers(void *data,
@@ -1105,24 +1004,12 @@ static void keyboard_modifiers(void *data,
                        group);
 }

-static void keyboard_repeat_info(void *data,
-                                 struct wl_keyboard * /*wl_keyboard*/,
-                                 int32_t rate,
-                                 int32_t delay)
-{
-  input_t *input = static_cast<input_t *>(data);
-
-  input->key_repeat.rate = rate;
-  input->key_repeat.delay = delay;
-}
-
 static const struct wl_keyboard_listener keyboard_listener = {
    keyboard_keymap,
    keyboard_enter,
    keyboard_leave,
    keyboard_key,
    keyboard_modifiers,
-    keyboard_repeat_info,
 };

 static void seat_capabilities(void *data, struct wl_seat *wl_seat, uint32_t capabilities)
@@ -1255,7 +1142,7 @@ static void global_add(void *data,
    input->relative_pointer = nullptr;
    input->locked_pointer = nullptr;
    input->seat = static_cast<wl_seat *>(
-        wl_registry_bind(wl_registry, name, &wl_seat_interface, 4));
+        wl_registry_bind(wl_registry, name, &wl_seat_interface, 2));
    display->inputs.push_back(input);
    wl_seat_add_listener(input->seat, &seat_listener, input);
  }
@@ -1352,18 +1239,10 @@ GHOST_SystemWayland::~GHOST_SystemWayland()
  display_destroy(d);
 }

-bool GHOST_SystemWayland::processEvents(bool waitForEvent)
+bool GHOST_SystemWayland::processEvents(bool /*waitForEvent*/)
 {
-  const bool fired = getTimerManager()->fireTimers(getMilliSeconds());
-
-  if (waitForEvent) {
-    wl_display_dispatch(d->display);
-  }
-  else {
-    wl_display_roundtrip(d->display);
-  }
-
-  return fired || (getEventManager()->getNumEvents() > 0);
+  wl_display_dispatch(d->display);
+  return true;
 }

 int GHOST_SystemWayland::toggleConsole(int /*action*/)
@@ -1452,7 +1331,7 @@ GHOST_TUns8 GHOST_SystemWayland::getNumDisplays() const

 GHOST_TSuccess GHOST_SystemWayland::getCursorPosition(GHOST_TInt32 &x, GHOST_TInt32 &y) const
 {
-  if (!d->inputs.empty() && (d->inputs[0]->focus_pointer != nullptr)) {
+  if (getWindowManager()->getActiveWindow() != nullptr && !d->inputs.empty()) {
    x = d->inputs[0]->x;
    y = d->inputs[0]->y;
    return GHOST_kSuccess;
@@ -1665,7 +1544,7 @@ GHOST_TSuccess GHOST_SystemWayland::setCustomCursorShape(GHOST_TUns8 *bitmap,
  wl_shm_pool_destroy(pool);
  close(fd);

-  wl_buffer_add_listener(buffer, &cursor_buffer_listener, cursor);
+  wl_buffer_add_listener(buffer, &buffer_listener, nullptr);

  static constexpr uint32_t black = 0xFF000000;
  static constexpr uint32_t white = 0xFFFFFFFF;
@@ -1714,17 +1593,15 @@ GHOST_TSuccess GHOST_SystemWayland::setCursorVisibility(bool visible)
    return GHOST_kFailure;
  }

-  input_t *input = d->inputs[0];
-
-  cursor_t *cursor = &input->cursor;
+  cursor_t *cursor = &d->inputs[0]->cursor;
  if (visible) {
    if (!cursor->visible) {
-      set_cursor_buffer(input, cursor->buffer);
+      set_cursor_buffer(d->inputs[0], d->inputs[0]->cursor.buffer);
    }
  }
  else {
    if (cursor->visible) {
-      set_cursor_buffer(input, nullptr);
+      set_cursor_buffer(d->inputs[0], nullptr);
    }
  }

--- a/intern/ghost/intern/GHOST_SystemX11.cpp
+++ b/intern/ghost/intern/GHOST_SystemX11.cpp
@@ -2172,24 +2172,14 @@ GHOST_TUns8 *GHOST_SystemX11::getClipboard(bool selection) const
  else if (owner == None)
    return (NULL);

-  /* Restore events so copy doesn't swallow other event types (keyboard/mouse). */
-  vector<XEvent> restore_events;
-
  while (1) {
    /* only get an event if xcout() is doing something */
-    bool restore_this_event = false;
-    if (context != XCLIB_XCOUT_NONE) {
+    if (context != XCLIB_XCOUT_NONE)
      XNextEvent(m_display, &evt);
-      restore_this_event = (evt.type != SelectionNotify);
-    }

    /* fetch the selection, or part of it */
    getClipboard_xcout(&evt, sseln, target, &sel_buf, &sel_len, &context);

-    if (restore_this_event) {
-      restore_events.push_back(evt);
-    }
-
    /* fallback is needed. set XA_STRING to target and restart the loop. */
    if (context == XCLIB_XCOUT_FALLBACK) {
      context = XCLIB_XCOUT_NONE;
@@ -2218,11 +2208,6 @@ GHOST_TUns8 *GHOST_SystemX11::getClipboard(bool selection) const
      break;
  }

-  while (!restore_events.empty()) {
-    XPutBackEvent(m_display, &restore_events.back());
-    restore_events.pop_back();
-  }
-
  if (sel_len) {
    /* only print the buffer out, and free it, if it's not
     * empty
--- a/intern/ghost/intern/GHOST_XrContext.cpp
+++ b/intern/ghost/intern/GHOST_XrContext.cpp
@@ -465,14 +465,7 @@ void GHOST_XrContext::startSession(const GHOST_XrSessionBeginInfo *begin_info)

 void GHOST_XrContext::endSession()
 {
-  if (m_session) {
-    if (m_session->isRunning()) {
-      m_session->requestEnd();
-    }
-    else {
-      m_session = nullptr;
-    }
-  }
+  m_session->requestEnd();
 }

 bool GHOST_XrContext::isSessionRunning() const
--- a/intern/ghost/intern/GHOST_XrSession.cpp
+++ b/intern/ghost/intern/GHOST_XrSession.cpp
@@ -203,17 +203,13 @@ void GHOST_XrSession::requestEnd()
  xrRequestExitSession(m_oxr->session);
 }

-void GHOST_XrSession::beginSession()
-{
-  XrSessionBeginInfo begin_info = {XR_TYPE_SESSION_BEGIN_INFO};
-  begin_info.primaryViewConfigurationType = m_oxr->view_type;
-  CHECK_XR(xrBeginSession(m_oxr->session, &begin_info), "Failed to cleanly begin the VR session.");
-}
-
-void GHOST_XrSession::endSession()
+void GHOST_XrSession::end()
 {
  assert(m_oxr->session != XR_NULL_HANDLE);
+
  CHECK_XR(xrEndSession(m_oxr->session), "Failed to cleanly end the VR session.");
+  unbindGraphicsContext();
+  m_draw_info = nullptr;
 }

 GHOST_XrSession::LifeExpectancy GHOST_XrSession::handleStateChangeEvent(
@@ -226,11 +222,16 @@ GHOST_XrSession::LifeExpectancy GHOST_XrSession::handleStateChangeEvent(

  switch (lifecycle->state) {
    case XR_SESSION_STATE_READY: {
-      beginSession();
+      XrSessionBeginInfo begin_info = {XR_TYPE_SESSION_BEGIN_INFO};
+
+      begin_info.primaryViewConfigurationType = m_oxr->view_type;
+      CHECK_XR(xrBeginSession(m_oxr->session, &begin_info),
+               "Failed to cleanly begin the VR session.");
      break;
    }
    case XR_SESSION_STATE_STOPPING:
-      endSession();
+      /* Runtime will change state to STATE_EXITING, don't destruct session yet. */
+      end();
      break;
    case XR_SESSION_STATE_EXITING:
    case XR_SESSION_STATE_LOSS_PENDING:
--- a/intern/ghost/intern/GHOST_XrSession.h
+++ b/intern/ghost/intern/GHOST_XrSession.h
@@ -68,8 +68,7 @@ class GHOST_XrSession {
  std::unique_ptr<GHOST_XrDrawInfo> m_draw_info;

  void initSystem();
-  void beginSession();
-  void endSession();
+  void end();

  void bindGraphicsContext();

--- a/intern/guardedalloc/CMakeLists.txt
+++ b/intern/guardedalloc/CMakeLists.txt
@@ -53,14 +53,6 @@ if(WIN32 AND NOT UNIX)

    mmap_win.h
  )
-
-  list(APPEND INC_SYS
-    ${PTHREADS_INC}
-  )
-
-  list(APPEND LIB
-    ${PTHREADS_LIBRARIES}
-  )
 endif()

 # Jemalloc 5.0.0+ needs extra configuration.
--- a/intern/guardedalloc/MEM_guardedalloc.h
+++ b/intern/guardedalloc/MEM_guardedalloc.h
@@ -145,6 +145,14 @@ extern void *(*MEM_mallocN_aligned)(size_t len,
                                    const char *str) /* ATTR_MALLOC */ ATTR_WARN_UNUSED_RESULT
    ATTR_ALLOC_SIZE(1) ATTR_NONNULL(3);

+/**
+ * Same as callocN, clears memory and uses mmap (disk cached) if supported.
+ * Can be free'd with MEM_freeN as usual.
+ * */
+extern void *(*MEM_mapallocN)(size_t len,
+                              const char *str) /* ATTR_MALLOC */ ATTR_WARN_UNUSED_RESULT
+    ATTR_ALLOC_SIZE(1) ATTR_NONNULL(2);
+
 /** Print a list of the names and sizes of all allocated memory
 * blocks. as a python dict for easy investigation */
 extern void (*MEM_printmemlist_pydict)(void);
@@ -168,11 +176,20 @@ extern void (*MEM_set_error_callback)(void (*func)(const char *));
 * @retval true for correct memory, false for corrupted memory. */
 extern bool (*MEM_consistency_check)(void);

+/** Set thread locking functions for safe memory allocation from multiple
+ * threads, pass NULL pointers to disable thread locking again. */
+extern void (*MEM_set_lock_callback)(void (*lock)(void), void (*unlock)(void));
+
 /** Attempt to enforce OSX (or other OS's) to have malloc and stack nonzero */
 extern void (*MEM_set_memory_debug)(void);

-/** Memory usage stats. */
+/**
+ * Memory usage stats
+ * - MEM_get_memory_in_use is all memory
+ * - MEM_get_mapped_memory_in_use is a subset of all memory */
 extern size_t (*MEM_get_memory_in_use)(void);
+/** Get mapped memory usage. */
+extern size_t (*MEM_get_mapped_memory_in_use)(void);
 /** Get amount of memory blocks in use. */
 extern unsigned int (*MEM_get_memory_blocks_in_use)(void);

--- a/intern/guardedalloc/cpp/mallocn.cpp
+++ b/intern/guardedalloc/cpp/mallocn.cpp
@@ -21,24 +21,24 @@
 #include "../MEM_guardedalloc.h"
 #include <new>

-void *operator new(size_t size, const char *str);
-void *operator new[](size_t size, const char *str);
+void *operator new(size_t size, const char *str) throw(std::bad_alloc);
+void *operator new[](size_t size, const char *str) throw(std::bad_alloc);

 /* not default but can be used when needing to set a string */
-void *operator new(size_t size, const char *str)
+void *operator new(size_t size, const char *str) throw(std::bad_alloc)
 {
  return MEM_mallocN(size, str);
 }
-void *operator new[](size_t size, const char *str)
+void *operator new[](size_t size, const char *str) throw(std::bad_alloc)
 {
  return MEM_mallocN(size, str);
 }

-void *operator new(size_t size)
+void *operator new(size_t size) throw(std::bad_alloc)
 {
  return MEM_mallocN(size, "C++/anonymous");
 }
-void *operator new[](size_t size)
+void *operator new[](size_t size) throw(std::bad_alloc)
 {
  return MEM_mallocN(size, "C++/anonymous[]");
 }
--- a/intern/guardedalloc/intern/mallocn.c
+++ b/intern/guardedalloc/intern/mallocn.c
@@ -48,14 +48,18 @@ void *(*MEM_malloc_arrayN)(size_t len, size_t size, const char *str) = MEM_lockf
 void *(*MEM_mallocN_aligned)(size_t len,
                             size_t alignment,
                             const char *str) = MEM_lockfree_mallocN_aligned;
+void *(*MEM_mapallocN)(size_t len, const char *str) = MEM_lockfree_mapallocN;
 void (*MEM_printmemlist_pydict)(void) = MEM_lockfree_printmemlist_pydict;
 void (*MEM_printmemlist)(void) = MEM_lockfree_printmemlist;
 void (*MEM_callbackmemlist)(void (*func)(void *)) = MEM_lockfree_callbackmemlist;
 void (*MEM_printmemlist_stats)(void) = MEM_lockfree_printmemlist_stats;
 void (*MEM_set_error_callback)(void (*func)(const char *)) = MEM_lockfree_set_error_callback;
 bool (*MEM_consistency_check)(void) = MEM_lockfree_consistency_check;
+void (*MEM_set_lock_callback)(void (*lock)(void),
+                              void (*unlock)(void)) = MEM_lockfree_set_lock_callback;
 void (*MEM_set_memory_debug)(void) = MEM_lockfree_set_memory_debug;
 size_t (*MEM_get_memory_in_use)(void) = MEM_lockfree_get_memory_in_use;
+size_t (*MEM_get_mapped_memory_in_use)(void) = MEM_lockfree_get_mapped_memory_in_use;
 unsigned int (*MEM_get_memory_blocks_in_use)(void) = MEM_lockfree_get_memory_blocks_in_use;
 void (*MEM_reset_peak_memory)(void) = MEM_lockfree_reset_peak_memory;
 size_t (*MEM_get_peak_memory)(void) = MEM_lockfree_get_peak_memory;
@@ -107,14 +111,17 @@ void MEM_use_guarded_allocator(void)
  MEM_mallocN = MEM_guarded_mallocN;
  MEM_malloc_arrayN = MEM_guarded_malloc_arrayN;
  MEM_mallocN_aligned = MEM_guarded_mallocN_aligned;
+  MEM_mapallocN = MEM_guarded_mapallocN;
  MEM_printmemlist_pydict = MEM_guarded_printmemlist_pydict;
  MEM_printmemlist = MEM_guarded_printmemlist;
  MEM_callbackmemlist = MEM_guarded_callbackmemlist;
  MEM_printmemlist_stats = MEM_guarded_printmemlist_stats;
  MEM_set_error_callback = MEM_guarded_set_error_callback;
  MEM_consistency_check = MEM_guarded_consistency_check;
+  MEM_set_lock_callback = MEM_guarded_set_lock_callback;
  MEM_set_memory_debug = MEM_guarded_set_memory_debug;
  MEM_get_memory_in_use = MEM_guarded_get_memory_in_use;
+  MEM_get_mapped_memory_in_use = MEM_guarded_get_mapped_memory_in_use;
  MEM_get_memory_blocks_in_use = MEM_guarded_get_memory_blocks_in_use;
  MEM_reset_peak_memory = MEM_guarded_reset_peak_memory;
  MEM_get_peak_memory = MEM_guarded_get_peak_memory;
--- a/intern/guardedalloc/intern/mallocn_guarded_impl.c
+++ b/intern/guardedalloc/intern/mallocn_guarded_impl.c
@@ -28,8 +28,6 @@
 #include <string.h> /* memcpy */
 #include <sys/types.h>

-#include <pthread.h>
-
 #include "MEM_guardedalloc.h"

 /* to ensure strict conversions */
@@ -52,6 +50,17 @@

 //#define DEBUG_MEMCOUNTER

+/* Only for debugging:
+ * defining DEBUG_THREADS will enable check whether memory manager
+ * is locked with a mutex when allocation is called from non-main
+ * thread.
+ *
+ * This helps troubleshooting memory issues caused by the fact
+ * guarded allocator is not thread-safe, however this check will
+ * fail to check allocations from openmp threads.
+ */
+//#define DEBUG_THREADS
+
 /* Only for debugging:
 * Defining DEBUG_BACKTRACE will store a backtrace from where
 * memory block was allocated and print this trace for all
@@ -95,7 +104,7 @@ typedef struct MemHead {
  const char *name;
  const char *nextname;
  int tag2;
-  short pad1;
+  short mmap;      /* if true, memory was mmapped */
  short alignment; /* if non-zero aligned alloc was used
                    * and alignment is stored here.
                    */
@@ -115,6 +124,24 @@ typedef struct MemHead {

 typedef MemHead MemHeadAligned;

+/* for openmp threading asserts, saves time troubleshooting
+ * we may need to extend this if blender code starts using MEM_
+ * functions inside OpenMP correctly with omp_set_lock() */
+
+#if 0 /* disable for now, only use to debug openmp code which doesn lock threads for malloc */
+#  if defined(_OPENMP) && defined(DEBUG)
+#    include <assert.h>
+#    include <omp.h>
+#    define DEBUG_OMP_MALLOC
+#  endif
+#endif
+
+#ifdef DEBUG_THREADS
+#  include <assert.h>
+#  include <pthread.h>
+static pthread_t mainid;
+#endif
+
 #ifdef DEBUG_BACKTRACE
 #  if defined(__linux__) || defined(__APPLE__)
 #    include <execinfo.h>
@@ -160,11 +187,13 @@ static const char *check_memlist(MemHead *memh);
 /* --------------------------------------------------------------------- */

 static unsigned int totblock = 0;
-static size_t mem_in_use = 0, peak_mem = 0;
+static size_t mem_in_use = 0, mmap_in_use = 0, peak_mem = 0;

 static volatile struct localListBase _membase;
 static volatile struct localListBase *membase = &_membase;
 static void (*error_callback)(const char *) = NULL;
+static void (*thread_lock_callback)(void) = NULL;
+static void (*thread_unlock_callback)(void) = NULL;

 static bool malloc_debug_memset = false;

@@ -204,16 +233,40 @@ print_error(const char *str, ...)
    fputs(buf, stderr);
 }

-static pthread_mutex_t thread_lock = PTHREAD_MUTEX_INITIALIZER;
-
 static void mem_lock_thread(void)
 {
-  pthread_mutex_lock(&thread_lock);
+#ifdef DEBUG_THREADS
+  static int initialized = 0;
+
+  if (initialized == 0) {
+    /* assume first allocation happens from main thread */
+    mainid = pthread_self();
+    initialized = 1;
+  }
+
+  if (!pthread_equal(pthread_self(), mainid) && thread_lock_callback == NULL) {
+    assert(!"Memory function is called from non-main thread without lock");
+  }
+#endif
+
+#ifdef DEBUG_OMP_MALLOC
+  assert(omp_in_parallel() == 0);
+#endif
+
+  if (thread_lock_callback)
+    thread_lock_callback();
 }

 static void mem_unlock_thread(void)
 {
-  pthread_mutex_unlock(&thread_lock);
+#ifdef DEBUG_THREADS
+  if (!pthread_equal(pthread_self(), mainid) && thread_lock_callback == NULL) {
+    assert(!"Thread lock was removed while allocation from thread is in progress");
+  }
+#endif
+
+  if (thread_unlock_callback)
+    thread_unlock_callback();
 }

 bool MEM_guarded_consistency_check(void)
@@ -234,6 +287,12 @@ void MEM_guarded_set_error_callback(void (*func)(const char *))
  error_callback = func;
 }

+void MEM_guarded_set_lock_callback(void (*lock)(void), void (*unlock)(void))
+{
+  thread_lock_callback = lock;
+  thread_unlock_callback = unlock;
+}
+
 void MEM_guarded_set_memory_debug(void)
 {
  malloc_debug_memset = true;
@@ -261,8 +320,10 @@ void *MEM_guarded_dupallocN(const void *vmemh)
    memh--;

 #ifndef DEBUG_MEMDUPLINAME
-    if (LIKELY(memh->alignment == 0))
-      newp = MEM_guarded_mallocN(memh->len, "dupli_alloc");
+    if (UNLIKELY(memh->mmap))
+      newp = MEM_guarded_mapallocN(memh->len, "dupli_mapalloc");
+    else if (LIKELY(memh->alignment == 0))
+      newp = MEM_guarded_mapallocN(memh->len, "dupli_mapalloc");
    else
      newp = MEM_guarded_mallocN_aligned(memh->len, (size_t)memh->alignment, "dupli_alloc");

@@ -273,7 +334,11 @@ void *MEM_guarded_dupallocN(const void *vmemh)
      MemHead *nmemh;
      char *name = malloc(strlen(memh->name) + 24);

-      if (LIKELY(memh->alignment == 0)) {
+      if (UNLIKELY(memh->mmap)) {
+        sprintf(name, "%s %s", "dupli_mapalloc", memh->name);
+        newp = MEM_guarded_mapallocN(memh->len, name);
+      }
+      else if (LIKELY(memh->alignment == 0)) {
        sprintf(name, "%s %s", "dupli_alloc", memh->name);
        newp = MEM_guarded_mallocN(memh->len, name);
      }
@@ -413,7 +478,7 @@ static void make_memhead_header(MemHead *memh, size_t len, const char *str)
  memh->name = str;
  memh->nextname = NULL;
  memh->len = len;
-  memh->pad1 = 0;
+  memh->mmap = 0;
  memh->alignment = 0;
  memh->tag2 = MEMTAG2;

@@ -581,6 +646,58 @@ void *MEM_guarded_calloc_arrayN(size_t len, size_t size, const char *str)
  return MEM_guarded_callocN(total_size, str);
 }

+/* note; mmap returns zero'd memory */
+void *MEM_guarded_mapallocN(size_t len, const char *str)
+{
+  MemHead *memh;
+
+  /* on 64 bit, simply use calloc instead, as mmap does not support
+   * allocating > 4 GB on Windows. the only reason mapalloc exists
+   * is to get around address space limitations in 32 bit OSes. */
+  if (sizeof(void *) >= 8)
+    return MEM_guarded_callocN(len, str);
+
+  len = SIZET_ALIGN_4(len);
+
+#if defined(WIN32)
+  /* our windows mmap implementation is not thread safe */
+  mem_lock_thread();
+#endif
+  memh = mmap(NULL,
+              len + sizeof(MemHead) + sizeof(MemTail),
+              PROT_READ | PROT_WRITE,
+              MAP_SHARED | MAP_ANON,
+              -1,
+              0);
+#if defined(WIN32)
+  mem_unlock_thread();
+#endif
+
+  if (memh != (MemHead *)-1) {
+    make_memhead_header(memh, len, str);
+    memh->mmap = 1;
+    atomic_add_and_fetch_z(&mmap_in_use, len);
+    mem_lock_thread();
+    peak_mem = mmap_in_use > peak_mem ? mmap_in_use : peak_mem;
+    mem_unlock_thread();
+#ifdef DEBUG_MEMCOUNTER
+    if (_mallocn_count == DEBUG_MEMCOUNTER_ERROR_VAL)
+      memcount_raise(__func__);
+    memh->_count = _mallocn_count++;
+#endif
+    return (++memh);
+  }
+  else {
+    print_error(
+        "Mapalloc returns null, fallback to regular malloc: "
+        "len=" SIZET_FORMAT " in %s, total %u\n",
+        SIZET_ARG(len),
+        str,
+        (unsigned int)mmap_in_use);
+    return MEM_guarded_callocN(len, str);
+  }
+}
+
 /* Memory statistics print */
 typedef struct MemPrintBlock {
  const char *name;
@@ -648,7 +765,7 @@ void MEM_guarded_printmemlist_stats(void)
    pb++;

 #ifdef USE_MALLOC_USABLE_SIZE
-    if (membl->alignment == 0) {
+    if (!membl->mmap && membl->alignment == 0) {
      mem_in_use_slop += (sizeof(MemHead) + sizeof(MemTail) + malloc_usable_size((void *)membl)) -
                         membl->len;
    }
@@ -981,13 +1098,27 @@ static void rem_memblock(MemHead *memh)
    free((char *)memh->name);
 #endif

-  if (UNLIKELY(malloc_debug_memset && memh->len))
-    memset(memh + 1, 255, memh->len);
-  if (LIKELY(memh->alignment == 0)) {
-    free(memh);
+  if (memh->mmap) {
+    atomic_sub_and_fetch_z(&mmap_in_use, memh->len);
+#if defined(WIN32)
+    /* our windows mmap implementation is not thread safe */
+    mem_lock_thread();
+#endif
+    if (munmap(memh, memh->len + sizeof(MemHead) + sizeof(MemTail)))
+      printf("Couldn't unmap memory %s\n", memh->name);
+#if defined(WIN32)
+    mem_unlock_thread();
+#endif
  }
  else {
-    aligned_free(MEMHEAD_REAL_PTR(memh));
+    if (UNLIKELY(malloc_debug_memset && memh->len))
+      memset(memh + 1, 255, memh->len);
+    if (LIKELY(memh->alignment == 0)) {
+      free(memh);
+    }
+    else {
+      aligned_free(MEMHEAD_REAL_PTR(memh));
+    }
  }
 }

@@ -1139,6 +1270,17 @@ size_t MEM_guarded_get_memory_in_use(void)
  return _mem_in_use;
 }

+size_t MEM_guarded_get_mapped_memory_in_use(void)
+{
+  size_t _mmap_in_use;
+
+  mem_lock_thread();
+  _mmap_in_use = mmap_in_use;
+  mem_unlock_thread();
+
+  return _mmap_in_use;
+}
+
 unsigned int MEM_guarded_get_memory_blocks_in_use(void)
 {
  unsigned int _totblock;
--- a/intern/guardedalloc/intern/mallocn_intern.h
+++ b/intern/guardedalloc/intern/mallocn_intern.h
@@ -24,6 +24,13 @@
 #ifndef __MALLOCN_INTERN_H__
 #define __MALLOCN_INTERN_H__

+/* mmap exception */
+#if defined(WIN32)
+#  include "mmap_win.h"
+#else
+#  include <sys/mman.h>
+#endif
+
 #ifdef __GNUC__
 #  define UNUSED(x) UNUSED_##x __attribute__((__unused__))
 #else
@@ -133,14 +140,19 @@ void *MEM_lockfree_mallocN_aligned(size_t len,
                                   size_t alignment,
                                   const char *UNUSED(str)) ATTR_MALLOC ATTR_WARN_UNUSED_RESULT
    ATTR_ALLOC_SIZE(1) ATTR_NONNULL(3);
+void *MEM_lockfree_mapallocN(size_t len,
+                             const char *UNUSED(str)) ATTR_MALLOC ATTR_WARN_UNUSED_RESULT
+    ATTR_ALLOC_SIZE(1) ATTR_NONNULL(2);
 void MEM_lockfree_printmemlist_pydict(void);
 void MEM_lockfree_printmemlist(void);
 void MEM_lockfree_callbackmemlist(void (*func)(void *));
 void MEM_lockfree_printmemlist_stats(void);
 void MEM_lockfree_set_error_callback(void (*func)(const char *));
 bool MEM_lockfree_consistency_check(void);
+void MEM_lockfree_set_lock_callback(void (*lock)(void), void (*unlock)(void));
 void MEM_lockfree_set_memory_debug(void);
 size_t MEM_lockfree_get_memory_in_use(void);
+size_t MEM_lockfree_get_mapped_memory_in_use(void);
 unsigned int MEM_lockfree_get_memory_blocks_in_use(void);
 void MEM_lockfree_reset_peak_memory(void);
 size_t MEM_lockfree_get_peak_memory(void) ATTR_WARN_UNUSED_RESULT;
@@ -176,14 +188,19 @@ void *MEM_guarded_mallocN_aligned(size_t len,
                                  size_t alignment,
                                  const char *UNUSED(str)) ATTR_MALLOC ATTR_WARN_UNUSED_RESULT
    ATTR_ALLOC_SIZE(1) ATTR_NONNULL(3);
+void *MEM_guarded_mapallocN(size_t len,
+                            const char *UNUSED(str)) ATTR_MALLOC ATTR_WARN_UNUSED_RESULT
+    ATTR_ALLOC_SIZE(1) ATTR_NONNULL(2);
 void MEM_guarded_printmemlist_pydict(void);
 void MEM_guarded_printmemlist(void);
 void MEM_guarded_callbackmemlist(void (*func)(void *));
 void MEM_guarded_printmemlist_stats(void);
 void MEM_guarded_set_error_callback(void (*func)(const char *));
 bool MEM_guarded_consistency_check(void);
+void MEM_guarded_set_lock_callback(void (*lock)(void), void (*unlock)(void));
 void MEM_guarded_set_memory_debug(void);
 size_t MEM_guarded_get_memory_in_use(void);
+size_t MEM_guarded_get_mapped_memory_in_use(void);
 unsigned int MEM_guarded_get_memory_blocks_in_use(void);
 void MEM_guarded_reset_peak_memory(void);
 size_t MEM_guarded_get_peak_memory(void) ATTR_WARN_UNUSED_RESULT;
--- a/intern/guardedalloc/intern/mallocn_lockfree_impl.c
+++ b/intern/guardedalloc/intern/mallocn_lockfree_impl.c
@@ -44,18 +44,22 @@ typedef struct MemHeadAligned {
 } MemHeadAligned;

 static unsigned int totblock = 0;
-static size_t mem_in_use = 0, peak_mem = 0;
+static size_t mem_in_use = 0, mmap_in_use = 0, peak_mem = 0;
 static bool malloc_debug_memset = false;

 static void (*error_callback)(const char *) = NULL;
+static void (*thread_lock_callback)(void) = NULL;
+static void (*thread_unlock_callback)(void) = NULL;

 enum {
-  MEMHEAD_ALIGN_FLAG = 1,
+  MEMHEAD_MMAP_FLAG = 1,
+  MEMHEAD_ALIGN_FLAG = 2,
 };

 #define MEMHEAD_FROM_PTR(ptr) (((MemHead *)ptr) - 1)
 #define PTR_FROM_MEMHEAD(memhead) (memhead + 1)
 #define MEMHEAD_ALIGNED_FROM_PTR(ptr) (((MemHeadAligned *)ptr) - 1)
+#define MEMHEAD_IS_MMAP(memhead) ((memhead)->len & (size_t)MEMHEAD_MMAP_FLAG)
 #define MEMHEAD_IS_ALIGNED(memhead) ((memhead)->len & (size_t)MEMHEAD_ALIGN_FLAG)

 /* Uncomment this to have proper peak counter. */
@@ -89,10 +93,24 @@ print_error(const char *str, ...)
  }
 }

+#if defined(WIN32)
+static void mem_lock_thread(void)
+{
+  if (thread_lock_callback)
+    thread_lock_callback();
+}
+
+static void mem_unlock_thread(void)
+{
+  if (thread_unlock_callback)
+    thread_unlock_callback();
+}
+#endif
+
 size_t MEM_lockfree_allocN_len(const void *vmemh)
 {
  if (vmemh) {
-    return MEMHEAD_FROM_PTR(vmemh)->len & ~((size_t)(MEMHEAD_ALIGN_FLAG));
+    return MEMHEAD_FROM_PTR(vmemh)->len & ~((size_t)(MEMHEAD_MMAP_FLAG | MEMHEAD_ALIGN_FLAG));
  }
  else {
    return 0;
@@ -115,15 +133,29 @@ void MEM_lockfree_freeN(void *vmemh)
  atomic_sub_and_fetch_u(&totblock, 1);
  atomic_sub_and_fetch_z(&mem_in_use, len);

-  if (UNLIKELY(malloc_debug_memset && len)) {
-    memset(memh + 1, 255, len);
-  }
-  if (UNLIKELY(MEMHEAD_IS_ALIGNED(memh))) {
-    MemHeadAligned *memh_aligned = MEMHEAD_ALIGNED_FROM_PTR(vmemh);
-    aligned_free(MEMHEAD_REAL_PTR(memh_aligned));
+  if (MEMHEAD_IS_MMAP(memh)) {
+    atomic_sub_and_fetch_z(&mmap_in_use, len);
+#if defined(WIN32)
+    /* our windows mmap implementation is not thread safe */
+    mem_lock_thread();
+#endif
+    if (munmap(memh, len + sizeof(MemHead)))
+      printf("Couldn't unmap memory\n");
+#if defined(WIN32)
+    mem_unlock_thread();
+#endif
  }
  else {
-    free(memh);
+    if (UNLIKELY(malloc_debug_memset && len)) {
+      memset(memh + 1, 255, len);
+    }
+    if (UNLIKELY(MEMHEAD_IS_ALIGNED(memh))) {
+      MemHeadAligned *memh_aligned = MEMHEAD_ALIGNED_FROM_PTR(vmemh);
+      aligned_free(MEMHEAD_REAL_PTR(memh_aligned));
+    }
+    else {
+      free(memh);
+    }
  }
 }

@@ -133,7 +165,10 @@ void *MEM_lockfree_dupallocN(const void *vmemh)
  if (vmemh) {
    MemHead *memh = MEMHEAD_FROM_PTR(vmemh);
    const size_t prev_size = MEM_lockfree_allocN_len(vmemh);
-    if (UNLIKELY(MEMHEAD_IS_ALIGNED(memh))) {
+    if (UNLIKELY(MEMHEAD_IS_MMAP(memh))) {
+      newp = MEM_lockfree_mapallocN(prev_size, "dupli_mapalloc");
+    }
+    else if (UNLIKELY(MEMHEAD_IS_ALIGNED(memh))) {
      MemHeadAligned *memh_aligned = MEMHEAD_ALIGNED_FROM_PTR(vmemh);
      newp = MEM_lockfree_mallocN_aligned(
          prev_size, (size_t)memh_aligned->alignment, "dupli_malloc");
@@ -362,6 +397,47 @@ void *MEM_lockfree_mallocN_aligned(size_t len, size_t alignment, const char *str
  return NULL;
 }

+void *MEM_lockfree_mapallocN(size_t len, const char *str)
+{
+  MemHead *memh;
+
+  /* on 64 bit, simply use calloc instead, as mmap does not support
+   * allocating > 4 GB on Windows. the only reason mapalloc exists
+   * is to get around address space limitations in 32 bit OSes. */
+  if (sizeof(void *) >= 8)
+    return MEM_lockfree_callocN(len, str);
+
+  len = SIZET_ALIGN_4(len);
+
+#if defined(WIN32)
+  /* our windows mmap implementation is not thread safe */
+  mem_lock_thread();
+#endif
+  memh = mmap(NULL, len + sizeof(MemHead), PROT_READ | PROT_WRITE, MAP_SHARED | MAP_ANON, -1, 0);
+#if defined(WIN32)
+  mem_unlock_thread();
+#endif
+
+  if (memh != (MemHead *)-1) {
+    memh->len = len | (size_t)MEMHEAD_MMAP_FLAG;
+    atomic_add_and_fetch_u(&totblock, 1);
+    atomic_add_and_fetch_z(&mem_in_use, len);
+    atomic_add_and_fetch_z(&mmap_in_use, len);
+
+    update_maximum(&peak_mem, mem_in_use);
+    update_maximum(&peak_mem, mmap_in_use);
+
+    return PTR_FROM_MEMHEAD(memh);
+  }
+  print_error(
+      "Mapalloc returns null, fallback to regular malloc: "
+      "len=" SIZET_FORMAT " in %s, total %u\n",
+      SIZET_ARG(len),
+      str,
+      (unsigned int)mmap_in_use);
+  return MEM_lockfree_callocN(len, str);
+}
+
 void MEM_lockfree_printmemlist_pydict(void)
 {
 }
@@ -400,6 +476,12 @@ bool MEM_lockfree_consistency_check(void)
  return true;
 }

+void MEM_lockfree_set_lock_callback(void (*lock)(void), void (*unlock)(void))
+{
+  thread_lock_callback = lock;
+  thread_unlock_callback = unlock;
+}
+
 void MEM_lockfree_set_memory_debug(void)
 {
  malloc_debug_memset = true;
@@ -410,6 +492,11 @@ size_t MEM_lockfree_get_memory_in_use(void)
  return mem_in_use;
 }

+size_t MEM_lockfree_get_mapped_memory_in_use(void)
+{
+  return mmap_in_use;
+}
+
 unsigned int MEM_lockfree_get_memory_blocks_in_use(void)
 {
  return totblock;
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Jacques Lucke	925166bed6	continue with derived node tree	2020-05-05 15:13:02 +02:00
Jacques Lucke	dd1a90353b	rename	2020-05-05 14:35:27 +02:00
Jacques Lucke	04c05a028c	DerivedNodeTree data structure	2020-05-05 14:31:43 +02:00
Jacques Lucke	22d68d0f26	fix memory leak	2020-05-04 17:41:29 +02:00
Jacques Lucke	220bcdaa15	Add dot exporter	2020-05-04 17:35:26 +02:00
Jacques Lucke	a9051f4106	initial NodeTreeRef generation	2020-05-04 17:15:20 +02:00
Jacques Lucke	a600e76a58	Merge branch 'master' into node-tree-ref	2020-05-04 17:13:12 +02:00
Jacques Lucke	723b12bbca	add inline functions	2020-05-04 16:03:00 +02:00
Jacques Lucke	7ca85442ad	initial node tree ref declarations	2020-04-28 15:03:44 +02:00