Painstakingly merge branch 'master' into temp_viewport_fx_merge

This branch will not make it into blender, however there are a few things that might be useful here for people who work on the viewport project and could be merged to master:

* Matrix stacks
There is working code for matrix stacks in gpu_matrix.c. Basically it works by overriding all draw calls with a special version that loads our custom stack in OpenGL. To refactor this out, coders should do this in three stages:

1) Inspect all calls to GPUBegin/GPUDraw*/GPURect* and only add one call to upload the matrices once to the relevant uniform of the shader, then substitute with the normal GL call.

2) Remove all hardcoded GL matrices from shaders. Shaders should use generic uniforms/uniform blocks for matrices

3) Do not use GL-style stacks. Actually stacks should be generic and local, not global. We do not care about a global projection/modelview etc matrix, rather we want to add matrix stacks locally in places where they are needed and upload those matrices to our shaders. The code can be generalized to do that by substituting the enum in the gpuMatrix* functions with a GPUMatrixStack* though it will be slightly tedious to do - like everything viewport related.

* GLU removal:

This branch also removes glu completely from blender by providing a helper library that handles drawing of some helper shapes. The way this is handled is not ideal, since it uses the GSOC code that is not the way we want to do things here, however it could be improved and made with a proper design.

There are some graphical glitches still here, but should be fixable.

CMakeLists.txt

@@ -172,7 +172,6 @@ if(UNIX AND NOT APPLE)
 	set(_init_CODEC_FFMPEG                   OFF)
 	set(_init_CYCLES_OSL                     OFF)
 	set(_init_IMAGE_OPENEXR                  OFF)
-	set(_init_INPUT_NDOF                     OFF)
 	set(_init_JACK                           OFF)
 	set(_init_OPENCOLLADA                    OFF)
 	set(_init_OPENCOLORIO                    OFF)
@@ -218,6 +217,10 @@ if(${CMAKE_VERSION} VERSION_LESS 2.8.8)
 	# add_library OBJECT arg unsupported
 	set(WITH_BUILDINFO OFF)
 endif()
+set(BUILDINFO_OVERRIDE_DATE "" CACHE STRING "Use instead of the current date for reproducable builds (empty string disables this option)")
+set(BUILDINFO_OVERRIDE_TIME "" CACHE STRING "Use instead of the current time for reproducable builds (empty string disables this option)")
+mark_as_advanced(BUILDINFO_OVERRIDE_DATE)
+mark_as_advanced(BUILDINFO_OVERRIDE_TIME)
 
 option(WITH_IK_ITASC      "Enable ITASC IK solver (only disable for development & for incompatible C++ compilers)" ON)
 option(WITH_IK_SOLVER     "Enable Legacy IK solver (only disable for development)" ON)
@@ -226,6 +229,11 @@ option(WITH_BULLET        "Enable Bullet (Physics Engine)" ON)
 option(WITH_SYSTEM_BULLET "Use the systems bullet library (currently unsupported due to missing features in upstream!)" )
 mark_as_advanced(WITH_SYSTEM_BULLET)
 option(WITH_GAMEENGINE    "Enable Game Engine" ${_init_GAMEENGINE})
+if(APPLE)
+	set(WITH_GAMEENGINE_DECKLINK OFF)
+else()
+	option(WITH_GAMEENGINE_DECKLINK "Support BlackMagicDesign DeckLink cards in the Game Engine" ON)
+endif()
 option(WITH_PLAYER        "Build Player" OFF)
 option(WITH_OPENCOLORIO   "Enable OpenColorIO color management" ${_init_OPENCOLORIO})
 
@@ -270,6 +278,7 @@ endif()
 if(WITH_X11)
 	option(WITH_X11_XINPUT    "Enable X11 Xinput (tablet support and unicode input)"  ON)
 	option(WITH_X11_XF86VMODE "Enable X11 video mode switching"                       ON)
+	option(WITH_X11_ALPHA     "Enable X11 transparent background"                     ON)
 endif()
 
 if(UNIX AND NOT APPLE)
@@ -452,12 +461,6 @@ mark_as_advanced(
 	WITH_GL_PROFILE_ES20
 )
 
-if(WITH_GL_PROFILE_COMPAT)
-	set(WITH_GLU ON)
-else()
-	set(WITH_GLU OFF)
-endif()
-
 if(WIN32)
 	option(WITH_GL_ANGLE "Link with the ANGLE library, an OpenGL ES 2.0 implementation based on Direct3D, instead of the system OpenGL library." OFF)
 	mark_as_advanced(WITH_GL_ANGLE)
@@ -474,9 +477,19 @@ if(WIN32)
 endif()
 
 # Experimental support of C11 and C++11
-option(WITH_C11 "Build with C11 standard enabled, for development use only!" OFF)
+#
+# We default options to whatever default standard in the current compiler.
+if(CMAKE_COMPILER_IS_GNUCC AND (NOT "${CMAKE_C_COMPILER_VERSION}" VERSION_LESS "6.0") AND (NOT WITH_CXX11))
+	set(_c11_init ON)
+	set(_cxx11_init ON)
+else()
+	set(_c11_init OFF)
+	set(_cxx11_init OFF)
+endif()
+
+option(WITH_C11 "Build with C11 standard enabled, for development use only!" ${_c11_init})
 mark_as_advanced(WITH_C11)
-option(WITH_CXX11 "Build with C++11 standard enabled, for development use only!" OFF)
+option(WITH_CXX11 "Build with C++11 standard enabled, for development use only!" ${_cxx11_init})
 mark_as_advanced(WITH_CXX11)
 
 # Dependency graph
@@ -515,8 +528,8 @@ if(APPLE)
 
 	if(NOT CMAKE_OSX_ARCHITECTURES)
 		set(CMAKE_OSX_ARCHITECTURES x86_64 CACHE STRING
-		"Choose the architecture you want to build Blender for: i386, x86_64 or ppc"
-		FORCE)
+			"Choose the architecture you want to build Blender for: i386, x86_64 or ppc"
+			FORCE)
 	endif()
 
 	if(NOT DEFINED OSX_SYSTEM)
@@ -526,15 +539,20 @@ if(APPLE)
 		        OUTPUT_STRIP_TRAILING_WHITESPACE)
 	endif()
 
-	# workaround for incorrect cmake xcode lookup for developer previews - XCODE_VERSION does not take xcode-select path into accout
-	# but would always look into /Applications/Xcode.app while dev versions are named Xcode<version>-DP<preview_number>
-	execute_process(COMMAND xcode-select --print-path  OUTPUT_VARIABLE XCODE_CHECK OUTPUT_STRIP_TRAILING_WHITESPACE)
+	# workaround for incorrect cmake xcode lookup for developer previews - XCODE_VERSION does not
+	# take xcode-select path into account but would always look  into /Applications/Xcode.app
+	# while dev versions are named Xcode<version>-DP<preview_number>
+	execute_process(
+	        COMMAND xcode-select --print-path
+	        OUTPUT_VARIABLE XCODE_CHECK OUTPUT_STRIP_TRAILING_WHITESPACE)
 	string(REPLACE "/Contents/Developer" "" XCODE_BUNDLE ${XCODE_CHECK}) # truncate to bundlepath in any case
 	
 	if(${CMAKE_GENERATOR} MATCHES "Xcode")
 	
-		if(${XCODE_VERSION} VERSION_GREATER 4.2) # earlier xcode has no bundled developer dir, no sense in getting xcode path from
-			string(SUBSTRING "${XCODE_CHECK}" 14 6 DP_NAME) # reduce to XCode name without dp extension
+		# earlier xcode has no bundled developer dir, no sense in getting xcode path from
+		if(${XCODE_VERSION} VERSION_GREATER 4.2) 
+			# reduce to XCode name without dp extension
+			string(SUBSTRING "${XCODE_CHECK}" 14 6 DP_NAME) 
 			if(${DP_NAME} MATCHES Xcode5)
 				set(XCODE_VERSION 5)
 			endif()
@@ -561,25 +579,30 @@ if(APPLE)
 	message(STATUS "Detected OS X ${OSX_SYSTEM} and Xcode ${XCODE_VERSION} at ${XCODE_BUNDLE}")
 
 	if(${XCODE_VERSION} VERSION_LESS 4.3)
-		set(CMAKE_OSX_SYSROOT /Developer/SDKs/MacOSX${OSX_SYSTEM}.sdk CACHE PATH "" FORCE)  # use guaranteed existing sdk
+		# use guaranteed existing sdk
+		set(CMAKE_OSX_SYSROOT /Developer/SDKs/MacOSX${OSX_SYSTEM}.sdk CACHE PATH "" FORCE)
 	else()
-		# note: xcode-select path could be ambigous, cause /Applications/Xcode.app/Contents/Developer or /Applications/Xcode.app would be allowed
+		# note: xcode-select path could be ambigous,
+		# cause /Applications/Xcode.app/Contents/Developer or /Applications/Xcode.app would be allowed
 		# so i use a selfcomposed bundlepath here  
 		set(OSX_SYSROOT_PREFIX ${XCODE_BUNDLE}/Contents/Developer/Platforms/MacOSX.platform)
 		message(STATUS "OSX_SYSROOT_PREFIX: " ${OSX_SYSROOT_PREFIX})
 		set(OSX_DEVELOPER_PREFIX /Developer/SDKs/MacOSX${OSX_SYSTEM}.sdk) # use guaranteed existing sdk
 		set(CMAKE_OSX_SYSROOT ${OSX_SYSROOT_PREFIX}/${OSX_DEVELOPER_PREFIX} CACHE PATH "" FORCE)
 		if(${CMAKE_GENERATOR} MATCHES "Xcode")
-			set(CMAKE_XCODE_ATTRIBUTE_SDKROOT macosx${OSX_SYSTEM}) # to silence sdk not found warning, just overrides CMAKE_OSX_SYSROOT
+			# to silence sdk not found warning, just overrides CMAKE_OSX_SYSROOT
+			set(CMAKE_XCODE_ATTRIBUTE_SDKROOT macosx${OSX_SYSTEM})
 		endif()
 	endif()
 
 	if(OSX_SYSTEM MATCHES 10.9)
-		set(CMAKE_FIND_ROOT_PATH ${CMAKE_OSX_SYSROOT}) # make sure syslibs and headers are looked up in sdk ( expecially for 10.9 openGL atm. )
+		# make sure syslibs and headers are looked up in sdk ( expecially for 10.9 openGL atm. )
+		set(CMAKE_FIND_ROOT_PATH ${CMAKE_OSX_SYSROOT})
 	endif()
 
 	if(NOT CMAKE_OSX_DEPLOYMENT_TARGET)
-		set(CMAKE_OSX_DEPLOYMENT_TARGET "10.6" CACHE STRING "" FORCE) # 10.6 is our min. target, if you use higher sdk, weak linking happens
+		# 10.6 is our min. target, if you use higher sdk, weak linking happens
+		set(CMAKE_OSX_DEPLOYMENT_TARGET "10.6" CACHE STRING "" FORCE)
 	endif()
 	
 	if(NOT ${CMAKE_GENERATOR} MATCHES "Xcode")
@@ -588,8 +611,6 @@ if(APPLE)
 		set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mmacosx-version-min=${CMAKE_OSX_DEPLOYMENT_TARGET}")
 		add_definitions("-DMACOSX_DEPLOYMENT_TARGET=${CMAKE_OSX_DEPLOYMENT_TARGET}")
 	endif()
-
-	option(WITH_LIBS10.5  "Use 10.5 libs (needed for 64bit builds)" OFF)
 endif()
 
 
@@ -597,7 +618,10 @@ endif()
 # Check for conflicting/unsupported configurations
 
 if(NOT WITH_BLENDER AND NOT WITH_PLAYER AND NOT WITH_CYCLES_STANDALONE)
-	message(FATAL_ERROR "At least one of WITH_BLENDER or WITH_PLAYER or WITH_CYCLES_STANDALONE must be enabled, nothing to do!")
+	message(FATAL_ERROR
+		"At least one of WITH_BLENDER or WITH_PLAYER or "
+		"WITH_CYCLES_STANDALONE must be enabled, nothing to do!"
+	)
 endif()
 
 if(NOT WITH_GAMEENGINE AND WITH_PLAYER)
@@ -661,7 +685,8 @@ if(NOT WITH_BOOST)
 	set_and_warn(WITH_OPENAL         OFF)  # depends on AUDASPACE
 	set_and_warn(WITH_GAMEENGINE     OFF)  # depends on AUDASPACE
 	set_and_warn(WITH_PLAYER         OFF)  # depends on GAMEENGINE
-elseif(WITH_CYCLES OR WITH_OPENIMAGEIO OR WITH_AUDASPACE OR WITH_INTERNATIONAL OR WITH_OPENVDB OR WITH_OPENCOLORIO OR WITH_MOD_BOOLEAN)
+elseif(WITH_CYCLES OR WITH_OPENIMAGEIO OR WITH_AUDASPACE OR WITH_INTERNATIONAL OR
+       WITH_OPENVDB OR WITH_OPENCOLORIO OR WITH_MOD_BOOLEAN)
 	# Keep enabled
 else()
 	# New dependency graph needs either Boost or C++11 for function bindings.
@@ -698,6 +723,7 @@ if(WITH_GHOST_SDL OR WITH_HEADLESS)
 	set(WITH_X11           OFF)
 	set(WITH_X11_XINPUT    OFF)
 	set(WITH_X11_XF86VMODE OFF)
+	set(WITH_X11_ALPHA     OFF)
 	set(WITH_GHOST_XDND    OFF)
 	set(WITH_INPUT_IME     OFF)
 endif()
@@ -758,18 +784,22 @@ endif()
 
 if(WITH_INTERNATIONAL)
 	if(NOT EXISTS "${CMAKE_SOURCE_DIR}/release/datafiles/locale/languages")
-		message(WARNING "Translation path '${CMAKE_SOURCE_DIR}/release/datafiles/locale' is missing, "
-						"This is a 'git submodule', which are known not to work with bridges to other version "
-						"control systems, disabling 'WITH_INTERNATIONAL'.")
+		message(WARNING
+			"Translation path '${CMAKE_SOURCE_DIR}/release/datafiles/locale' is missing, "
+			"This is a 'git submodule', which are known not to work with bridges to other version "
+			"control systems, disabling 'WITH_INTERNATIONAL'."
+		)
 		set(WITH_INTERNATIONAL OFF)
 	endif()
 endif()
 
 if(WITH_PYTHON)
 	if(NOT EXISTS "${CMAKE_SOURCE_DIR}/release/scripts/addons/modules")
-		message(WARNING "Addons path '${CMAKE_SOURCE_DIR}/release/scripts/addons' is missing, "
-						"This is a 'git submodule', which are known not to work with bridges to other version "
-						"control systems: * CONTINUING WITHOUT ADDONS *")
+		message(WARNING
+			"Addons path '${CMAKE_SOURCE_DIR}/release/scripts/addons' is missing, "
+			"This is a 'git submodule', which are known not to work with bridges to other version "
+			"control systems: * CONTINUING WITHOUT ADDONS *"
+		)
 	endif()
 endif()
 
@@ -803,21 +833,6 @@ set(PLATFORM_LINKFLAGS "")
 set(PLATFORM_LINKFLAGS_DEBUG "")
 
 
-# For alternate Python locations the commandline can be used to override detected/default cache settings, e.g:
-# On Unix:
-#   cmake ../blender \
-#         -D PYTHON_VERSION=3.5 \
-#         -D PYTHON_INCLUDE_DIR=/opt/py35/include/python3.5d \
-#         -D PYTHON_LIBRARY=/opt/py35/lib/libpython3.5d.so
-#
-# On Macs:
-#   cmake ../blender \
-#         -D PYTHON_INCLUDE_DIR=/System/Library/Frameworks/Python.framework/Versions/3.5/include/python3.5 \
-#         -D PYTHON_LIBPATH=/System/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/config \
-#         -G Xcode
-#
-# When changing any of this remember to update the notes in doc/build_systems/cmake.txt
-
 #-----------------------------------------------------------------------------
 #Platform specifics
 
@@ -848,6 +863,16 @@ if(WITH_X11)
 		endif()
 	endif()
 
+	if(WITH_X11_ALPHA)
+		find_library(X11_Xrender_LIB Xrender  ${X11_LIB_SEARCH_PATH})
+		mark_as_advanced(X11_Xrender_LIB)
+		if (X11_Xrender_LIB)
+			list(APPEND PLATFORM_LINKLIBS ${X11_Xrender_LIB})
+		else()
+			set(WITH_X11_ALPHA OFF)
+		endif()
+	endif()
+
 endif()
 
 
@@ -1019,14 +1044,12 @@ if(UNIX AND NOT APPLE)
 
 	if(WITH_INPUT_NDOF)
 		find_package_wrapper(Spacenav)
-		if(NOT SPACENAV_FOUND)
-			set(WITH_INPUT_NDOF OFF)
-		endif()
-
-		# use generic names within blenders buildsystem.
 		if(SPACENAV_FOUND)
+			# use generic names within blenders buildsystem.
 			set(NDOF_INCLUDE_DIRS ${SPACENAV_INCLUDE_DIRS})
 			set(NDOF_LIBRARIES ${SPACENAV_LIBRARIES})
+		else()
+			set(WITH_INPUT_NDOF OFF)
 		endif()
 	endif()
 
@@ -1040,7 +1063,11 @@ if(UNIX AND NOT APPLE)
 			if(${OSL_LIBRARY_VERSION_MAJOR} EQUAL "1" AND ${OSL_LIBRARY_VERSION_MINOR} LESS "6")
 				# Note: --whole-archive is needed to force loading of all symbols in liboslexec,
 				# otherwise LLVM is missing the osl_allocate_closure_component function
-				set(OSL_LIBRARIES ${OSL_OSLCOMP_LIBRARY} -Wl,--whole-archive ${OSL_OSLEXEC_LIBRARY} -Wl,--no-whole-archive ${OSL_OSLQUERY_LIBRARY})
+				set(OSL_LIBRARIES
+					${OSL_OSLCOMP_LIBRARY}
+					-Wl,--whole-archive ${OSL_OSLEXEC_LIBRARY}
+					-Wl,--no-whole-archive ${OSL_OSLQUERY_LIBRARY}
+				)
 			endif()
 		else()
 			message(STATUS "OSL not found, disabling it from Cycles")
@@ -1111,7 +1138,13 @@ if(UNIX AND NOT APPLE)
 			set(PUGIXML_LIBRARIES "")
 		endif()
 
-		set(OPENIMAGEIO_LIBRARIES ${OPENIMAGEIO_LIBRARIES} ${PNG_LIBRARIES} ${JPEG_LIBRARIES} ${ZLIB_LIBRARIES} ${BOOST_LIBRARIES})
+		set(OPENIMAGEIO_LIBRARIES
+			${OPENIMAGEIO_LIBRARIES}
+			${PNG_LIBRARIES}
+			${JPEG_LIBRARIES}
+			${ZLIB_LIBRARIES}
+			${BOOST_LIBRARIES}
+		)
 		set(OPENIMAGEIO_LIBPATH)  # TODO, remove and reference the absolute path everywhere
 		set(OPENIMAGEIO_DEFINITIONS "")
 
@@ -1152,7 +1185,9 @@ if(UNIX AND NOT APPLE)
 
 	if(WITH_LLVM OR WITH_SDL_DYNLOAD)
 		# Fix for conflict with Mesa llvmpipe
-		set(PLATFORM_LINKFLAGS "${PLATFORM_LINKFLAGS} -Wl,--version-script='${CMAKE_SOURCE_DIR}/source/creator/blender.map'")
+		set(PLATFORM_LINKFLAGS
+			"${PLATFORM_LINKFLAGS} -Wl,--version-script='${CMAKE_SOURCE_DIR}/source/creator/blender.map'"
+		)
 	endif()
 
 	if(WITH_OPENSUBDIV)
@@ -1255,7 +1290,10 @@ elseif(WIN32)
 		set(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} /SAFESEH:NO")
 		set(CMAKE_MODULE_LINKER_FLAGS "${CMAKE_MODULE_LINKER_FLAGS} /SAFESEH:NO")
 
-		list(APPEND PLATFORM_LINKLIBS ws2_32 vfw32 winmm kernel32 user32 gdi32 comdlg32 advapi32 shfolder shell32 ole32 oleaut32 uuid psapi Dbghelp)
+		list(APPEND PLATFORM_LINKLIBS
+			ws2_32 vfw32 winmm kernel32 user32 gdi32 comdlg32
+			advapi32 shfolder shell32 ole32 oleaut32 uuid psapi Dbghelp
+		)
 
 		if(WITH_INPUT_IME)
 			list(APPEND PLATFORM_LINKLIBS imm32)
@@ -1293,7 +1331,8 @@ elseif(WIN32)
 		set(CMAKE_CXX_FLAGS_RELWITHDEBINFO "${CMAKE_CXX_FLAGS_RELWITHDEBINFO} /MT")
 		set(CMAKE_C_FLAGS_RELWITHDEBINFO "${CMAKE_C_FLAGS_RELWITHDEBINFO} /MT")
 
-		set(PLATFORM_LINKFLAGS "/SUBSYSTEM:CONSOLE /STACK:2097152 /INCREMENTAL:NO /NODEFAULTLIB:msvcrt.lib /NODEFAULTLIB:msvcmrt.lib /NODEFAULTLIB:msvcurt.lib /NODEFAULTLIB:msvcrtd.lib")
+		set(PLATFORM_LINKFLAGS "/SUBSYSTEM:CONSOLE /STACK:2097152 /INCREMENTAL:NO ")
+		set(PLATFORM_LINKFLAGS "${PLATFORM_LINKFLAGS} /NODEFAULTLIB:msvcrt.lib /NODEFAULTLIB:msvcmrt.lib /NODEFAULTLIB:msvcurt.lib /NODEFAULTLIB:msvcrtd.lib ")
 
 		# Ignore meaningless for us linker warnings.
 		set(PLATFORM_LINKFLAGS "${PLATFORM_LINKFLAGS} /ignore:4049 /ignore:4217 /ignore:4221")
@@ -1347,9 +1386,9 @@ elseif(WIN32)
 		# Add each of our libraries to our cmake_prefix_path so find_package() could work
 		file(GLOB children RELATIVE ${LIBDIR} ${LIBDIR}/*)
 		foreach(child ${children})
-		if(IS_DIRECTORY ${LIBDIR}/${child})
-			list(APPEND CMAKE_PREFIX_PATH  ${LIBDIR}/${child})
-		endif()
+			if(IS_DIRECTORY ${LIBDIR}/${child})
+				list(APPEND CMAKE_PREFIX_PATH  ${LIBDIR}/${child})
+			endif()
 		endforeach()
 
 		set(ZLIB_INCLUDE_DIRS ${LIBDIR}/zlib/include)
@@ -1454,23 +1493,28 @@ elseif(WIN32)
 				set(OPENEXR_INCLUDE_DIRS ${OPENEXR_INCLUDE_DIR} ${OPENEXR}/include/OpenEXR)
 				set(OPENEXR_LIBPATH ${OPENEXR}/lib)
 				set(OPENEXR_LIBRARIES
-					optimized ${OPENEXR_LIBPATH}/Iex-2_2.lib debug ${OPENEXR_LIBPATH}/Iex-2_2_d.lib
-					optimized ${OPENEXR_LIBPATH}/Half.lib debug ${OPENEXR_LIBPATH}/Half_d.lib
-					optimized ${OPENEXR_LIBPATH}/IlmImf-2_2.lib debug ${OPENEXR_LIBPATH}/IlmImf-2_2_d.lib
-					optimized ${OPENEXR_LIBPATH}/Imath-2_2.lib debug ${OPENEXR_LIBPATH}/Imath-2_2_d.lib
-					optimized ${OPENEXR_LIBPATH}/IlmThread-2_2.lib debug ${OPENEXR_LIBPATH}/IlmThread-2_2_d.lib
+					optimized ${OPENEXR_LIBPATH}/Iex-2_2.lib
+					optimized ${OPENEXR_LIBPATH}/Half.lib
+					optimized ${OPENEXR_LIBPATH}/IlmImf-2_2.lib
+					optimized ${OPENEXR_LIBPATH}/Imath-2_2.lib
+					optimized ${OPENEXR_LIBPATH}/IlmThread-2_2.lib
+					debug ${OPENEXR_LIBPATH}/Iex-2_2_d.lib
+					debug ${OPENEXR_LIBPATH}/Half_d.lib
+					debug ${OPENEXR_LIBPATH}/IlmImf-2_2_d.lib
+					debug ${OPENEXR_LIBPATH}/Imath-2_2_d.lib
+					debug ${OPENEXR_LIBPATH}/IlmThread-2_2_d.lib
 				)
 			endif()
 		endif()
 
 		if(WITH_IMAGE_TIFF)
-		# Try to find tiff first then complain and set static and maybe wrong paths
-		find_package(TIFF)
-		if(NOT TIFF_FOUND)
-			message(WARNING "Using HARDCODED libtiff locations")
-			set(TIFF_LIBRARY ${LIBDIR}/tiff/lib/libtiff.lib)
-			set(TIFF_INCLUDE_DIR ${LIBDIR}/tiff/include)
-		endif()
+			# Try to find tiff first then complain and set static and maybe wrong paths
+			find_package(TIFF)
+			if(NOT TIFF_FOUND)
+				message(WARNING "Using HARDCODED libtiff locations")
+				set(TIFF_LIBRARY ${LIBDIR}/tiff/lib/libtiff.lib)
+				set(TIFF_INCLUDE_DIR ${LIBDIR}/tiff/include)
+			endif()
 		endif()
 
 		if(WITH_JACK)
@@ -1525,12 +1569,17 @@ elseif(WIN32)
 					set(BOOST_DEBUG_POSTFIX "vc140-mt-sgd-1_60.lib")
 				endif()
 				set(BOOST_LIBRARIES
-					optimized libboost_date_time-${BOOST_POSTFIX} optimized libboost_filesystem-${BOOST_POSTFIX}
+					optimized libboost_date_time-${BOOST_POSTFIX}
+					optimized libboost_filesystem-${BOOST_POSTFIX}
 					optimized libboost_regex-${BOOST_POSTFIX}
-					optimized libboost_system-${BOOST_POSTFIX} optimized libboost_thread-${BOOST_POSTFIX}
-					debug libboost_date_time-${BOOST_DEBUG_POSTFIX} debug libboost_filesystem-${BOOST_DEBUG_POSTFIX}
+					optimized libboost_system-${BOOST_POSTFIX}
+					optimized libboost_thread-${BOOST_POSTFIX}
+					debug libboost_date_time-${BOOST_DEBUG_POSTFIX}
+					debug libboost_filesystem-${BOOST_DEBUG_POSTFIX}
 					debug libboost_regex-${BOOST_DEBUG_POSTFIX}
-					debug libboost_system-${BOOST_DEBUG_POSTFIX} debug libboost_thread-${BOOST_DEBUG_POSTFIX})
+					debug libboost_system-${BOOST_DEBUG_POSTFIX}
+					debug libboost_thread-${BOOST_DEBUG_POSTFIX}
+				)
 				if(WITH_CYCLES_OSL)
 					set(BOOST_LIBRARIES ${BOOST_LIBRARIES}
 						optimized libboost_wave-${BOOST_POSTFIX}
@@ -1659,7 +1708,8 @@ elseif(WIN32)
 			#endif
 			int main(void) { return 0; }
 			" 
-			WITH_MINGW64)
+			WITH_MINGW64
+		)
 		
 		if(NOT DEFINED LIBDIR)
 			if(WITH_MINGW64)
@@ -1680,7 +1730,10 @@ elseif(WIN32)
 			message(FATAL_ERROR "Windows requires pre-compiled libs at: '${LIBDIR}'")
 		endif()
 
-		list(APPEND PLATFORM_LINKLIBS -lshell32 -lshfolder -lgdi32 -lmsvcrt -lwinmm -lmingw32 -lm -lws2_32 -lz -lstdc++ -lole32 -luuid -lwsock32 -lpsapi -ldbghelp)
+		list(APPEND PLATFORM_LINKLIBS
+			-lshell32 -lshfolder -lgdi32 -lmsvcrt -lwinmm -lmingw32 -lm -lws2_32
+			-lz -lstdc++ -lole32 -luuid -lwsock32 -lpsapi -ldbghelp
+		)
 
 		if(WITH_INPUT_IME)
 			list(APPEND PLATFORM_LINKLIBS -limm32)
@@ -1749,7 +1802,14 @@ elseif(WIN32)
 				${OPENCOLLADA}/include/opencollada/GeneratedSaxParser
 			)
 			set(OPENCOLLADA_LIBPATH ${OPENCOLLADA}/lib/opencollada)
-			set(OPENCOLLADA_LIBRARIES OpenCOLLADAStreamWriter OpenCOLLADASaxFrameworkLoader OpenCOLLADAFramework OpenCOLLADABaseUtils GeneratedSaxParser UTF MathMLSolver buffer ftoa xml)
+			set(OPENCOLLADA_LIBRARIES
+				OpenCOLLADAStreamWriter
+				OpenCOLLADASaxFrameworkLoader
+				OpenCOLLADAFramework
+				OpenCOLLADABaseUtils
+				GeneratedSaxParser
+				UTF MathMLSolver buffer ftoa xml
+			)
 			set(PCRE_LIBRARIES pcre)
 		endif()
 
@@ -1822,12 +1882,14 @@ elseif(WIN32)
 			if(WITH_INTERNATIONAL)
 				set(BOOST_LIBRARIES ${BOOST_LIBRARIES}
 					optimized boost_locale-${BOOST_POSTFIX}
-					debug boost_locale-${BOOST_DEBUG_POSTFIX}) 
+					debug boost_locale-${BOOST_DEBUG_POSTFIX}
+				)
 			endif()
 			if(WITH_CYCLES_OSL)
 				set(BOOST_LIBRARIES ${BOOST_LIBRARIES}
 					optimized boost_wave-${BOOST_POSTFIX}
-					debug boost_wave-${BOOST_DEBUG_POSTFIX}) 
+					debug boost_wave-${BOOST_DEBUG_POSTFIX}
+				)
 			endif()
 			set(BOOST_LIBPATH ${BOOST}/lib)
 			set(BOOST_DEFINITIONS "-DBOOST_ALL_NO_LIB -DBOOST_THREAD_USE_LIB ")
@@ -1908,7 +1970,7 @@ elseif(WIN32)
 		set(OPENAL ${LIBDIR}/openal)
 		set(OPENALDIR ${LIBDIR}/openal)
 		set(OPENAL_INCLUDE_DIR ${OPENAL}/include)
-		if(MSVC12)
+		if(MSVC)
 			set(OPENAL_LIBRARY openal32)
 		else()
 			set(OPENAL_LIBRARY wrap_oal)
@@ -1936,7 +1998,14 @@ elseif(WIN32)
 		find_library(OSL_LIB_EXEC_DEBUG NAMES oslexec_d PATHS ${CYCLES_OSL}/lib)
 		find_library(OSL_LIB_COMP_DEBUG NAMES oslcomp_d PATHS ${CYCLES_OSL}/lib)
 		find_library(OSL_LIB_QUERY_DEBUG NAMES oslquery_d PATHS ${CYCLES_OSL}/lib)
-		list(APPEND OSL_LIBRARIES optimized ${OSL_LIB_COMP} optimized ${OSL_LIB_EXEC} optimized ${OSL_LIB_QUERY} debug ${OSL_LIB_EXEC_DEBUG} debug ${OSL_LIB_COMP_DEBUG} debug ${OSL_LIB_QUERY_DEBUG})
+		list(APPEND OSL_LIBRARIES
+			optimized ${OSL_LIB_COMP}
+			optimized ${OSL_LIB_EXEC}
+			optimized ${OSL_LIB_QUERY}
+			debug ${OSL_LIB_EXEC_DEBUG}
+			debug ${OSL_LIB_COMP_DEBUG}
+			debug ${OSL_LIB_QUERY_DEBUG}
+		)
 		find_path(OSL_INCLUDE_DIR OSL/oslclosure.h PATHS ${CYCLES_OSL}/include)
 		find_program(OSL_COMPILER NAMES oslc PATHS ${CYCLES_OSL}/bin)
 	
@@ -1950,20 +2019,8 @@ elseif(WIN32)
 
 elseif(APPLE)
 
-	if(${CMAKE_OSX_DEPLOYMENT_TARGET} STREQUAL "10.5" OR ${CMAKE_OSX_DEPLOYMENT_TARGET} STRGREATER "10.5")
-		set(WITH_LIBS10.5 ON CACHE BOOL "Use 10.5 libs" FORCE) # valid also for 10.6/7/8/9
-	endif()
-
 	if(NOT DEFINED LIBDIR)
-		if(WITH_LIBS10.5)
-			set(LIBDIR ${CMAKE_SOURCE_DIR}/../lib/darwin-9.x.universal)
-		else()
-			if(CMAKE_OSX_ARCHITECTURES MATCHES i386)
-				set(LIBDIR ${CMAKE_SOURCE_DIR}/../lib/darwin-8.x.i386)
-			else()
-				set(LIBDIR ${CMAKE_SOURCE_DIR}/../lib/darwin-8.0.0-powerpc)
-			endif()
-		endif()
+		set(LIBDIR ${CMAKE_SOURCE_DIR}/../lib/darwin-9.x.universal)
 	else()
 		message(STATUS "Using pre-compiled LIBDIR: ${LIBDIR}")
 	endif()
@@ -2021,11 +2078,15 @@ elseif(APPLE)
 			# set(PYTHON_LINKFLAGS "-u _PyMac_Error")  # won't  build with this enabled
 		else()
 			# module must be compiled against Python framework
-			set(PYTHON_INCLUDE_DIR "/Library/Frameworks/Python.framework/Versions/${PYTHON_VERSION}/include/python${PYTHON_VERSION}m")
-			set(PYTHON_EXECUTABLE "/Library/Frameworks/Python.framework/Versions/${PYTHON_VERSION}/bin/python${PYTHON_VERSION}m")
+			set(_py_framework "/Library/Frameworks/Python.framework/Versions/${PYTHON_VERSION}")
+
+			set(PYTHON_INCLUDE_DIR "${_py_framework}/include/python${PYTHON_VERSION}m")
+			set(PYTHON_EXECUTABLE "${_py_framework}/bin/python${PYTHON_VERSION}m")
+			set(PYTHON_LIBPATH "${_py_framework}/lib/python${PYTHON_VERSION}/config-${PYTHON_VERSION}m")
 			#set(PYTHON_LIBRARY python${PYTHON_VERSION})
-			set(PYTHON_LIBPATH "/Library/Frameworks/Python.framework/Versions/${PYTHON_VERSION}/lib/python${PYTHON_VERSION}/config-${PYTHON_VERSION}m")
 			#set(PYTHON_LINKFLAGS "-u _PyMac_Error -framework Python")  # won't  build with this enabled
+
+			unset(_py_framework)
 		endif()
 		
 		# uncached vars
@@ -2067,7 +2128,10 @@ elseif(APPLE)
 	if(WITH_CODEC_FFMPEG)
 		set(FFMPEG ${LIBDIR}/ffmpeg)
 		set(FFMPEG_INCLUDE_DIRS ${FFMPEG}/include)
-		set(FFMPEG_LIBRARIES avcodec avdevice avformat avutil mp3lame swscale x264 xvidcore theora theoradec theoraenc vorbis vorbisenc vorbisfile ogg)
+		set(FFMPEG_LIBRARIES
+			avcodec avdevice avformat avutil
+			mp3lame swscale x264 xvidcore theora theoradec theoraenc vorbis vorbisenc vorbisfile ogg
+		)
 		set(FFMPEG_LIBPATH ${FFMPEG}/lib)
 	endif()
 
@@ -2078,13 +2142,13 @@ elseif(APPLE)
 	)
 	mark_as_advanced(SYSTEMSTUBS_LIBRARY)
 	if(SYSTEMSTUBS_LIBRARY)
-		list(APPEND PLATFORM_LINKLIBS stdc++ SystemStubs)
-	else()
-		list(APPEND PLATFORM_LINKLIBS stdc++)
+		list(APPEND PLATFORM_LINKLIBS SystemStubs)
 	endif()
 
 	set(PLATFORM_CFLAGS "-pipe -funsigned-char")
-	set(PLATFORM_LINKFLAGS "-fexceptions -framework CoreServices -framework Foundation -framework IOKit -framework AppKit -framework Cocoa -framework Carbon -framework AudioUnit -framework AudioToolbox -framework CoreAudio")
+	set(PLATFORM_LINKFLAGS
+		"-fexceptions -framework CoreServices -framework Foundation -framework IOKit -framework AppKit -framework Cocoa -framework Carbon -framework AudioUnit -framework AudioToolbox -framework CoreAudio"
+	)
 	if(WITH_CODEC_QUICKTIME)
 		set(PLATFORM_LINKFLAGS "${PLATFORM_LINKFLAGS} -framework QTKit")
 		if(CMAKE_OSX_ARCHITECTURES MATCHES i386)
@@ -2093,22 +2157,10 @@ elseif(APPLE)
 		endif()
 	endif()
 
-	# XXX - SOME MAC DEV PLEASE TEST WITH THE SDK INSTALLED!
-	# ALSO SHOULD BE MOVED INTO OWN MODULE WHEN FUNCTIONAL
-	if(WITH_INPUT_NDOF)
-		# This thread it *should* work and check the framework - campbell
-		# http://www.cmake.org/pipermail/cmake/2005-December/007740.html
-		find_library(3DCONNEXION_CLIENT_FRAMEWORK
-			NAMES 3DconnexionClient
-		)
-		if(NOT 3DCONNEXION_CLIENT_FRAMEWORK)
-			set(WITH_INPUT_NDOF OFF)
-		endif()
-
-		if(WITH_INPUT_NDOF)
-			set(PLATFORM_LINKFLAGS "${PLATFORM_LINKFLAGS} -F/Library/Frameworks -weak_framework 3DconnexionClient")
-			set(NDOF_INCLUDE_DIRS /Library/Frameworks/3DconnexionClient.framework/Headers )
-		endif()
+	if(WITH_CXX11)
+		list(APPEND PLATFORM_LINKLIBS c++)
+	else()
+		list(APPEND PLATFORM_LINKLIBS stdc++)
 	endif()
 
 	if(WITH_JACK)
@@ -2116,7 +2168,8 @@ elseif(APPLE)
 	endif()
 	
 	if(WITH_PYTHON_MODULE OR WITH_PYTHON_FRAMEWORK)
-		set(PLATFORM_LINKFLAGS "${PLATFORM_LINKFLAGS} /Library/Frameworks/Python.framework/Versions/${PYTHON_VERSION}/Python")# force cmake to link right framework
+		# force cmake to link right framework
+		set(PLATFORM_LINKFLAGS "${PLATFORM_LINKFLAGS} /Library/Frameworks/Python.framework/Versions/${PYTHON_VERSION}/Python")
 	endif()
 	
 	if(WITH_OPENCOLLADA)
@@ -2131,10 +2184,18 @@ elseif(APPLE)
 		)
 
 		set(OPENCOLLADA_LIBPATH ${OPENCOLLADA}/lib)
-		set(OPENCOLLADA_LIBRARIES "OpenCOLLADASaxFrameworkLoader -lOpenCOLLADAFramework -lOpenCOLLADABaseUtils -lOpenCOLLADAStreamWriter -lMathMLSolver -lGeneratedSaxParser -lxml2 -lbuffer -lftoa")
+		set(OPENCOLLADA_LIBRARIES
+			OpenCOLLADASaxFrameworkLoader
+			-lOpenCOLLADAFramework
+			-lOpenCOLLADABaseUtils
+			-lOpenCOLLADAStreamWriter
+			-lMathMLSolver
+			-lGeneratedSaxParser
+			-lxml2 -lbuffer -lftoa
+		)
 		# Use UTF functions from collada if LLVM is not enabled
 		if(NOT WITH_LLVM)
-			set(OPENCOLLADA_LIBRARIES "${OPENCOLLADA_LIBRARIES} -lUTF")
+			list(APPEND OPENCOLLADA_LIBRARIES -lUTF)
 		endif()
 		# pcre is bundled with openCollada
 		#set(PCRE ${LIBDIR}/pcre)
@@ -2169,14 +2230,17 @@ elseif(APPLE)
 		set(TIFF_LIBPATH ${TIFF}/lib)
 	endif()
 
-	if(WITH_INPUT_NDOF)
-		# linker needs "-weak_framework 3DconnexionClient"
-	endif()
-
 	if(WITH_BOOST)
 		set(BOOST ${LIBDIR}/boost)
 		set(BOOST_INCLUDE_DIR ${BOOST}/include)
-		set(BOOST_LIBRARIES boost_date_time-mt boost_filesystem-mt boost_regex-mt boost_system-mt boost_thread-mt boost_wave-mt)
+		set(BOOST_LIBRARIES
+			boost_date_time-mt
+			boost_filesystem-mt
+			boost_regex-mt
+			boost_system-mt
+			boost_thread-mt
+			boost_wave-mt
+		)
 		if(WITH_INTERNATIONAL)
 			list(APPEND BOOST_LIBRARIES boost_locale-mt)
 		endif()
@@ -2197,8 +2261,22 @@ elseif(APPLE)
 	if(WITH_OPENIMAGEIO)
 		set(OPENIMAGEIO ${LIBDIR}/openimageio)
 		set(OPENIMAGEIO_INCLUDE_DIRS ${OPENIMAGEIO}/include)
-		set(OPENIMAGEIO_LIBRARIES ${OPENIMAGEIO}/lib/libOpenImageIO.a ${PNG_LIBRARIES} ${JPEG_LIBRARIES} ${TIFF_LIBRARY} ${OPENEXR_LIBRARIES} ${ZLIB_LIBRARIES})
-		set(OPENIMAGEIO_LIBPATH ${OPENIMAGEIO}/lib ${JPEG_LIBPATH} ${PNG_LIBPATH} ${TIFF_LIBPATH} ${OPENEXR_LIBPATH} ${ZLIB_LIBPATH})
+		set(OPENIMAGEIO_LIBRARIES
+			${OPENIMAGEIO}/lib/libOpenImageIO.a
+			${PNG_LIBRARIES}
+			${JPEG_LIBRARIES}
+			${TIFF_LIBRARY}
+			${OPENEXR_LIBRARIES}
+			${ZLIB_LIBRARIES}
+		)
+		set(OPENIMAGEIO_LIBPATH
+			${OPENIMAGEIO}/lib
+			${JPEG_LIBPATH}
+			${PNG_LIBPATH}
+			${TIFF_LIBPATH}
+			${OPENEXR_LIBPATH}
+			${ZLIB_LIBPATH}
+		)
 		set(OPENIMAGEIO_DEFINITIONS "-DOIIO_STATIC_BUILD")
 		set(OPENIMAGEIO_IDIFF "${LIBDIR}/openimageio/bin/idiff")
 	endif()
@@ -2291,8 +2369,14 @@ elseif(APPLE)
 				include_directories(${LIBDIR}/openmp/include)
 				link_directories(${LIBDIR}/openmp/lib)
 				# This is a workaround for our helperbinaries ( datatoc, masgfmt, ... ),
-				# They are linked also to omp lib, so we need it in builddir for runtime exexcution, TODO: remove all unneeded dependencies from these
-				execute_process(COMMAND ditto -arch ${CMAKE_OSX_ARCHITECTURES} ${LIBDIR}/openmp/lib/libiomp5.dylib ${CMAKE_BINARY_DIR}/Resources/lib/libiomp5.dylib) # for intermediate binaries, in respect to lib ID
+				# They are linked also to omp lib, so we need it in builddir for runtime exexcution,
+				# TODO: remove all unneeded dependencies from these
+
+				# for intermediate binaries, in respect to lib ID
+				execute_process(
+				        COMMAND ditto -arch ${CMAKE_OSX_ARCHITECTURES}
+				        ${LIBDIR}/openmp/lib/libiomp5.dylib
+				        ${CMAKE_BINARY_DIR}/Resources/lib/libiomp5.dylib)
 			endif()
 		endif()
 	endif()
@@ -2318,7 +2402,14 @@ elseif(APPLE)
 		set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -ftemplate-depth=1024")
 	endif()
 	# Get rid of eventually clashes, we export some symbols explicite as local
-	set(PLATFORM_LINKFLAGS "${PLATFORM_LINKFLAGS} -Xlinker -unexported_symbols_list -Xlinker ${CMAKE_SOURCE_DIR}/source/creator/osx_locals.map")
+	set(PLATFORM_LINKFLAGS
+		"${PLATFORM_LINKFLAGS} -Xlinker -unexported_symbols_list -Xlinker ${CMAKE_SOURCE_DIR}/source/creator/osx_locals.map"
+	)
+
+	if(WITH_CXX11)
+		set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -stdlib=libc++")
+		set(PLATFORM_LINKFLAGS "${PLATFORM_LINKFLAGS} -stdlib=libc++")
+	endif()
 
 	# Suppress ranlib "has no symbols" warnings (workaround for T48250)
 	set(CMAKE_C_ARCHIVE_CREATE   "<CMAKE_AR> Scr <TARGET> <LINK_FLAGS> <OBJECTS>")
@@ -2336,38 +2427,50 @@ endif()
 
 if(WITH_CYCLES)
 	if(NOT WITH_OPENIMAGEIO)
-		message(FATAL_ERROR "Cycles requires WITH_OPENIMAGEIO, the library may not have been found. Configure OIIO or disable WITH_CYCLES")
+		message(FATAL_ERROR
+			"Cycles requires WITH_OPENIMAGEIO, the library may not have been found. "
+			"Configure OIIO or disable WITH_CYCLES"
+		)
 	endif()
 	if(NOT WITH_BOOST)
-		message(FATAL_ERROR "Cycles requires WITH_BOOST, the library may not have been found. Configure BOOST or disable WITH_CYCLES")
+		message(FATAL_ERROR
+			"Cycles requires WITH_BOOST, the library may not have been found. "
+			"Configure BOOST or disable WITH_CYCLES"
+		)
 	endif()
 
 	if(WITH_CYCLES_OSL)
 		if(NOT WITH_LLVM)
-			message(FATAL_ERROR "Cycles OSL requires WITH_LLVM, the library may not have been found. Configure LLVM or disable WITH_CYCLES_OSL")
+			message(FATAL_ERROR
+				"Cycles OSL requires WITH_LLVM, the library may not have been found. "
+				"Configure LLVM or disable WITH_CYCLES_OSL"
+			)
 		endif()
 	endif()
 endif()
 
 if(WITH_INTERNATIONAL)
 	if(NOT WITH_BOOST)
-		message(FATAL_ERROR "Internationalization requires WITH_BOOST, the library may not have been found. Configure BOOST or disable WITH_INTERNATIONAL")
+		message(FATAL_ERROR
+			"Internationalization requires WITH_BOOST, the library may not have been found. "
+			"Configure BOOST or disable WITH_INTERNATIONAL"
+		)
 	endif()
 endif()
 
 # See TEST_SSE_SUPPORT() for how this is defined.
 
-if(WITH_RAYOPTIMIZATION)
-	if(SUPPORT_SSE_BUILD)
-		set(PLATFORM_CFLAGS " ${COMPILER_SSE_FLAG} ${PLATFORM_CFLAGS}")
-		add_definitions(-D__SSE__ -D__MMX__)
-	endif()
-	if(SUPPORT_SSE2_BUILD)
-		set(PLATFORM_CFLAGS " ${COMPILER_SSE2_FLAG} ${PLATFORM_CFLAGS}")
-		add_definitions(-D__SSE2__)
-		if(NOT SUPPORT_SSE_BUILD) # dont double up
-			add_definitions(-D__MMX__)
-		endif()
+# Do it globally, SSE2 is required for quite some time now.
+# Doing it now allows to use SSE/SSE2 in inline headers.
+if(SUPPORT_SSE_BUILD)
+	set(PLATFORM_CFLAGS " ${COMPILER_SSE_FLAG} ${PLATFORM_CFLAGS}")
+	add_definitions(-D__SSE__ -D__MMX__)
+endif()
+if(SUPPORT_SSE2_BUILD)
+	set(PLATFORM_CFLAGS " ${COMPILER_SSE2_FLAG} ${PLATFORM_CFLAGS}")
+	add_definitions(-D__SSE2__)
+	if(NOT SUPPORT_SSE_BUILD) # dont double up
+		add_definitions(-D__MMX__)
 	endif()
 endif()
 
@@ -2422,11 +2525,6 @@ endif()
 find_package(OpenGL)
 blender_include_dirs_sys("${OPENGL_INCLUDE_DIR}")
 
-if(WITH_GLU)
-	list(APPEND BLENDER_GL_LIBRARIES "${OPENGL_glu_LIBRARY}")
-	list(APPEND GL_DEFINITIONS -DWITH_GLU)
-endif()
-
 if(WITH_SYSTEM_GLES)
 	find_package_wrapper(OpenGLES)
 endif()
@@ -2437,7 +2535,10 @@ if(WITH_GL_PROFILE_COMPAT OR WITH_GL_PROFILE_CORE)
 elseif(WITH_GL_PROFILE_ES20)
 	if(WITH_SYSTEM_GLES)
 		if(NOT OPENGLES_LIBRARY)
-			message(FATAL_ERROR "Unable to find OpenGL ES libraries.  Install them or disable WITH_SYSTEM_GLES.")
+			message(FATAL_ERROR
+				"Unable to find OpenGL ES libraries. "
+				"Install them or disable WITH_SYSTEM_GLES."
+			)
 		endif()
 
 		list(APPEND BLENDER_GL_LIBRARIES OPENGLES_LIBRARY)
@@ -2449,7 +2550,10 @@ elseif(WITH_GL_PROFILE_ES20)
 		list(APPEND BLENDER_GL_LIBRARIES "${OPENGLES_LIBRARY}")
 
 		if(NOT OPENGLES_LIBRARY)
-			message(FATAL_ERROR "To compile WITH_GL_EGL you need to set OPENGLES_LIBRARY to the file path of an OpenGL ES 2.0 library.")
+			message(FATAL_ERROR
+				"To compile WITH_GL_EGL you need to set OPENGLES_LIBRARY "
+				"to the file path of an OpenGL ES 2.0 library."
+			)
 		endif()
 
 	endif()
@@ -2461,7 +2565,10 @@ elseif(WITH_GL_PROFILE_ES20)
 		mark_as_advanced(OPENGLES_DLL)
 
 		if(NOT OPENGLES_DLL)
-			message(FATAL_ERROR "To compile WITH_GL_PROFILE_ES20 you need to set OPENGLES_DLL to the file path of an OpenGL ES 2.0 runtime dynamic link library (DLL).")
+			message(FATAL_ERROR
+				"To compile WITH_GL_PROFILE_ES20 you need to set OPENGLES_DLL to the file "
+				"path of an OpenGL ES 2.0 runtime dynamic link library (DLL)."
+			)
 		endif()
 
 		if(WITH_GL_ANGLE)
@@ -2475,7 +2582,10 @@ elseif(WITH_GL_PROFILE_ES20)
 			mark_as_advanced(D3DCOMPILER_DLL)
 
 			if(D3DCOMPILER_DLL STREQUAL "")
-				message(FATAL_ERROR "To compile WITH_GL_ANGLE you need to set D3DCOMPILER_DLL to the file path of a copy of the DirectX redistributable DLL file: D3DCompiler_46.dll")
+				message(FATAL_ERROR
+					"To compile WITH_GL_ANGLE you need to set D3DCOMPILER_DLL to the file "
+					"path of a copy of the DirectX redistributable DLL file: D3DCompiler_46.dll"
+				)
 			endif()
 
 		endif()
@@ -2489,7 +2599,10 @@ if(WITH_GL_EGL)
 
 	if(WITH_SYSTEM_GLES)
 		if(NOT OPENGLES_EGL_LIBRARY)
-			message(FATAL_ERROR "Unable to find OpenGL ES libraries.  Install them or disable WITH_SYSTEM_GLES.")
+			message(FATAL_ERROR
+				"Unable to find OpenGL ES libraries. "
+				"Install them or disable WITH_SYSTEM_GLES."
+			)
 		endif()
 
 		list(APPEND BLENDER_GL_LIBRARIES OPENGLES_EGL_LIBRARY)
@@ -2501,7 +2614,10 @@ if(WITH_GL_EGL)
 		list(APPEND BLENDER_GL_LIBRARIES "${OPENGLES_LIBRARY}" "${OPENGLES_EGL_LIBRARY}")
 
 		if(NOT OPENGLES_EGL_LIBRARY)
-			message(FATAL_ERROR "To compile WITH_GL_EGL you need to set OPENGLES_EGL_LIBRARY to the file path of an EGL library.")
+			message(FATAL_ERROR
+				"To compile WITH_GL_EGL you need to set OPENGLES_EGL_LIBRARY "
+				"to the file path of an EGL library."
+			)
 		endif()
 
 	endif()
@@ -2513,7 +2629,10 @@ if(WITH_GL_EGL)
 		mark_as_advanced(OPENGLES_EGL_DLL)
 
 		if(NOT OPENGLES_EGL_DLL)
-			message(FATAL_ERROR "To compile WITH_GL_EGL you need to set OPENGLES_EGL_DLL to the file path of an EGL runtime dynamic link library (DLL).")
+			message(FATAL_ERROR
+				"To compile WITH_GL_EGL you need to set OPENGLES_EGL_DLL "
+				"to the file path of an EGL runtime dynamic link library (DLL)."
+			)
 		endif()
 
 	endif()
@@ -2633,9 +2752,7 @@ else()
 
 endif()
 
-if(NOT WITH_GLU)
-	list(APPEND GL_DEFINITIONS -DGLEW_NO_GLU)
-endif()
+list(APPEND GL_DEFINITIONS -DGLEW_NO_GLU)
 
 #-----------------------------------------------------------------------------
 # Configure Bullet
@@ -2678,7 +2795,9 @@ endif()
 if(WITH_LIBMV)
 	set(CERES_DEFINES)
 
-	if(SHARED_PTR_FOUND)
+	if(WITH_CXX11)
+		# nothing to be done
+	elseif(SHARED_PTR_FOUND)
 		if(SHARED_PTR_TR1_MEMORY_HEADER)
 			list(APPEND CERES_DEFINES -DCERES_TR1_MEMORY_HEADER)
 		endif()
@@ -2689,7 +2808,9 @@ if(WITH_LIBMV)
 		message(FATAL_ERROR "Ceres: Unable to find shared_ptr.")
 	endif()
 
-	if(HAVE_STD_UNORDERED_MAP_HEADER)
+	if(WITH_CXX11)
+		list(APPEND CERES_DEFINES -DCERES_STD_UNORDERED_MAP)
+	elseif(HAVE_STD_UNORDERED_MAP_HEADER)
 		if(HAVE_UNORDERED_MAP_IN_STD_NAMESPACE)
 			list(APPEND CERES_DEFINES -DCERES_STD_UNORDERED_MAP)
 		else()
@@ -2891,10 +3012,12 @@ endif()
 # be most problematic.
 if(WITH_PYTHON)
 	if(NOT EXISTS "${PYTHON_INCLUDE_DIR}/Python.h")
-		message(FATAL_ERROR "Missing: \"${PYTHON_INCLUDE_DIR}/Python.h\",\n"
-							"Set the cache entry 'PYTHON_INCLUDE_DIR' to point "
-							"to a valid python include path. Containing "
-							"Python.h for python version \"${PYTHON_VERSION}\"")
+		message(FATAL_ERROR
+			"Missing: \"${PYTHON_INCLUDE_DIR}/Python.h\",\n"
+			"Set the cache entry 'PYTHON_INCLUDE_DIR' to point "
+			"to a valid python include path. Containing "
+			"Python.h for python version \"${PYTHON_VERSION}\""
+		)
 	endif()
 
 	if(WIN32 OR APPLE)
@@ -2912,12 +3035,22 @@ endif()
 
 if(WITH_CXX11)
 	if(CMAKE_COMPILER_IS_GNUCC OR CMAKE_C_COMPILER_ID MATCHES "Clang")
+		# TODO(sergey): Do we want c++11 or gnu-c++11 here?
 		set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
 	elseif(MSVC12)
 		# Nothing special is needed, C++11 features are available by default.
 	else()
 		message(FATAL_ERROR "Compiler ${CMAKE_C_COMPILER_ID} is not supported for C++11 build yet")
 	endif()
+else()
+	# GCC-6 switched to C++11 by default, which would break linking with existing libraries
+	# by default. So we explicitly disable C++11 for a new GCC so no linking issues happens.
+	if(CMAKE_COMPILER_IS_GNUCC AND (NOT "${CMAKE_C_COMPILER_VERSION}" VERSION_LESS "6.0"))
+		set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=gnu++98")
+		# We also disable any of C++11 ABI from usage, so we wouldn't even try to
+		# link to stuff from std::__cxx11 namespace.
+		add_definitions("-D_GLIBCXX_USE_CXX11_ABI=0")
+	endif()
 endif()
 
 # Visual Studio has all standards it supports available by default
@@ -3032,7 +3165,7 @@ if(FIRST_RUN)
 		string(LENGTH "${_msg}" _len)
 		while("32" GREATER "${_len}")
 			set(_msg "${_msg} ")
-			 math(EXPR _len "${_len} + 1")
+			math(EXPR _len "${_len} + 1")
 		endwhile()
 
 		set(_config_msg "${_config_msg}\n${_msg}${${_setting}}" PARENT_SCOPE)
@@ -3070,6 +3203,7 @@ if(FIRST_RUN)
 
 	info_cfg_text("System Options:")
 	info_cfg_option(WITH_INSTALL_PORTABLE)
+	info_cfg_option(WITH_X11_ALPHA)
 	info_cfg_option(WITH_X11_XF86VMODE)
 	info_cfg_option(WITH_X11_XINPUT)
 	info_cfg_option(WITH_MEM_JEMALLOC)
@@ -3117,7 +3251,6 @@ if(FIRST_RUN)
 
 	info_cfg_text("OpenGL:")
 	info_cfg_option(WITH_GLEW_ES)
-	info_cfg_option(WITH_GLU)
 	info_cfg_option(WITH_GL_EGL)
 	info_cfg_option(WITH_GL_PROFILE_COMPAT)
 	info_cfg_option(WITH_GL_PROFILE_CORE)

GNUmakefile

@@ -120,7 +120,7 @@ endif
 
 # -----------------------------------------------------------------------------
 # Build Blender
-all: FORCE
+all: .FORCE
 	@echo
 	@echo Configuring Blender in \"$(BUILD_DIR)\" ...
 
@@ -149,13 +149,13 @@ bpy: all
 
 # -----------------------------------------------------------------------------
 # Configuration (save some cd'ing around)
-config: FORCE
+config: .FORCE
 	$(CMAKE_CONFIG_TOOL) "$(BUILD_DIR)"
 
 
 # -----------------------------------------------------------------------------
 # Help for build targets
-help: FORCE
+help: .FORCE
 	@echo ""
 	@echo "Convenience targets provided for building blender, (multiple at once can be used)"
 	@echo "  * debug     - build a debug binary"
@@ -228,13 +228,13 @@ help: FORCE
 # -----------------------------------------------------------------------------
 # Packages
 #
-package_debian: FORCE
+package_debian: .FORCE
 	cd build_files/package_spec ; DEB_BUILD_OPTIONS="parallel=$(NPROCS)" sh ./build_debian.sh
 
-package_pacman: FORCE
+package_pacman: .FORCE
 	cd build_files/package_spec/pacman ; MAKEFLAGS="-j$(NPROCS)" makepkg
 
-package_archive: FORCE
+package_archive: .FORCE
 	make -C "$(BUILD_DIR)" -s package_archive
 	@echo archive in "$(BUILD_DIR)/release"
 
@@ -242,24 +242,24 @@ package_archive: FORCE
 # -----------------------------------------------------------------------------
 # Tests
 #
-test: FORCE
+test: .FORCE
 	cd $(BUILD_DIR) ; ctest . --output-on-failure
 
 # run pep8 check check on scripts we distribute.
-test_pep8: FORCE
+test_pep8: .FORCE
 	$(PYTHON) tests/python/pep8.py > test_pep8.log 2>&1
 	@echo "written: test_pep8.log"
 
 # run some checks on our cmakefiles.
-test_cmake: FORCE
+test_cmake: .FORCE
 	$(PYTHON) build_files/cmake/cmake_consistency_check.py > test_cmake_consistency.log 2>&1
 	@echo "written: test_cmake_consistency.log"
 
 # run deprecation tests, see if we have anything to remove.
-test_deprecated: FORCE
+test_deprecated: .FORCE
 	$(PYTHON) tests/check_deprecated.py
 
-test_style_c: FORCE
+test_style_c: .FORCE
 	# run our own checks on C/C++ style
 	PYTHONIOENCODING=utf_8 $(PYTHON) \
 	    "$(BLENDER_DIR)/source/tools/check_source/check_style_c.py" \
@@ -267,7 +267,7 @@ test_style_c: FORCE
 	    "$(BLENDER_DIR)/source/creator" \
 	    --no-length-check
 
-test_style_c_qtc: FORCE
+test_style_c_qtc: .FORCE
 	# run our own checks on C/C++ style
 	USE_QTC_TASK=1 \
 	PYTHONIOENCODING=utf_8 $(PYTHON) \
@@ -280,7 +280,7 @@ test_style_c_qtc: FORCE
 	@echo "written: test_style.tasks"
 
 
-test_style_osl: FORCE
+test_style_osl: .FORCE
 	# run our own checks on C/C++ style
 	PYTHONIOENCODING=utf_8 $(PYTHON) \
 	    "$(BLENDER_DIR)/source/tools/check_source/check_style_c.py" \
@@ -288,7 +288,7 @@ test_style_osl: FORCE
 	    "$(BLENDER_DIR)/release/scripts/templates_osl"
 
 
-test_style_osl_qtc: FORCE
+test_style_osl_qtc: .FORCE
 	# run our own checks on C/C++ style
 	USE_QTC_TASK=1 \
 	PYTHONIOENCODING=utf_8 $(PYTHON) \
@@ -303,13 +303,13 @@ test_style_osl_qtc: FORCE
 # Project Files
 #
 
-project_qtcreator: FORCE
+project_qtcreator: .FORCE
 	$(PYTHON) build_files/cmake/cmake_qtcreator_project.py "$(BUILD_DIR)"
 
-project_netbeans: FORCE
+project_netbeans: .FORCE
 	$(PYTHON) build_files/cmake/cmake_netbeans_project.py "$(BUILD_DIR)"
 
-project_eclipse: FORCE
+project_eclipse: .FORCE
 	cmake -G"Eclipse CDT4 - Unix Makefiles" -H"$(BLENDER_DIR)" -B"$(BUILD_DIR)"
 
 
@@ -317,40 +317,40 @@ project_eclipse: FORCE
 # Static Checking
 #
 
-check_cppcheck: FORCE
+check_cppcheck: .FORCE
 	$(CMAKE_CONFIG)
 	cd "$(BUILD_DIR)" ; \
 	$(PYTHON) "$(BLENDER_DIR)/build_files/cmake/cmake_static_check_cppcheck.py" 2> \
 	    "$(BLENDER_DIR)/check_cppcheck.txt"
 	@echo "written: check_cppcheck.txt"
 
-check_clang_array: FORCE
+check_clang_array: .FORCE
 	$(CMAKE_CONFIG)
 	cd "$(BUILD_DIR)" ; \
 	$(PYTHON) "$(BLENDER_DIR)/build_files/cmake/cmake_static_check_clang_array.py"
 
-check_splint: FORCE
+check_splint: .FORCE
 	$(CMAKE_CONFIG)
 	cd "$(BUILD_DIR)" ; \
 	$(PYTHON) "$(BLENDER_DIR)/build_files/cmake/cmake_static_check_splint.py"
 
-check_sparse: FORCE
+check_sparse: .FORCE
 	$(CMAKE_CONFIG)
 	cd "$(BUILD_DIR)" ; \
 	$(PYTHON) "$(BLENDER_DIR)/build_files/cmake/cmake_static_check_sparse.py"
 
-check_smatch: FORCE
+check_smatch: .FORCE
 	$(CMAKE_CONFIG)
 	cd "$(BUILD_DIR)" ; \
 	$(PYTHON) "$(BLENDER_DIR)/build_files/cmake/cmake_static_check_smatch.py"
 
-check_spelling_py: FORCE
+check_spelling_py: .FORCE
 	cd "$(BUILD_DIR)" ; \
 	PYTHONIOENCODING=utf_8 $(PYTHON) \
 	    "$(BLENDER_DIR)/source/tools/check_source/check_spelling.py" \
 	    "$(BLENDER_DIR)/release/scripts"
 
-check_spelling_c: FORCE
+check_spelling_c: .FORCE
 	cd "$(BUILD_DIR)" ; \
 	PYTHONIOENCODING=utf_8 $(PYTHON) \
 	    "$(BLENDER_DIR)/source/tools/check_source/check_spelling.py" \
@@ -359,7 +359,7 @@ check_spelling_c: FORCE
 	    "$(BLENDER_DIR)/intern/guardedalloc" \
 	    "$(BLENDER_DIR)/intern/ghost" \
 
-check_spelling_c_qtc: FORCE
+check_spelling_c_qtc: .FORCE
 	cd "$(BUILD_DIR)" ; USE_QTC_TASK=1 \
 	PYTHONIOENCODING=utf_8 $(PYTHON) \
 	    "$(BLENDER_DIR)/source/tools/check_source/check_spelling.py" \
@@ -370,13 +370,13 @@ check_spelling_c_qtc: FORCE
 	    > \
 	    "$(BLENDER_DIR)/check_spelling_c.tasks"
 
-check_spelling_osl: FORCE
+check_spelling_osl: .FORCE
 	cd "$(BUILD_DIR)" ;\
 	PYTHONIOENCODING=utf_8 $(PYTHON) \
 	    "$(BLENDER_DIR)/source/tools/check_source/check_spelling.py" \
 	    "$(BLENDER_DIR)/intern/cycles/kernel/shaders"
 
-check_descriptions: FORCE
+check_descriptions: .FORCE
 	"$(BUILD_DIR)/bin/blender" --background -noaudio --factory-startup --python \
 	    "$(BLENDER_DIR)/source/tools/check_source/check_descriptions.py"
 
@@ -384,14 +384,14 @@ check_descriptions: FORCE
 # Utilities
 #
 
-tgz: FORCE
+tgz: .FORCE
 	./build_files/utils/build_tgz.sh
 
-icons: FORCE
+icons: .FORCE
 	"$(BLENDER_DIR)/release/datafiles/blender_icons_update.py"
 	"$(BLENDER_DIR)/release/datafiles/prvicons_update.py"
 
-update: FORCE
+update: .FORCE
 	if [ -d "../lib" ]; then \
 		svn update ../lib/* ; \
 	fi
@@ -404,23 +404,23 @@ update: FORCE
 #
 
 # Simple version of ./doc/python_api/sphinx_doc_gen.sh with no PDF generation.
-doc_py: FORCE
+doc_py: .FORCE
 	"$(BUILD_DIR)/bin/blender" --background -noaudio --factory-startup --python doc/python_api/sphinx_doc_gen.py
 	cd doc/python_api ; sphinx-build -b html sphinx-in sphinx-out
 	@echo "docs written into: '$(BLENDER_DIR)/doc/python_api/sphinx-out/contents.html'"
 
-doc_doxy: FORCE
+doc_doxy: .FORCE
 	cd doc/doxygen; doxygen Doxyfile
 	@echo "docs written into: '$(BLENDER_DIR)/doc/doxygen/html/index.html'"
 
-doc_dna: FORCE
+doc_dna: .FORCE
 	"$(BUILD_DIR)/bin/blender" --background -noaudio --factory-startup --python doc/blender_file_format/BlendFileDnaExporter_25.py
 	@echo "docs written into: '$(BLENDER_DIR)/doc/blender_file_format/dna.html'"
 
-doc_man: FORCE
+doc_man: .FORCE
 	$(PYTHON) doc/manpage/blender.1.py "$(BUILD_DIR)/bin/blender"
 
-help_features: FORCE
+help_features: .FORCE
 	@$(PYTHON) -c \
 		"import re; \
 		print('\n'.join([ \
@@ -431,9 +431,9 @@ help_features: FORCE
 		if w.startswith('WITH_')]))" | uniq
 
 
-clean: FORCE
+clean: .FORCE
 	$(MAKE) -C "$(BUILD_DIR)" clean
 
 .PHONY: all
 
-FORCE:
+.FORCE:

build_files/buildbot/master.cfg

@@ -285,7 +285,7 @@ def generic_builder(id, libdir='', branch='', rsync=False):
                              maxsize=150 * 1024 * 1024,
                              workdir='install'))
     f.addStep(MasterShellCommand(name='unpack',
-                                 command=['python', unpack_script, filename],
+                                 command=['python2.7', unpack_script, filename],
                                  description='unpacking',
                                  descriptionDone='unpacked'))
     return f

build_files/buildbot/slave_compile.py

@@ -56,7 +56,6 @@ if 'cmake' in builder:
     chroot_name = None  # If not None command will be delegated to that chroot
     cuda_chroot_name = None  # If not None cuda compilationcommand will be delegated to that chroot
     build_cubins = True  # Whether to build Cycles CUDA kernels
-    remove_install_dir = False  # Remove installation folder before building
     bits = 64
 
     # Config file to be used (relative to blender's sources root)
@@ -70,19 +69,26 @@ if 'cmake' in builder:
     cuda_cmake_options = []
 
     if builder.startswith('mac'):
-        install_dir = None
         # Set up OSX architecture
         if builder.endswith('x86_64_10_6_cmake'):
             cmake_extra_options.append('-DCMAKE_OSX_ARCHITECTURES:STRING=x86_64')
         cmake_extra_options.append('-DCUDA_NVCC_EXECUTABLE=/usr/local/cuda-hack/bin/nvcc')
 
     elif builder.startswith('win'):
-        install_dir = None
-        if builder.startswith('win64'):
-            cmake_options.append(['-G', '"Visual Studio 12 2013 Win64"'])
-        elif builder.startswith('win32'):
-            bits = 32
-            cmake_options.append(['-G', '"Visual Studio 12 2013"'])
+        if builder.endswith('_vc2015'):
+            if builder.startswith('win64'):
+                cmake_options.extend(['-G', 'Visual Studio 14 2015 Win64'])
+            elif builder.startswith('win32'):
+                bits = 32
+                cmake_options.extend(['-G', 'Visual Studio 14 2015'])
+            cmake_extra_options.append('-DCUDA_NVCC_FLAGS=--cl-version;2013;' +
+                '--compiler-bindir;C:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\bin')
+        else:
+            if builder.startswith('win64'):
+                cmake_options.extend(['-G', 'Visual Studio 12 2013 Win64'])
+            elif builder.startswith('win32'):
+                bits = 32
+                cmake_options.extend(['-G', 'Visual Studio 12 2013'])
 
     elif builder.startswith('linux'):
         tokens = builder.split("_")
@@ -91,7 +97,6 @@ if 'cmake' in builder:
             deb_name = "jessie"
         elif glibc == 'glibc211':
             deb_name = "squeeze"
-        remove_install_dir = True
         cmake_config_file = "build_files/buildbot/config/blender_linux.cmake"
         cmake_player_config_file = "build_files/buildbot/config/blender_linux_player.cmake"
         if builder.endswith('x86_64_cmake'):
@@ -117,8 +122,7 @@ if 'cmake' in builder:
     if 'cuda' not in targets:
         cmake_options += cuda_cmake_options
 
-    if install_dir:
-        cmake_options.append("-DCMAKE_INSTALL_PREFIX=%s" % (install_dir))
+    cmake_options.append("-DCMAKE_INSTALL_PREFIX=%s" % (install_dir))
 
     cmake_options += cmake_extra_options
 
@@ -133,10 +137,8 @@ if 'cmake' in builder:
         cuda_chroot_prefix = chroot_prefix[:]
 
     # Make sure no garbage remained from the previous run
-    # (only do it if builder requested this)
-    if remove_install_dir:
-        if os.path.isdir(install_dir):
-            shutil.rmtree(install_dir)
+    if os.path.isdir(install_dir):
+        shutil.rmtree(install_dir)
 
     for target in targets:
         print("Building target %s" % (target))

build_files/buildbot/slave_pack.py

@@ -108,6 +108,8 @@ if builder.find('cmake') != -1:
                 platform += 'i386'
             elif builder.endswith('ppc_10_6_cmake'):
                 platform += 'ppc'
+        if builder.endswith('vc2015'):
+            platform += "-vc14"
         builderified_name = 'blender-{}-{}-{}'.format(blender_full_version, git_hash, platform)
         if branch != '':
             builderified_name = branch + "-" + builderified_name

build_files/cmake/buildinfo.cmake

@@ -1,5 +1,10 @@
-# This is called by cmake as an extermal process from
+# This is called by cmake as an external process from
 # ./source/creator/CMakeLists.txt to write ./source/creator/buildinfo.h
+# Caller must define:
+#   SOURCE_DIR
+# Optional overrides:
+#   BUILD_DATE
+#   BUILD_TIME
 
 # Extract working copy information for SOURCE_DIR into MY_XXX variables
 # with a default in case anything fails, for example when using git-svn
@@ -128,12 +133,19 @@ endif()
 # BUILD_PLATFORM and BUILD_PLATFORM are taken from CMake
 # but BUILD_DATE and BUILD_TIME are platform dependent
 if(UNIX)
-	execute_process(COMMAND date "+%Y-%m-%d" OUTPUT_VARIABLE BUILD_DATE OUTPUT_STRIP_TRAILING_WHITESPACE)
-	execute_process(COMMAND date "+%H:%M:%S" OUTPUT_VARIABLE BUILD_TIME OUTPUT_STRIP_TRAILING_WHITESPACE)
-endif()
-if(WIN32)
-	execute_process(COMMAND cmd /c date /t OUTPUT_VARIABLE BUILD_DATE OUTPUT_STRIP_TRAILING_WHITESPACE)
-	execute_process(COMMAND cmd /c time /t OUTPUT_VARIABLE BUILD_TIME OUTPUT_STRIP_TRAILING_WHITESPACE)
+	if(NOT BUILD_DATE)
+		execute_process(COMMAND date "+%Y-%m-%d" OUTPUT_VARIABLE BUILD_DATE OUTPUT_STRIP_TRAILING_WHITESPACE)
+	endif()
+	if(NOT BUILD_TIME)
+		execute_process(COMMAND date "+%H:%M:%S" OUTPUT_VARIABLE BUILD_TIME OUTPUT_STRIP_TRAILING_WHITESPACE)
+	endif()
+elseif(WIN32)
+	if(NOT BUILD_DATE)
+		execute_process(COMMAND cmd /c date /t OUTPUT_VARIABLE BUILD_DATE OUTPUT_STRIP_TRAILING_WHITESPACE)
+	endif()
+	if(NOT BUILD_TIME)
+		execute_process(COMMAND cmd /c time /t OUTPUT_VARIABLE BUILD_TIME OUTPUT_STRIP_TRAILING_WHITESPACE)
+	endif()
 endif()
 
 # Write a file with the BUILD_HASH define

build_files/cmake/config/blender_full.cmake

@@ -54,7 +54,7 @@ set(WITH_PLAYER              ON  CACHE BOOL "" FORCE)
 set(WITH_MEM_JEMALLOC        ON  CACHE BOOL "" FORCE)
 
 
-# platform dependant options
+# platform dependent options
 if(UNIX AND NOT APPLE)
 	set(WITH_JACK                ON  CACHE BOOL "" FORCE)
 	set(WITH_DOC_MANPAGE         ON  CACHE BOOL "" FORCE)

build_files/cmake/macros.cmake

@@ -435,9 +435,6 @@ function(setup_liblinks
 	if(WITH_MEM_JEMALLOC)
 		target_link_libraries(${target} ${JEMALLOC_LIBRARIES})
 	endif()
-	if(WITH_INPUT_NDOF)
-		target_link_libraries(${target} ${NDOF_LIBRARIES})
-	endif()
 	if(WITH_MOD_CLOTH_ELTOPO)
 		target_link_libraries(${target} ${LAPACK_LIBRARIES})
 	endif()
@@ -451,6 +448,9 @@ function(setup_liblinks
 		if(WITH_OPENMP_STATIC)
 			target_link_libraries(${target} ${OpenMP_LIBRARIES})
 		endif()
+		if(WITH_INPUT_NDOF)
+			target_link_libraries(${target} ${NDOF_LIBRARIES})
+		endif()
 	endif()
 
 	# We put CLEW and CUEW here because OPENSUBDIV_LIBRARIES dpeends on them..
@@ -487,6 +487,7 @@ function(SETUP_BLENDER_SORTED_LIBS)
 	if(WITH_CYCLES)
 		list(APPEND BLENDER_LINK_LIBS
 			cycles_render
+			cycles_graph
 			cycles_bvh
 			cycles_device
 			cycles_kernel
@@ -551,11 +552,11 @@ function(SETUP_BLENDER_SORTED_LIBS)
 		bf_modifiers
 		bf_bmesh
 		bf_gpu
+		bf_blenloader
 		bf_blenkernel
 		bf_physics
 		bf_nodes
 		bf_rna
-		bf_blenloader
 		bf_imbuf
 		bf_blenlib
 		bf_depsgraph
@@ -600,6 +601,7 @@ function(SETUP_BLENDER_SORTED_LIBS)
 		bf_intern_dualcon
 		bf_intern_cycles
 		cycles_render
+		cycles_graph
 		cycles_bvh
 		cycles_device
 		cycles_kernel
@@ -659,10 +661,6 @@ function(SETUP_BLENDER_SORTED_LIBS)
 		list(APPEND BLENDER_SORTED_LIBS bf_quicktime)
 	endif()
 
-	if(WITH_INPUT_NDOF)
-		list(APPEND BLENDER_SORTED_LIBS bf_intern_ghostndof3dconnexion)
-	endif()
-	
 	if(WITH_MOD_BOOLEAN)
 		list(APPEND BLENDER_SORTED_LIBS extern_carve)
 	endif()
@@ -687,6 +685,14 @@ function(SETUP_BLENDER_SORTED_LIBS)
 		list_insert_after(BLENDER_SORTED_LIBS "ge_logic_ngnetwork" "extern_bullet")
 	endif()
 
+	if(WITH_GAMEENGINE_DECKLINK)
+		list(APPEND BLENDER_SORTED_LIBS bf_intern_decklink)
+	endif()
+
+	if(WIN32)
+		list(APPEND BLENDER_SORTED_LIBS bf_intern_gpudirect)
+	endif()
+
 	if(WITH_OPENSUBDIV)
 		list(APPEND BLENDER_SORTED_LIBS bf_intern_opensubdiv)
 	endif()
@@ -803,7 +809,15 @@ macro(TEST_UNORDERED_MAP_SUPPORT)
 	#  UNORDERED_MAP_NAMESPACE, namespace for unordered_map, if found
 
 	include(CheckIncludeFileCXX)
-	CHECK_INCLUDE_FILE_CXX("unordered_map" HAVE_STD_UNORDERED_MAP_HEADER)
+
+	# Workaround for newer GCC (6.x+) where C++11 was enabled by default, which lead us
+	# to a situation when there is <unordered_map> include but which can't be used uless
+	# C++11 is enabled.
+	if(CMAKE_COMPILER_IS_GNUCC AND (NOT "${CMAKE_C_COMPILER_VERSION}" VERSION_LESS "6.0") AND (NOT WITH_CXX11))
+		set(HAVE_STD_UNORDERED_MAP_HEADER False)
+	else()
+		CHECK_INCLUDE_FILE_CXX("unordered_map" HAVE_STD_UNORDERED_MAP_HEADER)
+	endif()
 	if(HAVE_STD_UNORDERED_MAP_HEADER)
 		# Even so we've found unordered_map header file it doesn't
 		# mean unordered_map and unordered_set will be declared in
@@ -873,8 +887,16 @@ macro(TEST_SHARED_PTR_SUPPORT)
 	# otherwise it's assumed to be defined in std namespace.
 
 	include(CheckIncludeFileCXX)
+	include(CheckCXXSourceCompiles)
 	set(SHARED_PTR_FOUND FALSE)
-	CHECK_INCLUDE_FILE_CXX(memory HAVE_STD_MEMORY_HEADER)
+	# Workaround for newer GCC (6.x+) where C++11 was enabled by default, which lead us
+	# to a situation when there is <unordered_map> include but which can't be used uless
+	# C++11 is enabled.
+	if(CMAKE_COMPILER_IS_GNUCC AND (NOT "${CMAKE_C_COMPILER_VERSION}" VERSION_LESS "6.0") AND (NOT WITH_CXX11))
+		set(HAVE_STD_MEMORY_HEADER False)
+	else()
+		CHECK_INCLUDE_FILE_CXX(memory HAVE_STD_MEMORY_HEADER)
+	endif()
 	if(HAVE_STD_MEMORY_HEADER)
 		# Finding the memory header doesn't mean that shared_ptr is in std
 		# namespace.
@@ -882,7 +904,6 @@ macro(TEST_SHARED_PTR_SUPPORT)
 		# In particular, MSVC 2008 has shared_ptr declared in std::tr1.  In
 		# order to support this, we do an extra check to see which namespace
 		# should be used.
-		include(CheckCXXSourceCompiles)
 		CHECK_CXX_SOURCE_COMPILES("#include <memory>
 		                           int main() {
 		                             std::shared_ptr<int> int_ptr;
@@ -1050,6 +1071,19 @@ macro(remove_strict_flags_file
 
 endmacro()
 
+# External libs may need 'signed char' to be default.
+macro(remove_cc_flag_unsigned_char)
+	if(CMAKE_C_COMPILER_ID MATCHES "^(GNU|Clang|Intel)$")
+		remove_cc_flag("-funsigned-char")
+	elseif(MSVC)
+		remove_cc_flag("/J")
+	else()
+		message(WARNING
+			"Compiler '${CMAKE_C_COMPILER_ID}' failed to disable 'unsigned char' flag."
+			"Build files need updating."
+		)
+	endif()
+endmacro()
 
 function(ADD_CHECK_C_COMPILER_FLAG
 	_CFLAGS

build_files/scons/config/darwin-config.py

@@ -0,0 +1,263 @@
+import commands
+
+#############################################################################
+###################     Compiler & architecture settings      ##################
+#############################################################################
+
+MACOSX_ARCHITECTURE = 'x86_64' # valid archs: ppc, i386, ppc64, x86_64
+MACOSX_SDK='' # set an sdk name like '10.7' or leave empty for automatic choosing highest available
+MACOSX_DEPLOYMENT_TARGET = '10.6'
+
+# gcc always defaults to the system standard compiler linked by a shim or symlink
+CC = 'gcc'
+CXX = 'g++'
+LCGDIR = '#../lib/darwin-9.x.universal'
+LIBDIR = '${LCGDIR}'
+
+#############################################################################
+###################          Dependency settings           ##################
+#############################################################################
+
+# enable ffmpeg  support
+WITH_BF_FFMPEG = True
+BF_FFMPEG = LIBDIR + '/ffmpeg'
+BF_FFMPEG_INC = "${BF_FFMPEG}/include"
+BF_FFMPEG_LIBPATH='${BF_FFMPEG}/lib'
+BF_FFMPEG_LIB = 'avcodec avdevice avformat avutil mp3lame swscale x264 xvidcore theora theoradec theoraenc vorbis vorbisenc vorbisfile ogg bz2'
+#bz2 is a standard osx dynlib
+
+BF_PYTHON_VERSION = '3.4'
+WITH_OSX_STATICPYTHON = True
+
+# python 3.4 uses precompiled libraries in bf svn /lib by default
+BF_PYTHON = LIBDIR + '/python'
+BF_PYTHON_INC = '${BF_PYTHON}/include/python${BF_PYTHON_VERSION}m'
+# BF_PYTHON_BINARY = '${BF_PYTHON}/bin/python${BF_PYTHON_VERSION}'
+BF_PYTHON_LIB = 'python${BF_PYTHON_VERSION}m'
+BF_PYTHON_LIBPATH = '${BF_PYTHON}/lib/python${BF_PYTHON_VERSION}'
+# BF_PYTHON_LINKFLAGS = ['-u', '_PyMac_Error', '-framework', 'System']
+
+WITH_BF_OPENAL = True
+BF_OPENAL = LIBDIR + '/openal'
+
+WITH_BF_STATICOPENAL = False
+BF_OPENAL_INC = '${BF_OPENAL}/include' # only headers from libdir needed for proper use of framework !!!!
+#BF_OPENAL_LIB = 'openal'
+#BF_OPENAL_LIBPATH = '${BF_OPENAL}/lib'
+# Warning, this static lib configuration is untested! users of this OS please confirm.
+#BF_OPENAL_LIB_STATIC = '${BF_OPENAL}/lib/libopenal.a'
+
+# Warning, this static lib configuration is untested! users of this OS please confirm.
+BF_CXX = '/usr'
+WITH_BF_STATICCXX = False
+BF_CXX_LIB_STATIC = '${BF_CXX}/lib/libstdc++.a'
+
+# we use simply jack framework
+WITH_BF_JACK = True
+BF_JACK = '/Library/Frameworks/Jackmp.framework'
+BF_JACK_INC = '${BF_JACK}/headers'
+#BF_JACK_LIB = 'jack' # not used due framework
+BF_JACK_LIBPATH = '${BF_JACK}'
+
+WITH_BF_SNDFILE = True
+BF_SNDFILE = LIBDIR + '/sndfile'
+BF_SNDFILE_INC = '${BF_SNDFILE}/include'
+BF_SNDFILE_LIB = 'sndfile FLAC ogg vorbis vorbisenc'
+BF_SNDFILE_LIBPATH = '${BF_SNDFILE}/lib ${BF_FFMPEG}/lib' #ogg libs are stored in ffmpeg dir
+
+WITH_BF_SDL = True
+BF_SDL = LIBDIR + '/sdl' #$(shell sdl-config --prefix)
+BF_SDL_INC = '${BF_SDL}/include' #$(shell $(BF_SDL)/bin/sdl-config --cflags)
+BF_SDL_LIB = 'SDL2' #BF_SDL #$(shell $(BF_SDL)/bin/sdl-config --libs) -lSDL_mixer
+BF_SDL_LIBPATH = '${BF_SDL}/lib'
+
+WITH_BF_OPENEXR = True
+WITH_BF_STATICOPENEXR = False
+BF_OPENEXR = '${LCGDIR}/openexr'
+BF_OPENEXR_INC = '${BF_OPENEXR}/include ${BF_OPENEXR}/include/OpenEXR'
+BF_OPENEXR_LIB = ' Iex Half IlmImf Imath IlmThread'
+BF_OPENEXR_LIBPATH = '${BF_OPENEXR}/lib'
+# Warning, this static lib configuration is untested! users of this OS please confirm.
+BF_OPENEXR_LIB_STATIC = '${BF_OPENEXR}/lib/libHalf.a ${BF_OPENEXR}/lib/libIlmImf.a ${BF_OPENEXR}/lib/libIex.a ${BF_OPENEXR}/lib/libImath.a ${BF_OPENEXR}/lib/libIlmThread.a'
+
+WITH_BF_DDS = True
+
+WITH_BF_JPEG = True
+BF_JPEG = LIBDIR + '/jpeg'
+BF_JPEG_INC = '${BF_JPEG}/include'
+BF_JPEG_LIB = 'jpeg'
+BF_JPEG_LIBPATH = '${BF_JPEG}/lib'
+
+WITH_BF_OPENJPEG = True
+BF_OPENJPEG = '#extern/libopenjpeg'
+BF_OPENJPEG_LIB = ''
+BF_OPENJPEG_INC = '${BF_OPENJPEG}'
+BF_OPENJPEG_LIBPATH='${BF_OPENJPEG}/lib'
+
+WITH_BF_PNG = True
+BF_PNG = LIBDIR + '/png'
+BF_PNG_INC = '${BF_PNG}/include'
+BF_PNG_LIB = 'png'
+BF_PNG_LIBPATH = '${BF_PNG}/lib'
+
+WITH_BF_TIFF = True
+BF_TIFF = LIBDIR + '/tiff'
+BF_TIFF_INC = '${BF_TIFF}/include'
+BF_TIFF_LIB = 'tiff'
+BF_TIFF_LIBPATH = '${BF_TIFF}/lib'
+
+WITH_BF_ZLIB = True
+BF_ZLIB = '/usr'
+#BF_ZLIB_INC = '${BF_ZLIB}/include' # don't use this, it breaks -isysroot ${MACOSX_SDK}
+BF_ZLIB_LIB = 'z'
+
+WITH_BF_INTERNATIONAL = True
+
+WITH_BF_GAMEENGINE = True
+WITH_BF_PLAYER = True
+WITH_BF_OCEANSIM = True
+
+WITH_BF_BULLET = True
+BF_BULLET = '#extern/bullet2/src'
+BF_BULLET_INC = '${BF_BULLET}'
+BF_BULLET_LIB = 'extern_bullet'
+
+WITH_BF_FFTW3 = True
+BF_FFTW3 = LIBDIR + '/fftw3'
+BF_FFTW3_INC = '${BF_FFTW3}/include'
+BF_FFTW3_LIB = 'libfftw3'
+BF_FFTW3_LIBPATH = '${BF_FFTW3}/lib'
+
+BF_FREETYPE = LIBDIR + '/freetype'
+BF_FREETYPE_INC = '${BF_FREETYPE}/include ${BF_FREETYPE}/include/freetype2'
+BF_FREETYPE_LIB = 'freetype'
+BF_FREETYPE_LIBPATH = '${BF_FREETYPE}/lib'
+
+WITH_BF_QUICKTIME = True
+
+WITH_BF_ICONV = True
+BF_ICONV = '/usr'
+BF_ICONV_INC = '${BF_ICONV}/include'
+BF_ICONV_LIB = 'iconv'
+#BF_ICONV_LIBPATH = '${BF_ICONV}/lib'
+
+# Mesa Libs should go here if your using them as well....
+WITH_BF_STATICOPENGL = True
+BF_OPENGL_LIB = 'GL'
+BF_OPENGL_LIBPATH = '/System/Library/Frameworks/OpenGL.framework/Libraries'
+BF_OPENGL_LINKFLAGS = ['-framework', 'OpenGL']
+
+#OpenCollada flags
+WITH_BF_COLLADA = True
+BF_COLLADA = '#source/blender/collada'
+BF_COLLADA_INC = '${BF_COLLADA}'
+BF_COLLADA_LIB = 'bf_collada'
+BF_OPENCOLLADA = LIBDIR + '/opencollada'
+BF_OPENCOLLADA_INC = '${BF_OPENCOLLADA}/include'
+BF_OPENCOLLADA_LIB = 'OpenCOLLADASaxFrameworkLoader OpenCOLLADAFramework OpenCOLLADABaseUtils OpenCOLLADAStreamWriter MathMLSolver GeneratedSaxParser xml2 buffer ftoa'
+BF_OPENCOLLADA_LIBPATH = LIBDIR + '/opencollada'
+BF_PCRE = LIBDIR + '/opencollada'
+BF_PCRE_LIB = 'pcre'
+BF_PCRE_LIBPATH = '${BF_PCRE}/lib'
+#BF_EXPAT = '/usr'
+#BF_EXPAT_LIB = 'expat'
+#BF_EXPAT_LIBPATH = '/usr/lib'
+
+# Cycles
+WITH_BF_CYCLES = True
+
+#OSL
+
+WITH_BF_CYCLES_OSL = True
+BF_OSL = LIBDIR + '/osl'
+BF_OSL_INC = '${BF_OSL}/include'
+# note oslexec would passed via program linkflags, which is needed to
+# make llvm happy with osl_allocate_closure_component
+#BF_OSL_LIB = 'oslcomp oslquery'
+BF_OSL_LIBPATH = '${BF_OSL}/lib'
+BF_OSL_COMPILER = '${BF_OSL}/bin/oslc'
+
+WITH_BF_LLVM = True
+BF_LLVM = LIBDIR + '/llvm'
+BF_LLVM_LIB = 'LLVMBitReader LLVMJIT LLVMipo LLVMVectorize LLVMBitWriter LLVMX86CodeGen LLVMX86Desc LLVMX86Info LLVMX86AsmPrinter ' + \
+    'LLVMX86Utils LLVMSelectionDAG LLVMCodeGen LLVMScalarOpts LLVMInstCombine LLVMTransformUtils LLVMipa LLVMAnalysis LLVMExecutionEngine ' + \
+    'LLVMTarget LLVMMC LLVMCore LLVMSupport LLVMObject'
+BF_LLVM_LIBPATH = '${BF_LLVM}/lib'
+
+WITH_BF_OIIO = True
+BF_OIIO = LIBDIR + '/openimageio'
+BF_OIIO_INC = '${BF_OIIO}/include'
+BF_OIIO_LIB = 'OpenImageIO'
+BF_OIIO_LIBPATH = '${BF_OIIO}/lib'
+
+WITH_BF_OCIO = True
+BF_OCIO = LIBDIR + '/opencolorio'
+BF_OCIO_INC = '${BF_OCIO}/include'
+BF_OCIO_LIB = 'OpenColorIO tinyxml yaml-cpp'
+BF_OCIO_LIBPATH = '${BF_OCIO}/lib'
+
+WITH_BF_BOOST = True
+BF_BOOST = LIBDIR + '/boost'
+BF_BOOST_INC = '${BF_BOOST}/include'
+BF_BOOST_LIB = 'boost_date_time-mt boost_filesystem-mt boost_regex-mt boost_system-mt boost_thread-mt boost_wave-mt'
+BF_BOOST_LIB_INTERNATIONAL = 'boost_locale-mt'
+BF_BOOST_LIBPATH = '${BF_BOOST}/lib'
+
+WITH_BF_CYCLES_CUDA_BINARIES = False
+BF_CYCLES_CUDA_NVCC = '/usr/local/cuda/bin/nvcc'
+BF_CYCLES_CUDA_BINARIES_ARCH = ['sm_20', 'sm_21', 'sm_30', 'sm_35', 'sm_50', 'sm_52']
+
+#Freestyle
+WITH_BF_FREESTYLE = True
+
+#OpenMP ( will be checked for compiler support and turned off eventually )
+WITH_BF_OPENMP = True
+
+#Ray trace optimization
+WITH_BF_RAYOPTIMIZATION = True
+BF_RAYOPTIMIZATION_SSE_FLAGS = []
+
+# SpaceNavigator and related 3D mice, driver must be 3DxWare 10 Beta 4 (Mac OS X) or later !
+WITH_BF_3DMOUSE = True
+
+#############################################################################
+###################  various compile settings and flags    ##################
+#############################################################################
+
+BF_QUIET = '1' # suppress verbose output
+
+CFLAGS = []
+CXXFLAGS = []
+CCFLAGS = ['-pipe','-funsigned-char']
+CPPFLAGS = []
+
+PLATFORM_LINKFLAGS = ['-fexceptions','-framework','CoreServices','-framework','Foundation','-framework','IOKit','-framework','AppKit','-framework','Cocoa','-framework','Carbon','-framework','AudioUnit','-framework','AudioToolbox','-framework','CoreAudio','-framework','OpenAL']
+
+LLIBS = ['stdc++']
+
+REL_CFLAGS = []
+REL_CXXFLAGS = []
+REL_CCFLAGS = ['-O2']
+
+CC_WARN = ['-Wall']
+C_WARN = ['-Wno-char-subscripts', '-Wpointer-arith', '-Wcast-align', '-Wdeclaration-after-statement', '-Wno-unknown-pragmas', '-Wstrict-prototypes']
+CXX_WARN = ['-Wno-invalid-offsetof', '-Wno-sign-compare']
+
+##FIX_STUBS_WARNINGS = -Wno-unused
+
+##LOPTS = --dynamic
+##DYNLDFLAGS = -shared $(LDFLAGS)
+
+BF_PROFILE_CCFLAGS = ['-pg', '-g ']
+BF_PROFILE_LINKFLAGS = ['-pg']
+BF_PROFILE = False
+
+BF_DEBUG = False
+BF_DEBUG_CCFLAGS = ['-g']
+
+#############################################################################
+###################           Output directories           ##################
+#############################################################################
+
+BF_BUILDDIR='../build/darwin'
+BF_INSTALLDIR='../install/darwin'

build_files/scons/config/linux-config.py

@@ -0,0 +1,268 @@
+from FindPython import FindPython
+
+py = FindPython()
+
+BF_PYTHON_ABI_FLAGS = py['ABI_FLAGS']
+BF_PYTHON = py['PYTHON']
+BF_PYTHON_LIBPATH = py['LIBPATH']
+BF_PYTHON_LIBPATH_ARCH = py['LIBPATH_ARCH']
+BF_PYTHON_CONFIG = py['CONFIG']
+BF_PYTHON_VERSION = py['VERSION']
+WITH_BF_STATICPYTHON = False
+BF_PYTHON_INC = '${BF_PYTHON}/include/python${BF_PYTHON_VERSION}${BF_PYTHON_ABI_FLAGS} ' + BF_PYTHON_CONFIG
+BF_PYTHON_BINARY = '${BF_PYTHON}/bin/python${BF_PYTHON_VERSION}'
+BF_PYTHON_LIB = 'python${BF_PYTHON_VERSION}${BF_PYTHON_ABI_FLAGS}'  # BF_PYTHON+'/lib/python'+BF_PYTHON_VERSION+'/config/libpython'+BF_PYTHON_VERSION+'.a'
+BF_PYTHON_LINKFLAGS = ['-Xlinker', '-export-dynamic']
+BF_PYTHON_LIB_STATIC = '${BF_PYTHON_LIBPATH_ARCH}/libpython${BF_PYTHON_VERSION}${BF_PYTHON_ABI_FLAGS}.a'
+
+WITH_BF_OPENAL = True
+WITH_BF_STATICOPENAL = False
+BF_OPENAL = '/usr'
+BF_OPENAL_INC = '${BF_OPENAL}/include'
+BF_OPENAL_LIB = 'openal'
+BF_OPENAL_LIB_STATIC = '${BF_OPENAL}/lib/libopenal.a'
+
+BF_CXX = '/usr'
+WITH_BF_STATICCXX = False
+BF_CXX_LIB_STATIC = '${BF_CXX}/lib/libstdc++.a'
+
+WITH_BF_JACK = False
+BF_JACK = '/usr'
+BF_JACK_INC = '${BF_JACK}/include/jack'
+BF_JACK_LIB = 'jack'
+BF_JACK_LIBPATH = '${BF_JACK}/lib'
+
+WITH_BF_SNDFILE = False
+WITH_BF_STATICSNDFILE = False
+BF_SNDFILE = '/usr'
+BF_SNDFILE_INC = '${BF_SNDFILE}/include/sndfile'
+BF_SNDFILE_LIB = 'sndfile'
+BF_SNDFILE_LIBPATH = '${BF_SNDFILE}/lib'
+BF_SNDFILE_LIB_STATIC = '${BF_SNDFILE}/lib/libsndfile.a ${BF_OGG}/lib/libvorbis.a ${BF_OGG}/lib/libFLAC.a ${BF_OGG}/lib/libvorbisenc.a ${BF_OGG}/lib/libogg.a'
+
+WITH_BF_SDL = True
+BF_SDL = '/usr' #$(shell sdl-config --prefix)
+BF_SDL_INC = '${BF_SDL}/include/SDL' #$(shell $(BF_SDL)/bin/sdl-config --cflags)
+BF_SDL_LIB = 'SDL' #BF_SDL #$(shell $(BF_SDL)/bin/sdl-config --libs) -lSDL_mixer
+
+WITH_BF_OPENEXR = True
+WITH_BF_STATICOPENEXR = False
+BF_OPENEXR = '/usr'
+# when compiling with your own openexr lib you might need to set...
+# BF_OPENEXR_INC = '${BF_OPENEXR}/include/OpenEXR ${BF_OPENEXR}/include'
+
+BF_OPENEXR_INC = '${BF_OPENEXR}/include/OpenEXR'
+BF_OPENEXR_LIB = 'Half IlmImf Iex Imath '
+BF_OPENEXR_LIB_STATIC = '${BF_OPENEXR}/lib/libHalf.a ${BF_OPENEXR}/lib/libIlmImf.a ${BF_OPENEXR}/lib/libIex.a ${BF_OPENEXR}/lib/libImath.a ${BF_OPENEXR}/lib/libIlmThread.a'
+# BF_OPENEXR_LIBPATH = '${BF_OPENEXR}/lib'
+
+
+WITH_BF_DDS = True
+
+WITH_BF_JPEG = True
+BF_JPEG = '/usr'
+BF_JPEG_INC = '${BF_JPEG}/include'
+BF_JPEG_LIB = 'jpeg'
+
+WITH_BF_PNG = True
+BF_PNG = '/usr'
+BF_PNG_INC = '${BF_PNG}/include'
+BF_PNG_LIB = 'png'
+
+WITH_BF_TIFF = True
+BF_TIFF = '/usr'
+BF_TIFF_INC = '${BF_TIFF}/include'
+BF_TIFF_LIB = 'tiff'
+
+WITH_BF_ZLIB = True
+BF_ZLIB = '/usr'
+BF_ZLIB_INC = '${BF_ZLIB}/include'
+BF_ZLIB_LIB = 'z'
+
+WITH_BF_INTERNATIONAL = True
+
+WITH_BF_GAMEENGINE = True
+WITH_BF_PLAYER = True
+WITH_BF_OCEANSIM = True
+
+WITH_BF_BULLET = True
+BF_BULLET = '#extern/bullet2/src'
+BF_BULLET_INC = '${BF_BULLET}'
+BF_BULLET_LIB = 'extern_bullet'
+
+WITH_BF_ELTOPO = False
+BF_LAPACK = '/usr'
+BF_LAPACK_LIB = 'lapack3gf blas clapack'
+BF_LAPACK_LIBPATH = '${BF_LAPACK}/lib'
+
+BF_FREETYPE = '/usr'
+BF_FREETYPE_INC = '${BF_FREETYPE}/include ${BF_FREETYPE}/include/freetype2'
+BF_FREETYPE_LIB = 'freetype'
+#BF_FREETYPE_LIB_STATIC = '${BF_FREETYPE}/lib/libfreetype.a'
+
+WITH_BF_ICONV = False
+BF_ICONV = "/usr"
+BF_ICONV_INC = '${BF_ICONV}/include'
+BF_ICONV_LIB = 'iconv'
+BF_ICONV_LIBPATH = '${BF_ICONV}/lib'
+
+WITH_BF_BINRELOC = True
+
+# enable ffmpeg  support
+WITH_BF_FFMPEG = True
+BF_FFMPEG = '/usr'
+BF_FFMPEG_LIB = 'avformat avcodec swscale avutil avdevice'
+BF_FFMPEG_INC = '${BF_FFMPEG}/include'
+BF_FFMPEG_LIBPATH='${BF_FFMPEG}/lib'
+#WITH_BF_STATICFFMPEG = True
+#BF_FFMPEG_LIB_STATIC = '${BF_FFMPEG_LIBPATH}/libavformat.a ${BF_FFMPEG_LIBPATH/libavcodec.a ${BF_FFMPEG_LIBPATH}/libswscale.a ${BF_FFMPEG_LIBPATH}/libavutil.a ${BF_FFMPEG_LIBPATH}/libavdevice.a'
+
+# enable ogg, vorbis and theora in ffmpeg
+WITH_BF_OGG = False
+BF_OGG = '/usr'
+BF_OGG_INC = '${BF_OGG}/include'
+BF_OGG_LIB = 'ogg vorbis vorbisenc theoraenc theoradec'
+
+WITH_BF_OPENJPEG = True 
+BF_OPENJPEG = '#extern/libopenjpeg'
+BF_OPENJPEG_LIB = ''
+BF_OPENJPEG_INC = '${BF_OPENJPEG}'
+BF_OPENJPEG_LIBPATH='${BF_OPENJPEG}/lib'
+
+WITH_BF_FFTW3 = False
+WITH_BF_STATICFFTW3 = False
+BF_FFTW3 = '/usr'
+BF_FFTW3_INC = '${BF_FFTW3}/include'
+BF_FFTW3_LIB = 'fftw3'
+BF_FFTW3_LIBPATH = '${BF_FFTW3}/lib'
+BF_FFTW3_LIB_STATIC = '${BF_FFTW3_LIBPATH}/libfftw3.a'
+
+WITH_BF_REDCODE = False  
+BF_REDCODE = '#extern/libredcode'
+BF_REDCODE_LIB = ''
+# BF_REDCODE_INC = '${BF_REDCODE}/include'
+BF_REDCODE_INC = '${BF_REDCODE}/../' #C files request "libredcode/format.h" which is in "#extern/libredcode/format.h", stupid but compiles for now.
+BF_REDCODE_LIBPATH='${BF_REDCODE}/lib'
+
+# Mesa Libs should go here if you're using them as well....
+WITH_BF_STATICOPENGL = False
+BF_OPENGL = '/usr'
+BF_OPENGL_INC = '${BF_OPENGL}/include'
+BF_OPENGL_LIB = 'GL X11 Xi Xxf86vm'
+BF_OPENGL_LIBPATH = '/usr/X11R6/lib'
+BF_OPENGL_LIB_STATIC = '${BF_OPENGL_LIBPATH}/libGL.a ${BF_OPENGL_LIBPATH}/libXxf86vm.a ${BF_OPENGL_LIBPATH}/libX11.a ${BF_OPENGL_LIBPATH}/libXi.a ${BF_OPENGL_LIBPATH}/libXext.a ${BF_OPENGL_LIBPATH}/libXxf86vm.a'
+
+WITH_BF_COLLADA = False
+BF_COLLADA = '#source/blender/collada'
+BF_COLLADA_INC = '${BF_COLLADA}'
+BF_COLLADA_LIB = 'bf_collada'
+BF_OPENCOLLADA = '/usr'
+BF_OPENCOLLADA_INC = '${BF_OPENCOLLADA}/include/opencollada'
+BF_OPENCOLLADA_LIB = 'OpenCOLLADAStreamWriter OpenCOLLADASaxFrameworkLoader OpenCOLLADAFramework OpenCOLLADABaseUtils GeneratedSaxParser UTF MathMLSolver pcre buffer ftoa'
+BF_OPENCOLLADA_LIBPATH = '${BF_OPENCOLLADA}/lib/opencollada'
+BF_PCRE = ''
+BF_PCRE_LIB = 'pcre'
+BF_PCRE_LIBPATH = '/usr/lib'
+BF_EXPAT = '/usr'
+BF_EXPAT_LIB = 'expat'
+BF_EXPAT_LIBPATH = '/usr/lib'
+
+WITH_BF_JEMALLOC = False
+WITH_BF_STATICJEMALLOC = False
+BF_JEMALLOC = '/usr'
+BF_JEMALLOC_INC = '${BF_JEMALLOC}/include'
+BF_JEMALLOC_LIBPATH = '${BF_JEMALLOC}/lib'
+BF_JEMALLOC_LIB = 'jemalloc'
+BF_JEMALLOC_LIB_STATIC = '${BF_JEMALLOC_LIBPATH}/libjemalloc.a'
+
+WITH_BF_OIIO = False
+WITH_BF_STATICOIIO = False
+BF_OIIO = '/usr'
+BF_OIIO_INC = '${BF_OIIO}/include'
+BF_OIIO_LIB = 'OpenImageIO'
+BF_OIIO_LIB_STATIC = '${BF_OIIO_LIBPATH}/libOpenImageIO.a ${BF_OPENEXR}/lib/libIlmImf.a ${BF_JPEG}/lib/libjpeg.a'
+BF_OIIO_LIBPATH = '${BF_OIIO}/lib'
+
+WITH_BF_OCIO = False
+WITH_BF_STATICOCIO = False
+BF_OCIO = '/usr'
+BF_OCIO_INC = '${BF_OCIO}/include'
+BF_OCIO_LIB = 'OpenColorIO yaml-cpp tinyxml'
+BF_OCIO_LIB_STATIC = '${BF_OCIO_LIBPATH}/libOpenColorIO.a ${BF_OCIO_LIBPATH}/libtinyxml.a ${BF_OCIO_LIBPATH}/libyaml-cpp.a'
+BF_OCIO_LIBPATH = '${BF_OCIO}/lib'
+
+WITH_BF_BOOST = True
+WITH_BF_STATICBOOST = False
+BF_BOOST = '/usr'
+BF_BOOST_INC = '${BF_BOOST}/include'
+BF_BOOST_LIB = 'boost_filesystem boost_regex boost_system boost_thread boost_date_time'
+BF_BOOST_LIB_STATIC = '${BF_BOOST_LIBPATH}/libboost_filesystem.a ${BF_BOOST_LIBPATH}/libboost_date_time.a ' + \
+    '${BF_BOOST_LIBPATH}/libboost_regex.a ${BF_BOOST_LIBPATH}/libboost_locale.a ${BF_BOOST_LIBPATH}/libboost_system.a ' + \
+    '${BF_BOOST_LIBPATH}/libboost_thread.a'
+BF_BOOST_LIB_INTERNATIONAL = 'boost_locale'
+BF_BOOST_LIBPATH = '${BF_BOOST}/lib'
+
+WITH_BF_CYCLES = WITH_BF_OIIO and WITH_BF_BOOST
+
+WITH_BF_CYCLES_CUDA_BINARIES = False
+BF_CYCLES_CUDA_NVCC = '/usr/local/cuda/bin/nvcc'
+BF_CYCLES_CUDA_BINARIES_ARCH = ['sm_20', 'sm_21', 'sm_30', 'sm_35', 'sm_50']
+
+WITH_BF_OPENMP = True
+
+#Ray trace optimization
+WITH_BF_RAYOPTIMIZATION = True
+BF_RAYOPTIMIZATION_SSE_FLAGS = ['-msse','-pthread']
+
+#SpaceNavigator and friends
+WITH_BF_3DMOUSE = True
+WITH_BF_STATIC3DMOUSE = False
+BF_3DMOUSE = '/usr'
+BF_3DMOUSE_INC = '${BF_3DMOUSE}/include'
+BF_3DMOUSE_LIBPATH = '${BF_3DMOUSE}/lib'
+BF_3DMOUSE_LIB = 'spnav'
+BF_3DMOUSE_LIB_STATIC = '${BF_3DMOUSE_LIBPATH}/libspnav.a'
+
+#Freestyle
+WITH_BF_FREESTYLE = True
+
+##
+CC = 'gcc'
+CXX = 'g++'
+
+CCFLAGS = ['-pipe','-fPIC','-funsigned-char','-fno-strict-aliasing', '-D_LARGEFILE_SOURCE', '-D_FILE_OFFSET_BITS=64','-D_LARGEFILE64_SOURCE']
+CFLAGS = ['-std=gnu89']
+CXXFLAGS = []
+
+CPPFLAGS = []
+# g++ 4.6, only needed for bullet
+CXXFLAGS += ['-fpermissive']
+if WITH_BF_FFMPEG:
+    # libavutil needs UINT64_C()
+    CXXFLAGS += ['-D__STDC_CONSTANT_MACROS', ]
+REL_CFLAGS = []
+REL_CXXFLAGS = []
+REL_CCFLAGS = ['-O2']
+
+C_WARN = ['-Wno-char-subscripts', '-Wdeclaration-after-statement', '-Wunused-parameter', '-Wstrict-prototypes', '-Werror=declaration-after-statement', '-Werror=implicit-function-declaration', '-Werror=return-type']
+CC_WARN = ['-Wall']
+CXX_WARN = ['-Wno-invalid-offsetof', '-Wno-sign-compare']
+
+LLIBS = ['util', 'c', 'm', 'dl', 'pthread']
+
+BF_PROFILE = False
+BF_PROFILE_CCFLAGS = ['-pg','-g']
+BF_PROFILE_LINKFLAGS = ['-pg']
+
+BF_DEBUG = False
+BF_DEBUG_CCFLAGS = ['-g']
+
+BF_BUILDDIR = '../build/linux'
+BF_INSTALLDIR='../install/linux'
+
+#Link against pthread
+PLATFORM_LINKFLAGS = ['-pthread']
+
+#Fix for LLVM conflict with Mesa llvmpipe, SDL dynload also requires symbols to be hidden.
+# TODO(sergey): Move this to SConstruct, so we can have this line depended on user config.
+PLATFORM_LINKFLAGS += ['-Wl,--version-script=source/creator/blender.map']

build_files/scons/config/win32-mingw-config.py

@@ -0,0 +1,221 @@
+LCGDIR = '#../lib/mingw32'
+LIBDIR = "${LCGDIR}"
+
+BF_PYTHON = LIBDIR + '/python'
+BF_PYTHON_VERSION = '3.4'
+WITH_BF_STATICPYTHON = False
+BF_PYTHON_INC = '${BF_PYTHON}/include/python${BF_PYTHON_VERSION}'
+BF_PYTHON_BINARY = 'python'
+BF_PYTHON_LIB = 'python${BF_PYTHON_VERSION[0]}${BF_PYTHON_VERSION[2]}mw'
+BF_PYTHON_DLL = 'python${BF_PYTHON_VERSION[0]}${BF_PYTHON_VERSION[2]}'
+BF_PYTHON_LIBPATH = '${BF_PYTHON}/lib'
+BF_PYTHON_LIB_STATIC = '${BF_PYTHON}/lib/libpython${BF_PYTHON_VERSION[0]}${BF_PYTHON_VERSION[2]}.a'
+
+WITH_BF_OPENAL = True
+BF_OPENAL = LIBDIR + '/openal'
+BF_OPENAL_INC = '${BF_OPENAL}/include'
+BF_OPENAL_LIB = 'wrap_oal'
+BF_OPENAL_LIBPATH = '${BF_OPENAL}/lib'
+
+WITH_BF_FFMPEG = True
+BF_FFMPEG_LIB = 'avformat-55 avcodec-55 avdevice-55 avutil-52 swscale-2'
+BF_FFMPEG_LIBPATH = LIBDIR + '/ffmpeg/lib'
+BF_FFMPEG_INC =  LIBDIR + '/ffmpeg/include'
+BF_FFMPEG_DLL = '${BF_FFMPEG_LIBPATH}/avformat-55.dll ${BF_FFMPEG_LIBPATH}/avcodec-55.dll ${BF_FFMPEG_LIBPATH}/avdevice-55.dll ${BF_FFMPEG_LIBPATH}/avutil-52.dll ${BF_FFMPEG_LIBPATH}/swscale-2.dll'
+
+WITH_BF_JACK = False
+BF_JACK = LIBDIR + '/jack'
+BF_JACK_INC = '${BF_JACK}/include'
+BF_JACK_LIB = 'libjack'
+BF_JACK_LIBPATH = '${BF_JACK}/lib'
+
+WITH_BF_SNDFILE = False
+BF_SNDFILE = LIBDIR + '/sndfile'
+BF_SNDFILE_INC = '${BF_SNDFILE}/include'
+BF_SNDFILE_LIB = 'libsndfile-1'
+BF_SNDFILE_LIBPATH = '${BF_SNDFILE}/lib'
+
+WITH_BF_SDL = True
+BF_SDL = LIBDIR + '/sdl'
+BF_SDL_INC = '${BF_SDL}/include'
+BF_SDL_LIB = 'SDL'
+BF_SDL_LIBPATH = '${BF_SDL}/lib'
+
+BF_PTHREADS = LIBDIR + '/pthreads'
+BF_PTHREADS_INC = '${BF_PTHREADS}/include'
+BF_PTHREADS_LIB = 'pthreadGC2'
+BF_PTHREADS_LIBPATH = '${BF_PTHREADS}/lib'
+
+WITH_BF_OPENEXR = True
+WITH_BF_STATICOPENEXR = False
+BF_OPENEXR = LIBDIR + '/openexr'
+BF_OPENEXR_INC = '${BF_OPENEXR}/include ${BF_OPENEXR}/include/OpenEXR'
+BF_OPENEXR_LIB = 'Half IlmImf Imath IlmThread Iex'
+BF_OPENEXR_LIBPATH = '${BF_OPENEXR}/lib'
+# Warning, this static lib configuration is untested! users of this OS please confirm.
+BF_OPENEXR_LIB_STATIC = '${BF_OPENEXR}/lib/libHalf.a ${BF_OPENEXR}/lib/libIlmImf.a ${BF_OPENEXR}/lib/libIex.a ${BF_OPENEXR}/lib/libImath.a ${BF_OPENEXR}/lib/libIlmThread.a'
+
+WITH_BF_DDS = True
+
+WITH_BF_JPEG = True
+BF_JPEG = LIBDIR + '/jpeg'
+BF_JPEG_INC = '${BF_JPEG}/include'
+BF_JPEG_LIB = 'liblibjpeg'
+BF_JPEG_LIBPATH = '${BF_JPEG}/lib'
+
+WITH_BF_PNG = True
+BF_PNG = LIBDIR + '/png'
+BF_PNG_INC = '${BF_PNG}/include'
+BF_PNG_LIB = 'png'
+BF_PNG_LIBPATH = '${BF_PNG}/lib'
+
+WITH_BF_TIFF = True
+BF_TIFF = LIBDIR + '/tiff'
+BF_TIFF_INC = '${BF_TIFF}/include'
+BF_TIFF_LIB = 'tiff'
+BF_TIFF_LIBPATH = '${BF_TIFF}/lib'
+
+WITH_BF_ZLIB = True
+BF_ZLIB = LIBDIR + '/zlib'
+BF_ZLIB_INC = '${BF_ZLIB}/include'
+BF_ZLIB_LIB = 'z'
+BF_ZLIB_LIBPATH = '${BF_ZLIB}/lib'
+
+WITH_BF_INTERNATIONAL = True
+
+WITH_BF_OPENJPEG = True
+BF_OPENJPEG = '#extern/libopenjpeg'
+BF_OPENJPEG_LIB = ''
+BF_OPENJPEG_INC = '${BF_OPENJPEG}'
+BF_OPENJPEG_LIBPATH='${BF_OPENJPEG}/lib'
+
+WITH_BF_FFTW3 = True
+BF_FFTW3 = LIBDIR + '/fftw3'
+BF_FFTW3_INC = '${BF_FFTW3}/include'
+BF_FFTW3_LIB = 'fftw3'
+BF_FFTW3_LIBPATH = '${BF_FFTW3}/lib'
+
+WITH_BF_GAMEENGINE = True
+WITH_BF_PLAYER = True
+WITH_BF_OCEANSIM = True
+
+WITH_BF_BULLET = True
+BF_BULLET = '#extern/bullet2/src'
+BF_BULLET_INC = '${BF_BULLET}'
+BF_BULLET_LIB = 'extern_bullet'
+
+BF_WINTAB = LIBDIR + '/wintab'
+BF_WINTAB_INC = '${BF_WINTAB}/INCLUDE'
+
+BF_FREETYPE = LIBDIR + '/freetype'
+BF_FREETYPE_INC = '${BF_FREETYPE}/include ${BF_FREETYPE}/include/freetype2'
+BF_FREETYPE_LIB = 'freetype'
+BF_FREETYPE_LIBPATH = '${BF_FREETYPE}/lib'
+
+WITH_BF_ICONV = False
+BF_ICONV = LIBDIR + "/iconv"
+BF_ICONV_INC = '${BF_ICONV}/include'
+BF_ICONV_LIB = 'iconv'
+BF_ICONV_LIBPATH = '${BF_ICONV}/lib'
+
+WITH_BF_REDCODE = False
+BF_REDCODE_INC = '#extern'
+
+# Mesa Libs should go here if your using them as well....
+WITH_BF_STATICOPENGL = False
+BF_OPENGL = 'C:\\MingW'
+BF_OPENGL_INC = '${BF_OPENGL}/include'
+BF_OPENGL_LIBINC = '${BF_OPENGL}/lib'
+BF_OPENGL_LIB = 'opengl32'
+BF_OPENGL_LIB_STATIC = [ '${BF_OPENGL}/lib/libGL.a',
+             '${BF_OPENGL}/lib/libXmu.a', '${BF_OPENGL}/lib/libXext.a',
+             '${BF_OPENGL}/lib/libX11.a', '${BF_OPENGL}/lib/libXi.a' ]
+
+WITH_BF_COLLADA = True
+BF_COLLADA = '#source/blender/collada'
+BF_COLLADA_INC = '${BF_COLLADA}'
+BF_COLLADA_LIB = 'bf_collada'
+
+BF_OPENCOLLADA = LIBDIR + '/opencollada'
+BF_OPENCOLLADA_INC = '${BF_OPENCOLLADA}/include/opencollada'
+BF_OPENCOLLADA_LIB = 'OpenCOLLADAStreamWriter OpenCOLLADASaxFrameworkLoader OpenCOLLADAFramework OpenCOLLADABaseUtils GeneratedSaxParser UTF MathMLSolver pcre buffer ftoa xml'
+BF_OPENCOLLADA_LIBPATH = '${BF_OPENCOLLADA}/lib/opencollada'
+
+#Cycles
+WITH_BF_CYCLES = True
+WITH_BF_CYCLES_CUDA_BINARIES = False
+BF_CYCLES_CUDA_NVCC = "" # Path to the NVIDIA CUDA compiler
+BF_CYCLES_CUDA_BINARIES_ARCH = ['sm_20', 'sm_21', 'sm_30', 'sm_35', 'sm_50']
+
+WITH_BF_OIIO = True
+BF_OIIO = LIBDIR + '/openimageio'
+BF_OIIO_INC = '${BF_OIIO}/include'
+BF_OIIO_LIB = 'OpenImageIO'
+BF_OIIO_LIBPATH = '${BF_OIIO}/lib'
+
+WITH_BF_OCIO = True
+BF_OCIO = LIBDIR + '/opencolorio'
+BF_OCIO_INC = '${BF_OCIO}/include'
+BF_OCIO_LIB = 'OpenColorIO'
+BF_OCIO_LIBPATH = '${BF_OCIO}/lib'
+
+WITH_BF_BOOST = True
+BF_BOOST = LIBDIR + '/boost'
+BF_BOOST_INC = '${BF_BOOST}/include'
+BF_BOOST_LIB = 'boost_date_time-mgw46-mt-s-1_49 boost_filesystem-mgw46-mt-s-1_49 boost_regex-mgw46-mt-s-1_49 boost_system-mgw46-mt-s-1_49 boost_thread-mgw46-mt-s-1_49'
+BF_BOOST_LIB_INTERNATIONAL = 'boost_locale-mgw46-mt-s-1_49'
+BF_BOOST_LIBPATH = '${BF_BOOST}/lib'
+
+#Ray trace optimization
+WITH_BF_RAYOPTIMIZATION = True
+BF_RAYOPTIMIZATION_SSE_FLAGS = ['-msse']
+
+WITH_BF_IME = True
+
+WITH_BF_OPENMP = True
+
+#CUDA
+WITH_BF_CYCLES_CUDA_BINARIES = False
+#BF_CYCLES_CUDA_NVCC = "" # Path to the nvidia compiler
+BF_CYCLES_CUDA_BINARIES_ARCH = ['sm_20', 'sm_21', 'sm_30', 'sm_35', 'sm_50']
+
+#Freestyle
+WITH_BF_FREESTYLE = True
+
+##
+CC = 'gcc'
+CXX = 'g++'
+
+CCFLAGS = [ '-pipe', '-funsigned-char', '-fno-strict-aliasing' ]
+CXXFLAGS = []
+
+CPPFLAGS = ['-DWIN32', '-DFREE_WINDOWS', '-D_LARGEFILE_SOURCE', '-D_FILE_OFFSET_BITS=64', '-D_LARGEFILE64_SOURCE', '-DBOOST_ALL_NO_LIB', '-DBOOST_THREAD_USE_LIB', '-DGLEW_STATIC', '-DOPJ_STATIC']
+REL_CFLAGS = []
+REL_CXXFLAGS = []
+REL_CCFLAGS = ['-O2']
+
+C_WARN = ['-Wno-char-subscripts', '-Wdeclaration-after-statement', '-Wstrict-prototypes']
+
+CC_WARN = [ '-Wall' ]
+
+LLIBS = ['-lshell32', '-lshfolder', '-lgdi32', '-lmsvcrt', '-lwinmm', '-lmingw32', '-lm', '-lws2_32', '-lz', '-lstdc++','-lole32','-luuid', '-lwsock32', '-lpsapi', '-ldbghelp']
+
+if WITH_BF_IME:
+    LLIBS.append('-limm32')
+
+PLATFORM_LINKFLAGS = ['-Xlinker', '--stack=2097152']
+
+## DISABLED, causes linking errors!
+## for re-distribution, so users dont need mingw installed
+# PLATFORM_LINKFLAGS += ["-static-libgcc", "-static-libstdc++"]
+
+BF_DEBUG = False
+BF_DEBUG_CCFLAGS= ['-g']
+
+BF_PROFILE_CCFLAGS = ['-pg', '-g']
+BF_PROFILE_LINKFLAGS = ['-pg']
+BF_PROFILE_FLAGS = BF_PROFILE_CCFLAGS
+BF_PROFILE = False
+
+BF_BUILDDIR = '..\\build\\win32-mingw'
+BF_INSTALLDIR='..\\install\\win32-mingw'

build_files/scons/config/win32-vc-config.py

@@ -0,0 +1,256 @@
+import subprocess
+
+CL_OUT = subprocess.Popen(["cl.exe"],stdout=subprocess.PIPE,stderr=subprocess.PIPE)
+CL_STDOUT, CL_STDERR = CL_OUT.communicate()
+
+if "18.00." in CL_STDERR:
+    VC_VERSION = '12.0'
+    LCGDIR = '#../lib/windows_vc12'
+else:
+    import sys
+    print("Visual C version not supported {}\n".format(CL_STDERR))
+    sys.exit(1)
+
+LIBDIR = '${LCGDIR}'
+
+WITH_BF_FFMPEG = True
+BF_FFMPEG = LIBDIR +'/ffmpeg'
+BF_FFMPEG_INC = '${BF_FFMPEG}/include ${BF_FFMPEG}/include/msvc'
+BF_FFMPEG_LIBPATH='${BF_FFMPEG}/lib'
+BF_FFMPEG_LIB = 'avformat-55.lib avcodec-55.lib avdevice-55.lib avutil-52.lib swscale-2.lib'
+BF_FFMPEG_DLL = '${BF_FFMPEG_LIBPATH}/avformat-55.dll ${BF_FFMPEG_LIBPATH}/avcodec-55.dll ${BF_FFMPEG_LIBPATH}/avdevice-55.dll ${BF_FFMPEG_LIBPATH}/avutil-52.dll ${BF_FFMPEG_LIBPATH}/swscale-2.dll'
+
+BF_PYTHON = LIBDIR + '/python'
+BF_PYTHON_VERSION = '3.4'
+BF_PYTHON_INC = '${BF_PYTHON}/include/python${BF_PYTHON_VERSION}'
+BF_PYTHON_BINARY = 'python'
+BF_PYTHON_LIB = 'python${BF_PYTHON_VERSION[0]}${BF_PYTHON_VERSION[2]}'
+BF_PYTHON_DLL = '${BF_PYTHON_LIB}'
+BF_PYTHON_LIBPATH = '${BF_PYTHON}/lib'
+
+WITH_BF_PYTHON_INSTALL_NUMPY = True
+
+WITH_BF_OPENAL = True
+BF_OPENAL = LIBDIR + '/openal'
+BF_OPENAL_INC = '${BF_OPENAL}/include '
+BF_OPENAL_LIB = 'OpenAL32'
+BF_OPENAL_LIBPATH = '${BF_OPENAL}/lib'
+
+WITH_BF_ICONV = True
+BF_ICONV = LIBDIR + '/iconv'
+BF_ICONV_INC = '${BF_ICONV}/include'
+BF_ICONV_LIB = 'iconv'
+BF_ICONV_LIBPATH = '${BF_ICONV}/lib'
+
+WITH_BF_JACK = False
+BF_JACK = LIBDIR + '/jack'
+BF_JACK_INC = '${BF_JACK}/include ${BF_FFMPEG}/include/msvc'
+BF_JACK_LIB = 'libjack'
+BF_JACK_LIBPATH = '${BF_JACK}/lib'
+
+WITH_BF_SNDFILE = True
+BF_SNDFILE = LIBDIR + '/sndfile'
+BF_SNDFILE_INC = '${BF_SNDFILE}/include'
+BF_SNDFILE_LIB = 'libsndfile-1'
+BF_SNDFILE_LIBPATH = '${BF_SNDFILE}/lib'
+
+WITH_BF_SDL = True
+BF_SDL = LIBDIR + '/sdl'
+BF_SDL_INC = '${BF_SDL}/include'
+BF_SDL_LIB = 'SDL2.lib'
+BF_SDL_LIBPATH = '${BF_SDL}/lib'
+
+BF_PTHREADS = LIBDIR + '/pthreads'
+BF_PTHREADS_INC = '${BF_PTHREADS}/include'
+BF_PTHREADS_LIB = 'pthreadVC2'
+BF_PTHREADS_LIBPATH = '${BF_PTHREADS}/lib'
+
+WITH_BF_OPENEXR = True
+WITH_BF_STATICOPENEXR = False
+BF_OPENEXR = LIBDIR + '/openexr'
+BF_OPENEXR_INC = '${BF_OPENEXR}/include ${BF_OPENEXR}/include/OpenEXR '
+BF_OPENEXR_LIB = ' Iex-2_2 Half IlmImf-2_2 Imath-2_2 IlmThread-2_2 '
+BF_OPENEXR_LIBPATH = '${BF_OPENEXR}/lib'
+BF_OPENEXR_LIB_STATIC = '${BF_OPENEXR}/lib/libHalf.a ${BF_OPENEXR}/lib/libIlmImf.a ${BF_OPENEXR}/lib/libIex.a ${BF_OPENEXR}/lib/libImath.a ${BF_OPENEXR}/lib/libIlmThread.a'
+
+WITH_BF_DDS = True
+
+WITH_BF_JPEG = True
+BF_JPEG = LIBDIR + '/jpeg'
+BF_JPEG_INC = '${BF_JPEG}/include'
+BF_JPEG_LIB = 'libjpeg'
+BF_JPEG_LIBPATH = '${BF_JPEG}/lib'
+
+WITH_BF_PNG = True
+BF_PNG = LIBDIR + '/png'
+BF_PNG_INC = '${BF_PNG}/include'
+BF_PNG_LIB = 'libpng'
+BF_PNG_LIBPATH = '${BF_PNG}/lib'
+
+WITH_BF_TIFF = True
+BF_TIFF = LIBDIR + '/tiff'
+BF_TIFF_INC = '${BF_TIFF}/include'
+BF_TIFF_LIB = 'libtiff'
+BF_TIFF_LIBPATH = '${BF_TIFF}/lib'
+
+WITH_BF_ZLIB = True
+BF_ZLIB = LIBDIR + '/zlib'
+BF_ZLIB_INC = '${BF_ZLIB}/include'
+BF_ZLIB_LIB = 'libz_st'
+BF_ZLIB_LIBPATH = '${BF_ZLIB}/lib'
+
+WITH_BF_INTERNATIONAL = True
+
+WITH_BF_GAMEENGINE = True
+WITH_BF_PLAYER = True
+WITH_BF_OCEANSIM = True
+
+WITH_BF_BULLET = True
+BF_BULLET = '#extern/bullet2/src'
+BF_BULLET_INC = '${BF_BULLET}'
+BF_BULLET_LIB = 'extern_bullet'
+
+WITH_BF_ELTOPO = False
+BF_LAPACK = LIBDIR + '/lapack'
+BF_LAPACK_LIB = 'libf2c clapack_nowrap BLAS_nowrap'
+BF_LAPACK_LIBPATH = '${BF_LAPACK}/lib'
+
+BF_WINTAB = LIBDIR + '/wintab'
+BF_WINTAB_INC = '${BF_WINTAB}/INCLUDE'
+
+WITH_BF_BINRELOC = False
+
+BF_WITH_FREETYPE = True
+BF_FREETYPE = LIBDIR + '/freetype'
+BF_FREETYPE_INC = '${BF_FREETYPE}/include ${BF_FREETYPE}/include/freetype2'
+BF_FREETYPE_LIB = 'freetype2ST'
+BF_FREETYPE_LIBPATH = '${BF_FREETYPE}/lib'
+
+WITH_BF_OPENJPEG = True 
+BF_OPENJPEG = '#extern/libopenjpeg'
+BF_OPENJPEG_LIB = ''
+BF_OPENJPEG_INC = '${BF_OPENJPEG}'
+BF_OPENJPEG_LIBPATH='${BF_OPENJPEG}/lib'
+
+WITH_BF_FFTW3 = True
+BF_FFTW3 = LIBDIR + '/fftw3'
+BF_FFTW3_INC = '${BF_FFTW3}/include'
+BF_FFTW3_LIB = 'libfftw'
+BF_FFTW3_LIBPATH = '${BF_FFTW3}/lib'
+
+WITH_BF_REDCODE = False  
+BF_REDCODE_INC = '#extern'
+
+WITH_BF_COLLADA = True
+BF_COLLADA = '#source/blender/collada'
+BF_COLLADA_INC = '${BF_COLLADA}'
+BF_COLLADA_LIB = 'bf_collada'
+
+BF_OPENCOLLADA = LIBDIR + '/opencollada'
+BF_OPENCOLLADA_INC = '${BF_OPENCOLLADA}/include/opencollada'
+BF_OPENCOLLADA_LIB = 'OpenCOLLADAStreamWriter OpenCOLLADASaxFrameworkLoader OpenCOLLADAFramework OpenCOLLADABaseUtils GeneratedSaxParser MathMLSolver xml pcre buffer ftoa'
+BF_OPENCOLLADA_LIBPATH = '${BF_OPENCOLLADA}/lib/opencollada'
+
+WITH_BF_IME = True
+
+WITH_BF_3DMOUSE = True
+
+WITH_BF_OPENMP = True
+
+#Cycles
+WITH_BF_CYCLES = True 
+
+WITH_BF_CYCLES_OSL = True
+WITH_BF_STATICOSL = True
+BF_OSL = '${LIBDIR}/osl'
+BF_OSL_INC = '${BF_OSL}/include'
+BF_OSL_LIBPATH = '${BF_OSL}/lib'
+BF_OSL_LIB_STATIC = '${BF_OSL_LIBPATH}/oslcomp.lib ${BF_OSL_LIBPATH}/oslexec.lib ${BF_OSL_LIBPATH}/oslquery.lib '
+BF_OSL_COMPILER = '${BF_OSL}/bin/oslc'
+
+WITH_BF_LLVM = True
+BF_LLVM = LIBDIR + '/llvm'
+BF_LLVM_LIB = 'LLVMBitReader LLVMJIT LLVMipo LLVMVectorize LLVMBitWriter LLVMX86CodeGen LLVMX86Desc LLVMX86Info LLVMX86AsmPrinter ' + \
+    'LLVMX86Utils LLVMSelectionDAG LLVMCodeGen LLVMScalarOpts LLVMInstCombine LLVMTransformUtils LLVMipa LLVMAnalysis LLVMExecutionEngine ' + \
+    'LLVMTarget LLVMMC LLVMCore LLVMObject LLVMRuntimeDyld LLVMSupport'
+BF_LLVM_LIBPATH = '${BF_LLVM}/lib'
+
+WITH_BF_OIIO = True
+BF_OIIO = '${LIBDIR}/openimageio'
+BF_OIIO_INC = '${BF_OIIO}/include'
+BF_OIIO_LIBPATH = '${BF_OIIO}/lib'
+BF_OIIO_LIB_STATIC = '${BF_OIIO_LIBPATH}/OpenImageIO.lib'
+WITH_BF_STATICOIIO = True
+
+WITH_BF_OCIO = True
+BF_OCIO = '${LIBDIR}/opencolorio'
+BF_OCIO_INC = '${BF_OCIO}/include'
+BF_OCIO_LIBPATH = '${BF_OCIO}/lib'
+BF_OCIO_LIB_STATIC = '${BF_OCIO_LIBPATH}/OpenColorIO.lib'
+WITH_BF_STATICOCIO = True
+
+WITH_BF_BOOST = True
+BF_BOOST = '${LIBDIR}/boost'
+BF_BOOST_INC = '${BF_BOOST}/include'
+BF_BOOST_LIB = 'libboost_date_time-vc120-mt-s-1_55 libboost_filesystem-vc120-mt-s-1_55 libboost_regex-vc120-mt-s-1_55 libboost_system-vc120-mt-s-1_55 libboost_thread-vc120-mt-s-1_55 libboost_wave-vc120-mt-s-1_55'
+BF_BOOST_LIB_INTERNATIONAL = ' libboost_locale-vc120-mt-s-1_55'
+BF_BOOST_LIBPATH = '${BF_BOOST}/lib'
+
+#CUDA
+WITH_BF_CYCLES_CUDA_BINARIES = False
+#BF_CYCLES_CUDA_NVCC = "" # Path to the nvidia compiler
+BF_CYCLES_CUDA_BINARIES_ARCH = ['sm_20', 'sm_21', 'sm_30', 'sm_35', 'sm_50', 'sm_52']
+
+#Ray trace optimization
+WITH_BF_RAYOPTIMIZATION = True
+BF_RAYOPTIMIZATION_SSE_FLAGS = ['/arch:SSE']
+
+#Freestyle
+WITH_BF_FREESTYLE = True
+
+WITH_BF_STATICOPENGL = False
+BF_OPENGL_INC = '${BF_OPENGL}/include'
+BF_OPENGL_LIBINC = '${BF_OPENGL}/lib'
+BF_OPENGL_LIB = 'opengl32'
+BF_OPENGL_LIB_STATIC = [ '${BF_OPENGL}/lib/libGL.a',
+                         '${BF_OPENGL}/lib/libXmu.a', '${BF_OPENGL}/lib/libXext.a',
+                         '${BF_OPENGL}/lib/libX11.a', '${BF_OPENGL}/lib/libXi.a' ]
+                         
+CC = 'cl.exe'
+CXX = 'cl.exe'
+
+CCFLAGS = ['/nologo', '/J', '/W3', '/Gd', '/w34062', '/wd4018', '/wd4065', '/wd4127', '/wd4181', '/wd4200', '/wd4244', '/wd4267', '/wd4305', '/wd4800', '/we4013', '/we4431']
+CXXFLAGS = ['/EHsc']
+BGE_CXXFLAGS = ['/O2', '/Ob2', '/EHsc', '/GR', '/fp:fast', '/arch:SSE']
+
+if VC_VERSION == '12.0':
+    CCFLAGS.append('/DOIIO_STATIC_BUILD')  # OIIO api changed with 1.4 making this needed 
+
+BF_DEBUG_CCFLAGS = ['/Zi', '/Ob0', '/Od', '/FR${TARGET}.sbr']
+
+CPPFLAGS = ['-DWIN32','-D_CONSOLE', '-D_LIB', '-D_CRT_SECURE_NO_DEPRECATE', '-DOPJ_STATIC']
+REL_CFLAGS = []
+REL_CXXFLAGS = []
+REL_CCFLAGS = ['-O2', '/Ob2']
+
+C_WARN = []
+CC_WARN = []
+CXX_WARN = []
+
+LLIBS = ['ws2_32', 'vfw32', 'winmm', 'kernel32', 'user32', 'gdi32', 'comdlg32', 'advapi32', 'shfolder', 'shell32', 'ole32', 'oleaut32', 'uuid', 'psapi', 'Dbghelp']
+
+if WITH_BF_IME:
+    LLIBS.append('imm32')
+
+PLATFORM_LINKFLAGS = ['/SUBSYSTEM:CONSOLE','/MACHINE:IX86','/STACK:2097152','/INCREMENTAL:NO', '/LARGEADDRESSAWARE', '/NODEFAULTLIB:msvcrt.lib', '/NODEFAULTLIB:msvcmrt.lib', '/NODEFAULTLIB:msvcurt.lib', '/NODEFAULTLIB:msvcrtd.lib']
+
+# # Todo
+# BF_PROFILE_CCFLAGS = ['-pg', '-g ']
+# BF_PROFILE_LINKFLAGS = ['-pg']
+# BF_PROFILE = False
+
+BF_BSC=False
+
+BF_CYCLES_CUDA_ENV="C:\Program Files\Microsoft SDKs\Windows\v7.1\Bin\SetEnv.cmd"
+BF_BUILDDIR = '..\\build\\win32-vc'
+BF_INSTALLDIR='..\\install\\win32-vc'

build_files/scons/config/win64-mingw-config.py

@@ -0,0 +1,216 @@
+LCGDIR = '#../lib/mingw64'
+LIBDIR = "${LCGDIR}"
+
+BF_PYTHON = LIBDIR + '/python'
+BF_PYTHON_VERSION = '3.4'
+WITH_BF_STATICPYTHON = False
+BF_PYTHON_INC = '${BF_PYTHON}/include/python${BF_PYTHON_VERSION}'
+BF_PYTHON_BINARY = 'python'
+BF_PYTHON_LIB = 'python${BF_PYTHON_VERSION[0]}${BF_PYTHON_VERSION[2]}mw'
+BF_PYTHON_DLL = 'python${BF_PYTHON_VERSION[0]}${BF_PYTHON_VERSION[2]}'
+BF_PYTHON_LIBPATH = '${BF_PYTHON}/lib'
+
+WITH_BF_OPENAL = True
+BF_OPENAL = LIBDIR + '/openal'
+BF_OPENAL_INC = '${BF_OPENAL}/include'
+BF_OPENAL_LIB = 'wrap_oal'
+BF_OPENAL_LIBPATH = '${BF_OPENAL}/lib'
+
+WITH_BF_FFMPEG = True
+BF_FFMPEG_LIB = 'avformat.dll avcodec.dll avdevice.dll avutil.dll swscale.dll swresample.dll'
+BF_FFMPEG_LIBPATH = LIBDIR + '/ffmpeg/lib'
+BF_FFMPEG_INC =  LIBDIR + '/ffmpeg/include'
+BF_FFMPEG_DLL = '${BF_FFMPEG_LIBPATH}/avformat-53.dll ${BF_FFMPEG_LIBPATH}/avcodec-53.dll ${BF_FFMPEG_LIBPATH}/avdevice-53.dll ${BF_FFMPEG_LIBPATH}/avutil-51.dll ${BF_FFMPEG_LIBPATH}/swscale-2.dll ${BF_FFMPEG_LIBPATH}/swresample-0.dll ${BF_FFMPEG_LIBPATH}/xvidcore.dll'
+
+WITH_BF_JACK = False
+BF_JACK = LIBDIR + '/jack'
+BF_JACK_INC = '${BF_JACK}/include'
+BF_JACK_LIB = 'libjack'
+BF_JACK_LIBPATH = '${BF_JACK}/lib'
+
+WITH_BF_SNDFILE = False
+BF_SNDFILE = LIBDIR + '/sndfile'
+BF_SNDFILE_INC = '${BF_SNDFILE}/include'
+BF_SNDFILE_LIB = 'libsndfile-1'
+BF_SNDFILE_LIBPATH = '${BF_SNDFILE}/lib'
+
+WITH_BF_SDL = True
+BF_SDL = LIBDIR + '/sdl'
+BF_SDL_INC = '${BF_SDL}/include'
+BF_SDL_LIB = 'SDL'
+BF_SDL_LIBPATH = '${BF_SDL}/lib'
+
+BF_PTHREADS = '' # Part of MinGW-w64
+BF_PTHREADS_INC = ''
+BF_PTHREADS_LIB = ''
+BF_PTHREADS_LIBPATH = ''
+
+WITH_BF_OPENEXR = True
+WITH_BF_STATICOPENEXR = False
+BF_OPENEXR = LIBDIR + '/openexr'
+BF_OPENEXR_INC = '${BF_OPENEXR}/include ${BF_OPENEXR}/include/OpenEXR'
+BF_OPENEXR_LIB = 'Half IlmImf Imath IlmThread Iex'
+BF_OPENEXR_LIBPATH = '${BF_OPENEXR}/lib'
+
+WITH_BF_DDS = True
+
+WITH_BF_JPEG = True
+BF_JPEG = LIBDIR + '/jpeg'
+BF_JPEG_INC = '${BF_JPEG}/include'
+BF_JPEG_LIB = 'jpeg'
+BF_JPEG_LIBPATH = '${BF_JPEG}/lib'
+
+WITH_BF_PNG = True
+BF_PNG = LIBDIR + '/png'
+BF_PNG_INC = '${BF_PNG}/include'
+BF_PNG_LIB = 'png'
+BF_PNG_LIBPATH = '${BF_PNG}/lib'
+
+WITH_BF_TIFF = True
+BF_TIFF = LIBDIR + '/tiff'
+BF_TIFF_INC = '${BF_TIFF}/include'
+BF_TIFF_LIB = 'tiff'
+BF_TIFF_LIBPATH = '${BF_TIFF}/lib'
+
+WITH_BF_ZLIB = True
+BF_ZLIB = LIBDIR + '/zlib'
+BF_ZLIB_INC = '${BF_ZLIB}/include'
+BF_ZLIB_LIB = 'z'
+BF_ZLIB_LIBPATH = '${BF_ZLIB}/lib'
+
+WITH_BF_INTERNATIONAL = True
+
+WITH_BF_OPENJPEG = True
+BF_OPENJPEG = '#extern/libopenjpeg'
+BF_OPENJPEG_LIB = ''
+BF_OPENJPEG_INC = '${BF_OPENJPEG}'
+BF_OPENJPEG_LIBPATH='${BF_OPENJPEG}/lib'
+
+WITH_BF_FFTW3 = True
+BF_FFTW3 = LIBDIR + '/fftw3'
+BF_FFTW3_INC = '${BF_FFTW3}/include'
+BF_FFTW3_LIB = 'fftw3'
+BF_FFTW3_LIBPATH = '${BF_FFTW3}/lib'
+
+WITH_BF_GAMEENGINE = True
+WITH_BF_OCEANSIM = True
+WITH_BF_PLAYER = True
+WITH_BF_LIBMV = True
+
+WITH_BF_BULLET = True
+BF_BULLET = '#extern/bullet2/src'
+BF_BULLET_INC = '${BF_BULLET}'
+BF_BULLET_LIB = 'extern_bullet'
+
+BF_WINTAB = LIBDIR + '/wintab'
+BF_WINTAB_INC = '${BF_WINTAB}/INCLUDE'
+
+# enable freetype2 support for text objects
+BF_FREETYPE = LIBDIR + '/freetype'
+BF_FREETYPE_INC = '${BF_FREETYPE}/include ${BF_FREETYPE}/include/freetype2/'
+BF_FREETYPE_LIB = 'freetype'
+BF_FREETYPE_LIBPATH = '${BF_FREETYPE}/lib'
+
+WITH_BF_ICONV = False
+BF_ICONV = LIBDIR + "/iconv"
+BF_ICONV_INC = '${BF_ICONV}/include'
+BF_ICONV_LIB = 'iconv'
+BF_ICONV_LIBPATH = '${BF_ICONV}/lib'
+
+WITH_BF_REDCODE = False
+BF_REDCODE_INC = '#extern'
+
+# Mesa Libs should go here if your using them as well....
+WITH_BF_STATICOPENGL = False
+BF_OPENGL = 'C:\\MingW'
+BF_OPENGL_INC = '${BF_OPENGL}/include'
+BF_OPENGL_LIBINC = '${BF_OPENGL}/lib'
+BF_OPENGL_LIB = 'opengl32'
+BF_OPENGL_LIB_STATIC = [ '${BF_OPENGL}/lib/libGL.a',
+             '${BF_OPENGL}/lib/libXmu.a', '${BF_OPENGL}/lib/libXext.a',
+             '${BF_OPENGL}/lib/libX11.a', '${BF_OPENGL}/lib/libXi.a' ]
+
+WITH_BF_COLLADA = True
+BF_COLLADA = '#source/blender/collada'
+BF_COLLADA_INC = '${BF_COLLADA}'
+BF_COLLADA_LIB = 'bf_collada'
+
+BF_OPENCOLLADA = LIBDIR + '/opencollada'
+BF_OPENCOLLADA_INC = '${BF_OPENCOLLADA}/include/opencollada'
+BF_OPENCOLLADA_LIB = 'OpenCOLLADAStreamWriter OpenCOLLADASaxFrameworkLoader OpenCOLLADAFramework OpenCOLLADABaseUtils GeneratedSaxParser UTF MathMLSolver pcre buffer ftoa xml'
+BF_OPENCOLLADA_LIBPATH = '${BF_OPENCOLLADA}/lib/opencollada'
+
+#Cycles
+WITH_BF_CYCLES = True
+WITH_BF_CYCLES_CUDA_BINARIES = False
+BF_CYCLES_CUDA_NVCC = "" # Path to the NVIDIA CUDA compiler
+BF_CYCLES_CUDA_BINARIES_ARCH = ['sm_20', 'sm_21', 'sm_30', 'sm_35', 'sm_50']
+
+WITH_BF_OIIO = True
+BF_OIIO = LIBDIR + '/openimageio'
+BF_OIIO_INC = '${BF_OIIO}/include'
+BF_OIIO_LIB = 'OpenImageIO'
+BF_OIIO_LIBPATH = '${BF_OIIO}/lib'
+
+WITH_BF_OCIO = True
+BF_OCIO = LIBDIR + '/opencolorio'
+BF_OCIO_INC = '${BF_OCIO}/include'
+BF_OCIO_LIB = 'OpenColorIO'
+BF_OCIO_LIBPATH = '${BF_OCIO}/lib'
+
+WITH_BF_BOOST = True
+BF_BOOST = LIBDIR + '/boost'
+BF_BOOST_INC = '${BF_BOOST}/include'
+BF_BOOST_LIB = 'boost_date_time-mgw47-mt-s-1_49 boost_date_time-mgw47-mt-sd-1_49 boost_filesystem-mgw47-mt-s-1_49 boost_filesystem-mgw47-mt-sd-1_49 boost_regex-mgw47-mt-s-1_49 boost_regex-mgw47-mt-sd-1_49 boost_system-mgw47-mt-s-1_49 boost_system-mgw47-mt-sd-1_49 boost_thread-mgw47-mt-s-1_49 boost_thread-mgw47-mt-sd-1_49'
+BF_BOOST_LIB_INTERNATIONAL = ' boost_locale-mgw47-mt-s-1_49 boost_locale-mgw47-mt-sd-1_49'
+BF_BOOST_LIBPATH = '${BF_BOOST}/lib'
+
+#Ray trace optimization
+WITH_BF_RAYOPTIMIZATION = True
+BF_RAYOPTIMIZATION_SSE_FLAGS = ['-mmmx', '-msse', '-msse2']
+
+WITH_BF_IME = True
+
+WITH_BF_OPENMP = True
+
+#Freestyle
+WITH_BF_FREESTYLE = True
+
+##
+CC = 'gcc'
+CXX = 'g++'
+
+CCFLAGS = [ '-pipe', '-funsigned-char', '-fno-strict-aliasing' ]
+CXXFLAGS = [ '-fpermissive' ]
+
+CPPFLAGS = ['-DWIN32', '-DMS_WIN64', '-DFREE_WINDOWS', '-DFREE_WINDOWS64', '-D_LARGEFILE_SOURCE', '-D_FILE_OFFSET_BITS=64', '-D_LARGEFILE64_SOURCE', '-DBOOST_ALL_NO_LIB', '-DBOOST_THREAD_USE_LIB', '-DGLEW_STATIC', '-DOPJ_STATIC']
+REL_CFLAGS = []
+REL_CXXFLAGS = []
+REL_CCFLAGS = ['-O2', '-ftree-vectorize']
+
+# NOTE: C_WARN seems to get ignored - at least -Wno-char-subscripts doesn't work!
+C_WARN = ['-Wno-char-subscripts', '-Wdeclaration-after-statement', '-Wstrict-prototypes']
+
+CC_WARN = [ '-Wall', '-Wno-char-subscripts' ]
+
+LLIBS = ['-lshell32', '-lshfolder', '-lgdi32', '-lmsvcrt', '-lwinmm', '-lmingw32', '-lm', '-lws2_32', '-lz', '-lstdc++','-lole32','-luuid', '-lwsock32', '-lpsapi', '-lpthread', '-ldbghelp']
+
+if WITH_BF_IME:
+    LLIBS.append('-limm32')
+
+PLATFORM_LINKFLAGS = ['-Xlinker', '--stack=2097152']
+
+## DISABLED, causes linking errors!
+## for re-distribution, so users dont need mingw installed
+# PLATFORM_LINKFLAGS += ["-static-libgcc", "-static-libstdc++"]
+
+BF_DEBUG = False
+BF_DEBUG_CCFLAGS= ['-g']
+
+BF_PROFILE_CCFLAGS = ['-pg', '-g']
+BF_PROFILE_LINKFLAGS = ['-pg']
+BF_PROFILE_FLAGS = BF_PROFILE_CCFLAGS
+BF_PROFILE = False
+
+BF_BUILDDIR = '..\\build\\win64-mingw'
+BF_INSTALLDIR='..\\install\\win64-mingw'

build_files/scons/config/win64-vc-config.py

@@ -0,0 +1,256 @@
+import subprocess
+
+CL_OUT = subprocess.Popen(["cl.exe"],stdout=subprocess.PIPE,stderr=subprocess.PIPE)
+CL_STDOUT, CL_STDERR = CL_OUT.communicate()
+
+if "18.00." in CL_STDERR:
+    VC_VERSION = '12.0'
+    LCGDIR = '#../lib/win64_vc12'
+else:
+    import sys
+    print("Visual C version not supported {}\n".format(CL_STDERR))
+    sys.exit(1)
+ 
+LIBDIR = '${LCGDIR}'
+
+WITH_BF_FFMPEG = True
+BF_FFMPEG = LIBDIR +'/ffmpeg'
+BF_FFMPEG_INC = '${BF_FFMPEG}/include ${BF_FFMPEG}/include/msvc '
+BF_FFMPEG_LIBPATH='${BF_FFMPEG}/lib'
+BF_FFMPEG_LIB = 'avformat-55.lib avcodec-55.lib avdevice-55.lib avutil-52.lib swscale-2.lib'
+BF_FFMPEG_DLL = '${BF_FFMPEG_LIBPATH}/avformat-55.dll ${BF_FFMPEG_LIBPATH}/avcodec-55.dll ${BF_FFMPEG_LIBPATH}/avdevice-55.dll ${BF_FFMPEG_LIBPATH}/avutil-52.dll ${BF_FFMPEG_LIBPATH}/swscale-2.dll'
+    
+
+BF_PYTHON = LIBDIR + '/python'
+BF_PYTHON_VERSION = '3.4'
+BF_PYTHON_INC = '${BF_PYTHON}/include/python${BF_PYTHON_VERSION}'
+BF_PYTHON_BINARY = 'python'
+BF_PYTHON_LIB = 'python${BF_PYTHON_VERSION[0]}${BF_PYTHON_VERSION[2]}'
+BF_PYTHON_DLL = '${BF_PYTHON_LIB}'
+BF_PYTHON_LIBPATH = '${BF_PYTHON}/lib'
+
+WITH_BF_PYTHON_INSTALL_NUMPY = True
+
+WITH_BF_OPENAL = True
+BF_OPENAL = LIBDIR + '/openal'
+BF_OPENAL_INC = '${BF_OPENAL}/include '
+BF_OPENAL_LIB = 'OpenAL32'
+BF_OPENAL_LIBPATH = '${BF_OPENAL}/lib'
+
+WITH_BF_SNDFILE = True
+BF_SNDFILE = LIBDIR + '/sndfile'
+BF_SNDFILE_INC = '${BF_SNDFILE}/include'
+BF_SNDFILE_LIB = 'libsndfile-1'
+BF_SNDFILE_LIBPATH = '${BF_SNDFILE}/lib'
+
+WITH_BF_ICONV = True
+BF_ICONV = LIBDIR + '/iconv'
+BF_ICONV_INC = '${BF_ICONV}/include'
+BF_ICONV_LIB = 'iconv'
+BF_ICONV_LIBPATH = '${BF_ICONV}/lib'
+
+WITH_BF_SDL = True
+BF_SDL = LIBDIR + '/sdl'
+BF_SDL_INC = '${BF_SDL}/include'
+BF_SDL_LIB = 'SDL2.lib'
+BF_SDL_LIBPATH = '${BF_SDL}/lib'
+
+WITH_BF_JACK = False
+
+BF_PTHREADS = LIBDIR + '/pthreads'
+BF_PTHREADS_INC = '${BF_PTHREADS}/include'
+BF_PTHREADS_LIB = 'pthreadVC2'
+BF_PTHREADS_LIBPATH = '${BF_PTHREADS}/lib'
+
+WITH_BF_OPENEXR = True
+WITH_BF_STATICOPENEXR = False
+BF_OPENEXR = LIBDIR + '/openexr'
+BF_OPENEXR_INC = '${BF_OPENEXR}/include ${BF_OPENEXR}/include/OpenEXR '
+BF_OPENEXR_LIB = ' Iex-2_2 Half IlmImf-2_2 Imath-2_2 IlmThread-2_2 '
+BF_OPENEXR_LIBPATH = '${BF_OPENEXR}/lib'
+BF_OPENEXR_LIB_STATIC = '${BF_OPENEXR}/lib/libHalf.a ${BF_OPENEXR}/lib/libIlmImf.a ${BF_OPENEXR}/lib/libIex.a ${BF_OPENEXR}/lib/libImath.a ${BF_OPENEXR}/lib/libIlmThread.a'
+
+WITH_BF_DDS = True
+
+WITH_BF_JPEG = True
+BF_JPEG = LIBDIR + '/jpeg'
+BF_JPEG_INC = '${BF_JPEG}/include'
+BF_JPEG_LIB = 'libjpeg'
+BF_JPEG_LIBPATH = '${BF_JPEG}/lib'
+
+WITH_BF_PNG = True
+BF_PNG = LIBDIR + '/png'
+BF_PNG_INC = '${BF_PNG}/include'
+BF_PNG_LIB = 'libpng'
+BF_PNG_LIBPATH = '${BF_PNG}/lib'
+
+WITH_BF_TIFF = True
+BF_TIFF = LIBDIR + '/tiff'
+BF_TIFF_INC = '${BF_TIFF}/include'
+BF_TIFF_LIB = 'libtiff'
+BF_TIFF_LIBPATH = '${BF_TIFF}/lib'
+
+WITH_BF_ZLIB = True
+BF_ZLIB = LIBDIR + '/zlib'
+BF_ZLIB_INC = '${BF_ZLIB}/include'
+BF_ZLIB_LIB = 'libz_st'
+BF_ZLIB_LIBPATH = '${BF_ZLIB}/lib'
+
+WITH_BF_INTERNATIONAL = True
+
+WITH_BF_GAMEENGINE = True
+WITH_BF_PLAYER = True
+WITH_BF_OCEANSIM = True
+
+WITH_BF_BULLET = True
+BF_BULLET = '#extern/bullet2/src'
+BF_BULLET_INC = '${BF_BULLET}'
+BF_BULLET_LIB = 'extern_bullet'
+
+WITH_BF_ELTOPO = False
+BF_LAPACK = LIBDIR + '/lapack'
+BF_LAPACK_LIB = 'libf2c clapack_nowrap BLAS_nowrap'
+BF_LAPACK_LIBPATH = '${BF_LAPACK}/lib'
+
+BF_WINTAB = LIBDIR + '/wintab'
+BF_WINTAB_INC = '${BF_WINTAB}/INCLUDE'
+
+WITH_BF_BINRELOC = False
+
+BF_WITH_FREETYPE = True
+BF_FREETYPE = LIBDIR + '/freetype'
+BF_FREETYPE_INC = '${BF_FREETYPE}/include ${BF_FREETYPE}/include/freetype2'
+BF_FREETYPE_LIB = 'freetype2ST'
+BF_FREETYPE_LIBPATH = '${BF_FREETYPE}/lib'
+
+WITH_BF_QUICKTIME = False
+BF_QUICKTIME = LIBDIR + '/QTDevWin'
+BF_QUICKTIME_INC = '${BF_QUICKTIME}/CIncludes'
+BF_QUICKTIME_LIB = 'qtmlClient'
+BF_QUICKTIME_LIBPATH = '${BF_QUICKTIME}/Libraries'
+
+WITH_BF_OPENJPEG = True
+BF_OPENJPEG = '#extern/libopenjpeg'
+BF_OPENJPEG_LIB = ''
+BF_OPENJPEG_INC = '${BF_OPENJPEG}'
+BF_OPENJPEG_LIBPATH='${BF_OPENJPEG}/lib'
+
+WITH_BF_FFTW3 = True
+BF_FFTW3 = LIBDIR + '/fftw3'
+BF_FFTW3_INC = '${BF_FFTW3}/include'
+BF_FFTW3_LIB = 'libfftw'
+BF_FFTW3_LIBPATH = '${BF_FFTW3}/lib'
+
+WITH_BF_REDCODE = False  
+BF_REDCODE_INC = '#extern'
+
+WITH_BF_COLLADA = True
+BF_COLLADA = '#source/blender/collada'
+BF_COLLADA_INC = '${BF_COLLADA}'
+BF_COLLADA_LIB = 'bf_collada'
+
+BF_OPENCOLLADA = LIBDIR + '/opencollada'
+BF_OPENCOLLADA_INC = '${BF_OPENCOLLADA}/include/opencollada'
+BF_OPENCOLLADA_LIB = 'OpenCOLLADAStreamWriter OpenCOLLADASaxFrameworkLoader OpenCOLLADAFramework OpenCOLLADABaseUtils GeneratedSaxParser MathMLSolver xml pcre buffer ftoa'
+BF_OPENCOLLADA_LIBPATH = '${BF_OPENCOLLADA}/lib/opencollada'
+
+WITH_BF_IME = True
+
+WITH_BF_3DMOUSE = True
+
+WITH_BF_OPENMP = True
+
+#Cycles
+WITH_BF_CYCLES = True
+
+WITH_BF_CYCLES_OSL = True
+WITH_BF_STATICOSL = True
+BF_OSL = '${LIBDIR}/osl'
+BF_OSL_INC = '${BF_OSL}/include'
+BF_OSL_LIBPATH = '${BF_OSL}/lib'
+BF_OSL_LIB_STATIC = '${BF_OSL_LIBPATH}/oslcomp.lib ${BF_OSL_LIBPATH}/oslexec.lib ${BF_OSL_LIBPATH}/oslquery.lib '
+BF_OSL_COMPILER = '${BF_OSL}/bin/oslc'
+
+WITH_BF_LLVM = True
+BF_LLVM = LIBDIR + '/llvm'
+BF_LLVM_LIB = 'LLVMBitReader LLVMJIT LLVMipo LLVMVectorize LLVMBitWriter LLVMX86CodeGen LLVMX86Desc LLVMX86Info LLVMX86AsmPrinter ' + \
+    'LLVMX86Utils LLVMSelectionDAG LLVMCodeGen LLVMScalarOpts LLVMInstCombine LLVMTransformUtils LLVMipa LLVMAnalysis LLVMExecutionEngine ' + \
+    'LLVMTarget LLVMMC LLVMCore LLVMObject LLVMRuntimeDyld LLVMSupport'
+BF_LLVM_LIBPATH = '${BF_LLVM}/lib'
+
+WITH_BF_OIIO = True
+BF_OIIO = '${LIBDIR}/openimageio'
+BF_OIIO_INC = '${BF_OIIO}/include'
+BF_OIIO_LIBPATH = '${BF_OIIO}/lib'
+BF_OIIO_LIB_STATIC = '${BF_OIIO_LIBPATH}/OpenImageIO.lib ${BF_OIIO_LIBPATH}/OpenImageIO_Util.lib'
+WITH_BF_STATICOIIO = True
+
+WITH_BF_OCIO = True
+BF_OCIO = '${LIBDIR}/opencolorio'
+BF_OCIO_INC = '${BF_OCIO}/include'
+BF_OCIO_LIBPATH = '${BF_OCIO}/lib'
+BF_OCIO_LIB_STATIC = '${BF_OCIO_LIBPATH}/OpenColorIO.lib'
+WITH_BF_STATICOCIO = True
+
+WITH_BF_BOOST = True
+BF_BOOST = '${LIBDIR}/boost'
+BF_BOOST_INC = '${BF_BOOST}/include'
+BF_BOOST_LIB = 'libboost_date_time-vc120-mt-s-1_55 libboost_filesystem-vc120-mt-s-1_55 libboost_regex-vc120-mt-s-1_55 libboost_system-vc120-mt-s-1_55 libboost_thread-vc120-mt-s-1_55 libboost_wave-vc120-mt-s-1_55'
+BF_BOOST_LIB_INTERNATIONAL = ' libboost_locale-vc120-mt-s-1_55'
+BF_BOOST_LIBPATH = '${BF_BOOST}/lib'
+
+#CUDA
+WITH_BF_CYCLES_CUDA_BINARIES = False
+#BF_CYCLES_CUDA_NVCC = "" # Path to the nvidia compiler
+BF_CYCLES_CUDA_BINARIES_ARCH = ['sm_20', 'sm_21', 'sm_30', 'sm_35', 'sm_50', 'sm_52']
+
+#Ray trace optimization
+WITH_BF_RAYOPTIMIZATION = True
+# No need to manually specify SSE/SSE2 on x64 systems.
+BF_RAYOPTIMIZATION_SSE_FLAGS = ['']
+
+#Freestyle
+WITH_BF_FREESTYLE = True
+
+WITH_BF_STATICOPENGL = False
+BF_OPENGL_INC = '${BF_OPENGL}/include'
+BF_OPENGL_LIBINC = '${BF_OPENGL}/lib'
+BF_OPENGL_LIB = 'opengl32'
+BF_OPENGL_LIB_STATIC = [ '${BF_OPENGL}/lib/libGL.a',
+                         '${BF_OPENGL}/lib/libXmu.a', '${BF_OPENGL}/lib/libXext.a',
+                         '${BF_OPENGL}/lib/libX11.a', '${BF_OPENGL}/lib/libXi.a' ]
+CC = 'cl.exe'
+CXX = 'cl.exe'
+
+CFLAGS = []
+CCFLAGS = ['/nologo', '/J', '/W3', '/Gd', '/w34062', '/wd4018', '/wd4065', '/wd4127', '/wd4181', '/wd4200', '/wd4244', '/wd4267', '/wd4305', '/wd4800', '/we4013', '/we4431']
+
+# We want to support Vista level ABI for x64
+if VC_VERSION == '12.0':
+    CCFLAGS.append('/D_WIN32_WINNT=0x600')
+    CCFLAGS.append('/DOIIO_STATIC_BUILD')  # OIIO api changed with 1.4 making this needed 
+
+CXXFLAGS = ['/EHsc']
+BGE_CXXFLAGS = ['/O2', '/Ob2', '/EHsc', '/GR', '/fp:fast']
+
+BF_DEBUG_CCFLAGS = ['/Zi', '/FR${TARGET}.sbr', '/Od', '/Ob0']
+
+CPPFLAGS = ['-DWIN32', '-D_CONSOLE', '-D_LIB', '-D_CRT_SECURE_NO_DEPRECATE', '-DOPJ_STATIC']
+REL_CFLAGS = []
+REL_CXXFLAGS = []
+REL_CCFLAGS = ['-O2', '/Ob2']
+
+C_WARN = []
+CC_WARN = []
+CXX_WARN = []
+
+LLIBS = ['ws2_32', 'vfw32', 'winmm', 'kernel32', 'user32', 'gdi32', 'comdlg32', 'advapi32', 'shfolder', 'shell32', 'ole32', 'oleaut32', 'uuid', 'psapi', 'Dbghelp']
+
+if WITH_BF_IME:
+    LLIBS.append('imm32')
+
+PLATFORM_LINKFLAGS = ['/SUBSYSTEM:CONSOLE','/MACHINE:X64','/STACK:2097152','/OPT:NOREF','/INCREMENTAL:NO', '/NODEFAULTLIB:msvcrt.lib', '/NODEFAULTLIB:msvcmrt.lib', '/NODEFAULTLIB:msvcurt.lib', '/NODEFAULTLIB:msvcrtd.lib']
+
+BF_CYCLES_CUDA_ENV="C:\Program Files\Microsoft SDKs\Windows\v7.1\Bin\SetEnv.cmd"
+BF_BUILDDIR = '..\\build\\win64-vc'
+BF_INSTALLDIR='..\\install\\win64-vc'

doc/python_api/examples/bge.texture.2.py

@@ -0,0 +1,237 @@
+"""
+Video Capture with DeckLink
++++++++++++++++++++++++++++
+Video frames captured with DeckLink cards have pixel formats that are generally not directly
+usable by OpenGL, they must be processed by a shader. The three shaders presented here should
+cover all common video capture cases.
+
+This file reflects the current video transfer method implemented in the Decklink module:
+whenever possible the video images are transferred as float texture because this is more
+compatible with GPUs. Of course, only the pixel formats that have a correspondant GL format
+can be transferred as float. Look for fg_shaders in this file for an exhaustive list.
+
+Other pixel formats will be transferred as 32 bits integer red-channel texture but this
+won't work with certain GPU (Intel GMA); the corresponding shaders are not shown here.
+However, it should not be necessary to use any of them as the list below covers all practical
+cases of video capture with all types of Decklink product.
+
+In other words, only use one of the pixel format below and you will be fine. Note that depending
+on the video stream, only certain pixel formats will be allowed (others will throw an exception).
+For example, to capture a PAL video stream, you must use one of the YUV formats.
+
+To find which pixel format is suitable for a particular video stream, use the 'Media Express'
+utility that comes with the Decklink software : if you see the video in the 'Log and Capture'
+Window, you have selected the right pixel format and you can use the same in Blender.
+
+Notes: * these shaders only decode the RGB channel and set the alpha channel to a fixed
+value (look for color.a = ). It's up to you to add postprocessing to the color.
+       * these shaders are compatible with 2D and 3D video stream
+"""
+import bge
+from bge import logic
+from bge import texture as vt
+
+# The default vertex shader, because we need one
+#
+VertexShader = """
+#version 130
+   void main()
+   {
+      gl_Position = gl_ModelViewProjectionMatrix * gl_Vertex;
+      gl_TexCoord[0] = gl_MultiTexCoord0;
+   }
+
+"""
+
+# For use with RGB video stream: the pixel is directly usable
+#
+FragmentShader_R10l = """
+    #version 130
+    uniform sampler2D tex;
+    // stereo = 1.0 if 2D image, =0.5 if 3D (left eye below, right eye above)
+    uniform float stereo;
+    // eye = 0.0 for the left eye, 0.5 for the right eye
+    uniform float eye;
+
+    void main(void)
+    {
+        vec4 color;
+        float tx, ty;
+        tx = gl_TexCoord[0].x;
+        ty = eye+gl_TexCoord[0].y*stereo;
+        color = texture(tex, vec2(tx,ty));
+        color.a = 0.7;
+        gl_FragColor = color;
+    }
+"""
+
+# For use with YUV video stream
+#
+FragmentShader_2vuy = """
+    #version 130
+    uniform sampler2D tex;
+    // stereo = 1.0 if 2D image, =0.5 if 3D (left eye below, right eye above)
+    uniform float stereo;
+    // eye = 0.0 for the left eye, 0.5 for the right eye
+    uniform float eye;
+
+    void main(void)
+    {
+        vec4 color;
+        float tx, ty, width, Y, Cb, Cr;
+        int px;
+        tx = gl_TexCoord[0].x;
+        ty = eye+gl_TexCoord[0].y*stereo;
+        width = float(textureSize(tex, 0).x);
+        color = texture(tex, vec2(tx, ty));
+        px = int(floor(fract(tx*width)*2.0));
+        switch (px) {
+        case 0:
+            Y = color.g;
+            break;
+        case 1:
+            Y = color.a;
+            break;
+        }
+        Y = (Y - 0.0625) * 1.168949772;
+        Cb = (color.b - 0.0625) * 1.142857143 - 0.5;
+        Cr = (color.r - 0.0625) * 1.142857143 - 0.5;
+        color.r = Y + 1.5748 * Cr;
+        color.g = Y - 0.1873 * Cb - 0.4681 * Cr;
+        color.b = Y + 1.8556 * Cb;
+        color.a = 0.7;
+        gl_FragColor = color;
+    }
+"""
+
+# For use with high resolution YUV
+#
+FragmentShader_v210 = """
+    #version 130
+    uniform sampler2D tex;
+    // stereo = 1.0 if 2D image, =0.5 if 3D (left eye below, right eye above)
+    uniform float stereo;
+    // eye = 0.0 for the left eye, 0.5 for the right eye
+    uniform float eye;
+
+    void main(void)
+    {
+        vec4 color, color1, color2, color3;
+        int px;
+        float tx, ty, width, sx, dx, bx, Y, Cb, Cr;
+        tx = gl_TexCoord[0].x;
+        ty = eye+gl_TexCoord[0].y*stereo;
+        width = float(textureSize(tex, 0).x);
+        // to sample macro pixels (6 pixels in 4 words)
+        sx = tx*width*0.25+0.01;
+        // index of display pixel in the macro pixel 0..5
+        px = int(floor(fract(sx)*6.0));
+        // increment as we sample the macro pixel
+        dx = 1.0/width;
+        // base x coord of macro pixel
+        bx = (floor(sx)+0.01)*dx*4.0;
+        color = texture(tex, vec2(bx, ty));
+        color1 = texture(tex, vec2(bx+dx, ty));
+        color2 = texture(tex, vec2(bx+dx*2.0, ty));
+        color3 = texture(tex, vec2(bx+dx*3.0, ty));
+        switch (px) {
+        case 0:
+        case 1:
+            Cb = color.b;
+            Cr = color.r;
+            break;
+        case 2:
+        case 3:
+            Cb = color1.g;
+            Cr = color2.b;
+            break;
+        default:
+            Cb = color2.r;
+            Cr = color3.g;
+            break;
+        }
+        switch (px) {
+        case 0:
+            Y = color.g;
+            break;
+        case 1:
+            Y = color1.b;
+            break;
+        case 2:
+            Y = color1.r;
+            break;
+        case 3:
+            Y = color2.g;
+            break;
+        case 4:
+            Y = color3.b;
+            break;
+        default:
+            Y = color3.r;
+            break;
+        }
+        Y = (Y - 0.0625) * 1.168949772;
+        Cb = (Cb - 0.0625) * 1.142857143 - 0.5;
+        Cr = (Cr - 0.0625) * 1.142857143 - 0.5;
+        color.r = Y + 1.5748 * Cr;
+        color.g = Y - 0.1873 * Cb - 0.4681 * Cr;
+        color.b = Y + 1.8556 * Cb;
+        color.a = 0.7;
+        gl_FragColor = color;
+    }
+"""
+
+# The exhausitve list of pixel formats that are transferred as float texture
+# Only use those for greater efficiency and compatiblity.
+#
+fg_shaders = {
+    '2vuy'       :FragmentShader_2vuy,
+    '8BitYUV'    :FragmentShader_2vuy,
+    'v210'       :FragmentShader_v210,
+    '10BitYUV'   :FragmentShader_v210,
+    '8BitBGRA'   :FragmentShader_R10l,
+    'BGRA'       :FragmentShader_R10l,
+    '8BitARGB'   :FragmentShader_R10l,
+    '10BitRGBXLE':FragmentShader_R10l,
+    'R10l'       :FragmentShader_R10l
+    }
+
+
+#
+# Helper function to attach a pixel shader to the material that receives the video frame.
+#
+
+def config_video(obj, format, pixel, is3D=False, mat=0, card=0):
+    if pixel not in fg_shaders:
+        raise('Unsuported shader')
+    shader = obj.meshes[0].materials[mat].getShader()
+    if shader is not None and not shader.isValid():
+        shader.setSource(VertexShader, fg_shaders[pixel], True)
+        shader.setSampler('tex', 0)
+        shader.setUniformEyef("eye")
+        shader.setUniform1f("stereo", 0.5 if is3D else 1.0)
+    tex = vt.Texture(obj, mat)
+    tex.source = vt.VideoDeckLink(format + "/" + pixel + ("/3D" if is3D else ""), card)
+    print("frame rate: ", tex.source.framerate)
+    tex.source.play()
+    obj["video"] = tex
+
+#
+# Attach this function to an object that has a material with texture
+# and call it once to initialize the object
+#
+def init(cont):
+    # config_video(cont.owner, 'HD720p5994', '8BitBGRA')
+    # config_video(cont.owner, 'HD720p5994', '8BitYUV')
+    # config_video(cont.owner, 'pal ', '10BitYUV')
+    config_video(cont.owner, 'pal ', '8BitYUV')
+
+
+#
+# To be called on every frame
+#
+def play(cont):
+    obj = cont.owner
+    video = obj.get("video")
+    if video is not None:
+        video.refresh(True)

doc/python_api/rst/bge.constraints.rst

@@ -8,11 +8,11 @@ Physics Constraints (bge.constraints)
 Examples
 --------
 
-.. include:: ../examples/bge.constraints.py
+.. include:: __/examples/bge.constraints.py
    :start-line: 1
    :end-line: 4
 
-.. literalinclude:: ../examples/bge.constraints.py
+.. literalinclude:: __/examples/bge.constraints.py
    :lines: 6-
 
 

doc/python_api/rst/bge.events.rst

@@ -12,53 +12,53 @@ This module holds key constants for the SCA_KeyboardSensor.
 
 .. code-block:: python
 
-	# Set a connected keyboard sensor to accept F1
-	import bge
+   # Set a connected keyboard sensor to accept F1
+   import bge
 
-	co = bge.logic.getCurrentController()
-	# 'Keyboard' is a keyboard sensor
-	sensor = co.sensors["Keyboard"]
-	sensor.key = bge.events.F1KEY
+   co = bge.logic.getCurrentController()
+   # 'Keyboard' is a keyboard sensor
+   sensor = co.sensors["Keyboard"]
+   sensor.key = bge.events.F1KEY
 
-.. code-block:: python
+   code-block:: python
 
-	# Do the all keys thing
-	import bge
+   # Do the all keys thing
+   import bge
 
-	co = bge.logic.getCurrentController()
-	# 'Keyboard' is a keyboard sensor
-	sensor = co.sensors["Keyboard"]
+   co = bge.logic.getCurrentController()
+   # 'Keyboard' is a keyboard sensor
+   sensor = co.sensors["Keyboard"]
 
-	for key,status in sensor.events:
-		# key[0] == bge.events.keycode, key[1] = status
-		if status == bge.logic.KX_INPUT_JUST_ACTIVATED:
-			if key == bge.events.WKEY:
-				# Activate Forward!
-			if key == bge.events.SKEY:
-				# Activate Backward!
-			if key == bge.events.AKEY:
-				# Activate Left!
-			if key == bge.events.DKEY:
-				# Activate Right!
+   for key,status in sensor.events:
+   	# key[0] == bge.events.keycode, key[1] = status
+   	if status == bge.logic.KX_INPUT_JUST_ACTIVATED:
+   		if key == bge.events.WKEY:
+   			# Activate Forward!
+   		if key == bge.events.SKEY:
+   			# Activate Backward!
+   		if key == bge.events.AKEY:
+   			# Activate Left!
+   		if key == bge.events.DKEY:
+   			# Activate Right!
 
-.. code-block:: python
+   code-block:: python
 
-	# The all keys thing without a keyboard sensor (but you will
-	# need an always sensor with pulse mode on)
-	import bge
+   # The all keys thing without a keyboard sensor (but you will
+   # need an always sensor with pulse mode on)
+   import bge
 
-	# Just shortening names here
-	keyboard = bge.logic.keyboard
-	JUST_ACTIVATED = bge.logic.KX_INPUT_JUST_ACTIVATED
+   # Just shortening names here
+   keyboard = bge.logic.keyboard
+   JUST_ACTIVATED = bge.logic.KX_INPUT_JUST_ACTIVATED
 
-	if keyboard.events[bge.events.WKEY] == JUST_ACTIVATED:
-		print("Activate Forward!")
-	if keyboard.events[bge.events.SKEY] == JUST_ACTIVATED:
-		print("Activate Backward!")
-	if keyboard.events[bge.events.AKEY] == JUST_ACTIVATED:
-		print("Activate Left!")
-	if keyboard.events[bge.events.DKEY] == JUST_ACTIVATED:
-		print("Activate Right!")
+   if keyboard.events[bge.events.WKEY] == JUST_ACTIVATED:
+   	print("Activate Forward!")
+   if keyboard.events[bge.events.SKEY] == JUST_ACTIVATED:
+   	print("Activate Backward!")
+   if keyboard.events[bge.events.AKEY] == JUST_ACTIVATED:
+   	print("Activate Left!")
+   if keyboard.events[bge.events.DKEY] == JUST_ACTIVATED:
+   	print("Activate Right!")
 
 
 *********

doc/python_api/rst/bge.logic.rst

@@ -378,6 +378,28 @@ General functions
 
    Render next frame (if Python has control)
 
+.. function:: setRender(render)
+
+   Sets the global flag that controls the render of the scene. 
+   If True, the render is done after the logic frame.
+   If False, the render is skipped and another logic frame starts immediately.
+
+   .. note::
+
+      GPU VSync no longer limits the number of frame per second when render is off, 
+      but the *Use Frame Rate* option still regulates the fps. To run as many frames
+      as possible, untick this option (Render Properties, System panel).
+
+   :arg render: the render flag
+   :type render: bool
+
+.. function:: getRender()
+
+   Get the current value of the global render flag
+
+   :return: The flag value
+   :rtype: bool
+
 **********************
 Time related functions
 **********************

doc/python_api/rst/bge.render.rst

@@ -90,6 +90,48 @@ Constants
 
    Right eye being used during stereoscopic rendering.
 
+.. data:: RAS_OFS_RENDER_BUFFER
+
+   The pixel buffer for offscreen render is a RenderBuffer. Argument to :func:`offScreenCreate`
+
+.. data:: RAS_OFS_RENDER_TEXTURE
+
+   The pixel buffer for offscreen render is a Texture. Argument to :func:`offScreenCreate`
+
+
+*****
+Types
+*****
+
+.. class:: RASOffScreen
+
+   An off-screen render buffer object. 
+
+   Use :func:`offScreenCreate` to create it.
+   Currently it can only be used in the :class:`bge.texture.ImageRender`
+   constructor to render on a FBO rather than the default viewport.
+
+  .. attribute:: width
+
+     The width in pixel of the FBO
+
+     :type: integer
+
+  .. attribute:: height
+
+     The height in pixel of the FBO
+
+     :type: integer
+
+  .. attribute:: color
+
+     The underlying OpenGL bind code of the texture object that holds
+     the rendered image, 0 if the FBO is using RenderBuffer.
+     The choice between RenderBuffer and Texture is determined
+     by the target argument of :func:`offScreenCreate`.
+
+     :type: integer
+
 
 *********
 Functions
@@ -362,3 +404,22 @@ Functions
    Get the current vsync value
 
    :rtype: One of VSYNC_OFF, VSYNC_ON, VSYNC_ADAPTIVE
+
+.. function:: offScreenCreate(width,height[,samples=0][,target=bge.render.RAS_OFS_RENDER_BUFFER])
+
+   Create a Off-screen render buffer object.
+
+   :arg width: the width of the buffer in pixels
+   :type width: integer
+   :arg height: the height of the buffer in pixels
+   :type height: integer
+   :arg samples: the number of multisample for anti-aliasing (MSAA), 0 to disable MSAA
+   :type samples: integer
+   :arg target: the pixel storage: :data:`RAS_OFS_RENDER_BUFFER` to render on RenderBuffers (the default),
+      :data:`RAS_OFS_RENDER_TEXTURE` to render on texture. 
+      The later is interesting if you want to access the texture directly (see :attr:`RASOffScreen.color`).
+      Otherwise the default is preferable as it's more widely supported by GPUs and more efficient.
+      If the GPU does not support MSAA+Texture (e.g. Intel HD GPU), MSAA will be disabled.
+   :type target: integer
+   :rtype: :class:`RASOffScreen`
+

doc/python_api/rst/bge_types/bge.types.BL_Shader.rst

@@ -214,6 +214,16 @@ base class --- :class:`PyObjectPlus`
       :arg iList: a list (2, 3 or 4 elements) of integer values
       :type iList: list[integer]
 
+   .. method:: setUniformEyef(name)
+
+      Set a uniform with a float value that reflects the eye being render in stereo mode: 
+      0.0 for the left eye, 0.5 for the right eye. In non stereo mode, the value of the uniform
+      is fixed to 0.0. The typical use of this uniform is in stereo mode to sample stereo textures
+      containing the left and right eye images in a top-bottom order. 
+
+      :arg name: the uniform name
+      :type name: string
+
    .. method:: validate()
 
       Validate the shader object.

doc/python_api/rst/bge_types/bge.types.KX_LightObject.rst

@@ -60,37 +60,37 @@ base class --- :class:`KX_GameObject`
 
       :type: float (read only)
 
-   ..attribute:: shadowFrustumSize
+   .. attribute:: shadowFrustumSize
 
       Size of the frustum used for creating the shadowmap.
 
       :type: float (read only)
 
-   ..attribute:: shadowBindId
+   .. attribute:: shadowBindId
 
       The OpenGL shadow texture bind number/id.
 
       :type: int (read only)
 
-   ..attribute:: shadowMapType
+   .. attribute:: shadowMapType
 
       The shadow shadow map type (0 -> Simple; 1 -> Variance)
 
       :type: int (read only)
 
-   ..attribute:: shadowBias
+   .. attribute:: shadowBias
 
       The shadow buffer sampling bias.
 
       :type: float (read only)
 
-   ..attribute:: shadowBleedBias
+   .. attribute:: shadowBleedBias
 
       The bias for reducing light-bleed on variance shadow maps.
 
       :type: float (read only)
 
-   ..attribute:: useShadow
+   .. attribute:: useShadow
 
       Returns True if the light has Shadow option activated, else returns False.
 

doc/python_api/rst/bge_types/bge.types.SCA_PythonJoystick.rst

@@ -12,13 +12,13 @@ base class --- :class:`PyObjectPlus`
    .. attribute:: name
 
       The name assigned to the joystick by the operating system. (read-only)
-	  
+
       :type: string
 
    .. attribute:: activeButtons
 
       A list of active button values. (read-only)
-	  
+
       :type: list
 
    .. attribute:: axisValues
@@ -27,8 +27,10 @@ base class --- :class:`PyObjectPlus`
 
       :type: list of ints.
 
-      Each specifying the value of an axis between -1.0 and 1.0 depending on how far the axis is pushed, 0 for nothing.
-      The first 2 values are used by most joysticks and gamepads for directional control. 3rd and 4th values are only on some joysticks and can be used for arbitary controls.
+      Each specifying the value of an axis between -1.0 and 1.0
+      depending on how far the axis is pushed, 0 for nothing.
+      The first 2 values are used by most joysticks and gamepads for directional control.
+      3rd and 4th values are only on some joysticks and can be used for arbitary controls.
 
       * left:[-1.0, 0.0, ...]
       * right:[1.0, 0.0, ...]

doc/python_api/rst/include__bmesh.rst

@@ -1,7 +1,8 @@
 ..
    This document is appended to the auto generated bmesh api doc to avoid clogging up the C files with details.
    to test this run:
-   ./blender.bin -b -noaudio -P doc/python_api/sphinx_doc_gen.py -- --partial bmesh* ; cd doc/python_api ; sphinx-build sphinx-in sphinx-out ; cd ../../
+   ./blender.bin -b -noaudio -P doc/python_api/sphinx_doc_gen.py -- \
+                 --partial bmesh* ; cd doc/python_api ; sphinx-build sphinx-in sphinx-out ; cd ../../
 
 
 Submodules:
@@ -40,7 +41,7 @@ For an overview of BMesh data types and how they reference each other see:
 Example Script
 --------------
 
-.. literalinclude:: ../../../release/scripts/templates_py/bmesh_simple.py
+.. literalinclude:: __/__/__/release/scripts/templates_py/bmesh_simple.py
 
 
 Stand-Alone Module
@@ -59,9 +60,9 @@ There are 2 ways to access BMesh data, you can create a new BMesh by converting
 :class:`bpy.types.BlendData.meshes` or by accessing the current edit mode mesh.
 see: :class:`bmesh.types.BMesh.from_mesh` and :mod:`bmesh.from_edit_mesh` respectively.
 
-When explicitly converting from mesh data python **owns** the data, that is to say - that the mesh only exists while
-python holds a reference to it, and the script is responsible for putting it back into a mesh data-block when the edits
-are done.
+When explicitly converting from mesh data python **owns** the data, that is to say -
+that the mesh only exists while python holds a reference to it,
+and the script is responsible for putting it back into a mesh data-block when the edits are done.
 
 Note that unlike :mod:`bpy`, a BMesh does not necessarily correspond to data in the currently open blend file,
 a BMesh can be created, edited and freed without the user ever seeing or having access to it.

doc/python_api/rst/info_quickstart.rst

@@ -151,7 +151,7 @@ Data Creation/Removal
 ^^^^^^^^^^^^^^^^^^^^^
 
 Those of you familiar with other Python API's may be surprised that
-new datablocks in the bpy API can't be created by calling the class:
+new data-blocks in the bpy API can't be created by calling the class:
 
    >>> bpy.types.Mesh()
    Traceback (most recent call last):
@@ -305,7 +305,7 @@ In Python, this is done by defining a class, which is a subclass of an existing
 Example Operator
 ----------------
 
-.. literalinclude:: ../../../release/scripts/templates_py/operator_simple.py
+.. literalinclude:: __/__/__/release/scripts/templates_py/operator_simple.py
 
 Once this script runs, ``SimpleOperator`` is registered with Blender
 and can be called from the operator search popup or added to the toolbar.
@@ -336,7 +336,7 @@ Example Panel
 Panels register themselves as a class, like an operator.
 Notice the extra ``bl_`` variables used to set the context they display in.
 
-.. literalinclude:: ../../../release/scripts/templates_py/ui_panel_simple.py
+.. literalinclude:: __/__/__/release/scripts/templates_py/ui_panel_simple.py
 
 To run the script:
 
@@ -393,11 +393,11 @@ so these are accessed as normal Python types.
 Internal Types
 --------------
 
-Used for Blender datablocks and collections: :class:`bpy.types.bpy_struct`
+Used for Blender data-blocks and collections: :class:`bpy.types.bpy_struct`
 
 For data that contains its own attributes groups/meshes/bones/scenes... etc.
 
-There are 2 main types that wrap Blenders data, one for datablocks
+There are 2 main types that wrap Blenders data, one for data-blocks
 (known internally as ``bpy_struct``), another for properties.
 
    >>> bpy.context.object

doc/python_api/rst_from_bmesh_opdefines.py

@@ -57,7 +57,7 @@ Operator Example
 ++++++++++++++++
 This script shows how operators can be used to model a link of a chain.
 
-.. literalinclude:: ../examples/bmesh.ops.1.py
+.. literalinclude:: __/examples/bmesh.ops.1.py
 """
 
 

doc/python_api/sphinx_doc_gen.py

@@ -325,6 +325,24 @@ except ImportError:
                                                "freestyle.types",
                                                "freestyle.utils"]
 
+# Source files we use, and need to copy to the OUTPUT_DIR
+# to have working out-of-source builds.
+# Note that ".." is replaced by "__" in the RST files,
+# to avoid having to match Blender's source tree.
+EXTRA_SOURCE_FILES = (
+    "../../../release/scripts/templates_py/bmesh_simple.py",
+    "../../../release/scripts/templates_py/operator_simple.py",
+    "../../../release/scripts/templates_py/ui_panel_simple.py",
+    "../../../release/scripts/templates_py/ui_previews_custom_icon.py",
+    "../examples/bge.constraints.py",
+    "../examples/bge.texture.1.py",
+    "../examples/bge.texture.2.py",
+    "../examples/bge.texture.py",
+    "../examples/bmesh.ops.1.py",
+    "../examples/bpy.app.translations.py",
+    )
+
+
 # examples
 EXAMPLES_DIR = os.path.abspath(os.path.join(SCRIPT_DIR, "examples"))
 EXAMPLE_SET = set()
@@ -1890,6 +1908,21 @@ def copy_handwritten_rsts(basepath):
             shutil.copy2(os.path.join(RST_DIR, f), basepath)
 
 
+def copy_handwritten_extra(basepath):
+    for f_src in EXTRA_SOURCE_FILES:
+        if os.sep != "/":
+            f_src = os.sep.join(f_src.split("/"))
+
+        f_dst = f_src.replace("..", "__")
+
+        f_src = os.path.join(RST_DIR, f_src)
+        f_dst = os.path.join(basepath, f_dst)
+
+        os.makedirs(os.path.dirname(f_dst), exist_ok=True)
+
+        shutil.copy2(f_src, f_dst)
+
+
 def rna2sphinx(basepath):
 
     try:
@@ -1921,35 +1954,48 @@ def rna2sphinx(basepath):
     # copy the other rsts
     copy_handwritten_rsts(basepath)
 
+    # copy source files referenced
+    copy_handwritten_extra(basepath)
 
-def align_sphinx_in_to_sphinx_in_tmp():
+
+def align_sphinx_in_to_sphinx_in_tmp(dir_src, dir_dst):
     '''
     Move changed files from SPHINX_IN_TMP to SPHINX_IN
     '''
     import filecmp
 
-    sphinx_in_files = set(os.listdir(SPHINX_IN))
-    sphinx_in_tmp_files = set(os.listdir(SPHINX_IN_TMP))
+    # possible the dir doesn't exist when running recursively
+    os.makedirs(dir_dst, exist_ok=True)
+
+    sphinx_dst_files = set(os.listdir(dir_dst))
+    sphinx_src_files = set(os.listdir(dir_src))
 
     # remove deprecated files that have been removed
-    for f in sorted(sphinx_in_files):
-        if f not in sphinx_in_tmp_files:
+    for f in sorted(sphinx_dst_files):
+        if f not in sphinx_src_files:
             BPY_LOGGER.debug("\tdeprecated: %s" % f)
-            os.remove(os.path.join(SPHINX_IN, f))
+            f_dst = os.path.join(dir_dst, f)
+            if os.path.isdir(f_dst):
+                shutil.rmtree(f_dst, True)
+            else:
+                os.remove(f_dst)
 
     # freshen with new files.
-    for f in sorted(sphinx_in_tmp_files):
-        f_from = os.path.join(SPHINX_IN_TMP, f)
-        f_to = os.path.join(SPHINX_IN, f)
+    for f in sorted(sphinx_src_files):
+        f_src = os.path.join(dir_src, f)
+        f_dst = os.path.join(dir_dst, f)
 
-        do_copy = True
-        if f in sphinx_in_files:
-            if filecmp.cmp(f_from, f_to):
-                do_copy = False
+        if os.path.isdir(f_src):
+            align_sphinx_in_to_sphinx_in_tmp(f_src, f_dst)
+        else:
+            do_copy = True
+            if f in sphinx_dst_files:
+                if filecmp.cmp(f_src, f_dst):
+                    do_copy = False
 
-        if do_copy:
-            BPY_LOGGER.debug("\tupdating: %s" % f)
-            shutil.copy(f_from, f_to)
+            if do_copy:
+                BPY_LOGGER.debug("\tupdating: %s" % f)
+                shutil.copy(f_src, f_dst)
 
 
 def refactor_sphinx_log(sphinx_logfile):
@@ -2036,7 +2082,7 @@ def main():
             shutil.rmtree(SPHINX_OUT_PDF, True)
     else:
         # move changed files in SPHINX_IN
-        align_sphinx_in_to_sphinx_in_tmp()
+        align_sphinx_in_to_sphinx_in_tmp(SPHINX_IN_TMP, SPHINX_IN)
 
     # report which example files weren't used
     EXAMPLE_SET_UNUSED = EXAMPLE_SET - EXAMPLE_SET_USED

doc/python_api/sphinx_doc_gen.sh

@@ -3,11 +3,6 @@
 #   bash doc/python_api/sphinx_doc_gen.sh
 # ssh upload means you need an account on the server
 
-if [ "$1" == "" ] ; then
-	echo "Expected a single argument for the username on blender.org, aborting"
-	exit 1
-fi
-
 # ----------------------------------------------------------------------------
 # Upload vars
 
@@ -22,9 +17,15 @@ if [ -z $BLENDER_BIN ] ; then
 	BLENDER_BIN="./blender.bin"
 fi
 
-SSH_USER=$1
-SSH_HOST=$SSH_USER"@blender.org"
-SSH_UPLOAD="/data/www/vhosts/www.blender.org/api" # blender_python_api_VERSION, added after
+if [ "$1" == "" ] ; then
+	echo "Expected a single argument for the username on blender.org, skipping upload step!"
+	DO_UPLOAD=false
+else
+	SSH_USER=$1
+	SSH_HOST=$SSH_USER"@blender.org"
+	SSH_UPLOAD="/data/www/vhosts/www.blender.org/api" # blender_python_api_VERSION, added after
+fi
+
 
 # ----------------------------------------------------------------------------
 # Blender Version & Info
@@ -33,10 +34,12 @@ SSH_UPLOAD="/data/www/vhosts/www.blender.org/api" # blender_python_api_VERSION,
 # "_".join(str(v) for v in bpy.app.version)
 # custom blender vars
 blender_srcdir=$(dirname -- $0)/../..
-blender_version=$(grep "BLENDER_VERSION\s" "$blender_srcdir/source/blender/blenkernel/BKE_blender_version.h" | awk '{print $3}')
-blender_version_char=$(grep "BLENDER_VERSION_CHAR\s" "$blender_srcdir/source/blender/blenkernel/BKE_blender_version.h" | awk '{print $3}')
-blender_version_cycle=$(grep "BLENDER_VERSION_CYCLE\s" "$blender_srcdir/source/blender/blenkernel/BKE_blender_version.h" | awk '{print $3}')
-blender_subversion=$(grep "BLENDER_SUBVERSION\s" "$blender_srcdir/source/blender/blenkernel/BKE_blender_version.h" | awk '{print $3}')
+blender_version_header="$blender_srcdir/source/blender/blenkernel/BKE_blender_version.h"
+blender_version=$(grep "BLENDER_VERSION\s" "$blender_version_header" | awk '{print $3}')
+blender_version_char=$(grep "BLENDER_VERSION_CHAR\s" "$blender_version_header" | awk '{print $3}')
+blender_version_cycle=$(grep "BLENDER_VERSION_CYCLE\s" "$blender_version_header" | awk '{print $3}')
+blender_subversion=$(grep "BLENDER_SUBVERSION\s" "$blender_version_header" | awk '{print $3}')
+unset blender_version_header
 
 if [ "$blender_version_cycle" = "release" ] ; then
 	BLENDER_VERSION=$(expr $blender_version / 100)_$(expr $blender_version % 100)$blender_version_char"_release"
@@ -48,6 +51,8 @@ SSH_UPLOAD_FULL=$SSH_UPLOAD/"blender_python_api_"$BLENDER_VERSION
 
 SPHINXBASE=doc/python_api
 
+SPHINX_WORKDIR="$(mktemp --directory --suffix=.sphinx)"
+
 
 # ----------------------------------------------------------------------------
 # Generate reStructuredText (blender/python only)
@@ -59,23 +64,25 @@ if $DO_EXE_BLENDER ; then
 		-noaudio \
 		--factory-startup \
 		--python-exit-code 1 \
-		--python $SPHINXBASE/sphinx_doc_gen.py
+		--python $SPHINXBASE/sphinx_doc_gen.py \
+		-- \
+		--output=$SPHINX_WORKDIR
 
-	if (($? == 1)) ; then
+
+	if (($? != 0)) ; then
 		echo "Generating documentation failed, aborting"
 		exit 1
 	fi
 fi
 
-
 # ----------------------------------------------------------------------------
 # Generate HTML (sphinx)
 
 if $DO_OUT_HTML ; then
-	# sphinx-build -n -b html $SPHINXBASE/sphinx-in $SPHINXBASE/sphinx-out
+	# sphinx-build -n -b html $SPHINX_WORKDIR/sphinx-in $SPHINX_WORKDIR/sphinx-out
 
 	# annoying bug in sphinx makes it very slow unless we do this. should report.
-	cd $SPHINXBASE
+	cd $SPHINX_WORKDIR
 	sphinx-build -b html sphinx-in sphinx-out
 
 	# XXX, saves space on upload and zip, should move HTML outside
@@ -103,20 +110,21 @@ fi
 # Generate PDF (sphinx/laytex)
 
 if $DO_OUT_PDF ; then
-	sphinx-build -n -b latex $SPHINXBASE/sphinx-in $SPHINXBASE/sphinx-out
-	make -C $SPHINXBASE/sphinx-out
-	mv $SPHINXBASE/sphinx-out/contents.pdf $SPHINXBASE/sphinx-out/blender_python_reference_$BLENDER_VERSION.pdf
+	cd $SPHINX_WORKDIR
+	sphinx-build -n -b latex $SPHINX_WORKDIR/sphinx-in $SPHINX_WORKDIR/sphinx-out
+	make -C $SPHINX_WORKDIR/sphinx-out
+	mv $SPHINX_WORKDIR/sphinx-out/contents.pdf \
+	   $SPHINX_WORKDIR/sphinx-out/blender_python_reference_$BLENDER_VERSION.pdf
 fi
 
-
 # ----------------------------------------------------------------------------
 # Upload to blender servers, comment this section for testing
 
 if $DO_UPLOAD ; then
 
-	cp $SPHINXBASE/sphinx-out/contents.html $SPHINXBASE/sphinx-out/index.html
+	cp $SPHINX_WORKDIR/sphinx-out/contents.html $SPHINX_WORKDIR/sphinx-out/index.html
 	ssh $SSH_USER@blender.org 'rm -rf '$SSH_UPLOAD_FULL'/*'
-	rsync --progress -ave "ssh -p 22" $SPHINXBASE/sphinx-out/* $SSH_HOST:$SSH_UPLOAD_FULL/
+	rsync --progress -ave "ssh -p 22" $SPHINX_WORKDIR/sphinx-out/* $SSH_HOST:$SSH_UPLOAD_FULL/
 
 	## symlink the dir to a static URL
 	#ssh $SSH_USER@blender.org 'rm '$SSH_UPLOAD'/250PythonDoc && ln -s '$SSH_UPLOAD_FULL' '$SSH_UPLOAD'/250PythonDoc'
@@ -134,11 +142,15 @@ if $DO_UPLOAD ; then
 
 	if $DO_OUT_PDF ; then
 		# rename so local PDF has matching name.
-		rsync --progress -ave "ssh -p 22" $SPHINXBASE/sphinx-out/blender_python_reference_$BLENDER_VERSION.pdf $SSH_HOST:$SSH_UPLOAD_FULL/blender_python_reference_$BLENDER_VERSION.pdf
+		rsync --progress -ave "ssh -p 22" \
+		      $SPHINX_WORKDIR/sphinx-out/blender_python_reference_$BLENDER_VERSION.pdf \
+		      $SSH_HOST:$SSH_UPLOAD_FULL/blender_python_reference_$BLENDER_VERSION.pdf
 	fi
 
 	if $DO_OUT_HTML_ZIP ; then
-		rsync --progress -ave "ssh -p 22" $SPHINXBASE/blender_python_reference_$BLENDER_VERSION.zip $SSH_HOST:$SSH_UPLOAD_FULL/blender_python_reference_$BLENDER_VERSION.zip
+		rsync --progress -ave "ssh -p 22" \
+		      $SPHINX_WORKDIR/blender_python_reference_$BLENDER_VERSION.zip \
+		      $SSH_HOST:$SSH_UPLOAD_FULL/blender_python_reference_$BLENDER_VERSION.zip
 	fi
 
 fi
@@ -149,5 +161,5 @@ fi
 
 echo ""
 echo "Finished! view the docs from: "
-if $DO_OUT_HTML ; then echo "  html:" $SPHINXBASE/sphinx-out/contents.html ; fi
-if $DO_OUT_PDF ; then  echo "   pdf:" $SPHINXBASE/sphinx-out/blender_python_reference_$BLENDER_VERSION.pdf ; fi
+if $DO_OUT_HTML ; then echo "  html:" $SPHINX_WORKDIR/sphinx-out/contents.html ; fi
+if $DO_OUT_PDF ; then  echo "   pdf:" $SPHINX_WORKDIR/sphinx-out/blender_python_reference_$BLENDER_VERSION.pdf ; fi

extern/Eigen3/README.blender

@@ -0,0 +1,6 @@
+Project: Eigen, template library for linear algebra: matrices, vectors, numerical solvers, and related algorithms
+URL: http://eigen.tuxfamily.org/index.php?title=Main_Page
+License: GPLv3+
+Upstream version: 3.2.7
+Local modifications:
+- OpenMP fix for MSVC2015, see http://eigen.tuxfamily.org/bz/show_bug.cgi?id=1131

extern/binreloc/README.blender

@@ -0,0 +1,6 @@
+Project: AutoPackage
+URL: http://autopackage.org/docs/binreloc (original, defunct)
+     http://alien.cern.ch/cache/autopackage-1.0/site/docs/binreloc/ (cache)
+License: Public Domain
+Upstream version: Unknown (Last Release)
+Local modifications: None

extern/carve/README.blender

@@ -0,0 +1,4 @@
+Project: Carve, CSG library
+URL: https://code.google.com/archive/p/carve/
+Upstream version 9a85d733a43d
+Local modifications: See patches/ folder

extern/ceres/README.blender

@@ -0,0 +1,4 @@
+Project: Ceres Solver
+URL: http://ceres-solver.org/
+Upstream version 1.11 (aef9c9563b08d5f39eee1576af133a84749d1b48)
+Local modifications: None

extern/clew/README.blender

@@ -0,0 +1,5 @@
+Project: OpenCL Wrangler
+URL: https://github.com/OpenCLWrangler/clew
+License: Apache 2.0
+Upstream version: 277db43
+Local modifications: None

extern/cuew/README.blender

@@ -0,0 +1,5 @@
+Project: Cuda Wrangler
+URL: https://github.com/CudaWrangler/cuew
+License: Apache 2.0
+Upstream version: e2e0315
+Local modifications: None

extern/cuew/include/cuew.h

@@ -131,8 +131,8 @@ typedef struct CUsurfref_st* CUsurfref;
 typedef struct CUevent_st* CUevent;
 typedef struct CUstream_st* CUstream;
 typedef struct CUgraphicsResource_st* CUgraphicsResource;
-typedef unsigned CUtexObject;
-typedef unsigned CUsurfObject;
+typedef unsigned long long CUtexObject;
+typedef unsigned long long CUsurfObject;
 
 typedef struct CUuuid_st {
   char bytes[16];
@@ -603,7 +603,7 @@ typedef struct CUDA_ARRAY_DESCRIPTOR_st {
   size_t Width;
   size_t Height;
   CUarray_format Format;
-  unsigned NumChannels;
+  unsigned int NumChannels;
 } CUDA_ARRAY_DESCRIPTOR;
 
 typedef struct CUDA_ARRAY3D_DESCRIPTOR_st {
@@ -611,8 +611,8 @@ typedef struct CUDA_ARRAY3D_DESCRIPTOR_st {
   size_t Height;
   size_t Depth;
   CUarray_format Format;
-  unsigned NumChannels;
-  unsigned Flags;
+  unsigned int NumChannels;
+  unsigned int Flags;
 } CUDA_ARRAY3D_DESCRIPTOR;
 
 typedef struct CUDA_RESOURCE_DESC_st {
@@ -627,13 +627,13 @@ typedef struct CUDA_RESOURCE_DESC_st {
     struct {
       CUdeviceptr devPtr;
       CUarray_format format;
-      unsigned numChannels;
+      unsigned int numChannels;
       size_t sizeInBytes;
     } linear;
     struct {
       CUdeviceptr devPtr;
       CUarray_format format;
-      unsigned numChannels;
+      unsigned int numChannels;
       size_t width;
       size_t height;
       size_t pitchInBytes;
@@ -642,14 +642,14 @@ typedef struct CUDA_RESOURCE_DESC_st {
       int reserved[32];
     } reserved;
   } res;
-  unsigned flags;
+  unsigned int flags;
 } CUDA_RESOURCE_DESC;
 
 typedef struct CUDA_TEXTURE_DESC_st {
   CUaddress_mode addressMode[3];
   CUfilter_mode filterMode;
-  unsigned flags;
-  unsigned maxAnisotropy;
+  unsigned int flags;
+  unsigned int maxAnisotropy;
   CUfilter_mode mipmapFilterMode;
   float mipmapLevelBias;
   float minMipmapLevelClamp;
@@ -700,19 +700,19 @@ typedef struct CUDA_RESOURCE_VIEW_DESC_st {
   size_t width;
   size_t height;
   size_t depth;
-  unsigned firstMipmapLevel;
-  unsigned lastMipmapLevel;
-  unsigned firstLayer;
-  unsigned lastLayer;
-  unsigned reserved[16];
+  unsigned int firstMipmapLevel;
+  unsigned int lastMipmapLevel;
+  unsigned int firstLayer;
+  unsigned int lastLayer;
+  unsigned int reserved[16];
 } CUDA_RESOURCE_VIEW_DESC;
 
 typedef struct CUDA_POINTER_ATTRIBUTE_P2P_TOKENS_st {
-  unsigned p2pToken;
-  unsigned vaSpaceToken;
+  unsigned long long p2pToken;
+  unsigned int vaSpaceToken;
 } CUDA_POINTER_ATTRIBUTE_P2P_TOKENS;
-typedef unsigned GLenum;
-typedef unsigned GLuint;
+typedef unsigned int GLenum;
+typedef unsigned int GLuint;
 typedef int GLint;
 
 typedef enum CUGLDeviceList_enum {
@@ -751,7 +751,7 @@ typedef struct _nvrtcProgram* nvrtcProgram;
 /* Function types. */
 typedef CUresult CUDAAPI tcuGetErrorString(CUresult error, const char* pStr);
 typedef CUresult CUDAAPI tcuGetErrorName(CUresult error, const char* pStr);
-typedef CUresult CUDAAPI tcuInit(unsigned Flags);
+typedef CUresult CUDAAPI tcuInit(unsigned int Flags);
 typedef CUresult CUDAAPI tcuDriverGetVersion(int* driverVersion);
 typedef CUresult CUDAAPI tcuDeviceGet(CUdevice* device, int ordinal);
 typedef CUresult CUDAAPI tcuDeviceGetCount(int* count);
@@ -762,17 +762,17 @@ typedef CUresult CUDAAPI tcuDeviceGetProperties(CUdevprop* prop, CUdevice dev);
 typedef CUresult CUDAAPI tcuDeviceComputeCapability(int* major, int* minor, CUdevice dev);
 typedef CUresult CUDAAPI tcuDevicePrimaryCtxRetain(CUcontext* pctx, CUdevice dev);
 typedef CUresult CUDAAPI tcuDevicePrimaryCtxRelease(CUdevice dev);
-typedef CUresult CUDAAPI tcuDevicePrimaryCtxSetFlags(CUdevice dev, unsigned flags);
-typedef CUresult CUDAAPI tcuDevicePrimaryCtxGetState(CUdevice dev, unsigned* flags, int* active);
+typedef CUresult CUDAAPI tcuDevicePrimaryCtxSetFlags(CUdevice dev, unsigned int flags);
+typedef CUresult CUDAAPI tcuDevicePrimaryCtxGetState(CUdevice dev, unsigned int* flags, int* active);
 typedef CUresult CUDAAPI tcuDevicePrimaryCtxReset(CUdevice dev);
-typedef CUresult CUDAAPI tcuCtxCreate_v2(CUcontext* pctx, unsigned flags, CUdevice dev);
+typedef CUresult CUDAAPI tcuCtxCreate_v2(CUcontext* pctx, unsigned int flags, CUdevice dev);
 typedef CUresult CUDAAPI tcuCtxDestroy_v2(CUcontext ctx);
 typedef CUresult CUDAAPI tcuCtxPushCurrent_v2(CUcontext ctx);
 typedef CUresult CUDAAPI tcuCtxPopCurrent_v2(CUcontext* pctx);
 typedef CUresult CUDAAPI tcuCtxSetCurrent(CUcontext ctx);
 typedef CUresult CUDAAPI tcuCtxGetCurrent(CUcontext* pctx);
 typedef CUresult CUDAAPI tcuCtxGetDevice(CUdevice* device);
-typedef CUresult CUDAAPI tcuCtxGetFlags(unsigned* flags);
+typedef CUresult CUDAAPI tcuCtxGetFlags(unsigned int* flags);
 typedef CUresult CUDAAPI tcuCtxSynchronize(void);
 typedef CUresult CUDAAPI tcuCtxSetLimit(CUlimit limit, size_t value);
 typedef CUresult CUDAAPI tcuCtxGetLimit(size_t* pvalue, CUlimit limit);
@@ -780,43 +780,43 @@ typedef CUresult CUDAAPI tcuCtxGetCacheConfig(CUfunc_cache* pconfig);
 typedef CUresult CUDAAPI tcuCtxSetCacheConfig(CUfunc_cache config);
 typedef CUresult CUDAAPI tcuCtxGetSharedMemConfig(CUsharedconfig* pConfig);
 typedef CUresult CUDAAPI tcuCtxSetSharedMemConfig(CUsharedconfig config);
-typedef CUresult CUDAAPI tcuCtxGetApiVersion(CUcontext ctx, unsigned* version);
+typedef CUresult CUDAAPI tcuCtxGetApiVersion(CUcontext ctx, unsigned int* version);
 typedef CUresult CUDAAPI tcuCtxGetStreamPriorityRange(int* leastPriority, int* greatestPriority);
-typedef CUresult CUDAAPI tcuCtxAttach(CUcontext* pctx, unsigned flags);
+typedef CUresult CUDAAPI tcuCtxAttach(CUcontext* pctx, unsigned int flags);
 typedef CUresult CUDAAPI tcuCtxDetach(CUcontext ctx);
 typedef CUresult CUDAAPI tcuModuleLoad(CUmodule* module, const char* fname);
 typedef CUresult CUDAAPI tcuModuleLoadData(CUmodule* module, const void* image);
-typedef CUresult CUDAAPI tcuModuleLoadDataEx(CUmodule* module, const void* image, unsigned numOptions, CUjit_option* options, void* optionValues);
+typedef CUresult CUDAAPI tcuModuleLoadDataEx(CUmodule* module, const void* image, unsigned int numOptions, CUjit_option* options, void* optionValues);
 typedef CUresult CUDAAPI tcuModuleLoadFatBinary(CUmodule* module, const void* fatCubin);
 typedef CUresult CUDAAPI tcuModuleUnload(CUmodule hmod);
 typedef CUresult CUDAAPI tcuModuleGetFunction(CUfunction* hfunc, CUmodule hmod, const char* name);
 typedef CUresult CUDAAPI tcuModuleGetGlobal_v2(CUdeviceptr* dptr, size_t* bytes, CUmodule hmod, const char* name);
 typedef CUresult CUDAAPI tcuModuleGetTexRef(CUtexref* pTexRef, CUmodule hmod, const char* name);
 typedef CUresult CUDAAPI tcuModuleGetSurfRef(CUsurfref* pSurfRef, CUmodule hmod, const char* name);
-typedef CUresult CUDAAPI tcuLinkCreate_v2(unsigned numOptions, CUjit_option* options, void* optionValues, CUlinkState* stateOut);
-typedef CUresult CUDAAPI tcuLinkAddData_v2(CUlinkState state, CUjitInputType type, void* data, size_t size, const char* name, unsigned numOptions, CUjit_option* options, void* optionValues);
-typedef CUresult CUDAAPI tcuLinkAddFile_v2(CUlinkState state, CUjitInputType type, const char* path, unsigned numOptions, CUjit_option* options, void* optionValues);
+typedef CUresult CUDAAPI tcuLinkCreate_v2(unsigned int numOptions, CUjit_option* options, void* optionValues, CUlinkState* stateOut);
+typedef CUresult CUDAAPI tcuLinkAddData_v2(CUlinkState state, CUjitInputType type, void* data, size_t size, const char* name, unsigned int numOptions, CUjit_option* options, void* optionValues);
+typedef CUresult CUDAAPI tcuLinkAddFile_v2(CUlinkState state, CUjitInputType type, const char* path, unsigned int numOptions, CUjit_option* options, void* optionValues);
 typedef CUresult CUDAAPI tcuLinkComplete(CUlinkState state, void* cubinOut, size_t* sizeOut);
 typedef CUresult CUDAAPI tcuLinkDestroy(CUlinkState state);
 typedef CUresult CUDAAPI tcuMemGetInfo_v2(size_t* free, size_t* total);
 typedef CUresult CUDAAPI tcuMemAlloc_v2(CUdeviceptr* dptr, size_t bytesize);
-typedef CUresult CUDAAPI tcuMemAllocPitch_v2(CUdeviceptr* dptr, size_t* pPitch, size_t WidthInBytes, size_t Height, unsigned ElementSizeBytes);
+typedef CUresult CUDAAPI tcuMemAllocPitch_v2(CUdeviceptr* dptr, size_t* pPitch, size_t WidthInBytes, size_t Height, unsigned int ElementSizeBytes);
 typedef CUresult CUDAAPI tcuMemFree_v2(CUdeviceptr dptr);
 typedef CUresult CUDAAPI tcuMemGetAddressRange_v2(CUdeviceptr* pbase, size_t* psize, CUdeviceptr dptr);
 typedef CUresult CUDAAPI tcuMemAllocHost_v2(void* pp, size_t bytesize);
 typedef CUresult CUDAAPI tcuMemFreeHost(void* p);
-typedef CUresult CUDAAPI tcuMemHostAlloc(void* pp, size_t bytesize, unsigned Flags);
-typedef CUresult CUDAAPI tcuMemHostGetDevicePointer_v2(CUdeviceptr* pdptr, void* p, unsigned Flags);
-typedef CUresult CUDAAPI tcuMemHostGetFlags(unsigned* pFlags, void* p);
-typedef CUresult CUDAAPI tcuMemAllocManaged(CUdeviceptr* dptr, size_t bytesize, unsigned flags);
+typedef CUresult CUDAAPI tcuMemHostAlloc(void* pp, size_t bytesize, unsigned int Flags);
+typedef CUresult CUDAAPI tcuMemHostGetDevicePointer_v2(CUdeviceptr* pdptr, void* p, unsigned int Flags);
+typedef CUresult CUDAAPI tcuMemHostGetFlags(unsigned int* pFlags, void* p);
+typedef CUresult CUDAAPI tcuMemAllocManaged(CUdeviceptr* dptr, size_t bytesize, unsigned int flags);
 typedef CUresult CUDAAPI tcuDeviceGetByPCIBusId(CUdevice* dev, const char* pciBusId);
 typedef CUresult CUDAAPI tcuDeviceGetPCIBusId(char* pciBusId, int len, CUdevice dev);
 typedef CUresult CUDAAPI tcuIpcGetEventHandle(CUipcEventHandle* pHandle, CUevent event);
 typedef CUresult CUDAAPI tcuIpcOpenEventHandle(CUevent* phEvent, CUipcEventHandle handle);
 typedef CUresult CUDAAPI tcuIpcGetMemHandle(CUipcMemHandle* pHandle, CUdeviceptr dptr);
-typedef CUresult CUDAAPI tcuIpcOpenMemHandle(CUdeviceptr* pdptr, CUipcMemHandle handle, unsigned Flags);
+typedef CUresult CUDAAPI tcuIpcOpenMemHandle(CUdeviceptr* pdptr, CUipcMemHandle handle, unsigned int Flags);
 typedef CUresult CUDAAPI tcuIpcCloseMemHandle(CUdeviceptr dptr);
-typedef CUresult CUDAAPI tcuMemHostRegister_v2(void* p, size_t bytesize, unsigned Flags);
+typedef CUresult CUDAAPI tcuMemHostRegister_v2(void* p, size_t bytesize, unsigned int Flags);
 typedef CUresult CUDAAPI tcuMemHostUnregister(void* p);
 typedef CUresult CUDAAPI tcuMemcpy(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount);
 typedef CUresult CUDAAPI tcuMemcpyPeer(CUdeviceptr dstDevice, CUcontext dstContext, CUdeviceptr srcDevice, CUcontext srcContext, size_t ByteCount);
@@ -842,40 +842,40 @@ typedef CUresult CUDAAPI tcuMemcpyAtoHAsync_v2(void* dstHost, CUarray srcArray,
 typedef CUresult CUDAAPI tcuMemcpy2DAsync_v2(const CUDA_MEMCPY2D* pCopy, CUstream hStream);
 typedef CUresult CUDAAPI tcuMemcpy3DAsync_v2(const CUDA_MEMCPY3D* pCopy, CUstream hStream);
 typedef CUresult CUDAAPI tcuMemcpy3DPeerAsync(const CUDA_MEMCPY3D_PEER* pCopy, CUstream hStream);
-typedef CUresult CUDAAPI tcuMemsetD8_v2(CUdeviceptr dstDevice, unsigned uc, size_t N);
-typedef CUresult CUDAAPI tcuMemsetD16_v2(CUdeviceptr dstDevice, unsigned us, size_t N);
-typedef CUresult CUDAAPI tcuMemsetD32_v2(CUdeviceptr dstDevice, unsigned ui, size_t N);
-typedef CUresult CUDAAPI tcuMemsetD2D8_v2(CUdeviceptr dstDevice, size_t dstPitch, unsigned uc, size_t Width, size_t Height);
-typedef CUresult CUDAAPI tcuMemsetD2D16_v2(CUdeviceptr dstDevice, size_t dstPitch, unsigned us, size_t Width, size_t Height);
-typedef CUresult CUDAAPI tcuMemsetD2D32_v2(CUdeviceptr dstDevice, size_t dstPitch, unsigned ui, size_t Width, size_t Height);
-typedef CUresult CUDAAPI tcuMemsetD8Async(CUdeviceptr dstDevice, unsigned uc, size_t N, CUstream hStream);
-typedef CUresult CUDAAPI tcuMemsetD16Async(CUdeviceptr dstDevice, unsigned us, size_t N, CUstream hStream);
-typedef CUresult CUDAAPI tcuMemsetD32Async(CUdeviceptr dstDevice, unsigned ui, size_t N, CUstream hStream);
-typedef CUresult CUDAAPI tcuMemsetD2D8Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned uc, size_t Width, size_t Height, CUstream hStream);
-typedef CUresult CUDAAPI tcuMemsetD2D16Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned us, size_t Width, size_t Height, CUstream hStream);
-typedef CUresult CUDAAPI tcuMemsetD2D32Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned ui, size_t Width, size_t Height, CUstream hStream);
+typedef CUresult CUDAAPI tcuMemsetD8_v2(CUdeviceptr dstDevice, unsigned char uc, size_t N);
+typedef CUresult CUDAAPI tcuMemsetD16_v2(CUdeviceptr dstDevice, unsigned short us, size_t N);
+typedef CUresult CUDAAPI tcuMemsetD32_v2(CUdeviceptr dstDevice, unsigned int ui, size_t N);
+typedef CUresult CUDAAPI tcuMemsetD2D8_v2(CUdeviceptr dstDevice, size_t dstPitch, unsigned char uc, size_t Width, size_t Height);
+typedef CUresult CUDAAPI tcuMemsetD2D16_v2(CUdeviceptr dstDevice, size_t dstPitch, unsigned short us, size_t Width, size_t Height);
+typedef CUresult CUDAAPI tcuMemsetD2D32_v2(CUdeviceptr dstDevice, size_t dstPitch, unsigned int ui, size_t Width, size_t Height);
+typedef CUresult CUDAAPI tcuMemsetD8Async(CUdeviceptr dstDevice, unsigned char uc, size_t N, CUstream hStream);
+typedef CUresult CUDAAPI tcuMemsetD16Async(CUdeviceptr dstDevice, unsigned short us, size_t N, CUstream hStream);
+typedef CUresult CUDAAPI tcuMemsetD32Async(CUdeviceptr dstDevice, unsigned int ui, size_t N, CUstream hStream);
+typedef CUresult CUDAAPI tcuMemsetD2D8Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned char uc, size_t Width, size_t Height, CUstream hStream);
+typedef CUresult CUDAAPI tcuMemsetD2D16Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned short us, size_t Width, size_t Height, CUstream hStream);
+typedef CUresult CUDAAPI tcuMemsetD2D32Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned int ui, size_t Width, size_t Height, CUstream hStream);
 typedef CUresult CUDAAPI tcuArrayCreate_v2(CUarray* pHandle, const CUDA_ARRAY_DESCRIPTOR* pAllocateArray);
 typedef CUresult CUDAAPI tcuArrayGetDescriptor_v2(CUDA_ARRAY_DESCRIPTOR* pArrayDescriptor, CUarray hArray);
 typedef CUresult CUDAAPI tcuArrayDestroy(CUarray hArray);
 typedef CUresult CUDAAPI tcuArray3DCreate_v2(CUarray* pHandle, const CUDA_ARRAY3D_DESCRIPTOR* pAllocateArray);
 typedef CUresult CUDAAPI tcuArray3DGetDescriptor_v2(CUDA_ARRAY3D_DESCRIPTOR* pArrayDescriptor, CUarray hArray);
-typedef CUresult CUDAAPI tcuMipmappedArrayCreate(CUmipmappedArray* pHandle, const CUDA_ARRAY3D_DESCRIPTOR* pMipmappedArrayDesc, unsigned numMipmapLevels);
-typedef CUresult CUDAAPI tcuMipmappedArrayGetLevel(CUarray* pLevelArray, CUmipmappedArray hMipmappedArray, unsigned level);
+typedef CUresult CUDAAPI tcuMipmappedArrayCreate(CUmipmappedArray* pHandle, const CUDA_ARRAY3D_DESCRIPTOR* pMipmappedArrayDesc, unsigned int numMipmapLevels);
+typedef CUresult CUDAAPI tcuMipmappedArrayGetLevel(CUarray* pLevelArray, CUmipmappedArray hMipmappedArray, unsigned int level);
 typedef CUresult CUDAAPI tcuMipmappedArrayDestroy(CUmipmappedArray hMipmappedArray);
 typedef CUresult CUDAAPI tcuPointerGetAttribute(void* data, CUpointer_attribute attribute, CUdeviceptr ptr);
 typedef CUresult CUDAAPI tcuPointerSetAttribute(const void* value, CUpointer_attribute attribute, CUdeviceptr ptr);
-typedef CUresult CUDAAPI tcuPointerGetAttributes(unsigned numAttributes, CUpointer_attribute* attributes, void* data, CUdeviceptr ptr);
-typedef CUresult CUDAAPI tcuStreamCreate(CUstream* phStream, unsigned Flags);
-typedef CUresult CUDAAPI tcuStreamCreateWithPriority(CUstream* phStream, unsigned flags, int priority);
+typedef CUresult CUDAAPI tcuPointerGetAttributes(unsigned int numAttributes, CUpointer_attribute* attributes, void* data, CUdeviceptr ptr);
+typedef CUresult CUDAAPI tcuStreamCreate(CUstream* phStream, unsigned int Flags);
+typedef CUresult CUDAAPI tcuStreamCreateWithPriority(CUstream* phStream, unsigned int flags, int priority);
 typedef CUresult CUDAAPI tcuStreamGetPriority(CUstream hStream, int* priority);
-typedef CUresult CUDAAPI tcuStreamGetFlags(CUstream hStream, unsigned* flags);
-typedef CUresult CUDAAPI tcuStreamWaitEvent(CUstream hStream, CUevent hEvent, unsigned Flags);
-typedef CUresult CUDAAPI tcuStreamAddCallback(CUstream hStream, CUstreamCallback callback, void* userData, unsigned flags);
-typedef CUresult CUDAAPI tcuStreamAttachMemAsync(CUstream hStream, CUdeviceptr dptr, size_t length, unsigned flags);
+typedef CUresult CUDAAPI tcuStreamGetFlags(CUstream hStream, unsigned int* flags);
+typedef CUresult CUDAAPI tcuStreamWaitEvent(CUstream hStream, CUevent hEvent, unsigned int Flags);
+typedef CUresult CUDAAPI tcuStreamAddCallback(CUstream hStream, CUstreamCallback callback, void* userData, unsigned int flags);
+typedef CUresult CUDAAPI tcuStreamAttachMemAsync(CUstream hStream, CUdeviceptr dptr, size_t length, unsigned int flags);
 typedef CUresult CUDAAPI tcuStreamQuery(CUstream hStream);
 typedef CUresult CUDAAPI tcuStreamSynchronize(CUstream hStream);
 typedef CUresult CUDAAPI tcuStreamDestroy_v2(CUstream hStream);
-typedef CUresult CUDAAPI tcuEventCreate(CUevent* phEvent, unsigned Flags);
+typedef CUresult CUDAAPI tcuEventCreate(CUevent* phEvent, unsigned int Flags);
 typedef CUresult CUDAAPI tcuEventRecord(CUevent hEvent, CUstream hStream);
 typedef CUresult CUDAAPI tcuEventQuery(CUevent hEvent);
 typedef CUresult CUDAAPI tcuEventSynchronize(CUevent hEvent);
@@ -884,23 +884,23 @@ typedef CUresult CUDAAPI tcuEventElapsedTime(float* pMilliseconds, CUevent hStar
 typedef CUresult CUDAAPI tcuFuncGetAttribute(int* pi, CUfunction_attribute attrib, CUfunction hfunc);
 typedef CUresult CUDAAPI tcuFuncSetCacheConfig(CUfunction hfunc, CUfunc_cache config);
 typedef CUresult CUDAAPI tcuFuncSetSharedMemConfig(CUfunction hfunc, CUsharedconfig config);
-typedef CUresult CUDAAPI tcuLaunchKernel(CUfunction f, unsigned gridDimX, unsigned gridDimY, unsigned gridDimZ, unsigned blockDimX, unsigned blockDimY, unsigned blockDimZ, unsigned sharedMemBytes, CUstream hStream, void* kernelParams, void* extra);
+typedef CUresult CUDAAPI tcuLaunchKernel(CUfunction f, unsigned int gridDimX, unsigned int gridDimY, unsigned int gridDimZ, unsigned int blockDimX, unsigned int blockDimY, unsigned int blockDimZ, unsigned int sharedMemBytes, CUstream hStream, void* kernelParams, void* extra);
 typedef CUresult CUDAAPI tcuFuncSetBlockShape(CUfunction hfunc, int x, int y, int z);
-typedef CUresult CUDAAPI tcuFuncSetSharedSize(CUfunction hfunc, unsigned bytes);
-typedef CUresult CUDAAPI tcuParamSetSize(CUfunction hfunc, unsigned numbytes);
-typedef CUresult CUDAAPI tcuParamSeti(CUfunction hfunc, int offset, unsigned value);
+typedef CUresult CUDAAPI tcuFuncSetSharedSize(CUfunction hfunc, unsigned int bytes);
+typedef CUresult CUDAAPI tcuParamSetSize(CUfunction hfunc, unsigned int numbytes);
+typedef CUresult CUDAAPI tcuParamSeti(CUfunction hfunc, int offset, unsigned int value);
 typedef CUresult CUDAAPI tcuParamSetf(CUfunction hfunc, int offset, float value);
-typedef CUresult CUDAAPI tcuParamSetv(CUfunction hfunc, int offset, void* ptr, unsigned numbytes);
+typedef CUresult CUDAAPI tcuParamSetv(CUfunction hfunc, int offset, void* ptr, unsigned int numbytes);
 typedef CUresult CUDAAPI tcuLaunch(CUfunction f);
 typedef CUresult CUDAAPI tcuLaunchGrid(CUfunction f, int grid_width, int grid_height);
 typedef CUresult CUDAAPI tcuLaunchGridAsync(CUfunction f, int grid_width, int grid_height, CUstream hStream);
 typedef CUresult CUDAAPI tcuParamSetTexRef(CUfunction hfunc, int texunit, CUtexref hTexRef);
 typedef CUresult CUDAAPI tcuOccupancyMaxActiveBlocksPerMultiprocessor(int* numBlocks, CUfunction func, int blockSize, size_t dynamicSMemSize);
-typedef CUresult CUDAAPI tcuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int* numBlocks, CUfunction func, int blockSize, size_t dynamicSMemSize, unsigned flags);
+typedef CUresult CUDAAPI tcuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int* numBlocks, CUfunction func, int blockSize, size_t dynamicSMemSize, unsigned int flags);
 typedef CUresult CUDAAPI tcuOccupancyMaxPotentialBlockSize(int* minGridSize, int* blockSize, CUfunction func, CUoccupancyB2DSize blockSizeToDynamicSMemSize, size_t dynamicSMemSize, int blockSizeLimit);
-typedef CUresult CUDAAPI tcuOccupancyMaxPotentialBlockSizeWithFlags(int* minGridSize, int* blockSize, CUfunction func, CUoccupancyB2DSize blockSizeToDynamicSMemSize, size_t dynamicSMemSize, int blockSizeLimit, unsigned flags);
-typedef CUresult CUDAAPI tcuTexRefSetArray(CUtexref hTexRef, CUarray hArray, unsigned Flags);
-typedef CUresult CUDAAPI tcuTexRefSetMipmappedArray(CUtexref hTexRef, CUmipmappedArray hMipmappedArray, unsigned Flags);
+typedef CUresult CUDAAPI tcuOccupancyMaxPotentialBlockSizeWithFlags(int* minGridSize, int* blockSize, CUfunction func, CUoccupancyB2DSize blockSizeToDynamicSMemSize, size_t dynamicSMemSize, int blockSizeLimit, unsigned int flags);
+typedef CUresult CUDAAPI tcuTexRefSetArray(CUtexref hTexRef, CUarray hArray, unsigned int Flags);
+typedef CUresult CUDAAPI tcuTexRefSetMipmappedArray(CUtexref hTexRef, CUmipmappedArray hMipmappedArray, unsigned int Flags);
 typedef CUresult CUDAAPI tcuTexRefSetAddress_v2(size_t* ByteOffset, CUtexref hTexRef, CUdeviceptr dptr, size_t bytes);
 typedef CUresult CUDAAPI tcuTexRefSetAddress2D_v3(CUtexref hTexRef, const CUDA_ARRAY_DESCRIPTOR* desc, CUdeviceptr dptr, size_t Pitch);
 typedef CUresult CUDAAPI tcuTexRefSetFormat(CUtexref hTexRef, CUarray_format fmt, int NumPackedComponents);
@@ -909,8 +909,8 @@ typedef CUresult CUDAAPI tcuTexRefSetFilterMode(CUtexref hTexRef, CUfilter_mode
 typedef CUresult CUDAAPI tcuTexRefSetMipmapFilterMode(CUtexref hTexRef, CUfilter_mode fm);
 typedef CUresult CUDAAPI tcuTexRefSetMipmapLevelBias(CUtexref hTexRef, float bias);
 typedef CUresult CUDAAPI tcuTexRefSetMipmapLevelClamp(CUtexref hTexRef, float minMipmapLevelClamp, float maxMipmapLevelClamp);
-typedef CUresult CUDAAPI tcuTexRefSetMaxAnisotropy(CUtexref hTexRef, unsigned maxAniso);
-typedef CUresult CUDAAPI tcuTexRefSetFlags(CUtexref hTexRef, unsigned Flags);
+typedef CUresult CUDAAPI tcuTexRefSetMaxAnisotropy(CUtexref hTexRef, unsigned int maxAniso);
+typedef CUresult CUDAAPI tcuTexRefSetFlags(CUtexref hTexRef, unsigned int Flags);
 typedef CUresult CUDAAPI tcuTexRefGetAddress_v2(CUdeviceptr* pdptr, CUtexref hTexRef);
 typedef CUresult CUDAAPI tcuTexRefGetArray(CUarray* phArray, CUtexref hTexRef);
 typedef CUresult CUDAAPI tcuTexRefGetMipmappedArray(CUmipmappedArray* phMipmappedArray, CUtexref hTexRef);
@@ -921,10 +921,10 @@ typedef CUresult CUDAAPI tcuTexRefGetMipmapFilterMode(CUfilter_mode* pfm, CUtexr
 typedef CUresult CUDAAPI tcuTexRefGetMipmapLevelBias(float* pbias, CUtexref hTexRef);
 typedef CUresult CUDAAPI tcuTexRefGetMipmapLevelClamp(float* pminMipmapLevelClamp, float* pmaxMipmapLevelClamp, CUtexref hTexRef);
 typedef CUresult CUDAAPI tcuTexRefGetMaxAnisotropy(int* pmaxAniso, CUtexref hTexRef);
-typedef CUresult CUDAAPI tcuTexRefGetFlags(unsigned* pFlags, CUtexref hTexRef);
+typedef CUresult CUDAAPI tcuTexRefGetFlags(unsigned int* pFlags, CUtexref hTexRef);
 typedef CUresult CUDAAPI tcuTexRefCreate(CUtexref* pTexRef);
 typedef CUresult CUDAAPI tcuTexRefDestroy(CUtexref hTexRef);
-typedef CUresult CUDAAPI tcuSurfRefSetArray(CUsurfref hSurfRef, CUarray hArray, unsigned Flags);
+typedef CUresult CUDAAPI tcuSurfRefSetArray(CUsurfref hSurfRef, CUarray hArray, unsigned int Flags);
 typedef CUresult CUDAAPI tcuSurfRefGetArray(CUarray* phArray, CUsurfref hSurfRef);
 typedef CUresult CUDAAPI tcuTexObjectCreate(CUtexObject* pTexObject, const CUDA_RESOURCE_DESC* pResDesc, const CUDA_TEXTURE_DESC* pTexDesc, const CUDA_RESOURCE_VIEW_DESC* pResViewDesc);
 typedef CUresult CUDAAPI tcuTexObjectDestroy(CUtexObject texObject);
@@ -935,27 +935,27 @@ typedef CUresult CUDAAPI tcuSurfObjectCreate(CUsurfObject* pSurfObject, const CU
 typedef CUresult CUDAAPI tcuSurfObjectDestroy(CUsurfObject surfObject);
 typedef CUresult CUDAAPI tcuSurfObjectGetResourceDesc(CUDA_RESOURCE_DESC* pResDesc, CUsurfObject surfObject);
 typedef CUresult CUDAAPI tcuDeviceCanAccessPeer(int* canAccessPeer, CUdevice dev, CUdevice peerDev);
-typedef CUresult CUDAAPI tcuCtxEnablePeerAccess(CUcontext peerContext, unsigned Flags);
+typedef CUresult CUDAAPI tcuCtxEnablePeerAccess(CUcontext peerContext, unsigned int Flags);
 typedef CUresult CUDAAPI tcuCtxDisablePeerAccess(CUcontext peerContext);
 typedef CUresult CUDAAPI tcuGraphicsUnregisterResource(CUgraphicsResource resource);
-typedef CUresult CUDAAPI tcuGraphicsSubResourceGetMappedArray(CUarray* pArray, CUgraphicsResource resource, unsigned arrayIndex, unsigned mipLevel);
+typedef CUresult CUDAAPI tcuGraphicsSubResourceGetMappedArray(CUarray* pArray, CUgraphicsResource resource, unsigned int arrayIndex, unsigned int mipLevel);
 typedef CUresult CUDAAPI tcuGraphicsResourceGetMappedMipmappedArray(CUmipmappedArray* pMipmappedArray, CUgraphicsResource resource);
 typedef CUresult CUDAAPI tcuGraphicsResourceGetMappedPointer_v2(CUdeviceptr* pDevPtr, size_t* pSize, CUgraphicsResource resource);
-typedef CUresult CUDAAPI tcuGraphicsResourceSetMapFlags_v2(CUgraphicsResource resource, unsigned flags);
-typedef CUresult CUDAAPI tcuGraphicsMapResources(unsigned count, CUgraphicsResource* resources, CUstream hStream);
-typedef CUresult CUDAAPI tcuGraphicsUnmapResources(unsigned count, CUgraphicsResource* resources, CUstream hStream);
+typedef CUresult CUDAAPI tcuGraphicsResourceSetMapFlags_v2(CUgraphicsResource resource, unsigned int flags);
+typedef CUresult CUDAAPI tcuGraphicsMapResources(unsigned int count, CUgraphicsResource* resources, CUstream hStream);
+typedef CUresult CUDAAPI tcuGraphicsUnmapResources(unsigned int count, CUgraphicsResource* resources, CUstream hStream);
 typedef CUresult CUDAAPI tcuGetExportTable(const void* ppExportTable, const CUuuid* pExportTableId);
 
-typedef CUresult CUDAAPI tcuGraphicsGLRegisterBuffer(CUgraphicsResource* pCudaResource, GLuint buffer, unsigned Flags);
-typedef CUresult CUDAAPI tcuGraphicsGLRegisterImage(CUgraphicsResource* pCudaResource, GLuint image, GLenum target, unsigned Flags);
-typedef CUresult CUDAAPI tcuGLGetDevices_v2(unsigned* pCudaDeviceCount, CUdevice* pCudaDevices, unsigned cudaDeviceCount, CUGLDeviceList deviceList);
-typedef CUresult CUDAAPI tcuGLCtxCreate_v2(CUcontext* pCtx, unsigned Flags, CUdevice device);
+typedef CUresult CUDAAPI tcuGraphicsGLRegisterBuffer(CUgraphicsResource* pCudaResource, GLuint buffer, unsigned int Flags);
+typedef CUresult CUDAAPI tcuGraphicsGLRegisterImage(CUgraphicsResource* pCudaResource, GLuint image, GLenum target, unsigned int Flags);
+typedef CUresult CUDAAPI tcuGLGetDevices_v2(unsigned int* pCudaDeviceCount, CUdevice* pCudaDevices, unsigned int cudaDeviceCount, CUGLDeviceList deviceList);
+typedef CUresult CUDAAPI tcuGLCtxCreate_v2(CUcontext* pCtx, unsigned int Flags, CUdevice device);
 typedef CUresult CUDAAPI tcuGLInit(void);
 typedef CUresult CUDAAPI tcuGLRegisterBufferObject(GLuint buffer);
 typedef CUresult CUDAAPI tcuGLMapBufferObject_v2(CUdeviceptr* dptr, size_t* size, GLuint buffer);
 typedef CUresult CUDAAPI tcuGLUnmapBufferObject(GLuint buffer);
 typedef CUresult CUDAAPI tcuGLUnregisterBufferObject(GLuint buffer);
-typedef CUresult CUDAAPI tcuGLSetBufferObjectMapFlags(GLuint buffer, unsigned Flags);
+typedef CUresult CUDAAPI tcuGLSetBufferObjectMapFlags(GLuint buffer, unsigned int Flags);
 typedef CUresult CUDAAPI tcuGLMapBufferObjectAsync_v2(CUdeviceptr* dptr, size_t* size, GLuint buffer, CUstream hStream);
 typedef CUresult CUDAAPI tcuGLUnmapBufferObjectAsync(GLuint buffer, CUstream hStream);
 

extern/curve_fit_nd/curve_fit_nd.h

@@ -25,8 +25,8 @@
  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-#ifndef __SPLINE_FIT__
-#define __SPLINE_FIT__
+#ifndef __CURVE_FIT_ND_H__
+#define __CURVE_FIT_ND_H__
 
 /** \file curve_fit_nd.h
  *  \ingroup curve_fit
@@ -60,6 +60,7 @@ int curve_fit_cubic_to_points_db(
         const unsigned int  points_len,
         const unsigned int  dims,
         const double        error_threshold,
+        const unsigned int  calc_flag,
         const unsigned int *corners,
         unsigned int        corners_len,
 
@@ -72,6 +73,7 @@ int curve_fit_cubic_to_points_fl(
         const unsigned int  points_len,
         const unsigned int  dims,
         const float         error_threshold,
+        const unsigned int  calc_flag,
         const unsigned int *corners,
         const unsigned int  corners_len,
 
@@ -79,6 +81,47 @@ int curve_fit_cubic_to_points_fl(
         unsigned int **r_cubic_orig_index,
         unsigned int **r_corners_index_array, unsigned int *r_corners_index_len);
 
+/**
+ * Takes a flat array of points and evalues that to calculate handle lengths.
+ *
+ * \param points, points_len: The array of points to calculate a cubics from.
+ * \param dims: The number of dimensions for for each element in \a points.
+ * \param error_threshold: the error threshold to allow for,
+ * \param tan_l, tan_r: Normalized tangents the handles will be aligned to.
+ * Note that tangents must both point along the direction of the \a points,
+ * so \a tan_l points in the same direction of the resulting handle,
+ * where \a tan_r will point the opposite direction of its handle.
+ *
+ * \param r_handle_l, r_handle_r: Resulting calculated handles.
+ * \param r_error_sq: The maximum distance  (squared) this curve diverges from \a points.
+ */
+int curve_fit_cubic_to_points_single_db(
+        const double      *points,
+        const unsigned int points_len,
+        const unsigned int dims,
+        const double       error_threshold,
+        const double       tan_l[],
+        const double       tan_r[],
+
+        double  r_handle_l[],
+        double  r_handle_r[],
+        double *r_error_sq);
+
+int curve_fit_cubic_to_points_single_fl(
+        const float       *points,
+        const unsigned int points_len,
+        const unsigned int dims,
+        const float        error_threshold,
+        const float        tan_l[],
+        const float        tan_r[],
+
+        float   r_handle_l[],
+        float   r_handle_r[],
+        float  *r_error_sq);
+
+enum {
+	CURVE_FIT_CALC_HIGH_QUALIY          = (1 << 0),
+};
 
 /* curve_fit_corners_detect.c */
 
@@ -122,4 +165,4 @@ int curve_fit_corners_detect_fl(
         unsigned int **r_corners,
         unsigned int  *r_corners_len);
 
-#endif  /* __SPLINE_FIT__ */
+#endif  /* __CURVE_FIT_ND_H__ */

extern/curve_fit_nd/intern/curve_fit_corners_detect.c

@@ -382,9 +382,9 @@ int curve_fit_corners_detect_db(
 				uint i_best = i_span_start;
 				while (i_next < points_len) {
 					if ((points_angle[i_next] == 0.0) ||
-					   (len_squared_vnvn(
-					        &points[(i_next - 1) * dims],
-					        &points[i_next * dims], dims) > radius_min_sq))
+					    (len_squared_vnvn(
+					         &points[(i_next - 1) * dims],
+					         &points[i_next * dims], dims) > radius_min_sq))
 					{
 						break;
 					}

extern/curve_fit_nd/intern/curve_fit_cubic.c

@@ -29,6 +29,10 @@
  *  \ingroup curve_fit
  */
 
+#ifdef _MSC_VER
+#  define _USE_MATH_DEFINES
+#endif
+
 #include <math.h>
 #include <float.h>
 #include <stdbool.h>
@@ -39,11 +43,20 @@
 
 #include "../curve_fit_nd.h"
 
+/* Take curvature into account when calculating the least square solution isn't usable. */
+#define USE_CIRCULAR_FALLBACK
+
+/* Use the maximum distance of any points from the direct line between 2 points
+ * to calculate how long the handles need to be.
+ * Can do a 'perfect' reversal of subdivision when for curve has symmetrical handles and doesn't change direction
+ * (as with an 'S' shape). */
+#define USE_OFFSET_FALLBACK
+
 /* avoid re-calculating lengths multiple times */
 #define USE_LENGTH_CACHE
 
 /* store the indices in the cubic data so we can return the original indices,
- * useful when the caller has data assosiated with the curve. */
+ * useful when the caller has data associated with the curve. */
 #define USE_ORIG_INDEX_DATA
 
 typedef unsigned int uint;
@@ -109,9 +122,19 @@ typedef struct Cubic {
 	*_p3 = _p2 + (dims); ((void)0)
 
 
+static size_t cubic_alloc_size(const uint dims)
+{
+	return sizeof(Cubic) + (sizeof(double) * 4 * dims);
+}
+
 static Cubic *cubic_alloc(const uint dims)
 {
-	return malloc(sizeof(Cubic) + (sizeof(double) * 4 * dims));
+	return malloc(cubic_alloc_size(dims));
+}
+
+static void cubic_copy(Cubic *cubic_dst, const Cubic *cubic_src, const uint dims)
+{
+	memcpy(cubic_dst, cubic_src, cubic_alloc_size(dims));
 }
 
 static void cubic_init(
@@ -278,7 +301,7 @@ static void cubic_calc_acceleration(
         double r_v[])
 {
 	CUBIC_VARS_CONST(cubic, dims, p0, p1, p2, p3);
-    const double s = 1.0 - t;
+	const double s = 1.0 - t;
 	for (uint j = 0; j < dims; j++) {
 		r_v[j] = 6.0 * ((p2[j] - 2.0 * p1[j] + p0[j]) * s +
 		                (p3[j] - 2.0 * p2[j] + p1[j]) * t);
@@ -286,20 +309,19 @@ static void cubic_calc_acceleration(
 }
 
 /**
- * Returns a 'measure' of the maximal discrepancy of the points specified
+ * Returns a 'measure' of the maximum distance (squared) of the points specified
  * by points_offset from the corresponding cubic(u[]) points.
  */
-static void cubic_calc_error(
+static double cubic_calc_error(
         const Cubic *cubic,
         const double *points_offset,
         const uint points_offset_len,
         const double *u,
         const uint dims,
 
-        double *r_error_sq_max,
         uint *r_error_index)
 {
-	double error_sq_max = 0.0;
+	double error_max_sq = 0.0;
 	uint   error_index = 0;
 
 	const double *pt_real = points_offset + dims;
@@ -313,16 +335,54 @@ static void cubic_calc_error(
 		cubic_evaluate(cubic, u[i], dims, pt_eval);
 
 		const double err_sq = len_squared_vnvn(pt_real, pt_eval, dims);
-		if (err_sq >= error_sq_max) {
-			error_sq_max = err_sq;
+		if (err_sq >= error_max_sq) {
+			error_max_sq = err_sq;
 			error_index = i;
 		}
 	}
 
-	*r_error_sq_max   = error_sq_max;
 	*r_error_index = error_index;
+	return error_max_sq;
 }
 
+#ifdef USE_OFFSET_FALLBACK
+/**
+ * A version #cubic_calc_error where we don't need the split-index and can exit early when over the limit.
+ */
+static double cubic_calc_error_simple(
+        const Cubic *cubic,
+        const double *points_offset,
+        const uint points_offset_len,
+        const double *u,
+        const double error_threshold_sq,
+        const uint dims)
+
+{
+	double error_max_sq = 0.0;
+
+	const double *pt_real = points_offset + dims;
+#ifdef USE_VLA
+	double        pt_eval[dims];
+#else
+	double       *pt_eval = alloca(sizeof(double) * dims);
+#endif
+
+	for (uint i = 1; i < points_offset_len - 1; i++, pt_real += dims) {
+		cubic_evaluate(cubic, u[i], dims, pt_eval);
+
+		const double err_sq = len_squared_vnvn(pt_real, pt_eval, dims);
+		if (err_sq >= error_threshold_sq) {
+			return error_threshold_sq;
+		}
+		else if (err_sq >= error_max_sq) {
+			error_max_sq = err_sq;
+		}
+	}
+
+	return error_max_sq;
+}
+#endif
+
 /**
  * Bezier multipliers
  */
@@ -388,12 +448,220 @@ static void points_calc_center_weighted(
 	}
 }
 
+#ifdef USE_CIRCULAR_FALLBACK
+
+/**
+ * Return a scale value, used to calculate how much the curve handles should be increased,
+ *
+ * This works by placing each end-point on an imaginary circle,
+ * the placement on the circle is based on the tangent vectors,
+ * where larger differences in tangent angle cover a larger part of the circle.
+ *
+ * Return the scale representing how much larger the distance around the circle is.
+ */
+static double points_calc_circumference_factor(
+        const double  tan_l[],
+        const double  tan_r[],
+        const uint dims)
+{
+	const double dot = dot_vnvn(tan_l, tan_r, dims);
+	const double len_tangent = dot < 0.0 ? len_vnvn(tan_l, tan_r, dims) : len_negated_vnvn(tan_l, tan_r, dims);
+	if (len_tangent > DBL_EPSILON) {
+		/* only clamp to avoid precision error */
+		double angle = acos(max(-fabs(dot), -1.0));
+		/* Angle may be less than the length when the tangents define >180 degrees of the circle,
+		 * (tangents that point away from each other).
+		 * We could try support this but will likely cause extreme >1 scales which could cause other issues. */
+		// assert(angle >= len_tangent);
+		double factor = (angle / len_tangent);
+		assert(factor < (M_PI / 2) + DBL_EPSILON);
+		return factor;
+	}
+	else {
+		/* tangents are exactly aligned (think two opposite sides of a circle). */
+		return (M_PI / 2);
+	}
+}
+
+/**
+ * Return the value which the distance between points will need to be scaled by,
+ * to define a handle, given both points are on a perfect circle.
+ *
+ * \note the return value will need to be multiplied by 1.3... for correct results.
+ */
+static double points_calc_circle_tangent_factor(
+        const double  tan_l[],
+        const double  tan_r[],
+        const uint dims)
+{
+	const double eps = 1e-8;
+	const double tan_dot = dot_vnvn(tan_l, tan_r, dims);
+	if (tan_dot > 1.0 - eps) {
+		/* no angle difference (use fallback, length wont make any difference) */
+		return (1.0 / 3.0) * 0.75;
+	}
+	else if (tan_dot < -1.0 + eps) {
+		/* parallele tangents (half-circle) */
+		return (1.0 / 2.0);
+	}
+	else {
+		/* non-aligned tangents, calculate handle length */
+		const double angle = acos(tan_dot) / 2.0;
+
+		/* could also use 'angle_sin = len_vnvn(tan_l, tan_r, dims) / 2.0' */
+		const double angle_sin = sin(angle);
+		const double angle_cos = cos(angle);
+		return ((1.0 - angle_cos) / (angle_sin * 2.0)) / angle_sin;
+	}
+}
+
+/**
+ * Calculate the scale the handles, which serves as a best-guess
+ * used as a fallback when the least-square solution fails.
+ */
+static double points_calc_cubic_scale(
+        const double v_l[], const double v_r[],
+        const double  tan_l[],
+        const double  tan_r[],
+        const double coords_length, uint dims)
+{
+	const double len_direct = len_vnvn(v_l, v_r, dims);
+	const double len_circle_factor = points_calc_circle_tangent_factor(tan_l, tan_r, dims);
+
+	/* if this curve is a circle, this value doesn't need modification */
+	const double len_circle_handle = (len_direct * (len_circle_factor / 0.75));
+
+	/* scale by the difference from the circumference distance */
+	const double len_circle = len_direct * points_calc_circumference_factor(tan_l, tan_r, dims);
+	double scale_handle = (coords_length / len_circle);
+
+	/* Could investigate an accurate calculation here,
+	 * though this gives close results */
+	scale_handle = ((scale_handle - 1.0) * 1.75) + 1.0;
+
+	return len_circle_handle * scale_handle;
+}
+
+static void cubic_from_points_fallback(
+        const double *points_offset,
+        const uint    points_offset_len,
+        const double  tan_l[],
+        const double  tan_r[],
+        const uint dims,
+
+        Cubic *r_cubic)
+{
+	const double *p0 = &points_offset[0];
+	const double *p3 = &points_offset[(points_offset_len - 1) * dims];
+
+	double alpha = len_vnvn(p0, p3, dims) / 3.0;
+
+	double *p1 = CUBIC_PT(r_cubic, 1, dims);
+	double *p2 = CUBIC_PT(r_cubic, 2, dims);
+
+	copy_vnvn(CUBIC_PT(r_cubic, 0, dims), p0, dims);
+	copy_vnvn(CUBIC_PT(r_cubic, 3, dims), p3, dims);
+
+#ifdef USE_ORIG_INDEX_DATA
+	r_cubic->orig_span = (points_offset_len - 1);
+#endif
+
+	/* p1 = p0 - (tan_l * alpha_l);
+	 * p2 = p3 + (tan_r * alpha_r);
+	 */
+	msub_vn_vnvn_fl(p1, p0, tan_l, alpha, dims);
+	madd_vn_vnvn_fl(p2, p3, tan_r, alpha, dims);
+}
+#endif  /* USE_CIRCULAR_FALLBACK */
+
+
+#ifdef USE_OFFSET_FALLBACK
+
+static void cubic_from_points_offset_fallback(
+        const double *points_offset,
+        const uint    points_offset_len,
+        const double  tan_l[],
+        const double  tan_r[],
+        const uint dims,
+
+        Cubic *r_cubic)
+{
+	const double *p0 = &points_offset[0];
+	const double *p3 = &points_offset[(points_offset_len - 1) * dims];
+
+#ifdef USE_VLA
+	double dir_unit[dims];
+	double a[2][dims];
+	double tmp[dims];
+#else
+	double *dir_unit = alloca(sizeof(double) * dims);
+	double *a[2] = {
+	    alloca(sizeof(double) * dims),
+	    alloca(sizeof(double) * dims),
+	};
+	double *tmp = alloca(sizeof(double) * dims);
+#endif
+
+	const double dir_dist = normalize_vn_vnvn(dir_unit, p3, p0, dims);
+	project_plane_vn_vnvn_normalized(a[0], tan_l, dir_unit, dims);
+	project_plane_vn_vnvn_normalized(a[1], tan_r, dir_unit, dims);
+
+	/* only for better accuracy, not essential */
+	normalize_vn(a[0], dims);
+	normalize_vn(a[1], dims);
+
+	mul_vnvn_fl(a[1], a[1], -1, dims);
+
+	double dists[2] = {0, 0};
+
+	const double *pt = points_offset;
+	for (uint i = 1; i < points_offset_len - 1; i++, pt += dims) {
+		for (uint k = 0; k < 2; k++) {
+			sub_vn_vnvn(tmp, p0, pt, dims);
+			project_vn_vnvn_normalized(tmp, tmp, a[k], dims);
+			dists[k] = max(dists[k], dot_vnvn(tmp, a[k], dims));
+		}
+	}
+
+	float alpha_l = (dists[0] / 0.75) /  dot_vnvn(tan_l, a[0], dims);
+	float alpha_r = (dists[1] / 0.75) / -dot_vnvn(tan_r, a[1], dims);
+
+	if (!(alpha_l > 0.0f)) {
+		alpha_l = dir_dist / 3.0;
+	}
+	if (!(alpha_r > 0.0f)) {
+		alpha_r = dir_dist / 3.0;
+	}
+
+	double *p1 = CUBIC_PT(r_cubic, 1, dims);
+	double *p2 = CUBIC_PT(r_cubic, 2, dims);
+
+	copy_vnvn(CUBIC_PT(r_cubic, 0, dims), p0, dims);
+	copy_vnvn(CUBIC_PT(r_cubic, 3, dims), p3, dims);
+
+#ifdef USE_ORIG_INDEX_DATA
+	r_cubic->orig_span = (points_offset_len - 1);
+#endif
+
+	/* p1 = p0 - (tan_l * alpha_l);
+	 * p2 = p3 + (tan_r * alpha_r);
+	 */
+	msub_vn_vnvn_fl(p1, p0, tan_l, alpha_l, dims);
+	madd_vn_vnvn_fl(p2, p3, tan_r, alpha_r, dims);
+}
+
+#endif  /* USE_OFFSET_FALLBACK */
+
+
 /**
  * Use least-squares method to find Bezier control points for region.
  */
 static void cubic_from_points(
         const double *points_offset,
         const uint    points_offset_len,
+#ifdef USE_CIRCULAR_FALLBACK
+        const double  points_offset_coords_length,
+#endif
         const double *u_prime,
         const double  tan_l[],
         const double  tan_r[],
@@ -467,11 +735,20 @@ static void cubic_from_points(
 	 * so only problems absurd of approximation and not for bugs in the code.
 	 */
 
+	bool use_clamp = true;
+
 	/* flip check to catch nan values */
 	if (!(alpha_l >= 0.0) ||
 	    !(alpha_r >= 0.0))
 	{
+#ifdef USE_CIRCULAR_FALLBACK
+		alpha_l = alpha_r = points_calc_cubic_scale(p0, p3, tan_l, tan_r, points_offset_coords_length, dims);
+#else
 		alpha_l = alpha_r = len_vnvn(p0, p3, dims) / 3.0;
+#endif
+
+		/* skip clamping when we're using default handles */
+		use_clamp = false;
 	}
 
 	double *p1 = CUBIC_PT(r_cubic, 1, dims);
@@ -493,64 +770,69 @@ static void cubic_from_points(
 	/* ------------------------------------
 	 * Clamping (we could make it optional)
 	 */
+	if (use_clamp) {
 #ifdef USE_VLA
-	double center[dims];
+		double center[dims];
 #else
-	double *center = alloca(sizeof(double) * dims);
+		double *center = alloca(sizeof(double) * dims);
 #endif
-	points_calc_center_weighted(points_offset, points_offset_len, dims, center);
+		points_calc_center_weighted(points_offset, points_offset_len, dims, center);
 
-	const double clamp_scale = 3.0;  /* clamp to 3x */
-	double dist_sq_max = 0.0;
+		const double clamp_scale = 3.0;  /* clamp to 3x */
+		double dist_sq_max = 0.0;
 
-	{
-		const double *pt = points_offset;
-		for (uint i = 0; i < points_offset_len; i++, pt += dims) {
+		{
+			const double *pt = points_offset;
+			for (uint i = 0; i < points_offset_len; i++, pt += dims) {
 #if 0
-			double dist_sq_test = sq(len_vnvn(center, pt, dims) * clamp_scale);
+				double dist_sq_test = sq(len_vnvn(center, pt, dims) * clamp_scale);
 #else
-			/* do inline */
-			double dist_sq_test = 0.0;
-			for (uint j = 0; j < dims; j++) {
-				dist_sq_test += sq((pt[j] - center[j]) * clamp_scale);
-			}
+				/* do inline */
+				double dist_sq_test = 0.0;
+				for (uint j = 0; j < dims; j++) {
+					dist_sq_test += sq((pt[j] - center[j]) * clamp_scale);
+				}
 #endif
-			dist_sq_max = max(dist_sq_max, dist_sq_test);
-		}
-	}
-
-	double p1_dist_sq = len_squared_vnvn(center, p1, dims);
-	double p2_dist_sq = len_squared_vnvn(center, p2, dims);
-
-	if (p1_dist_sq > dist_sq_max ||
-	    p2_dist_sq > dist_sq_max)
-	{
-
-		alpha_l = alpha_r = len_vnvn(p0, p3, dims) / 3.0;
-
-		/*
-		 * p1 = p0 - (tan_l * alpha_l);
-		 * p2 = p3 + (tan_r * alpha_r);
-		 */
-		for (uint j = 0; j < dims; j++) {
-			p1[j] = p0[j] - (tan_l[j] * alpha_l);
-			p2[j] = p3[j] + (tan_r[j] * alpha_r);
+				dist_sq_max = max(dist_sq_max, dist_sq_test);
+			}
 		}
 
-		p1_dist_sq = len_squared_vnvn(center, p1, dims);
-		p2_dist_sq = len_squared_vnvn(center, p2, dims);
-	}
+		double p1_dist_sq = len_squared_vnvn(center, p1, dims);
+		double p2_dist_sq = len_squared_vnvn(center, p2, dims);
 
-	/* clamp within the 3x radius */
-	if (p1_dist_sq > dist_sq_max) {
-		isub_vnvn(p1, center, dims);
-		imul_vn_fl(p1, sqrt(dist_sq_max) / sqrt(p1_dist_sq), dims);
-		iadd_vnvn(p1, center, dims);
-	}
-	if (p2_dist_sq > dist_sq_max) {
-		isub_vnvn(p2, center, dims);
-		imul_vn_fl(p2, sqrt(dist_sq_max) / sqrt(p2_dist_sq), dims);
-		iadd_vnvn(p2, center, dims);
+		if (p1_dist_sq > dist_sq_max ||
+		    p2_dist_sq > dist_sq_max)
+		{
+#ifdef USE_CIRCULAR_FALLBACK
+			alpha_l = alpha_r = points_calc_cubic_scale(p0, p3, tan_l, tan_r, points_offset_coords_length, dims);
+#else
+			alpha_l = alpha_r = len_vnvn(p0, p3, dims) / 3.0;
+#endif
+
+			/*
+			 * p1 = p0 - (tan_l * alpha_l);
+			 * p2 = p3 + (tan_r * alpha_r);
+			 */
+			for (uint j = 0; j < dims; j++) {
+				p1[j] = p0[j] - (tan_l[j] * alpha_l);
+				p2[j] = p3[j] + (tan_r[j] * alpha_r);
+			}
+
+			p1_dist_sq = len_squared_vnvn(center, p1, dims);
+			p2_dist_sq = len_squared_vnvn(center, p2, dims);
+		}
+
+		/* clamp within the 3x radius */
+		if (p1_dist_sq > dist_sq_max) {
+			isub_vnvn(p1, center, dims);
+			imul_vn_fl(p1, sqrt(dist_sq_max) / sqrt(p1_dist_sq), dims);
+			iadd_vnvn(p1, center, dims);
+		}
+		if (p2_dist_sq > dist_sq_max) {
+			isub_vnvn(p2, center, dims);
+			imul_vn_fl(p2, sqrt(dist_sq_max) / sqrt(p2_dist_sq), dims);
+			iadd_vnvn(p2, center, dims);
+		}
 	}
 	/* end clamping */
 }
@@ -574,8 +856,10 @@ static void points_calc_coord_length_cache(
 }
 #endif  /* USE_LENGTH_CACHE */
 
-
-static void points_calc_coord_length(
+/**
+ * \return the accumulated length of \a points_offset.
+ */
+static double points_calc_coord_length(
         const double *points_offset,
         const uint    points_offset_len,
         const uint    dims,
@@ -608,6 +892,7 @@ static void points_calc_coord_length(
 	for (uint i = 0; i < points_offset_len; i++) {
 		r_u[i] /= w;
 	}
+	return w;
 }
 
 /**
@@ -620,10 +905,10 @@ static void points_calc_coord_length(
  * \note Return value may be `nan` caller must check for this.
  */
 static double cubic_find_root(
-		const Cubic *cubic,
-		const double p[],
-		const double u,
-		const uint dims)
+        const Cubic *cubic,
+        const double p[],
+        const double u,
+        const uint dims)
 {
 	/* Newton-Raphson Method. */
 	/* all vectors */
@@ -695,7 +980,7 @@ static bool cubic_reparameterize(
 }
 
 
-static void fit_cubic_to_points(
+static bool fit_cubic_to_points(
         const double *points_offset,
         const uint    points_offset_len,
 #ifdef USE_LENGTH_CACHE
@@ -703,19 +988,15 @@ static void fit_cubic_to_points(
 #endif
         const double  tan_l[],
         const double  tan_r[],
-        const double  error_threshold,
+        const double  error_threshold_sq,
         const uint    dims,
-        /* fill in the list */
-        CubicList *clist)
+
+        Cubic *r_cubic, double *r_error_max_sq, uint *r_split_index)
 {
 	const uint iteration_max = 4;
-	const double error_sq = sq(error_threshold);
-
-	Cubic *cubic;
 
 	if (points_offset_len == 2) {
-		cubic = cubic_alloc(dims);
-		CUBIC_VARS(cubic, dims, p0, p1, p2, p3);
+		CUBIC_VARS(r_cubic, dims, p0, p1, p2, p3);
 
 		copy_vnvn(p0, &points_offset[0 * dims], dims);
 		copy_vnvn(p3, &points_offset[1 * dims], dims);
@@ -725,14 +1006,16 @@ static void fit_cubic_to_points(
 		madd_vn_vnvn_fl(p2, p3, tan_r, dist, dims);
 
 #ifdef USE_ORIG_INDEX_DATA
-		cubic->orig_span = 1;
+		r_cubic->orig_span = 1;
 #endif
-
-		cubic_list_prepend(clist, cubic);
-		return;
+		return true;
 	}
 
 	double *u = malloc(sizeof(double) * points_offset_len);
+
+#ifdef USE_CIRCULAR_FALLBACK
+	const double points_offset_coords_length  =
+#endif
 	points_calc_coord_length(
 	        points_offset, points_offset_len, dims,
 #ifdef USE_LENGTH_CACHE
@@ -740,55 +1023,154 @@ static void fit_cubic_to_points(
 #endif
 	        u);
 
-	cubic = cubic_alloc(dims);
-
-	double error_sq_max;
+	double error_max_sq;
 	uint split_index;
 
 	/* Parameterize points, and attempt to fit curve */
 	cubic_from_points(
-	        points_offset, points_offset_len, u, tan_l, tan_r, dims, cubic);
+	        points_offset, points_offset_len,
+#ifdef USE_CIRCULAR_FALLBACK
+	        points_offset_coords_length,
+#endif
+	        u, tan_l, tan_r, dims, r_cubic);
 
 	/* Find max deviation of points to fitted curve */
-	cubic_calc_error(
-	        cubic, points_offset, points_offset_len, u, dims,
-	        &error_sq_max, &split_index);
+	error_max_sq = cubic_calc_error(
+	        r_cubic, points_offset, points_offset_len, u, dims,
+	        &split_index);
 
-	if (error_sq_max < error_sq) {
-		free(u);
-		cubic_list_prepend(clist, cubic);
-		return;
+	Cubic *cubic_test = alloca(cubic_alloc_size(dims));
+
+	/* Run this so we use the non-circular calculation when the circular-fallback
+	 * in 'cubic_from_points' failed to give a close enough result. */
+#ifdef USE_CIRCULAR_FALLBACK
+	if (!(error_max_sq < error_threshold_sq)) {
+		/* Don't use the cubic calculated above, instead calculate a new fallback cubic,
+		 * since this tends to give more balanced split_index along the curve.
+		 * This is because the attempt to calcualte the cubic may contain spikes
+		 * along the curve which may give a lop-sided maximum distance. */
+		cubic_from_points_fallback(
+		        points_offset, points_offset_len,
+		        tan_l, tan_r, dims, cubic_test);
+		const double error_max_sq_test = cubic_calc_error(
+		        cubic_test, points_offset, points_offset_len, u, dims,
+		        &split_index);
+
+		/* intentionally use the newly calculated 'split_index',
+		 * even if the 'error_max_sq_test' is worse. */
+		if (error_max_sq > error_max_sq_test) {
+			error_max_sq = error_max_sq_test;
+			cubic_copy(r_cubic, cubic_test, dims);
+		}
 	}
-	else {
+#endif
+
+	/* Test the offset fallback */
+#ifdef USE_OFFSET_FALLBACK
+	if (!(error_max_sq < error_threshold_sq)) {
+		/* Using the offset from the curve to calculate cubic handle length may give better results
+		 * try this as a second fallback. */
+		cubic_from_points_offset_fallback(
+		        points_offset, points_offset_len,
+		        tan_l, tan_r, dims, cubic_test);
+		const double error_max_sq_test = cubic_calc_error_simple(
+		        cubic_test, points_offset, points_offset_len, u, error_max_sq, dims);
+
+		if (error_max_sq > error_max_sq_test) {
+			error_max_sq = error_max_sq_test;
+			cubic_copy(r_cubic, cubic_test, dims);
+		}
+	}
+#endif
+
+	*r_error_max_sq = error_max_sq;
+	*r_split_index  = split_index;
+
+	if (!(error_max_sq < error_threshold_sq)) {
+		cubic_copy(cubic_test, r_cubic, dims);
+
 		/* If error not too large, try some reparameterization and iteration */
 		double *u_prime = malloc(sizeof(double) * points_offset_len);
 		for (uint iter = 0; iter < iteration_max; iter++) {
 			if (!cubic_reparameterize(
-			        cubic, points_offset, points_offset_len, u, dims, u_prime))
+			        cubic_test, points_offset, points_offset_len, u, dims, u_prime))
 			{
 				break;
 			}
 
 			cubic_from_points(
-			        points_offset, points_offset_len, u_prime,
-			        tan_l, tan_r, dims, cubic);
-			cubic_calc_error(
-			        cubic, points_offset, points_offset_len, u_prime, dims,
-			        &error_sq_max, &split_index);
+			        points_offset, points_offset_len,
+#ifdef USE_CIRCULAR_FALLBACK
+			        points_offset_coords_length,
+#endif
+			        u_prime, tan_l, tan_r, dims, cubic_test);
 
-			if (error_sq_max < error_sq) {
+			const double error_max_sq_test = cubic_calc_error(
+			        cubic_test, points_offset, points_offset_len, u_prime, dims,
+			        &split_index);
+
+			if (error_max_sq > error_max_sq_test) {
+				error_max_sq = error_max_sq_test;
+				cubic_copy(r_cubic, cubic_test, dims);
+				*r_error_max_sq = error_max_sq;
+				*r_split_index = split_index;
+			}
+
+			if (!(error_max_sq < error_threshold_sq)) {
+				/* continue */
+			}
+			else {
+				assert((error_max_sq < error_threshold_sq));
 				free(u_prime);
 				free(u);
-				cubic_list_prepend(clist, cubic);
-				return;
+				return true;
 			}
 
 			SWAP(double *, u, u_prime);
 		}
 		free(u_prime);
-	}
+		free(u);
 
-	free(u);
+		return false;
+	}
+	else {
+		free(u);
+		return true;
+	}
+}
+
+static void fit_cubic_to_points_recursive(
+        const double *points_offset,
+        const uint    points_offset_len,
+#ifdef USE_LENGTH_CACHE
+        const double *points_length_cache,
+#endif
+        const double  tan_l[],
+        const double  tan_r[],
+        const double  error_threshold_sq,
+        const uint    calc_flag,
+        const uint    dims,
+        /* fill in the list */
+        CubicList *clist)
+{
+	Cubic *cubic = cubic_alloc(dims);
+	uint split_index;
+	double error_max_sq;
+
+	if (fit_cubic_to_points(
+	        points_offset, points_offset_len,
+#ifdef USE_LENGTH_CACHE
+	        points_length_cache,
+#endif
+	        tan_l, tan_r,
+	        (calc_flag & CURVE_FIT_CALC_HIGH_QUALIY) ? DBL_EPSILON : error_threshold_sq,
+	        dims,
+	        cubic, &error_max_sq, &split_index) ||
+	    (error_max_sq < error_threshold_sq))
+	{
+		cubic_list_prepend(clist, cubic);
+		return;
+	}
 	cubic_free(cubic);
 
 
@@ -814,21 +1196,35 @@ static void fit_cubic_to_points(
 		pt_a += dims;
 	}
 
-	/* tan_center = (pt_a - pt_b).normalized() */
-	normalize_vn_vnvn(tan_center, pt_a, pt_b, dims);
+	{
+#ifdef USE_VLA
+		double tan_center_a[dims];
+		double tan_center_b[dims];
+#else
+		double *tan_center_a = alloca(sizeof(double) * dims);
+		double *tan_center_b = alloca(sizeof(double) * dims);
+#endif
+		const double *pt   = &points_offset[split_index * dims];
 
-	fit_cubic_to_points(
+		/* tan_center = ((pt_a - pt).normalized() + (pt - pt_b).normalized()).normalized() */
+		normalize_vn_vnvn(tan_center_a, pt_a, pt, dims);
+		normalize_vn_vnvn(tan_center_b, pt, pt_b, dims);
+		add_vn_vnvn(tan_center, tan_center_a, tan_center_b, dims);
+		normalize_vn(tan_center, dims);
+	}
+
+	fit_cubic_to_points_recursive(
 	        points_offset, split_index + 1,
 #ifdef USE_LENGTH_CACHE
 	        points_length_cache,
 #endif
-	        tan_l, tan_center, error_threshold, dims, clist);
-	fit_cubic_to_points(
+	        tan_l, tan_center, error_threshold_sq, calc_flag, dims, clist);
+	fit_cubic_to_points_recursive(
 	        &points_offset[split_index * dims], points_offset_len - split_index,
 #ifdef USE_LENGTH_CACHE
 	        points_length_cache + split_index,
 #endif
-	        tan_center, tan_r, error_threshold, dims, clist);
+	        tan_center, tan_r, error_threshold_sq, calc_flag, dims, clist);
 
 }
 
@@ -851,6 +1247,7 @@ int curve_fit_cubic_to_points_db(
         const uint    points_len,
         const uint    dims,
         const double  error_threshold,
+        const uint    calc_flag,
         const uint   *corners,
         uint          corners_len,
 
@@ -890,6 +1287,8 @@ int curve_fit_cubic_to_points_db(
 		corner_index_array[corner_index++] = corners[0];
 	}
 
+	const double error_threshold_sq = sq(error_threshold);
+
 	for (uint i = 1; i < corners_len; i++) {
 		const uint points_offset_len = corners[i] - corners[i - 1] + 1;
 		const uint first_point = corners[i - 1];
@@ -919,12 +1318,12 @@ int curve_fit_cubic_to_points_db(
 			        points_length_cache);
 #endif
 
-			fit_cubic_to_points(
+			fit_cubic_to_points_recursive(
 			        &points[first_point * dims], points_offset_len,
 #ifdef USE_LENGTH_CACHE
 			        points_length_cache,
 #endif
-			        tan_l, tan_r, error_threshold, dims, &clist);
+			        tan_l, tan_r, error_threshold_sq, calc_flag, dims, &clist);
 		}
 		else if (points_len == 1) {
 			assert(points_offset_len == 1);
@@ -991,6 +1390,7 @@ int curve_fit_cubic_to_points_fl(
         const uint    points_len,
         const uint    dims,
         const float   error_threshold,
+        const uint    calc_flag,
         const uint   *corners,
         const uint    corners_len,
 
@@ -1001,16 +1401,14 @@ int curve_fit_cubic_to_points_fl(
 	const uint points_flat_len = points_len * dims;
 	double *points_db = malloc(sizeof(double) * points_flat_len);
 
-	for (uint i = 0; i < points_flat_len; i++) {
-		points_db[i] = (double)points[i];
-	}
+	copy_vndb_vnfl(points_db, points, points_flat_len);
 
 	double *cubic_array_db = NULL;
 	float  *cubic_array_fl = NULL;
 	uint    cubic_array_len = 0;
 
 	int result = curve_fit_cubic_to_points_db(
-	        points_db, points_len, dims, error_threshold, corners, corners_len,
+	        points_db, points_len, dims, error_threshold, calc_flag, corners, corners_len,
 	        &cubic_array_db, &cubic_array_len,
 	        r_cubic_orig_index,
 	        r_corner_index_array, r_corner_index_len);
@@ -1031,4 +1429,100 @@ int curve_fit_cubic_to_points_fl(
 	return result;
 }
 
+/**
+ * Fit a single cubic to points.
+ */
+int curve_fit_cubic_to_points_single_db(
+        const double *points,
+        const uint    points_len,
+        const uint    dims,
+        const double  error_threshold,
+        const double tan_l[],
+        const double tan_r[],
+
+        double  r_handle_l[],
+        double  r_handle_r[],
+        double  *r_error_max_sq)
+{
+	Cubic *cubic = alloca(cubic_alloc_size(dims));
+
+	uint split_index;
+
+	/* in this instance theres no advantage in using length cache,
+	 * since we're not recursively calculating values. */
+#ifdef USE_LENGTH_CACHE
+	double *points_length_cache = malloc(sizeof(double) * points_len);
+	points_calc_coord_length_cache(
+	        points, points_len, dims,
+	        points_length_cache);
+#endif
+
+	fit_cubic_to_points(
+	        points, points_len,
+#ifdef USE_LENGTH_CACHE
+	        points_length_cache,
+#endif
+	        tan_l, tan_r, error_threshold, dims,
+
+	        cubic, r_error_max_sq, &split_index);
+
+#ifdef USE_LENGTH_CACHE
+	free(points_length_cache);
+#endif
+
+	copy_vnvn(r_handle_l, CUBIC_PT(cubic, 1, dims), dims);
+	copy_vnvn(r_handle_r, CUBIC_PT(cubic, 2, dims), dims);
+
+	return 0;
+}
+
+int curve_fit_cubic_to_points_single_fl(
+        const float  *points,
+        const uint    points_len,
+        const uint    dims,
+        const float   error_threshold,
+        const float   tan_l[],
+        const float   tan_r[],
+
+        float   r_handle_l[],
+        float   r_handle_r[],
+        float  *r_error_sq)
+{
+	const uint points_flat_len = points_len * dims;
+	double *points_db = malloc(sizeof(double) * points_flat_len);
+
+	copy_vndb_vnfl(points_db, points, points_flat_len);
+
+#ifdef USE_VLA
+	double tan_l_db[dims];
+	double tan_r_db[dims];
+	double r_handle_l_db[dims];
+	double r_handle_r_db[dims];
+#else
+	double *tan_l_db = alloca(sizeof(double) * dims);
+	double *tan_r_db = alloca(sizeof(double) * dims);
+	double *r_handle_l_db = alloca(sizeof(double) * dims);
+	double *r_handle_r_db = alloca(sizeof(double) * dims);
+#endif
+	double r_error_sq_db;
+
+	copy_vndb_vnfl(tan_l_db, tan_l, dims);
+	copy_vndb_vnfl(tan_r_db, tan_r, dims);
+
+	int result = curve_fit_cubic_to_points_single_db(
+	        points_db, points_len, dims,
+	        (double)error_threshold,
+	        tan_l_db, tan_r_db,
+	        r_handle_l_db, r_handle_r_db,
+	        &r_error_sq_db);
+
+	free(points_db);
+
+	copy_vnfl_vndb(r_handle_l, r_handle_l_db, dims);
+	copy_vnfl_vndb(r_handle_r, r_handle_r_db, dims);
+	*r_error_sq = (float)r_error_sq_db;
+
+	return result;
+}
+
 /** \} */

extern/curve_fit_nd/intern/curve_fit_inline.h

@@ -57,7 +57,7 @@ MINLINE double max(const double a, const double b)
 #endif
 
 MINLINE void zero_vn(
-		double v0[], const uint dims)
+        double v0[], const uint dims)
 {
 	for (uint j = 0; j < dims; j++) {
 		v0[j] = 0.0;
@@ -65,7 +65,7 @@ MINLINE void zero_vn(
 }
 
 MINLINE void flip_vn_vnvn(
-		double v_out[], const double v0[], const double v1[], const uint dims)
+        double v_out[], const double v0[], const double v1[], const uint dims)
 {
 	for (uint j = 0; j < dims; j++) {
 		v_out[j] = v0[j] + (v0[j] - v1[j]);
@@ -80,6 +80,22 @@ MINLINE void copy_vnvn(
 	}
 }
 
+MINLINE void copy_vnfl_vndb(
+        float v0[], const double v1[], const uint dims)
+{
+	for (uint j = 0; j < dims; j++) {
+		v0[j] = (float)v1[j];
+	}
+}
+
+MINLINE void copy_vndb_vnfl(
+        double v0[], const float v1[], const uint dims)
+{
+	for (uint j = 0; j < dims; j++) {
+		v0[j] = (double)v1[j];
+	}
+}
+
 MINLINE double dot_vnvn(
         const double v0[], const double v1[], const uint dims)
 {
@@ -178,7 +194,7 @@ MINLINE void imul_vn_fl(double v0[], const double f, const uint dims)
 
 
 MINLINE double len_squared_vnvn(
-		const double v0[], const double v1[], const uint dims)
+        const double v0[], const double v1[], const uint dims)
 {
 	double d = 0.0;
 	for (uint j = 0; j < dims; j++) {
@@ -203,13 +219,29 @@ MINLINE double len_vnvn(
 	return sqrt(len_squared_vnvn(v0, v1, dims));
 }
 
-#if 0
-static double len_vn(
-		const double v0[], const uint dims)
+MINLINE double len_vn(
+        const double v0[], const uint dims)
 {
 	return sqrt(len_squared_vn(v0, dims));
 }
 
+/* special case, save us negating a copy, then getting the length */
+MINLINE double len_squared_negated_vnvn(
+        const double v0[], const double v1[], const uint dims)
+{
+	double d = 0.0;
+	for (uint j = 0; j < dims; j++) {
+		d += sq(v0[j] + v1[j]);
+	}
+	return d;
+}
+
+MINLINE double len_negated_vnvn(
+        const double v0[], const double v1[], const uint dims)
+{
+	return sqrt(len_squared_negated_vnvn(v0, v1, dims));
+}
+
 MINLINE double normalize_vn(
         double v0[], const uint dims)
 {
@@ -219,7 +251,6 @@ MINLINE double normalize_vn(
 	}
 	return d;
 }
-#endif
 
 /* v_out = (v0 - v1).normalized() */
 MINLINE double normalize_vn_vnvn(
@@ -259,4 +290,28 @@ MINLINE bool equals_vnvn(
 	return true;
 }
 
+#if 0
+MINLINE void project_vn_vnvn(
+        double v_out[], const double p[], const double v_proj[], const uint dims)
+{
+	const double mul = dot_vnvn(p, v_proj, dims) / dot_vnvn(v_proj, v_proj, dims);
+	mul_vnvn_fl(v_out, v_proj, mul, dims);
+}
+#endif
+
+MINLINE void project_vn_vnvn_normalized(
+        double v_out[], const double p[], const double v_proj[], const uint dims)
+{
+	const double mul = dot_vnvn(p, v_proj, dims);
+	mul_vnvn_fl(v_out, v_proj, mul, dims);
+}
+
+MINLINE void project_plane_vn_vnvn_normalized(
+        double v_out[], const double v[], const double v_plane[], const uint dims)
+{
+	assert(v != v_out);
+	project_vn_vnvn_normalized(v_out, v, v_plane, dims);
+	sub_vn_vnvn(v_out, v, v_out, dims);
+}
+
 /** \} */

extern/gflags/README.blender

@@ -1,5 +1,5 @@
 Project: Google Flags
-URL: http://code.google.com/p/google-gflags/
+URL: https://github.com/gflags/gflags
 License: New BSD
 Upstream version: 2.2.0 (9db82895)
 Local modifications:
@@ -17,3 +17,7 @@ Local modifications:
 
 - Applied some modifications from fork https://github.com/Nazg-Gul/gflags.git
   (see https://github.com/gflags/gflags/pull/129)
+
+- Avoid attemot of acquiring mutex lock in FlagRegistry::GlobalRegistry when
+  doing static flags initialization. See d81dd2d in Blender repository.
+

extern/glog/README.blender

@@ -1,5 +1,5 @@
 Project: Google Logging
-URL: http://code.google.com/p/google-glog/
+URL: https://github.com/google/glog
 License: New BSD
 Upstream version: 0.3.4, 4d391fe
 Local modifications:

extern/gtest/README.blender

@@ -1,7 +1,5 @@
 Project: Google C++ Testing Framework
-URL: http://code.google.com/p/googletest
+URL: https://github.com/google/googletest
 License: New BSD
 Upstream version: 1.7.0
-Local modifications:
-
-None.
+Local modifications:None

extern/libopenjpeg/README.blender

@@ -0,0 +1,5 @@
+Project: OpenJPEG
+URL: http://www.openjpeg.org
+License: BSD 2-Clause
+Upstream version: 1.5.2
+Local modifications:

extern/rangetree/README.blender

@@ -0,0 +1,5 @@
+Project: RangeTree
+URL: https://github.com/nicholasbishop/RangeTree
+License: GPLv2+
+Upstream version: c4ecf6bb7dfd
+Local modifications: None

extern/recastnavigation/CMakeLists.txt

@@ -23,9 +23,11 @@
 #
 # ***** END GPL LICENSE BLOCK *****
 
+remove_cc_flag_unsigned_char()
+
 set(INC 
-		Recast/Include
-		Detour/Include
+	Recast/Include
+	Detour/Include
 )
 
 set(INC_SYS
@@ -33,38 +35,38 @@ set(INC_SYS
 )
 
 set(SRC 
-		recast-capi.cpp
-		recast-capi.h
+	recast-capi.cpp
+	recast-capi.h
 
 
-		Detour/Source/DetourCommon.cpp
-		Detour/Source/DetourNode.cpp
-		Detour/Source/DetourStatNavMesh.cpp
-		Detour/Source/DetourStatNavMeshBuilder.cpp
-		Detour/Source/DetourTileNavMesh.cpp
-		Detour/Source/DetourTileNavMeshBuilder.cpp
+	Detour/Source/DetourCommon.cpp
+	Detour/Source/DetourNode.cpp
+	Detour/Source/DetourStatNavMesh.cpp
+	Detour/Source/DetourStatNavMeshBuilder.cpp
+	Detour/Source/DetourTileNavMesh.cpp
+	Detour/Source/DetourTileNavMeshBuilder.cpp
 
-		Detour/Include/DetourCommon.h
-		Detour/Include/DetourNode.h
-		Detour/Include/DetourStatNavMesh.h
-		Detour/Include/DetourStatNavMeshBuilder.h
-		Detour/Include/DetourTileNavMesh.h
-		Detour/Include/DetourTileNavMeshBuilder.h
+	Detour/Include/DetourCommon.h
+	Detour/Include/DetourNode.h
+	Detour/Include/DetourStatNavMesh.h
+	Detour/Include/DetourStatNavMeshBuilder.h
+	Detour/Include/DetourTileNavMesh.h
+	Detour/Include/DetourTileNavMeshBuilder.h
 
-		Recast/Source/Recast.cpp
-		Recast/Source/RecastAlloc.cpp
-		Recast/Source/RecastArea.cpp
-		Recast/Source/RecastContour.cpp
-		Recast/Source/RecastFilter.cpp
-		Recast/Source/RecastLayers.cpp
-		Recast/Source/RecastMesh.cpp
-		Recast/Source/RecastMeshDetail.cpp
-		Recast/Source/RecastRasterization.cpp
-		Recast/Source/RecastRegion.cpp
+	Recast/Source/Recast.cpp
+	Recast/Source/RecastAlloc.cpp
+	Recast/Source/RecastArea.cpp
+	Recast/Source/RecastContour.cpp
+	Recast/Source/RecastFilter.cpp
+	Recast/Source/RecastLayers.cpp
+	Recast/Source/RecastMesh.cpp
+	Recast/Source/RecastMeshDetail.cpp
+	Recast/Source/RecastRasterization.cpp
+	Recast/Source/RecastRegion.cpp
 
-		Recast/Include/Recast.h
-		Recast/Include/RecastAlloc.h
-		Recast/Include/RecastAssert.h
+	Recast/Include/Recast.h
+	Recast/Include/RecastAlloc.h
+	Recast/Include/RecastAssert.h
 )
 
 blender_add_lib(extern_recastnavigation "${SRC}" "${INC}" "${INC_SYS}")

extern/sdlew/README.blender

@@ -0,0 +1,5 @@
+Project: SDL Extension Wrangler
+URL: https://github.com/SDLWrangler/sdlew
+License: Apache 2.0
+Upstream version: 15edf8e 
+Local modifications: None

extern/wcwidth/README.blender

@@ -0,0 +1,5 @@
+Project: WC Width
+URL: http://www.cl.cam.ac.uk/~mgk25/ucs/wcwidth.c
+License: ICS
+Upstream version: 2007-05-26
+Local modifications: None

extern/xdnd/README.blender

@@ -0,0 +1,8 @@
+Project: X Drag and Drop
+URL: http://www.newplanetsoftware.com/xdnd/ (defunct)
+     https://freedesktop.org/wiki/Specifications/XDND/ (cache)
+License: GPLv2+
+Upstream version: 2000-08-08
+Local modifications:
+* Fix T33192
+  Opening Blender breaks drag-and-drop support on the KDE desktop.

intern/CMakeLists.txt

@@ -34,6 +34,10 @@ add_subdirectory(mikktspace)
 add_subdirectory(glew-mx)
 add_subdirectory(eigen)
 
+if (WITH_GAMEENGINE_DECKLINK)
+	add_subdirectory(decklink)
+endif()
+
 if(WITH_AUDASPACE)
 	add_subdirectory(audaspace)
 endif()
@@ -79,8 +83,10 @@ if(WITH_OPENSUBDIV)
 endif()
 
 # only windows needs utf16 converter
+# gpudirect is a runtime interface to the nVidia's DVP driver, only for windows
 if(WIN32)
 	add_subdirectory(utfconv)
+	add_subdirectory(gpudirect)
 endif()
 
 if(WITH_OPENVDB)

intern/atomic/atomic_ops.h

@@ -1,11 +1,11 @@
 /*
- * Adopted from jemalloc with this license:
+ * Original code from jemalloc with this license:
  *
  * Copyright (C) 2002-2013 Jason Evans <jasone@canonware.com>.
  * All rights reserved.
  * Copyright (C) 2007-2012 Mozilla Foundation.  All rights reserved.
  * Copyright (C) 2009-2013 Facebook, Inc.  All rights reserved.
-
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
  * 1. Redistributions of source code must retain the above copyright notice(s),
@@ -13,7 +13,7 @@
  * 2. Redistributions in binary form must reproduce the above copyright notice(s),
  *    this list of conditions and the following disclaimer in the documentation
  *    and/or other materials provided with the distribution.
-
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY EXPRESS
  * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO
@@ -24,63 +24,59 @@
  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
  * OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
  * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ * ***** 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.
+ *
+ * The Original Code is Copyright (C) 2016 Blender Foundation.
+ * All rights reserved.
+ *
+ * The Original Code is: adapted from jemalloc.
+ *
+ * ***** END GPL LICENSE BLOCK *****
+ */
+
+/**
+ * \file atomic_ops.h
+ * \ingroup Atomic
+ *
+ * \author Copyright (C) 2016 Blender Foundation, adapted from jemalloc.
+ * \brief Provides wrapper around system-specific atomic primitives, and some extensions (faked-atomic operations
+ *        over float numbers).
  */
 
 #ifndef __ATOMIC_OPS_H__
 #define __ATOMIC_OPS_H__
 
-#include <assert.h>
-
-#if defined (__APPLE__)
-#  include <libkern/OSAtomic.h>
-#elif defined(_MSC_VER)
-#  define NOGDI
-#  ifndef NOMINMAX
-#    define NOMINMAX
-#  endif
-#  define WIN32_LEAN_AND_MEAN
-#  include <windows.h>
-#elif defined(__arm__)
+#if defined(__arm__)
 /* Attempt to fix compilation error on Debian armel kernel.
  * arm7 architecture does have both 32 and 64bit atomics, however
  * it's gcc doesn't have __GCC_HAVE_SYNC_COMPARE_AND_SWAP_n defined.
  */
-#  define JE_FORCE_SYNC_COMPARE_AND_SWAP_8
+#  define JE_FORCE_SYNC_COMPARE_AND_SWAP_1
 #  define JE_FORCE_SYNC_COMPARE_AND_SWAP_4
+#  define JE_FORCE_SYNC_COMPARE_AND_SWAP_8
 #endif
 
-/* needed for int types */
-#include "../../source/blender/blenlib/BLI_sys_types.h"
-#include <stdlib.h>
-#include <stddef.h>
+#include "intern/atomic_ops_utils.h"
 
-/* little macro so inline keyword works */
-#if defined(_MSC_VER)
-#  define ATOMIC_INLINE static __forceinline
-#else
-#  if (defined(__APPLE__) && defined(__ppc__))
-/* static inline __attribute__ here breaks osx ppc gcc42 build */
-#    define ATOMIC_INLINE static __attribute__((always_inline))
-#  else
-#    define ATOMIC_INLINE static inline __attribute__((always_inline))
-#  endif
-#endif
-
-/* This is becoming a bit nastier that it was originally foreseen,
- * consider using autoconfig detection instead.
- */
-#if defined(_M_X64) || defined(__amd64__) || defined(__x86_64__) || defined(__s390x__) || defined(__powerpc64__) || defined(__aarch64__) || (defined(__sparc__) && defined(__arch64__)) || defined(__alpha__) || defined(__mips64)
-#  define LG_SIZEOF_PTR 3
-#  define LG_SIZEOF_INT 2
-#else
-#  define LG_SIZEOF_PTR 2
-#  define LG_SIZEOF_INT 2
-#endif
-
-/************************/
+/******************************************************************************/
 /* Function prototypes. */
 
-#if (LG_SIZEOF_PTR == 3 || LG_SIZEOF_INT == 3)
+#if (LG_SIZEOF_PTR == 8 || LG_SIZEOF_INT == 8)
 ATOMIC_INLINE uint64_t atomic_add_uint64(uint64_t *p, uint64_t x);
 ATOMIC_INLINE uint64_t atomic_sub_uint64(uint64_t *p, uint64_t x);
 ATOMIC_INLINE uint64_t atomic_cas_uint64(uint64_t *v, uint64_t old, uint64_t _new);
@@ -90,6 +86,9 @@ ATOMIC_INLINE uint32_t atomic_add_uint32(uint32_t *p, uint32_t x);
 ATOMIC_INLINE uint32_t atomic_sub_uint32(uint32_t *p, uint32_t x);
 ATOMIC_INLINE uint32_t atomic_cas_uint32(uint32_t *v, uint32_t old, uint32_t _new);
 
+ATOMIC_INLINE uint32_t atomic_fetch_and_add_uint32(uint32_t *p, uint32_t x);
+
+ATOMIC_INLINE uint8_t atomic_fetch_and_or_uint8(uint8_t *p, uint8_t b);
 ATOMIC_INLINE uint8_t atomic_fetch_and_and_uint8(uint8_t *p, uint8_t b);
 
 ATOMIC_INLINE size_t atomic_add_z(size_t *p, size_t x);
@@ -100,395 +99,22 @@ ATOMIC_INLINE unsigned atomic_add_u(unsigned *p, unsigned x);
 ATOMIC_INLINE unsigned atomic_sub_u(unsigned *p, unsigned x);
 ATOMIC_INLINE unsigned atomic_cas_u(unsigned *v, unsigned old, unsigned _new);
 
-/******************************************************************************/
-/* 64-bit operations. */
-#if (LG_SIZEOF_PTR == 3 || LG_SIZEOF_INT == 3)
-#  ifdef __GCC_HAVE_SYNC_COMPARE_AND_SWAP_8
-ATOMIC_INLINE uint64_t
-atomic_add_uint64(uint64_t *p, uint64_t x)
-{
-	return __sync_add_and_fetch(p, x);
-}
-
-ATOMIC_INLINE uint64_t
-atomic_sub_uint64(uint64_t *p, uint64_t x)
-{
-	return __sync_sub_and_fetch(p, x);
-}
-
-ATOMIC_INLINE uint64_t
-atomic_cas_uint64(uint64_t *v, uint64_t old, uint64_t _new)
-{
-	return __sync_val_compare_and_swap(v, old, _new);
-}
-#elif (defined(_MSC_VER))
-ATOMIC_INLINE uint64_t
-atomic_add_uint64(uint64_t *p, uint64_t x)
-{
-	return InterlockedExchangeAdd64((int64_t *)p, (int64_t)x) + x;
-}
-
-ATOMIC_INLINE uint64_t
-atomic_sub_uint64(uint64_t *p, uint64_t x)
-{
-	return InterlockedExchangeAdd64((int64_t *)p, -((int64_t)x)) - x;
-}
-
-ATOMIC_INLINE uint64_t
-atomic_cas_uint64(uint64_t *v, uint64_t old, uint64_t _new)
-{
-	return InterlockedCompareExchange64((int64_t *)v, _new, old);
-}
-#elif (defined(__APPLE__))
-ATOMIC_INLINE uint64_t
-atomic_add_uint64(uint64_t *p, uint64_t x)
-{
-	return (uint64_t)OSAtomicAdd64((int64_t)x, (int64_t *)p);
-}
-
-ATOMIC_INLINE uint64_t
-atomic_sub_uint64(uint64_t *p, uint64_t x)
-{
-	return (uint64_t)OSAtomicAdd64(-((int64_t)x), (int64_t *)p);
-}
-
-ATOMIC_INLINE uint64_t
-atomic_cas_uint64(uint64_t *v, uint64_t old, uint64_t _new)
-{
-	uint64_t init_val = *v;
-	OSAtomicCompareAndSwap64((int64_t)old, (int64_t)_new, (int64_t *)v);
-	return init_val;
-}
-#  elif (defined(__amd64__) || defined(__x86_64__))
-ATOMIC_INLINE uint64_t
-atomic_add_uint64(uint64_t *p, uint64_t x)
-{
-	asm volatile (
-	    "lock; xaddq %0, %1;"
-	    : "+r" (x), "=m" (*p) /* Outputs. */
-	    : "m" (*p) /* Inputs. */
-	    );
-	return x;
-}
-
-ATOMIC_INLINE uint64_t
-atomic_sub_uint64(uint64_t *p, uint64_t x)
-{
-	x = (uint64_t)(-(int64_t)x);
-	asm volatile (
-	    "lock; xaddq %0, %1;"
-	    : "+r" (x), "=m" (*p) /* Outputs. */
-	    : "m" (*p) /* Inputs. */
-	    );
-	return x;
-}
-
-ATOMIC_INLINE uint64_t
-atomic_cas_uint64(uint64_t *v, uint64_t old, uint64_t _new)
-{
-	uint64_t ret;
-	asm volatile (
-	    "lock; cmpxchgq %2,%1"
-	    : "=a" (ret), "+m" (*v)
-	    : "r" (_new), "0" (old)
-	    : "memory");
-	return ret;
-}
-
-#  elif (defined(JEMALLOC_ATOMIC9))
-ATOMIC_INLINE uint64_t
-atomic_add_uint64(uint64_t *p, uint64_t x)
-{
-	/*
-	 * atomic_fetchadd_64() doesn't exist, but we only ever use this
-	 * function on LP64 systems, so atomic_fetchadd_long() will do.
-	 */
-	assert(sizeof(uint64_t) == sizeof(unsigned long));
-
-	return atomic_fetchadd_long(p, (unsigned long)x) + x;
-}
-
-ATOMIC_INLINE uint64_t
-atomic_sub_uint64(uint64_t *p, uint64_t x)
-{
-	assert(sizeof(uint64_t) == sizeof(unsigned long));
-
-	return atomic_fetchadd_long(p, (unsigned long)(-(long)x)) - x;
-}
-
-ATOMIC_INLINE uint64_t
-atomic_cas_uint64(uint64_t *v, uint64_t old, uint64_t _new)
-{
-	assert(sizeof(uint64_t) == sizeof(unsigned long));
-
-	return atomic_cmpset_long(v, old, _new);
-}
-#  elif (defined(JE_FORCE_SYNC_COMPARE_AND_SWAP_8))
-ATOMIC_INLINE uint64_t
-atomic_add_uint64(uint64_t *p, uint64_t x)
-{
-	return __sync_add_and_fetch(p, x);
-}
-
-ATOMIC_INLINE uint64_t
-atomic_sub_uint64(uint64_t *p, uint64_t x)
-{
-	return __sync_sub_and_fetch(p, x);
-}
-
-ATOMIC_INLINE uint64_t
-atomic_cas_uint64(uint64_t *v, uint64_t old, uint64_t _new)
-{
-	return __sync_val_compare_and_swap(v, old, _new);
-}
-#  else
-#    error "Missing implementation for 64-bit atomic operations"
-#  endif
-#endif
+/* WARNING! Float 'atomics' are really faked ones, those are actually closer to some kind of spinlock-sync'ed operation,
+ *          which means they are only efficient if collisions are highly unlikely (i.e. if probability of two threads
+ *          working on the same pointer at the same time is very low). */
+ATOMIC_INLINE float atomic_add_fl(float *p, const float x);
 
 /******************************************************************************/
-/* 32-bit operations. */
-#ifdef __GCC_HAVE_SYNC_COMPARE_AND_SWAP_4
-ATOMIC_INLINE uint32_t
-atomic_add_uint32(uint32_t *p, uint32_t x)
-{
-	return __sync_add_and_fetch(p, x);
-}
+/* Include system-dependent implementations. */
 
-ATOMIC_INLINE uint32_t
-atomic_sub_uint32(uint32_t *p, uint32_t x)
-{
-	return __sync_sub_and_fetch(p, x);
-}
-
-ATOMIC_INLINE uint32_t
-atomic_cas_uint32(uint32_t *v, uint32_t old, uint32_t _new)
-{
-   return __sync_val_compare_and_swap(v, old, _new);
-}
-#elif (defined(_MSC_VER))
-ATOMIC_INLINE uint32_t
-atomic_add_uint32(uint32_t *p, uint32_t x)
-{
-	return InterlockedExchangeAdd(p, x) + x;
-}
-
-ATOMIC_INLINE uint32_t
-atomic_sub_uint32(uint32_t *p, uint32_t x)
-{
-	return InterlockedExchangeAdd(p, -((int32_t)x)) - x;
-}
-
-ATOMIC_INLINE uint32_t
-atomic_cas_uint32(uint32_t *v, uint32_t old, uint32_t _new)
-{
-	return InterlockedCompareExchange((long *)v, _new, old);
-}
-#elif (defined(__APPLE__))
-ATOMIC_INLINE uint32_t
-atomic_add_uint32(uint32_t *p, uint32_t x)
-{
-	return (uint32_t)OSAtomicAdd32((int32_t)x, (int32_t *)p);
-}
-
-ATOMIC_INLINE uint32_t
-atomic_sub_uint32(uint32_t *p, uint32_t x)
-{
-	return (uint32_t)OSAtomicAdd32(-((int32_t)x), (int32_t *)p);
-}
-
-ATOMIC_INLINE uint32_t
-atomic_cas_uint32(uint32_t *v, uint32_t old, uint32_t _new)
-{
-	uint32_t init_val = *v;
-	OSAtomicCompareAndSwap32((int32_t)old, (int32_t)_new, (int32_t *)v);
-	return init_val;
-}
-#elif (defined(__i386__) || defined(__amd64__) || defined(__x86_64__))
-ATOMIC_INLINE uint32_t
-atomic_add_uint32(uint32_t *p, uint32_t x)
-{
-	asm volatile (
-	    "lock; xaddl %0, %1;"
-	    : "+r" (x), "=m" (*p) /* Outputs. */
-	    : "m" (*p) /* Inputs. */
-	    );
-	return x;
-}
-
-ATOMIC_INLINE uint32_t
-atomic_sub_uint32(uint32_t *p, uint32_t x)
-{
-	x = (uint32_t)(-(int32_t)x);
-	asm volatile (
-	    "lock; xaddl %0, %1;"
-	    : "+r" (x), "=m" (*p) /* Outputs. */
-	    : "m" (*p) /* Inputs. */
-	    );
-	return x;
-}
-
-ATOMIC_INLINE uint32_t
-atomic_cas_uint32(uint32_t *v, uint32_t old, uint32_t _new)
-{
-	uint32_t ret;
-	asm volatile (
-	    "lock; cmpxchgl %2,%1"
-	    : "=a" (ret), "+m" (*v)
-	    : "r" (_new), "0" (old)
-	    : "memory");
-	return ret;
-}
-#elif (defined(JEMALLOC_ATOMIC9))
-ATOMIC_INLINE uint32_t
-atomic_add_uint32(uint32_t *p, uint32_t x)
-{
-	return atomic_fetchadd_32(p, x) + x;
-}
-
-ATOMIC_INLINE uint32_t
-atomic_sub_uint32(uint32_t *p, uint32_t x)
-{
-	return atomic_fetchadd_32(p, (uint32_t)(-(int32_t)x)) - x;
-}
-
-ATOMIC_INLINE uint32_t
-atomic_cas_uint32(uint32_t *v, uint32_t old, uint32_t _new)
-{
-	return atomic_cmpset_32(v, old, _new);
-}
-#elif defined(JE_FORCE_SYNC_COMPARE_AND_SWAP_4)
-ATOMIC_INLINE uint32_t
-atomic_add_uint32(uint32_t *p, uint32_t x)
-{
-	return __sync_add_and_fetch(p, x);
-}
-
-ATOMIC_INLINE uint32_t
-atomic_sub_uint32(uint32_t *p, uint32_t x)
-{
-	return __sync_sub_and_fetch(p, x);
-}
-
-ATOMIC_INLINE uint32_t
-atomic_cas_uint32(uint32_t *v, uint32_t old, uint32_t _new)
-{
-	return __sync_val_compare_and_swap(v, old, _new);
-}
+/* Note that we are using _unix flavor as fallback here (it will raise precompiler errors as needed). */
+#if defined(_MSC_VER)
+#  include "intern/atomic_ops_msvc.h"
 #else
-#  error "Missing implementation for 32-bit atomic operations"
+#  include "intern/atomic_ops_unix.h"
 #endif
 
-/******************************************************************************/
-/* 8-bit operations. */
-#ifdef __GCC_HAVE_SYNC_COMPARE_AND_SWAP_1
-ATOMIC_INLINE uint8_t
-atomic_fetch_and_and_uint8(uint8_t *p, uint8_t b)
-{
-	return __sync_fetch_and_and(p, b);
-}
-#elif (defined(_MSC_VER))
-#include <intrin.h>
-#pragma intrinsic(_InterlockedAnd8)
-ATOMIC_INLINE uint8_t
-atomic_fetch_and_and_uint8(uint8_t *p, uint8_t b)
-{
-#if (LG_SIZEOF_PTR == 3 || LG_SIZEOF_INT == 3)
-	return InterlockedAnd8((char *)p, (char)b);
-#else
-	return _InterlockedAnd8((char *)p, (char)b);
-#endif
-}
-#else
-#  error "Missing implementation for 8-bit atomic operations"
-#endif
-
-/******************************************************************************/
-/* size_t operations. */
-ATOMIC_INLINE size_t
-atomic_add_z(size_t *p, size_t x)
-{
-	assert(sizeof(size_t) == 1 << LG_SIZEOF_PTR);
-
-#if (LG_SIZEOF_PTR == 3)
-	return (size_t)atomic_add_uint64((uint64_t *)p, (uint64_t)x);
-#elif (LG_SIZEOF_PTR == 2)
-	return (size_t)atomic_add_uint32((uint32_t *)p, (uint32_t)x);
-#endif
-}
-
-ATOMIC_INLINE size_t
-atomic_sub_z(size_t *p, size_t x)
-{
-	assert(sizeof(size_t) == 1 << LG_SIZEOF_PTR);
-
-#if (LG_SIZEOF_PTR == 3)
-	return (size_t)atomic_add_uint64((uint64_t *)p,
-	                                 (uint64_t)-((int64_t)x));
-#elif (LG_SIZEOF_PTR == 2)
-	return (size_t)atomic_add_uint32((uint32_t *)p,
-	                                 (uint32_t)-((int32_t)x));
-#endif
-}
-
-ATOMIC_INLINE size_t
-atomic_cas_z(size_t *v, size_t old, size_t _new)
-{
-	assert(sizeof(size_t) == 1 << LG_SIZEOF_PTR);
-
-#if (LG_SIZEOF_PTR == 3)
-	return (size_t)atomic_cas_uint64((uint64_t *)v,
-	                                 (uint64_t)old,
-	                                 (uint64_t)_new);
-#elif (LG_SIZEOF_PTR == 2)
-	return (size_t)atomic_cas_uint32((uint32_t *)v,
-	                                 (uint32_t)old,
-	                                 (uint32_t)_new);
-#endif
-}
-
-/******************************************************************************/
-/* unsigned operations. */
-ATOMIC_INLINE unsigned
-atomic_add_u(unsigned *p, unsigned x)
-{
-	assert(sizeof(unsigned) == 1 << LG_SIZEOF_INT);
-
-#if (LG_SIZEOF_INT == 3)
-	return (unsigned)atomic_add_uint64((uint64_t *)p, (uint64_t)x);
-#elif (LG_SIZEOF_INT == 2)
-	return (unsigned)atomic_add_uint32((uint32_t *)p, (uint32_t)x);
-#endif
-}
-
-ATOMIC_INLINE unsigned
-atomic_sub_u(unsigned *p, unsigned x)
-{
-	assert(sizeof(unsigned) == 1 << LG_SIZEOF_INT);
-
-#if (LG_SIZEOF_INT == 3)
-	return (unsigned)atomic_add_uint64((uint64_t *)p,
-	                                   (uint64_t)-((int64_t)x));
-#elif (LG_SIZEOF_INT == 2)
-	return (unsigned)atomic_add_uint32((uint32_t *)p,
-	                                   (uint32_t)-((int32_t)x));
-#endif
-}
-
-ATOMIC_INLINE unsigned
-atomic_cas_u(unsigned *v, unsigned old, unsigned _new)
-{
-	assert(sizeof(unsigned) == 1 << LG_SIZEOF_INT);
-
-#if (LG_SIZEOF_PTR == 3)
-	return (unsigned)atomic_cas_uint64((uint64_t *)v,
-	                                   (uint64_t)old,
-	                                   (uint64_t)_new);
-#elif (LG_SIZEOF_PTR == 2)
-	return (unsigned)atomic_cas_uint32((uint32_t *)v,
-	                                   (uint32_t)old,
-	                                   (uint32_t)_new);
-#endif
-}
+/* Include 'fake' atomic extensions, built over real atomic primitives. */
+#include "intern/atomic_ops_ext.h"
 
 #endif /* __ATOMIC_OPS_H__ */

intern/atomic/intern/atomic_ops_ext.h

@@ -0,0 +1,146 @@
+/*
+ * Original code from jemalloc with this license:
+ *
+ * Copyright (C) 2002-2013 Jason Evans <jasone@canonware.com>.
+ * All rights reserved.
+ * Copyright (C) 2007-2012 Mozilla Foundation.  All rights reserved.
+ * Copyright (C) 2009-2013 Facebook, Inc.  All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ * 1. Redistributions of source code must retain the above copyright notice(s),
+ *    this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice(s),
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY EXPRESS
+ * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO
+ * EVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+ * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
+ * OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
+ * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ * ***** 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.
+ *
+ * The Original Code is Copyright (C) 2016 Blender Foundation.
+ * All rights reserved.
+ *
+ * The Original Code is: adapted from jemalloc.
+ *
+ * ***** END GPL LICENSE BLOCK *****
+ */
+
+#ifndef __ATOMIC_OPS_EXT_H__
+#define __ATOMIC_OPS_EXT_H__
+
+#include "atomic_ops_utils.h"
+
+/******************************************************************************/
+/* size_t operations. */
+ATOMIC_INLINE size_t atomic_add_z(size_t *p, size_t x)
+{
+	assert(sizeof(size_t) == LG_SIZEOF_PTR);
+
+#if (LG_SIZEOF_PTR == 8)
+	return (size_t)atomic_add_uint64((uint64_t *)p, (uint64_t)x);
+#elif (LG_SIZEOF_PTR == 4)
+	return (size_t)atomic_add_uint32((uint32_t *)p, (uint32_t)x);
+#endif
+}
+
+ATOMIC_INLINE size_t atomic_sub_z(size_t *p, size_t x)
+{
+	assert(sizeof(size_t) == LG_SIZEOF_PTR);
+
+#if (LG_SIZEOF_PTR == 8)
+	return (size_t)atomic_add_uint64((uint64_t *)p, (uint64_t)-((int64_t)x));
+#elif (LG_SIZEOF_PTR == 4)
+	return (size_t)atomic_add_uint32((uint32_t *)p, (uint32_t)-((int32_t)x));
+#endif
+}
+
+ATOMIC_INLINE size_t atomic_cas_z(size_t *v, size_t old, size_t _new)
+{
+	assert(sizeof(size_t) == LG_SIZEOF_PTR);
+
+#if (LG_SIZEOF_PTR == 8)
+	return (size_t)atomic_cas_uint64((uint64_t *)v, (uint64_t)old, (uint64_t)_new);
+#elif (LG_SIZEOF_PTR == 4)
+	return (size_t)atomic_cas_uint32((uint32_t *)v, (uint32_t)old, (uint32_t)_new);
+#endif
+}
+
+/******************************************************************************/
+/* unsigned operations. */
+ATOMIC_INLINE unsigned atomic_add_u(unsigned *p, unsigned x)
+{
+	assert(sizeof(unsigned) == LG_SIZEOF_INT);
+
+#if (LG_SIZEOF_INT == 8)
+	return (unsigned)atomic_add_uint64((uint64_t *)p, (uint64_t)x);
+#elif (LG_SIZEOF_INT == 4)
+	return (unsigned)atomic_add_uint32((uint32_t *)p, (uint32_t)x);
+#endif
+}
+
+ATOMIC_INLINE unsigned atomic_sub_u(unsigned *p, unsigned x)
+{
+	assert(sizeof(unsigned) == LG_SIZEOF_INT);
+
+#if (LG_SIZEOF_INT == 8)
+	return (unsigned)atomic_add_uint64((uint64_t *)p, (uint64_t)-((int64_t)x));
+#elif (LG_SIZEOF_INT == 4)
+	return (unsigned)atomic_add_uint32((uint32_t *)p, (uint32_t)-((int32_t)x));
+#endif
+}
+
+ATOMIC_INLINE unsigned atomic_cas_u(unsigned *v, unsigned old, unsigned _new)
+{
+	assert(sizeof(unsigned) == LG_SIZEOF_INT);
+
+#if (LG_SIZEOF_INT == 8)
+	return (unsigned)atomic_cas_uint64((uint64_t *)v, (uint64_t)old, (uint64_t)_new);
+#elif (LG_SIZEOF_INT == 4)
+	return (unsigned)atomic_cas_uint32((uint32_t *)v, (uint32_t)old, (uint32_t)_new);
+#endif
+}
+
+/******************************************************************************/
+/* float operations. */
+
+ATOMIC_INLINE float atomic_add_fl(float *p, const float x)
+{
+	assert(sizeof(float) == sizeof(uint32_t));
+
+	float oldval, newval;
+	uint32_t prevval;
+
+	do {  /* Note that since collisions are unlikely, loop will nearly always run once. */
+		oldval = *p;
+		newval = oldval + x;
+		prevval = atomic_cas_uint32((uint32_t *)p, *(uint32_t *)(&oldval), *(uint32_t *)(&newval));
+	} while (UNLIKELY(prevval != *(uint32_t *)(&oldval)));
+
+	return newval;
+}
+
+#endif /* __ATOMIC_OPS_EXT_H__ */

intern/atomic/intern/atomic_ops_msvc.h

@@ -0,0 +1,107 @@
+/*
+ * Adopted from jemalloc with this license:
+ *
+ * Copyright (C) 2002-2013 Jason Evans <jasone@canonware.com>.
+ * All rights reserved.
+ * Copyright (C) 2007-2012 Mozilla Foundation.  All rights reserved.
+ * Copyright (C) 2009-2013 Facebook, Inc.  All rights reserved.
+
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ * 1. Redistributions of source code must retain the above copyright notice(s),
+ *    this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice(s),
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY EXPRESS
+ * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO
+ * EVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+ * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
+ * OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
+ * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __ATOMIC_OPS_MSVC_H__
+#define __ATOMIC_OPS_MSVC_H__
+
+#include "atomic_ops_utils.h"
+
+#define NOGDI
+#ifndef NOMINMAX
+#  define NOMINMAX
+#endif
+#define WIN32_LEAN_AND_MEAN
+
+#include <windows.h>
+#include <intrin.h>
+
+/******************************************************************************/
+/* 64-bit operations. */
+#if (LG_SIZEOF_PTR == 8 || LG_SIZEOF_INT == 8)
+ATOMIC_INLINE uint64_t atomic_add_uint64(uint64_t *p, uint64_t x)
+{
+	return InterlockedExchangeAdd64((int64_t *)p, (int64_t)x) + x;
+}
+
+ATOMIC_INLINE uint64_t atomic_sub_uint64(uint64_t *p, uint64_t x)
+{
+	return InterlockedExchangeAdd64((int64_t *)p, -((int64_t)x)) - x;
+}
+
+ATOMIC_INLINE uint64_t atomic_cas_uint64(uint64_t *v, uint64_t old, uint64_t _new)
+{
+	return InterlockedCompareExchange64((int64_t *)v, _new, old);
+}
+#endif
+
+/******************************************************************************/
+/* 32-bit operations. */
+ATOMIC_INLINE uint32_t atomic_add_uint32(uint32_t *p, uint32_t x)
+{
+	return InterlockedExchangeAdd(p, x) + x;
+}
+
+ATOMIC_INLINE uint32_t atomic_sub_uint32(uint32_t *p, uint32_t x)
+{
+	return InterlockedExchangeAdd(p, -((int32_t)x)) - x;
+}
+
+ATOMIC_INLINE uint32_t atomic_cas_uint32(uint32_t *v, uint32_t old, uint32_t _new)
+{
+	return InterlockedCompareExchange((long *)v, _new, old);
+}
+
+ATOMIC_INLINE uint32_t atomic_fetch_and_add_uint32(uint32_t *p, uint32_t x)
+{
+	return InterlockedExchangeAdd(p, x);
+}
+
+/******************************************************************************/
+/* 8-bit operations. */
+
+#pragma intrinsic(_InterlockedAnd8)
+ATOMIC_INLINE uint8_t atomic_fetch_and_and_uint8(uint8_t *p, uint8_t b)
+{
+#if (LG_SIZEOF_PTR == 8 || LG_SIZEOF_INT == 8)
+	return InterlockedAnd8((char *)p, (char)b);
+#else
+	return _InterlockedAnd8((char *)p, (char)b);
+#endif
+}
+
+#pragma intrinsic(_InterlockedOr8)
+ATOMIC_INLINE uint8_t atomic_fetch_and_or_uint8(uint8_t *p, uint8_t b)
+{
+#if (LG_SIZEOF_PTR == 8 || LG_SIZEOF_INT == 8)
+	return InterlockedOr8((char *)p, (char)b);
+#else
+	return _InterlockedOr8((char *)p, (char)b);
+#endif
+}
+
+#endif /* __ATOMIC_OPS_MSVC_H__ */

intern/atomic/intern/atomic_ops_unix.h

@@ -0,0 +1,191 @@
+/*
+ * Original code from jemalloc with this license:
+ *
+ * Copyright (C) 2002-2013 Jason Evans <jasone@canonware.com>.
+ * All rights reserved.
+ * Copyright (C) 2007-2012 Mozilla Foundation.  All rights reserved.
+ * Copyright (C) 2009-2013 Facebook, Inc.  All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ * 1. Redistributions of source code must retain the above copyright notice(s),
+ *    this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice(s),
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY EXPRESS
+ * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO
+ * EVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+ * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
+ * OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
+ * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ * ***** 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.
+ *
+ * The Original Code is Copyright (C) 2016 Blender Foundation.
+ * All rights reserved.
+ *
+ * The Original Code is: adapted from jemalloc.
+ *
+ * ***** END GPL LICENSE BLOCK *****
+ */
+
+#ifndef __ATOMIC_OPS_UNIX_H__
+#define __ATOMIC_OPS_UNIX_H__
+
+#include "atomic_ops_utils.h"
+
+/******************************************************************************/
+/* 64-bit operations. */
+#if (LG_SIZEOF_PTR == 8 || LG_SIZEOF_INT == 8)
+#  if (defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_8) || defined(JE_FORCE_SYNC_COMPARE_AND_SWAP_8))
+ATOMIC_INLINE uint64_t atomic_add_uint64(uint64_t *p, uint64_t x)
+{
+	return __sync_add_and_fetch(p, x);
+}
+
+ATOMIC_INLINE uint64_t atomic_sub_uint64(uint64_t *p, uint64_t x)
+{
+	return __sync_sub_and_fetch(p, x);
+}
+
+ATOMIC_INLINE uint64_t atomic_cas_uint64(uint64_t *v, uint64_t old, uint64_t _new)
+{
+	return __sync_val_compare_and_swap(v, old, _new);
+}
+#  elif (defined(__amd64__) || defined(__x86_64__))
+ATOMIC_INLINE uint64_t atomic_add_uint64(uint64_t *p, uint64_t x)
+{
+	asm volatile (
+	    "lock; xaddq %0, %1;"
+	    : "+r" (x), "=m" (*p) /* Outputs. */
+	    : "m" (*p) /* Inputs. */
+	    );
+	return x;
+}
+
+ATOMIC_INLINE uint64_t atomic_sub_uint64(uint64_t *p, uint64_t x)
+{
+	x = (uint64_t)(-(int64_t)x);
+	asm volatile (
+	    "lock; xaddq %0, %1;"
+	    : "+r" (x), "=m" (*p) /* Outputs. */
+	    : "m" (*p) /* Inputs. */
+	    );
+	return x;
+}
+
+ATOMIC_INLINE uint64_t atomic_cas_uint64(uint64_t *v, uint64_t old, uint64_t _new)
+{
+	uint64_t ret;
+	asm volatile (
+	    "lock; cmpxchgq %2,%1"
+	    : "=a" (ret), "+m" (*v)
+	    : "r" (_new), "0" (old)
+	    : "memory");
+	return ret;
+}
+#  else
+#    error "Missing implementation for 64-bit atomic operations"
+#  endif
+#endif
+
+/******************************************************************************/
+/* 32-bit operations. */
+#if (defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_4) || defined(JE_FORCE_SYNC_COMPARE_AND_SWAP_4))
+ATOMIC_INLINE uint32_t atomic_add_uint32(uint32_t *p, uint32_t x)
+{
+	return __sync_add_and_fetch(p, x);
+}
+
+ATOMIC_INLINE uint32_t atomic_sub_uint32(uint32_t *p, uint32_t x)
+{
+	return __sync_sub_and_fetch(p, x);
+}
+
+ATOMIC_INLINE uint32_t atomic_cas_uint32(uint32_t *v, uint32_t old, uint32_t _new)
+{
+   return __sync_val_compare_and_swap(v, old, _new);
+}
+#elif (defined(__i386__) || defined(__amd64__) || defined(__x86_64__))
+ATOMIC_INLINE uint32_t atomic_add_uint32(uint32_t *p, uint32_t x)
+{
+	uint32_t ret = x;
+	asm volatile (
+	    "lock; xaddl %0, %1;"
+	    : "+r" (ret), "=m" (*p) /* Outputs. */
+	    : "m" (*p) /* Inputs. */
+	    );
+	return ret+x;
+}
+
+ATOMIC_INLINE uint32_t atomic_sub_uint32(uint32_t *p, uint32_t x)
+{
+	ret = (uint32_t)(-(int32_t)x);
+	asm volatile (
+	    "lock; xaddl %0, %1;"
+	    : "+r" (ret), "=m" (*p) /* Outputs. */
+	    : "m" (*p) /* Inputs. */
+	    );
+	return ret-x;
+}
+
+ATOMIC_INLINE uint32_t atomic_cas_uint32(uint32_t *v, uint32_t old, uint32_t _new)
+{
+	uint32_t ret;
+	asm volatile (
+	    "lock; cmpxchgl %2,%1"
+	    : "=a" (ret), "+m" (*v)
+	    : "r" (_new), "0" (old)
+	    : "memory");
+	return ret;
+}
+#else
+#  error "Missing implementation for 32-bit atomic operations"
+#endif
+
+#if (defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_4) || defined(JE_FORCE_SYNC_COMPARE_AND_SWAP_4))
+ATOMIC_INLINE uint32_t atomic_fetch_and_add_uint32(uint32_t *p, uint32_t x)
+{
+	return __sync_fetch_and_add(p, x);
+}
+
+#else
+#  error "Missing implementation for 32-bit atomic operations"
+#endif
+
+/******************************************************************************/
+/* 8-bit operations. */
+#if (defined(__GCC_HAVE_SYNC_COMPARE_AND_SWAP_1) || defined(JE_FORCE_SYNC_COMPARE_AND_SWAP_1))
+ATOMIC_INLINE uint8_t atomic_fetch_and_and_uint8(uint8_t *p, uint8_t b)
+{
+	return __sync_fetch_and_and(p, b);
+}
+ATOMIC_INLINE uint8_t atomic_fetch_and_or_uint8(uint8_t *p, uint8_t b)
+{
+	return __sync_fetch_and_or(p, b);
+}
+#else
+#  error "Missing implementation for 8-bit atomic operations"
+#endif
+
+#endif /* __ATOMIC_OPS_UNIX_H__ */

intern/atomic/intern/atomic_ops_utils.h

@@ -0,0 +1,110 @@
+/*
+ * Original code from jemalloc with this license:
+ *
+ * Copyright (C) 2002-2013 Jason Evans <jasone@canonware.com>.
+ * All rights reserved.
+ * Copyright (C) 2007-2012 Mozilla Foundation.  All rights reserved.
+ * Copyright (C) 2009-2013 Facebook, Inc.  All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ * 1. Redistributions of source code must retain the above copyright notice(s),
+ *    this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice(s),
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY EXPRESS
+ * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO
+ * EVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+ * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+ * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
+ * OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
+ * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ * ***** 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.
+ *
+ * The Original Code is Copyright (C) 2016 Blender Foundation.
+ * All rights reserved.
+ *
+ * The Original Code is: adapted from jemalloc.
+ *
+ * ***** END GPL LICENSE BLOCK *****
+ */
+
+#ifndef __ATOMIC_OPS_UTILS_H__
+#define __ATOMIC_OPS_UTILS_H__
+
+/* needed for int types */
+#include "../../../source/blender/blenlib/BLI_sys_types.h"
+#include <stdlib.h>
+#include <limits.h>
+
+#include <assert.h>
+
+/* little macro so inline keyword works */
+#if defined(_MSC_VER)
+#  define ATOMIC_INLINE static __forceinline
+#else
+#  if (defined(__APPLE__) && defined(__ppc__))
+/* static inline __attribute__ here breaks osx ppc gcc42 build */
+#    define ATOMIC_INLINE static __attribute__((always_inline))
+#  else
+#    define ATOMIC_INLINE static inline __attribute__((always_inline))
+#  endif
+#endif
+
+#ifndef LIKELY
+#  ifdef __GNUC__
+#    define LIKELY(x)       __builtin_expect(!!(x), 1)
+#    define UNLIKELY(x)     __builtin_expect(!!(x), 0)
+#  else
+#    define LIKELY(x)       (x)
+#    define UNLIKELY(x)     (x)
+#  endif
+#endif
+
+#ifdef UINTPTR_MAX
+#  if (UINTPTR_MAX == 0xFFFFFFFF)
+#    define LG_SIZEOF_PTR 4
+#  elif (UINTPTR_MAX == 0xFFFFFFFFFFFFFFFF)
+#    define LG_SIZEOF_PTR 8
+#  endif
+#elif defined(__WORDSIZE)  /* Fallback for older glibc and cpp */
+#  if (__WORDSIZE == 32)
+#    define LG_SIZEOF_PTR 4
+#  elif (__WORDSIZE == 64)
+#    define LG_SIZEOF_PTR 8
+#  endif
+#endif
+
+#ifndef LG_SIZEOF_PTR
+#  error "Cannot find pointer size"
+#endif
+
+#if (UINT_MAX == 0xFFFFFFFF)
+#  define LG_SIZEOF_INT 4
+#elif (UINT_MAX == 0xFFFFFFFFFFFFFFFF)
+#  define LG_SIZEOF_INT 8
+#else
+#  error "Cannot find int size"
+#endif
+
+#endif /* __ATOMIC_OPS_UTILS_H__ */

intern/cycles/CMakeLists.txt

@@ -153,7 +153,9 @@ set(WITH_CYCLES_DEVICE_MULTI TRUE)
 if(CYCLES_STANDALONE_REPOSITORY)
 	TEST_UNORDERED_MAP_SUPPORT()
 endif()
-if(HAVE_STD_UNORDERED_MAP_HEADER)
+if(WITH_CXX11)
+	add_definitions(-DCYCLES_STD_UNORDERED_MAP)
+elseif(HAVE_STD_UNORDERED_MAP_HEADER)
 	if(HAVE_UNORDERED_MAP_IN_STD_NAMESPACE)
 		add_definitions(-DCYCLES_STD_UNORDERED_MAP)
 	else()
@@ -235,6 +237,7 @@ endif()
 add_subdirectory(bvh)
 add_subdirectory(device)
 add_subdirectory(doc)
+add_subdirectory(graph)
 add_subdirectory(kernel)
 add_subdirectory(render)
 add_subdirectory(subd)

intern/cycles/app/CMakeLists.txt

@@ -1,13 +1,14 @@
 
 set(INC
 	.
+	../bvh
 	../device
+	../graph
 	../kernel
 	../kernel/svm
-	../bvh
-	../util
 	../render
 	../subd
+	../util
 )
 set(INC_SYS
 )
@@ -20,6 +21,7 @@ set(LIBRARIES
 	cycles_render
 	cycles_bvh
 	cycles_subd
+	cycles_graph
 	cycles_util
 	${BLENDER_GL_LIBRARIES}
 	${CYCLES_APP_GLEW_LIBRARY}

intern/cycles/app/cycles_xml.cpp

@@ -20,6 +20,8 @@
 #include <algorithm>
 #include <iterator>
 
+#include "node_xml.h"
+
 #include "background.h"
 #include "camera.h"
 #include "film.h"
@@ -29,6 +31,7 @@
 #include "mesh.h"
 #include "nodes.h"
 #include "object.h"
+#include "osl.h"
 #include "shader.h"
 #include "scene.h"
 
@@ -48,11 +51,11 @@ CCL_NAMESPACE_BEGIN
 
 /* XML reading state */
 
-struct XMLReadState {
+struct XMLReadState : public XMLReader {
 	Scene *scene;		/* scene pointer */
 	Transform tfm;		/* current transform state */
 	bool smooth;		/* smooth normal state */
-	int shader;			/* current shader */
+	Shader *shader;		/* current shader */
 	string base;		/* base path to current file*/
 	float dicing_rate;	/* current dicing rate */
 	Mesh::DisplacementMethod displacement_method;
@@ -60,7 +63,7 @@ struct XMLReadState {
 	XMLReadState()
 	  : scene(NULL),
 	    smooth(false),
-	    shader(0),
+	    shader(NULL),
 	    dicing_rate(0.0f),
 	    displacement_method(Mesh::DISPLACE_BUMP)
 	{
@@ -70,18 +73,6 @@ struct XMLReadState {
 
 /* Attribute Reading */
 
-static bool xml_read_bool(bool *value, pugi::xml_node node, const char *name)
-{
-	pugi::xml_attribute attr = node.attribute(name);
-
-	if(attr) {
-		*value = (string_iequals(attr.value(), "true")) || (atoi(attr.value()) != 0);
-		return true;
-	}
-
-	return false;
-}
-
 static bool xml_read_int(int *value, pugi::xml_node node, const char *name)
 {
 	pugi::xml_attribute attr = node.attribute(name);
@@ -190,18 +181,6 @@ static bool xml_read_string(string *str, pugi::xml_node node, const char *name)
 	return false;
 }
 
-static bool xml_read_ustring(ustring *str, pugi::xml_node node, const char *name)
-{
-	pugi::xml_attribute attr = node.attribute(name);
-
-	if(attr) {
-		*str = ustring(attr.value());
-		return true;
-	}
-
-	return false;
-}
-
 static bool xml_equal_string(pugi::xml_node node, const char *name, const char *value)
 {
 	pugi::xml_attribute attr = node.attribute(name);
@@ -212,159 +191,16 @@ static bool xml_equal_string(pugi::xml_node node, const char *name, const char *
 	return false;
 }
 
-static bool xml_read_enum(ustring *str, ShaderEnum& enm, pugi::xml_node node, const char *name)
-{
-	pugi::xml_attribute attr = node.attribute(name);
-
-	if(attr) {
-		ustring ustr(attr.value());
-
-		if(enm.exists(ustr)) {
-			*str = ustr;
-			return true;
-		}
-		else
-			fprintf(stderr, "Unknown value \"%s\" for attribute \"%s\".\n", ustr.c_str(), name);
-	}
-
-	return false;
-}
-
-static ShaderSocketType xml_read_socket_type(pugi::xml_node node, const char *name)
-{
-	pugi::xml_attribute attr = node.attribute(name);
-
-	if(attr) {
-		string value = attr.value();
-		if(string_iequals(value, "float"))
-			return SHADER_SOCKET_FLOAT;
-		else if(string_iequals(value, "int"))
-			return SHADER_SOCKET_INT;
-		else if(string_iequals(value, "color"))
-			return SHADER_SOCKET_COLOR;
-		else if(string_iequals(value, "vector"))
-			return SHADER_SOCKET_VECTOR;
-		else if(string_iequals(value, "point"))
-			return SHADER_SOCKET_POINT;
-		else if(string_iequals(value, "normal"))
-			return SHADER_SOCKET_NORMAL;
-		else if(string_iequals(value, "closure color"))
-			return SHADER_SOCKET_CLOSURE;
-		else if(string_iequals(value, "string"))
-			return SHADER_SOCKET_STRING;
-		else
-			fprintf(stderr, "Unknown shader socket type \"%s\" for attribute \"%s\".\n", value.c_str(), name);
-	}
-	
-	return SHADER_SOCKET_UNDEFINED;
-}
-
-/* Film */
-
-static void xml_read_film(const XMLReadState& state, pugi::xml_node node)
-{
-	Film *film = state.scene->film;
-	
-	xml_read_float(&film->exposure, node, "exposure");
-
-	/* ToDo: Filter Type */
-	xml_read_float(&film->filter_width, node, "filter_width");
-}
-
-/* Integrator */
-
-static void xml_read_integrator(const XMLReadState& state, pugi::xml_node node)
-{
-	Integrator *integrator = state.scene->integrator;
-	
-	/* Branched Path */
-	bool branched = false;
-	xml_read_bool(&branched, node, "branched");
-
-	if(branched) {
-		integrator->method = Integrator::BRANCHED_PATH;
-
-		xml_read_int(&integrator->diffuse_samples, node, "diffuse_samples");
-		xml_read_int(&integrator->glossy_samples, node, "glossy_samples");
-		xml_read_int(&integrator->transmission_samples, node, "transmission_samples");
-		xml_read_int(&integrator->ao_samples, node, "ao_samples");
-		xml_read_int(&integrator->mesh_light_samples, node, "mesh_light_samples");
-		xml_read_int(&integrator->subsurface_samples, node, "subsurface_samples");
-		xml_read_int(&integrator->volume_samples, node, "volume_samples");
-		xml_read_bool(&integrator->sample_all_lights_direct, node, "sample_all_lights_direct");
-		xml_read_bool(&integrator->sample_all_lights_indirect, node, "sample_all_lights_indirect");
-	}
-	
-	/* Bounces */
-	xml_read_int(&integrator->min_bounce, node, "min_bounce");
-	xml_read_int(&integrator->max_bounce, node, "max_bounce");
-	
-	xml_read_int(&integrator->max_diffuse_bounce, node, "max_diffuse_bounce");
-	xml_read_int(&integrator->max_glossy_bounce, node, "max_glossy_bounce");
-	xml_read_int(&integrator->max_transmission_bounce, node, "max_transmission_bounce");
-	xml_read_int(&integrator->max_volume_bounce, node, "max_volume_bounce");
-	
-	/* Transparency */
-	xml_read_int(&integrator->transparent_min_bounce, node, "transparent_min_bounce");
-	xml_read_int(&integrator->transparent_max_bounce, node, "transparent_max_bounce");
-	xml_read_bool(&integrator->transparent_shadows, node, "transparent_shadows");
-	
-	/* Volume */
-	xml_read_float(&integrator->volume_step_size, node, "volume_step_size");
-	xml_read_int(&integrator->volume_max_steps, node, "volume_max_steps");
-	
-	/* Various Settings */
-	xml_read_bool(&integrator->caustics_reflective, node, "caustics_reflective");
-	xml_read_bool(&integrator->caustics_refractive, node, "caustics_refractive");
-	xml_read_float(&integrator->filter_glossy, node, "filter_glossy");
-	
-	xml_read_int(&integrator->seed, node, "seed");
-	xml_read_float(&integrator->sample_clamp_direct, node, "sample_clamp_direct");
-	xml_read_float(&integrator->sample_clamp_indirect, node, "sample_clamp_indirect");
-}
-
 /* Camera */
 
-static void xml_read_camera(const XMLReadState& state, pugi::xml_node node)
+static void xml_read_camera(XMLReadState& state, pugi::xml_node node)
 {
 	Camera *cam = state.scene->camera;
 
 	xml_read_int(&cam->width, node, "width");
 	xml_read_int(&cam->height, node, "height");
 
-	if(xml_read_float(&cam->fov, node, "fov"))
-		cam->fov = DEG2RADF(cam->fov);
-
-	xml_read_float(&cam->nearclip, node, "nearclip");
-	xml_read_float(&cam->farclip, node, "farclip");
-	xml_read_float(&cam->aperturesize, node, "aperturesize"); // 0.5*focallength/fstop
-	xml_read_float(&cam->focaldistance, node, "focaldistance");
-	xml_read_float(&cam->shuttertime, node, "shuttertime");
-	xml_read_float(&cam->aperture_ratio, node, "aperture_ratio");
-
-	if(xml_equal_string(node, "type", "orthographic"))
-		cam->type = CAMERA_ORTHOGRAPHIC;
-	else if(xml_equal_string(node, "type", "perspective"))
-		cam->type = CAMERA_PERSPECTIVE;
-	else if(xml_equal_string(node, "type", "panorama"))
-		cam->type = CAMERA_PANORAMA;
-
-	if(xml_equal_string(node, "panorama_type", "equirectangular"))
-		cam->panorama_type = PANORAMA_EQUIRECTANGULAR;
-	else if(xml_equal_string(node, "panorama_type", "fisheye_equidistant"))
-		cam->panorama_type = PANORAMA_FISHEYE_EQUIDISTANT;
-	else if(xml_equal_string(node, "panorama_type", "fisheye_equisolid"))
-		cam->panorama_type = PANORAMA_FISHEYE_EQUISOLID;
-
-	xml_read_float(&cam->fisheye_fov, node, "fisheye_fov");
-	xml_read_float(&cam->fisheye_lens, node, "fisheye_lens");
-
-	xml_read_bool(&cam->use_spherical_stereo, node, "use_spherical_stereo");
-	xml_read_float(&cam->interocular_distance, node, "interocular_distance");
-	xml_read_float(&cam->convergence_distance, node, "convergence_distance");
-
-	xml_read_float(&cam->sensorwidth, node, "sensorwidth");
-	xml_read_float(&cam->sensorheight, node, "sensorheight");
+	xml_read_node(state, cam, node);
 
 	cam->matrix = state.tfm;
 
@@ -385,364 +221,20 @@ static string xml_socket_name(const char *name)
 	return sname;
 }
 
-static void xml_read_shader_graph(const XMLReadState& state, Shader *shader, pugi::xml_node graph_node)
+static void xml_read_shader_graph(XMLReadState& state, Shader *shader, pugi::xml_node graph_node)
 {
+	xml_read_node(state, shader, graph_node);
+
 	ShaderGraph *graph = new ShaderGraph();
 
-	map<string, ShaderNode*> nodemap;
-
-	nodemap["output"] = graph->output();
+	/* local state, shader nodes can't link to nodes outside the shader graph */
+	XMLReader graph_reader;
+	graph_reader.node_map[ustring("output")] = graph->output();
 
 	for(pugi::xml_node node = graph_node.first_child(); node; node = node.next_sibling()) {
-		ShaderNode *snode = NULL;
+		ustring node_name(node.name());
 
-		/* ToDo: Add missing nodes
-		 * RGBCurvesNode, VectorCurvesNode, RGBRampNode and ConvertNode (RGB -> BW).
-		 */
-
-		if(string_iequals(node.name(), "image_texture")) {
-			ImageTextureNode *img = new ImageTextureNode();
-
-			xml_read_string(&img->filename, node, "src");
-			img->filename = path_join(state.base, img->filename);
-			
-			xml_read_enum(&img->color_space, ImageTextureNode::color_space_enum, node, "color_space");
-			xml_read_enum(&img->projection, ImageTextureNode::projection_enum, node, "projection");
-			xml_read_float(&img->projection_blend, node, "projection_blend");
-
-			/* ToDo: Interpolation */
-
-			snode = img;
-		}
-		else if(string_iequals(node.name(), "environment_texture")) {
-			EnvironmentTextureNode *env = new EnvironmentTextureNode();
-
-			xml_read_string(&env->filename, node, "src");
-			env->filename = path_join(state.base, env->filename);
-			
-			xml_read_enum(&env->color_space, EnvironmentTextureNode::color_space_enum, node, "color_space");
-			xml_read_enum(&env->projection, EnvironmentTextureNode::projection_enum, node, "projection");
-
-			snode = env;
-		}
-		else if(string_iequals(node.name(), "osl_shader")) {
-			OSLScriptNode *osl = new OSLScriptNode();
-
-			/* Source */
-			xml_read_string(&osl->filepath, node, "src");
-			if(path_is_relative(osl->filepath)) {
-				osl->filepath = path_join(state.base, osl->filepath);
-			}
-
-			/* Generate inputs/outputs from node sockets
-			 *
-			 * Note: ShaderInput/ShaderOutput store shallow string copies only!
-			 * Socket names must be stored in the extra lists instead. */
-			/* read input values */
-			for(pugi::xml_node param = node.first_child(); param; param = param.next_sibling()) {
-				if(string_iequals(param.name(), "input")) {
-					string name;
-					if(!xml_read_string(&name, param, "name"))
-						continue;
-					
-					ShaderSocketType type = xml_read_socket_type(param, "type");
-					if(type == SHADER_SOCKET_UNDEFINED)
-						continue;
-					
-					osl->input_names.push_back(ustring(name));
-					osl->add_input(osl->input_names.back().c_str(), type);
-				}
-				else if(string_iequals(param.name(), "output")) {
-					string name;
-					if(!xml_read_string(&name, param, "name"))
-						continue;
-					
-					ShaderSocketType type = xml_read_socket_type(param, "type");
-					if(type == SHADER_SOCKET_UNDEFINED)
-						continue;
-					
-					osl->output_names.push_back(ustring(name));
-					osl->add_output(osl->output_names.back().c_str(), type);
-				}
-			}
-			
-			snode = osl;
-		}
-		else if(string_iequals(node.name(), "sky_texture")) {
-			SkyTextureNode *sky = new SkyTextureNode();
-			
-			xml_read_enum(&sky->type, SkyTextureNode::type_enum, node, "type");
-			xml_read_float3(&sky->sun_direction, node, "sun_direction");
-			xml_read_float(&sky->turbidity, node, "turbidity");
-			xml_read_float(&sky->ground_albedo, node, "ground_albedo");
-			
-			snode = sky;
-		}
-		else if(string_iequals(node.name(), "noise_texture")) {
-			snode = new NoiseTextureNode();
-		}
-		else if(string_iequals(node.name(), "checker_texture")) {
-			snode = new CheckerTextureNode();
-		}
-		else if(string_iequals(node.name(), "brick_texture")) {
-			BrickTextureNode *brick = new BrickTextureNode();
-
-			xml_read_float(&brick->offset, node, "offset");
-			xml_read_int(&brick->offset_frequency, node, "offset_frequency");
-			xml_read_float(&brick->squash, node, "squash");
-			xml_read_int(&brick->squash_frequency, node, "squash_frequency");
-
-			snode = brick;
-		}
-		else if(string_iequals(node.name(), "gradient_texture")) {
-			GradientTextureNode *blend = new GradientTextureNode();
-			xml_read_enum(&blend->type, GradientTextureNode::type_enum, node, "type");
-			snode = blend;
-		}
-		else if(string_iequals(node.name(), "voronoi_texture")) {
-			VoronoiTextureNode *voronoi = new VoronoiTextureNode();
-			xml_read_enum(&voronoi->coloring, VoronoiTextureNode::coloring_enum, node, "coloring");
-			snode = voronoi;
-		}
-		else if(string_iequals(node.name(), "musgrave_texture")) {
-			MusgraveTextureNode *musgrave = new MusgraveTextureNode();
-			xml_read_enum(&musgrave->type, MusgraveTextureNode::type_enum, node, "type");
-			snode = musgrave;
-		}
-		else if(string_iequals(node.name(), "magic_texture")) {
-			MagicTextureNode *magic = new MagicTextureNode();
-			xml_read_int(&magic->depth, node, "depth");
-			snode = magic;
-		}
-		else if(string_iequals(node.name(), "wave_texture")) {
-			WaveTextureNode *wave = new WaveTextureNode();
-			xml_read_enum(&wave->type, WaveTextureNode::type_enum, node, "type");
-			xml_read_enum(&wave->profile, WaveTextureNode::profile_enum, node, "profile");
-			snode = wave;
-		}
-		else if(string_iequals(node.name(), "normal")) {
-			NormalNode *normal = new NormalNode();
-			xml_read_float3(&normal->direction, node, "direction");
-			snode = normal;
-		}
-		else if(string_iequals(node.name(), "bump")) {
-			BumpNode *bump = new BumpNode();
-			xml_read_bool(&bump->invert, node, "invert");
-			snode = bump;
-		}
-		else if(string_iequals(node.name(), "mapping")) {
-			snode = new MappingNode();
-		}
-		else if(string_iequals(node.name(), "anisotropic_bsdf")) {
-			AnisotropicBsdfNode *aniso = new AnisotropicBsdfNode();
-			xml_read_enum(&aniso->distribution, AnisotropicBsdfNode::distribution_enum, node, "distribution");
-			snode = aniso;
-		}
-		else if(string_iequals(node.name(), "diffuse_bsdf")) {
-			snode = new DiffuseBsdfNode();
-		}
-		else if(string_iequals(node.name(), "translucent_bsdf")) {
-			snode = new TranslucentBsdfNode();
-		}
-		else if(string_iequals(node.name(), "transparent_bsdf")) {
-			snode = new TransparentBsdfNode();
-		}
-		else if(string_iequals(node.name(), "velvet_bsdf")) {
-			snode = new VelvetBsdfNode();
-		}
-		else if(string_iequals(node.name(), "toon_bsdf")) {
-			ToonBsdfNode *toon = new ToonBsdfNode();
-			xml_read_enum(&toon->component, ToonBsdfNode::component_enum, node, "component");
-			snode = toon;
-		}
-		else if(string_iequals(node.name(), "glossy_bsdf")) {
-			GlossyBsdfNode *glossy = new GlossyBsdfNode();
-			xml_read_enum(&glossy->distribution, GlossyBsdfNode::distribution_enum, node, "distribution");
-			snode = glossy;
-		}
-		else if(string_iequals(node.name(), "glass_bsdf")) {
-			GlassBsdfNode *diel = new GlassBsdfNode();
-			xml_read_enum(&diel->distribution, GlassBsdfNode::distribution_enum, node, "distribution");
-			snode = diel;
-		}
-		else if(string_iequals(node.name(), "refraction_bsdf")) {
-			RefractionBsdfNode *diel = new RefractionBsdfNode();
-			xml_read_enum(&diel->distribution, RefractionBsdfNode::distribution_enum, node, "distribution");
-			snode = diel;
-		}
-		else if(string_iequals(node.name(), "hair_bsdf")) {
-			HairBsdfNode *hair = new HairBsdfNode();
-			xml_read_enum(&hair->component, HairBsdfNode::component_enum, node, "component");
-			snode = hair;
-		}
-		else if(string_iequals(node.name(), "emission")) {
-			snode = new EmissionNode();
-		}
-		else if(string_iequals(node.name(), "ambient_occlusion")) {
-			snode = new AmbientOcclusionNode();
-		}
-		else if(string_iequals(node.name(), "background")) {
-			snode = new BackgroundNode();
-		}
-		else if(string_iequals(node.name(), "holdout")) {
-			snode = new HoldoutNode();
-		}
-		else if(string_iequals(node.name(), "absorption_volume")) {
-			snode = new AbsorptionVolumeNode();
-		}
-		else if(string_iequals(node.name(), "scatter_volume")) {
-			snode = new ScatterVolumeNode();
-		}
-		else if(string_iequals(node.name(), "subsurface_scattering")) {
-			SubsurfaceScatteringNode *sss = new SubsurfaceScatteringNode();
-
-			string falloff;
-			xml_read_string(&falloff, node, "falloff");
-			if(falloff == "cubic")
-				sss->closure = CLOSURE_BSSRDF_CUBIC_ID;
-			else if(falloff == "gaussian")
-				sss->closure = CLOSURE_BSSRDF_GAUSSIAN_ID;
-			else /*if(falloff == "burley")*/
-				sss->closure = CLOSURE_BSSRDF_BURLEY_ID;
-
-			snode = sss;
-		}
-		else if(string_iequals(node.name(), "geometry")) {
-			snode = new GeometryNode();
-		}
-		else if(string_iequals(node.name(), "texture_coordinate")) {
-			snode = new TextureCoordinateNode();
-		}
-		else if(string_iequals(node.name(), "light_path")) {
-			snode = new LightPathNode();
-		}
-		else if(string_iequals(node.name(), "light_falloff")) {
-			snode = new LightFalloffNode();
-		}
-		else if(string_iequals(node.name(), "object_info")) {
-			snode = new ObjectInfoNode();
-		}
-		else if(string_iequals(node.name(), "particle_info")) {
-			snode = new ParticleInfoNode();
-		}
-		else if(string_iequals(node.name(), "hair_info")) {
-			snode = new HairInfoNode();
-		}
-		else if(string_iequals(node.name(), "value")) {
-			ValueNode *value = new ValueNode();
-			xml_read_float(&value->value, node, "value");
-			snode = value;
-		}
-		else if(string_iequals(node.name(), "color")) {
-			ColorNode *color = new ColorNode();
-			xml_read_float3(&color->value, node, "value");
-			snode = color;
-		}
-		else if(string_iequals(node.name(), "mix_closure")) {
-			snode = new MixClosureNode();
-		}
-		else if(string_iequals(node.name(), "add_closure")) {
-			snode = new AddClosureNode();
-		}
-		else if(string_iequals(node.name(), "invert")) {
-			snode = new InvertNode();
-		}
-		else if(string_iequals(node.name(), "mix")) {
-			/* ToDo: Tag Mix case for optimization */
-			MixNode *mix = new MixNode();
-			xml_read_enum(&mix->type, MixNode::type_enum, node, "type");
-			xml_read_bool(&mix->use_clamp, node, "use_clamp");
-			snode = mix;
-		}
-		else if(string_iequals(node.name(), "gamma")) {
-			snode = new GammaNode();
-		}
-		else if(string_iequals(node.name(), "brightness")) {
-			snode = new BrightContrastNode();
-		}
-		else if(string_iequals(node.name(), "combine_rgb")) {
-			snode = new CombineRGBNode();
-		}
-		else if(string_iequals(node.name(), "separate_rgb")) {
-			snode = new SeparateRGBNode();
-		}
-		else if(string_iequals(node.name(), "combine_hsv")) {
-			snode = new CombineHSVNode();
-		}
-		else if(string_iequals(node.name(), "separate_hsv")) {
-			snode = new SeparateHSVNode();
-		}
-		else if(string_iequals(node.name(), "combine_xyz")) {
-			snode = new CombineXYZNode();
-		}
-		else if(string_iequals(node.name(), "separate_xyz")) {
-			snode = new SeparateXYZNode();
-		}
-		else if(string_iequals(node.name(), "hsv")) {
-			snode = new HSVNode();
-		}
-		else if(string_iequals(node.name(), "wavelength")) {
-			snode = new WavelengthNode();
-		}
-		else if(string_iequals(node.name(), "blackbody")) {
-			snode = new BlackbodyNode();
-		}
-		else if(string_iequals(node.name(), "attribute")) {
-			AttributeNode *attr = new AttributeNode();
-			xml_read_ustring(&attr->attribute, node, "attribute");
-			snode = attr;
-		}
-		else if(string_iequals(node.name(), "uv_map")) {
-			UVMapNode *uvm = new UVMapNode();
-			xml_read_ustring(&uvm->attribute, node, "uv_map");
-			snode = uvm;
-		}
-		else if(string_iequals(node.name(), "camera")) {
-			snode = new CameraNode();
-		}
-		else if(string_iequals(node.name(), "fresnel")) {
-			snode = new FresnelNode();
-		}
-		else if(string_iequals(node.name(), "layer_weight")) {
-			snode = new LayerWeightNode();
-		}
-		else if(string_iequals(node.name(), "wireframe")) {
-			WireframeNode *wire = new WireframeNode;
-			xml_read_bool(&wire->use_pixel_size, node, "use_pixel_size");
-			snode = wire;
-		}
-		else if(string_iequals(node.name(), "normal_map")) {
-			NormalMapNode *nmap = new NormalMapNode;
-			xml_read_ustring(&nmap->attribute, node, "attribute");
-			xml_read_enum(&nmap->space, NormalMapNode::space_enum, node, "space");
-			snode = nmap;
-		}
-		else if(string_iequals(node.name(), "tangent")) {
-			TangentNode *tangent = new TangentNode;
-			xml_read_ustring(&tangent->attribute, node, "attribute");
-			xml_read_enum(&tangent->direction_type, TangentNode::direction_type_enum, node, "direction_type");
-			xml_read_enum(&tangent->axis, TangentNode::axis_enum, node, "axis");
-			snode = tangent;
-		}
-		else if(string_iequals(node.name(), "math")) {
-			MathNode *math = new MathNode();
-			xml_read_enum(&math->type, MathNode::type_enum, node, "type");
-			xml_read_bool(&math->use_clamp, node, "use_clamp");
-			snode = math;
-		}
-		else if(string_iequals(node.name(), "vector_math")) {
-			VectorMathNode *vmath = new VectorMathNode();
-			xml_read_enum(&vmath->type, VectorMathNode::type_enum, node, "type");
-			snode = vmath;
-		}
-		else if(string_iequals(node.name(), "vector_transform")) {
-			VectorTransformNode *vtransform = new VectorTransformNode();
-			xml_read_enum(&vtransform->type, VectorTransformNode::type_enum, node, "type");
-			xml_read_enum(&vtransform->convert_from, VectorTransformNode::convert_space_enum, node, "convert_from");
-			xml_read_enum(&vtransform->convert_to, VectorTransformNode::convert_space_enum, node, "convert_to");
-			snode = vtransform;
-		}
-		else if(string_iequals(node.name(), "connect")) {
+		if(node_name == "connect") {
 			/* connect nodes */
 			vector<string> from_tokens, to_tokens;
 
@@ -750,35 +242,40 @@ static void xml_read_shader_graph(const XMLReadState& state, Shader *shader, pug
 			string_split(to_tokens, node.attribute("to").value());
 
 			if(from_tokens.size() == 2 && to_tokens.size() == 2) {
+				ustring from_node_name(from_tokens[0]);
+				ustring from_socket_name(from_tokens[1]);
+				ustring to_node_name(to_tokens[0]);
+				ustring to_socket_name(to_tokens[1]);
+
 				/* find nodes and sockets */
 				ShaderOutput *output = NULL;
 				ShaderInput *input = NULL;
 
-				if(nodemap.find(from_tokens[0]) != nodemap.end()) {
-					ShaderNode *fromnode = nodemap[from_tokens[0]];
+				if(graph_reader.node_map.find(from_node_name) != graph_reader.node_map.end()) {
+					ShaderNode *fromnode = (ShaderNode*)graph_reader.node_map[from_node_name];
 
 					foreach(ShaderOutput *out, fromnode->outputs)
-						if(string_iequals(xml_socket_name(out->name), from_tokens[1]))
+						if(string_iequals(xml_socket_name(out->name().c_str()), from_socket_name.c_str()))
 							output = out;
 
 					if(!output)
-						fprintf(stderr, "Unknown output socket name \"%s\" on \"%s\".\n", from_tokens[1].c_str(), from_tokens[0].c_str());
+						fprintf(stderr, "Unknown output socket name \"%s\" on \"%s\".\n", from_node_name.c_str(), from_socket_name.c_str());
 				}
 				else
-					fprintf(stderr, "Unknown shader node name \"%s\".\n", from_tokens[0].c_str());
+					fprintf(stderr, "Unknown shader node name \"%s\".\n", from_node_name.c_str());
 
-				if(nodemap.find(to_tokens[0]) != nodemap.end()) {
-					ShaderNode *tonode = nodemap[to_tokens[0]];
+				if(graph_reader.node_map.find(to_node_name) != graph_reader.node_map.end()) {
+					ShaderNode *tonode = (ShaderNode*)graph_reader.node_map[to_node_name];
 
 					foreach(ShaderInput *in, tonode->inputs)
-						if(string_iequals(xml_socket_name(in->name), to_tokens[1]))
+						if(string_iequals(xml_socket_name(in->name().c_str()), to_socket_name.c_str()))
 							input = in;
 
 					if(!input)
-						fprintf(stderr, "Unknown input socket name \"%s\" on \"%s\".\n", to_tokens[1].c_str(), to_tokens[0].c_str());
+						fprintf(stderr, "Unknown input socket name \"%s\" on \"%s\".\n", to_socket_name.c_str(), to_node_name.c_str());
 				}
 				else
-					fprintf(stderr, "Unknown shader node name \"%s\".\n", to_tokens[0].c_str());
+					fprintf(stderr, "Unknown shader node name \"%s\".\n", to_node_name.c_str());
 
 				/* connect */
 				if(output && input)
@@ -786,44 +283,76 @@ static void xml_read_shader_graph(const XMLReadState& state, Shader *shader, pug
 			}
 			else
 				fprintf(stderr, "Invalid from or to value for connect node.\n");
+
+			continue;
+		}
+
+		ShaderNode *snode = NULL;
+
+#ifdef WITH_OSL
+		if(node_name == "osl_shader") {
+			ShaderManager *manager = state.scene->shader_manager;
+
+			if(manager->use_osl()) {
+				std::string filepath;
+
+				if(xml_read_string(&filepath, node, "src")) {
+					if(path_is_relative(filepath)) {
+						filepath = path_join(state.base, filepath);
+					}
+
+					snode = ((OSLShaderManager*)manager)->osl_node(filepath);
+
+					if(!snode) {
+						fprintf(stderr, "Failed to create OSL node from \"%s\".\n", filepath.c_str());
+						continue;
+					}
+				}
+				else {
+					fprintf(stderr, "OSL node missing \"src\" attribute.\n");
+					continue;
+				}
+			}
+			else {
+				fprintf(stderr, "OSL node without using --shadingsys osl.\n");
+				continue;
+			}
 		}
 		else
-			fprintf(stderr, "Unknown shader node \"%s\".\n", node.name());
+#endif
+		{
+			/* exception for name collision */
+			if(node_name == "background")
+				node_name = "background_shader";
+
+			const NodeType *node_type = NodeType::find(node_name);
+
+			if(!node_type) {
+				fprintf(stderr, "Unknown shader node \"%s\".\n", node.name());
+				continue;
+			}
+			else if(node_type->type != NodeType::SHADER) {
+				fprintf(stderr, "Node type \"%s\" is not a shader node.\n", node_type->name.c_str());
+				continue;
+			}
+
+			snode = (ShaderNode*) node_type->create(node_type);
+		}
+
+		xml_read_node(graph_reader, snode, node);
+
+		if(node_name == "image_texture") {
+			ImageTextureNode *img = (ImageTextureNode*) snode;
+			img->filename = path_join(state.base, img->filename.string());
+		}
+		else if(node_name == "environment_texture") {
+			EnvironmentTextureNode *env = (EnvironmentTextureNode*) snode;
+			env->filename = path_join(state.base, env->filename.string());
+		}
 
 		if(snode) {
 			/* add to graph */
 			graph->add(snode);
-
-			/* add to map for name lookups */
-			string name = "";
-			xml_read_string(&name, node, "name");
-
-			nodemap[name] = snode;
-
-			/* read input values */
-			for(pugi::xml_attribute attr = node.first_attribute(); attr; attr = attr.next_attribute()) {
-				foreach(ShaderInput *in, snode->inputs) {
-					if(string_iequals(in->name, attr.name())) {
-						switch(in->type) {
-							case SHADER_SOCKET_FLOAT:
-							case SHADER_SOCKET_INT:
-								xml_read_float(&in->value.x, node, attr.name());
-								break;
-							case SHADER_SOCKET_COLOR:
-							case SHADER_SOCKET_VECTOR:
-							case SHADER_SOCKET_POINT:
-							case SHADER_SOCKET_NORMAL:
-								xml_read_float3(&in->value, node, attr.name());
-								break;
-							case SHADER_SOCKET_STRING:
-								xml_read_ustring( &in->value_string, node, attr.name() );
-								break;
-							default:
-								break;
-						}
-					}
-				}
-			}
 		}
 	}
 
@@ -831,54 +360,22 @@ static void xml_read_shader_graph(const XMLReadState& state, Shader *shader, pug
 	shader->tag_update(state.scene);
 }
 
-static void xml_read_shader(const XMLReadState& state, pugi::xml_node node)
+static void xml_read_shader(XMLReadState& state, pugi::xml_node node)
 {
 	Shader *shader = new Shader();
-
-	xml_read_string(&shader->name, node, "name");
-	xml_read_bool(&shader->use_mis, node, "use_mis");
-	xml_read_bool(&shader->use_transparent_shadow, node, "use_transparent_shadow");
-
-	/* Volume */
-	xml_read_bool(&shader->heterogeneous_volume, node, "heterogeneous_volume");
-	xml_read_int(&shader->volume_interpolation_method, node, "volume_interpolation_method");
-
-	if(xml_equal_string(node, "volume_sampling_method", "distance"))
-		shader->volume_sampling_method = VOLUME_SAMPLING_DISTANCE;
-	else if(xml_equal_string(node, "volume_sampling_method", "equiangular"))
-		shader->volume_sampling_method = VOLUME_SAMPLING_EQUIANGULAR;
-	else if(xml_equal_string(node, "volume_sampling_method", "multiple_importance"))
-		shader->volume_sampling_method = VOLUME_SAMPLING_MULTIPLE_IMPORTANCE;
-
 	xml_read_shader_graph(state, shader, node);
 	state.scene->shaders.push_back(shader);
 }
 
 /* Background */
 
-static void xml_read_background(const XMLReadState& state, pugi::xml_node node)
+static void xml_read_background(XMLReadState& state, pugi::xml_node node)
 {
 	/* Background Settings */
-	Background *bg = state.scene->background;
-
-	xml_read_float(&bg->ao_distance, node, "ao_distance");
-	xml_read_float(&bg->ao_factor, node, "ao_factor");
-
-	xml_read_bool(&bg->transparent, node, "transparent");
+	xml_read_node(state, state.scene->background, node);
 
 	/* Background Shader */
-	Shader *shader = state.scene->shaders[state.scene->default_background];
-	
-	xml_read_bool(&shader->heterogeneous_volume, node, "heterogeneous_volume");
-	xml_read_int(&shader->volume_interpolation_method, node, "volume_interpolation_method");
-
-	if(xml_equal_string(node, "volume_sampling_method", "distance"))
-		shader->volume_sampling_method = VOLUME_SAMPLING_DISTANCE;
-	else if(xml_equal_string(node, "volume_sampling_method", "equiangular"))
-		shader->volume_sampling_method = VOLUME_SAMPLING_EQUIANGULAR;
-	else if(xml_equal_string(node, "volume_sampling_method", "multiple_importance"))
-		shader->volume_sampling_method = VOLUME_SAMPLING_MULTIPLE_IMPORTANCE;
-
+	Shader *shader = state.scene->default_background;
 	xml_read_shader_graph(state, shader, node);
 }
 
@@ -906,7 +403,7 @@ static void xml_read_mesh(const XMLReadState& state, pugi::xml_node node)
 	mesh->used_shaders.push_back(state.shader);
 
 	/* read state */
-	int shader = state.shader;
+	int shader = 0;
 	bool smooth = state.smooth;
 
 	mesh->displacement_method = state.displacement_method;
@@ -967,6 +464,11 @@ static void xml_read_mesh(const XMLReadState& state, pugi::xml_node node)
 		/* create vertices */
 		mesh->verts = P;
 
+		size_t num_triangles = 0;
+		for(size_t i = 0; i < nverts.size(); i++)
+			num_triangles += nverts[i]-2;
+		mesh->reserve_mesh(mesh->verts.size(), num_triangles);
+
 		/* create triangles */
 		int index_offset = 0;
 
@@ -995,9 +497,9 @@ static void xml_read_mesh(const XMLReadState& state, pugi::xml_node node)
 			index_offset = 0;
 			for(size_t i = 0; i < nverts.size(); i++) {
 				for(int j = 0; j < nverts[i]-2; j++) {
-					int v0 = verts[index_offset];
-					int v1 = verts[index_offset + j + 1];
-					int v2 = verts[index_offset + j + 2];
+					int v0 = index_offset;
+					int v1 = index_offset + j + 1;
+					int v2 = index_offset + j + 2;
 
 					assert(v0*2+1 < (int)UV.size());
 					assert(v1*2+1 < (int)UV.size());
@@ -1066,7 +568,7 @@ static void xml_read_patch(const XMLReadState& state, pugi::xml_node node)
 		mesh->used_shaders.push_back(state.shader);
 
 		/* split */
-		SubdParams sdparams(mesh, state.shader, state.smooth);
+		SubdParams sdparams(mesh, 0, state.smooth);
 		xml_read_float(&sdparams.dicing_rate, node, "dicing_rate");
 
 		DiagSplit dsplit(sdparams);
@@ -1081,47 +583,12 @@ static void xml_read_patch(const XMLReadState& state, pugi::xml_node node)
 
 /* Light */
 
-static void xml_read_light(const XMLReadState& state, pugi::xml_node node)
+static void xml_read_light(XMLReadState& state, pugi::xml_node node)
 {
 	Light *light = new Light();
+
 	light->shader = state.shader;
-
-	/* Light Type
-	 * 0: Point, 1: Sun, 3: Area, 5: Spot */
-	int type = 0;
-	xml_read_int(&type, node, "type");
-	light->type = (LightType)type;
-
-	/* Spot Light */
-	xml_read_float(&light->spot_angle, node, "spot_angle");
-	xml_read_float(&light->spot_smooth, node, "spot_smooth");
-
-	/* Area Light */
-	xml_read_float(&light->sizeu, node, "sizeu");
-	xml_read_float(&light->sizev, node, "sizev");
-	xml_read_float3(&light->axisu, node, "axisu");
-	xml_read_float3(&light->axisv, node, "axisv");
-
-	/* Portal? (Area light only) */
-	xml_read_bool(&light->is_portal, node, "is_portal");
-
-	/* Generic */
-	xml_read_float(&light->size, node, "size");
-	xml_read_float3(&light->dir, node, "dir");
-	xml_read_float3(&light->co, node, "P");
-	light->co = transform_point(&state.tfm, light->co);
-
-	/* Settings */
-	xml_read_bool(&light->cast_shadow, node, "cast_shadow");
-	xml_read_bool(&light->use_mis, node, "use_mis");
-	xml_read_int(&light->samples, node, "samples");
-	xml_read_int(&light->max_bounces, node, "max_bounces");
-
-	/* Ray Visibility */
-	xml_read_bool(&light->use_diffuse, node, "use_diffuse");
-	xml_read_bool(&light->use_glossy, node, "use_glossy");
-	xml_read_bool(&light->use_transmission, node, "use_transmission");
-	xml_read_bool(&light->use_scatter, node, "use_scatter");
+	xml_read_node(state, light, node);
 
 	state.scene->lights.push_back(light);
 }
@@ -1163,17 +630,14 @@ static void xml_read_state(XMLReadState& state, pugi::xml_node node)
 	string shadername;
 
 	if(xml_read_string(&shadername, node, "shader")) {
-		int i = 0;
 		bool found = false;
 
 		foreach(Shader *shader, state.scene->shaders) {
 			if(shader->name == shadername) {
-				state.shader = i;
+				state.shader = shader;
 				found = true;
 				break;
 			}
-
-			i++;
 		}
 
 		if(!found)
@@ -1199,16 +663,16 @@ static void xml_read_state(XMLReadState& state, pugi::xml_node node)
 
 /* Scene */
 
-static void xml_read_include(const XMLReadState& state, const string& src);
+static void xml_read_include(XMLReadState& state, const string& src);
 
-static void xml_read_scene(const XMLReadState& state, pugi::xml_node scene_node)
+static void xml_read_scene(XMLReadState& state, pugi::xml_node scene_node)
 {
 	for(pugi::xml_node node = scene_node.first_child(); node; node = node.next_sibling()) {
 		if(string_iequals(node.name(), "film")) {
-			xml_read_film(state, node);
+			xml_read_node(state, state.scene->film, node);
 		}
 		else if(string_iequals(node.name(), "integrator")) {
-			xml_read_integrator(state, node);
+			xml_read_node(state, state.scene->integrator, node);
 		}
 		else if(string_iequals(node.name(), "camera")) {
 			xml_read_camera(state, node);
@@ -1253,7 +717,7 @@ static void xml_read_scene(const XMLReadState& state, pugi::xml_node scene_node)
 
 /* Include */
 
-static void xml_read_include(const XMLReadState& state, const string& src)
+static void xml_read_include(XMLReadState& state, const string& src)
 {
 	/* open XML document */
 	pugi::xml_document doc;

intern/cycles/blender/CMakeLists.txt

@@ -1,5 +1,6 @@
 
 set(INC
+	../graph
 	../render
 	../device
 	../kernel

intern/cycles/blender/addon/properties.py

@@ -193,57 +193,57 @@ class CyclesRenderSettings(bpy.types.PropertyGroup):
         cls.aa_samples = IntProperty(
                 name="AA Samples",
                 description="Number of antialiasing samples to render for each pixel",
-                min=1, max=10000,
+                min=1, max=2097151,
                 default=4,
                 )
         cls.preview_aa_samples = IntProperty(
                 name="AA Samples",
                 description="Number of antialiasing samples to render in the viewport, unlimited if 0",
-                min=0, max=10000,
+                min=0, max=2097151,
                 default=4,
                 )
         cls.diffuse_samples = IntProperty(
                 name="Diffuse Samples",
                 description="Number of diffuse bounce samples to render for each AA sample",
-                min=1, max=10000,
+                min=1, max=1024,
                 default=1,
                 )
         cls.glossy_samples = IntProperty(
                 name="Glossy Samples",
                 description="Number of glossy bounce samples to render for each AA sample",
-                min=1, max=10000,
+                min=1, max=1024,
                 default=1,
                 )
         cls.transmission_samples = IntProperty(
                 name="Transmission Samples",
                 description="Number of transmission bounce samples to render for each AA sample",
-                min=1, max=10000,
+                min=1, max=1024,
                 default=1,
                 )
         cls.ao_samples = IntProperty(
                 name="Ambient Occlusion Samples",
                 description="Number of ambient occlusion samples to render for each AA sample",
-                min=1, max=10000,
+                min=1, max=1024,
                 default=1,
                 )
         cls.mesh_light_samples = IntProperty(
                 name="Mesh Light Samples",
                 description="Number of mesh emission light samples to render for each AA sample",
-                min=1, max=10000,
+                min=1, max=1024,
                 default=1,
                 )
 
         cls.subsurface_samples = IntProperty(
                 name="Subsurface Samples",
                 description="Number of subsurface scattering samples to render for each AA sample",
-                min=1, max=10000,
+                min=1, max=1024,
                 default=1,
                 )
 
         cls.volume_samples = IntProperty(
                 name="Volume Samples",
                 description="Number of volume scattering samples to render for each AA sample",
-                min=1, max=10000,
+                min=1, max=1024,
                 default=1,
                 )
 
@@ -359,7 +359,7 @@ class CyclesRenderSettings(bpy.types.PropertyGroup):
                 description="Distance between volume shader samples when rendering the volume "
                             "(lower values give more accurate and detailed results, but also increased render time)",
                 default=0.1,
-                min=0.0000001, max=100000.0, soft_min=0.01, soft_max=1.0
+                min=0.0000001, max=100000.0, soft_min=0.01, soft_max=1.0, precision=4
                 )
 
         cls.volume_max_steps = IntProperty(
@@ -594,6 +594,8 @@ class CyclesRenderSettings(bpy.types.PropertyGroup):
         cls.debug_use_cpu_sse2 = BoolProperty(name="SSE2", default=True)
         cls.debug_use_qbvh = BoolProperty(name="QBVH", default=True)
 
+        cls.debug_use_cuda_adaptive_compile = BoolProperty(name="Adaptive Compile", default=False)
+
         cls.debug_opencl_kernel_type = EnumProperty(
             name="OpenCL Kernel Type",
             default='DEFAULT',

intern/cycles/blender/addon/ui.py

@@ -76,9 +76,8 @@ def use_cuda(context):
 
 def use_branched_path(context):
     cscene = context.scene.cycles
-    device_type = context.user_preferences.system.compute_device_type
 
-    return (cscene.progressive == 'BRANCHED_PATH' and device_type != 'OPENCL')
+    return (cscene.progressive == 'BRANCHED_PATH' and not use_opencl(context))
 
 
 def use_sample_all_lights(context):
@@ -704,7 +703,7 @@ class Cycles_PT_mesh_displacement(CyclesButtonsPanel, Panel):
 
         col = split.column()
         sub = col.column(align=True)
-        sub.label(text="Displacment:")
+        sub.label(text="Displacement:")
         sub.prop(cdata, "displacement_method", text="")
 
         col = split.column()
@@ -1553,6 +1552,10 @@ class CyclesRender_PT_debug(CyclesButtonsPanel, Panel):
         row.prop(cscene, "debug_use_cpu_avx2", toggle=True)
         col.prop(cscene, "debug_use_qbvh")
 
+        col = layout.column()
+        col.label('CUDA Flags:')
+        col.prop(cscene, "debug_use_cuda_adaptive_compile")
+
         col = layout.column()
         col.label('OpenCL Flags:')
         col.prop(cscene, "debug_opencl_kernel_type", text="Kernel")

intern/cycles/blender/blender_camera.cpp

@@ -37,7 +37,7 @@ struct BlenderCamera {
 	float lens;
 	float shuttertime;
 	Camera::MotionPosition motion_position;
-	float shutter_curve[RAMP_TABLE_SIZE];
+	array<float> shutter_curve;
 
 	Camera::RollingShutterType rolling_shutter_type;
 	float rolling_shutter_duration;
@@ -65,6 +65,9 @@ struct BlenderCamera {
 	bool use_spherical_stereo;
 	float interocular_distance;
 	float convergence_distance;
+	bool use_pole_merge;
+	float pole_merge_angle_from;
+	float pole_merge_angle_to;
 
 	enum { AUTO, HORIZONTAL, VERTICAL } sensor_fit;
 	float sensor_width;
@@ -105,10 +108,6 @@ static void blender_camera_init(BlenderCamera *bcam,
 	/* render resolution */
 	bcam->full_width = render_resolution_x(b_render);
 	bcam->full_height = render_resolution_y(b_render);
-
-	/* pixel aspect */
-	bcam->pixelaspect.x = b_render.pixel_aspect_x();
-	bcam->pixelaspect.y = b_render.pixel_aspect_y();
 }
 
 static float blender_camera_focal_distance(BL::RenderEngine& b_engine,
@@ -183,6 +182,10 @@ static void blender_camera_from_object(BlenderCamera *bcam,
 		}
 		bcam->use_spherical_stereo = b_engine.use_spherical_stereo(b_ob);
 
+		bcam->use_pole_merge = b_camera.stereo().use_pole_merge();
+		bcam->pole_merge_angle_from = b_camera.stereo().pole_merge_angle_from();
+		bcam->pole_merge_angle_to = b_camera.stereo().pole_merge_angle_to();
+
 		bcam->ortho_scale = b_camera.ortho_scale();
 
 		bcam->lens = b_camera.lens();
@@ -427,6 +430,10 @@ static void blender_camera_sync(Camera *cam, BlenderCamera *bcam, int width, int
 			cam->stereo_eye = Camera::STEREO_NONE;
 	}
 
+	cam->use_pole_merge = bcam->use_pole_merge;
+	cam->pole_merge_angle_from = bcam->pole_merge_angle_from;
+	cam->pole_merge_angle_to = bcam->pole_merge_angle_to;
+
 	/* anamorphic lens bokeh */
 	cam->aperture_ratio = bcam->aperture_ratio;
 
@@ -453,7 +460,7 @@ static void blender_camera_sync(Camera *cam, BlenderCamera *bcam, int width, int
 	cam->rolling_shutter_type = bcam->rolling_shutter_type;
 	cam->rolling_shutter_duration = bcam->rolling_shutter_duration;
 
-	memcpy(cam->shutter_curve, bcam->shutter_curve, sizeof(cam->shutter_curve));
+	cam->shutter_curve = bcam->shutter_curve;
 
 	/* border */
 	cam->border = bcam->border;
@@ -552,6 +559,10 @@ void BlenderSync::sync_camera_motion(BL::RenderSettings& b_render,
 		float aspectratio, sensor_size;
 		blender_camera_init(&bcam, b_render);
 
+		/* TODO(sergey): Consider making it a part of blender_camera_init(). */
+		bcam.pixelaspect.x = b_render.pixel_aspect_x();
+		bcam.pixelaspect.y = b_render.pixel_aspect_y();
+
 		blender_camera_from_object(&bcam, b_engine, b_ob);
 		blender_camera_viewplane(&bcam,
 		                         width, height,

intern/cycles/blender/blender_curves.cpp

@@ -138,8 +138,7 @@ bool ObtainCacheParticleData(Mesh *mesh, BL::Mesh *b_mesh, BL::Object *b_ob, Par
 			BL::ParticleSettings b_part((const PointerRNA)b_psys.settings().ptr);
 
 			if((b_part.render_type() == BL::ParticleSettings::render_type_PATH) && (b_part.type() == BL::ParticleSettings::type_HAIR)) {
-				int mi = clamp(b_part.material()-1, 0, mesh->used_shaders.size()-1);
-				int shader = mesh->used_shaders[mi];
+				int shader = clamp(b_part.material()-1, 0, mesh->used_shaders.size()-1);
 				int draw_step = background ? b_part.render_step() : b_part.draw_step();
 				int totparts = b_psys.particles.length();
 				int totchild = background ? b_psys.child_particles.length() : (int)((float)b_psys.child_particles.length() * (float)b_part.draw_percentage() / 100.0f);
@@ -157,16 +156,16 @@ bool ObtainCacheParticleData(Mesh *mesh, BL::Mesh *b_mesh, BL::Object *b_ob, Par
 
 				PointerRNA cpsys = RNA_pointer_get(&b_part.ptr, "cycles");
 
-				CData->psys_firstcurve.push_back(curvenum);
-				CData->psys_curvenum.push_back(totcurves);
-				CData->psys_shader.push_back(shader);
+				CData->psys_firstcurve.push_back_slow(curvenum);
+				CData->psys_curvenum.push_back_slow(totcurves);
+				CData->psys_shader.push_back_slow(shader);
 
 				float radius = get_float(cpsys, "radius_scale") * 0.5f;
 	
-				CData->psys_rootradius.push_back(radius * get_float(cpsys, "root_width"));
-				CData->psys_tipradius.push_back(radius * get_float(cpsys, "tip_width"));
-				CData->psys_shape.push_back(get_float(cpsys, "shape"));
-				CData->psys_closetip.push_back(get_boolean(cpsys, "use_closetip"));
+				CData->psys_rootradius.push_back_slow(radius * get_float(cpsys, "root_width"));
+				CData->psys_tipradius.push_back_slow(radius * get_float(cpsys, "tip_width"));
+				CData->psys_shape.push_back_slow(get_float(cpsys, "shape"));
+				CData->psys_closetip.push_back_slow(get_boolean(cpsys, "use_closetip"));
 
 				int pa_no = 0;
 				if(!(b_part.child_type() == 0) && totchild != 0)
@@ -181,7 +180,7 @@ bool ObtainCacheParticleData(Mesh *mesh, BL::Mesh *b_mesh, BL::Object *b_ob, Par
 
 				for(; pa_no < totparts+totchild; pa_no++) {
 					int keynum = 0;
-					CData->curve_firstkey.push_back(keyno);
+					CData->curve_firstkey.push_back_slow(keyno);
 					
 					float curve_length = 0.0f;
 					float3 pcKey;
@@ -196,15 +195,15 @@ bool ObtainCacheParticleData(Mesh *mesh, BL::Mesh *b_mesh, BL::Object *b_ob, Par
 								continue;
 							curve_length += step_length;
 						}
-						CData->curvekey_co.push_back(cKey);
-						CData->curvekey_time.push_back(curve_length);
+						CData->curvekey_co.push_back_slow(cKey);
+						CData->curvekey_time.push_back_slow(curve_length);
 						pcKey = cKey;
 						keynum++;
 					}
 					keyno += keynum;
 
-					CData->curve_keynum.push_back(keynum);
-					CData->curve_length.push_back(curve_length);
+					CData->curve_keynum.push_back_slow(keynum);
+					CData->curve_length.push_back_slow(curve_length);
 					curvenum++;
 				}
 			}
@@ -256,7 +255,7 @@ bool ObtainCacheParticleUV(Mesh *mesh, BL::Mesh *b_mesh, BL::Object *b_ob, Parti
 					float3 uv = make_float3(0.0f, 0.0f, 0.0f);
 					if(b_mesh->tessface_uv_textures.length())
 						b_psys.uv_on_emitter(psmd, *b_pa, pa_no, uv_num, &uv.x);
-					CData->curve_uv.push_back(uv);
+					CData->curve_uv.push_back_slow(uv);
 
 					if(pa_no < totparts && b_pa != b_psys.particles.end())
 						++b_pa;
@@ -310,7 +309,7 @@ bool ObtainCacheParticleVcol(Mesh *mesh, BL::Mesh *b_mesh, BL::Object *b_ob, Par
 					float3 vcol = make_float3(0.0f, 0.0f, 0.0f);
 					if(b_mesh->tessface_vertex_colors.length())
 						b_psys.mcol_on_emitter(psmd, *b_pa, pa_no, vcol_num, &vcol.x);
-					CData->curve_vcol.push_back(vcol);
+					CData->curve_vcol.push_back_slow(vcol);
 
 					if(pa_no < totparts && b_pa != b_psys.particles.end())
 						++b_pa;
@@ -352,10 +351,7 @@ void ExportCurveTrianglePlanes(Mesh *mesh, ParticleCurveData *CData,
 		}
 	}
 
-	mesh->verts.reserve(mesh->verts.size() + numverts);
-	mesh->triangles.reserve(mesh->triangles.size() + numtris);
-	mesh->shader.reserve(mesh->shader.size() + numtris);
-	mesh->smooth.reserve(mesh->smooth.size() + numtris);
+	mesh->reserve_mesh(mesh->verts.size() + numverts, mesh->num_triangles() + numtris);
 
 	/* actually export */
 	for(int sys = 0; sys < CData->psys_firstcurve.size() ; sys++) {
@@ -375,8 +371,8 @@ void ExportCurveTrianglePlanes(Mesh *mesh, ParticleCurveData *CData,
 				xbasis = normalize(cross(RotCam - ickey_loc, v1));
 			float3 ickey_loc_shfl = ickey_loc - radius * xbasis;
 			float3 ickey_loc_shfr = ickey_loc + radius * xbasis;
-			mesh->verts.push_back(ickey_loc_shfl);
-			mesh->verts.push_back(ickey_loc_shfr);
+			mesh->add_vertex(ickey_loc_shfl);
+			mesh->add_vertex(ickey_loc_shfr);
 			vertexindex += 2;
 
 			for(int curvekey = CData->curve_firstkey[curve] + 1; curvekey < CData->curve_firstkey[curve] + CData->curve_keynum[curve]; curvekey++) {
@@ -402,8 +398,8 @@ void ExportCurveTrianglePlanes(Mesh *mesh, ParticleCurveData *CData,
 					xbasis = normalize(cross(RotCam - ickey_loc, v1));
 				float3 ickey_loc_shfl = ickey_loc - radius * xbasis;
 				float3 ickey_loc_shfr = ickey_loc + radius * xbasis;
-				mesh->verts.push_back(ickey_loc_shfl);
-				mesh->verts.push_back(ickey_loc_shfr);
+				mesh->add_vertex(ickey_loc_shfl);
+				mesh->add_vertex(ickey_loc_shfr);
 				mesh->add_triangle(vertexindex-2, vertexindex, vertexindex-1, CData->psys_shader[sys], true);
 				mesh->add_triangle(vertexindex+1, vertexindex-1, vertexindex, CData->psys_shader[sys], true);
 				vertexindex += 2;
@@ -411,7 +407,6 @@ void ExportCurveTrianglePlanes(Mesh *mesh, ParticleCurveData *CData,
 		}
 	}
 
-	mesh->reserve(mesh->verts.size(), mesh->triangles.size(), 0, 0);
 	mesh->attributes.remove(ATTR_STD_VERTEX_NORMAL);
 	mesh->attributes.remove(ATTR_STD_FACE_NORMAL);
 	mesh->add_face_normals();
@@ -438,10 +433,7 @@ void ExportCurveTriangleGeometry(Mesh *mesh, ParticleCurveData *CData, int resol
 		}
 	}
 
-	mesh->verts.reserve(mesh->verts.size() + numverts);
-	mesh->triangles.reserve(mesh->triangles.size() + numtris);
-	mesh->shader.reserve(mesh->shader.size() + numtris);
-	mesh->smooth.reserve(mesh->smooth.size() + numtris);
+	mesh->reserve_mesh(mesh->verts.size() + numverts, mesh->num_triangles() + numtris);
 
 	/* actually export */
 	for(int sys = 0; sys < CData->psys_firstcurve.size() ; sys++) {
@@ -530,7 +522,7 @@ void ExportCurveTriangleGeometry(Mesh *mesh, ParticleCurveData *CData, int resol
 					float angle = M_2PI_F / (float)resolution;
 					for(int section = 0; section < resolution; section++) {
 						float3 ickey_loc_shf = ickey_loc + radius * (cosf(angle * section) * xbasis + sinf(angle * section) * ybasis);
-						mesh->verts.push_back(ickey_loc_shf);
+						mesh->add_vertex(ickey_loc_shf);
 					}
 
 					if(subv != 0) {
@@ -547,7 +539,6 @@ void ExportCurveTriangleGeometry(Mesh *mesh, ParticleCurveData *CData, int resol
 		}
 	}
 
-	mesh->reserve(mesh->verts.size(), mesh->triangles.size(), 0, 0);
 	mesh->attributes.remove(ATTR_STD_VERTEX_NORMAL);
 	mesh->attributes.remove(ATTR_STD_FACE_NORMAL);
 	mesh->add_face_normals();
@@ -562,7 +553,7 @@ void ExportCurveSegments(Scene *scene, Mesh *mesh, ParticleCurveData *CData)
 	int num_keys = 0;
 	int num_curves = 0;
 
-	if(!(mesh->curves.empty() && mesh->curve_keys.empty()))
+	if(mesh->num_curves())
 		return;
 
 	Attribute *attr_intercept = NULL;
@@ -585,8 +576,7 @@ void ExportCurveSegments(Scene *scene, Mesh *mesh, ParticleCurveData *CData)
 		VLOG(1) << "Exporting curve segments for mesh " << mesh->name;
 	}
 
-	mesh->curve_keys.reserve(mesh->curve_keys.size() + num_keys);
-	mesh->curves.reserve(mesh->curves.size() + num_curves);
+	mesh->reserve_curves(mesh->num_curves() + num_curves, mesh->curve_keys.size() + num_keys);
 
 	num_keys = 0;
 	num_curves = 0;
@@ -614,18 +604,16 @@ void ExportCurveSegments(Scene *scene, Mesh *mesh, ParticleCurveData *CData)
 				num_curve_keys++;
 			}
 
-			mesh->add_curve(num_keys, num_curve_keys, CData->psys_shader[sys]);
+			mesh->add_curve(num_keys, CData->psys_shader[sys]);
 			num_keys += num_curve_keys;
 			num_curves++;
 		}
 	}
 
 	/* check allocation */
-	if((mesh->curve_keys.size() != num_keys) || (mesh->curves.size() != num_curves)) {
+	if((mesh->curve_keys.size() != num_keys) || (mesh->num_curves() != num_curves)) {
 		VLOG(1) << "Allocation failed, clearing data";
-		mesh->curve_keys.clear();
-		mesh->curves.clear();
-		mesh->curve_attributes.clear();
+		mesh->clear();
 	}
 }
 
@@ -668,13 +656,16 @@ static void ExportCurveSegmentsMotion(Mesh *mesh, ParticleCurveData *CData, int
 					if(CData->psys_closetip[sys] && (curvekey == CData->curve_firstkey[curve] + CData->curve_keynum[curve] - 1))
 						radius = 0.0f;
 
+					/* curve motion keys store both position and radius in float4 */
 					mP[i] = float3_to_float4(ickey_loc);
 					mP[i].w = radius;
 
 					/* unlike mesh coordinates, these tend to be slightly different
 					 * between frames due to particle transforms into/out of object
 					 * space, so we use an epsilon to detect actual changes */
-					if(len_squared(mP[i] - mesh->curve_keys[i]) > 1e-5f*1e-5f)
+					float4 curve_key = float3_to_float4(mesh->curve_keys[i]);
+					curve_key.w = mesh->curve_radius[i];
+					if(len_squared(mP[i] - curve_key) > 1e-5f*1e-5f)
 						have_motion = true;
 				}
 
@@ -698,8 +689,10 @@ static void ExportCurveSegmentsMotion(Mesh *mesh, ParticleCurveData *CData, int
 			for(int step = 0; step < time_index; step++) {
 				float4 *mP = attr_mP->data_float4() + step*numkeys;
 
-				for(int key = 0; key < numkeys; key++)
-					mP[key] = mesh->curve_keys[key];
+				for(int key = 0; key < numkeys; key++) {
+					mP[key] = float3_to_float4(mesh->curve_keys[key]);
+					mP[key].w = mesh->curve_radius[key];
+				}
 			}
 		}
 	}
@@ -873,7 +866,9 @@ void BlenderSync::sync_curves(Mesh *mesh,
 	if(!motion) {
 		/* Clear stored curve data */
 		mesh->curve_keys.clear();
-		mesh->curves.clear();
+		mesh->curve_radius.clear();
+		mesh->curve_first_key.clear();
+		mesh->curve_shader.clear();
 		mesh->curve_attributes.clear();
 	}
 
@@ -890,7 +885,7 @@ void BlenderSync::sync_curves(Mesh *mesh,
 	int triangle_method = scene->curve_system_manager->triangle_method;
 	int resolution = scene->curve_system_manager->resolution;
 	size_t vert_num = mesh->verts.size();
-	size_t tri_num = mesh->triangles.size();
+	size_t tri_num = mesh->num_triangles();
 	int used_res = 1;
 
 	/* extract particle hair data - should be combined with connecting to mesh later*/
@@ -951,11 +946,10 @@ void BlenderSync::sync_curves(Mesh *mesh,
 			else {
 				Attribute *attr_generated = mesh->curve_attributes.add(ATTR_STD_GENERATED);
 				float3 *generated = attr_generated->data_float3();
-				size_t i = 0;
 
-				foreach(Mesh::Curve& curve, mesh->curves) {
-					float3 co = float4_to_float3(mesh->curve_keys[curve.first_key]);
-					generated[i++] = co*size - loc;
+				for(size_t i = 0; i < mesh->num_curves(); i++) {
+					float3 co = mesh->curve_keys[mesh->get_curve(i).first_key];
+					generated[i] = co*size - loc;
 				}
 			}
 		}

intern/cycles/blender/blender_mesh.cpp

@@ -532,7 +532,7 @@ static void attr_create_pointiness(Scene *scene,
 static void create_mesh(Scene *scene,
                         Mesh *mesh,
                         BL::Mesh& b_mesh,
-                        const vector<uint>& used_shaders)
+                        const vector<Shader*>& used_shaders)
 {
 	/* count vertices and faces */
 	int numverts = b_mesh.vertices.length();
@@ -548,13 +548,12 @@ static void create_mesh(Scene *scene,
 		numtris += (vi[3] == 0)? 1: 2;
 	}
 
-	/* reserve memory */
-	mesh->reserve(numverts, numtris, 0, 0);
+	/* allocate memory */
+	mesh->reserve_mesh(numverts, numtris);
 
 	/* create vertex coordinates and normals */
-	int i = 0;
-	for(b_mesh.vertices.begin(v); v != b_mesh.vertices.end(); ++v, ++i)
-		mesh->verts[i] = get_float3(v->co());
+	for(b_mesh.vertices.begin(v); v != b_mesh.vertices.end(); ++v)
+		mesh->add_vertex(get_float3(v->co()));
 
 	Attribute *attr_N = mesh->attributes.add(ATTR_STD_VERTEX_NORMAL);
 	float3 *N = attr_N->data_float3();
@@ -583,13 +582,12 @@ static void create_mesh(Scene *scene,
 	/* create faces */
 	vector<int> nverts(numfaces);
 	vector<int> face_flags(numfaces, FACE_FLAG_NONE);
-	int fi = 0, ti = 0;
+	int fi = 0;
 
 	for(b_mesh.tessfaces.begin(f); f != b_mesh.tessfaces.end(); ++f, ++fi) {
 		int4 vi = get_int4(f->vertices_raw());
 		int n = (vi[3] == 0)? 3: 4;
-		int mi = clamp(f->material_index(), 0, used_shaders.size()-1);
-		int shader = used_shaders[mi];
+		int shader = clamp(f->material_index(), 0, used_shaders.size()-1);
 		bool smooth = f->use_smooth() || use_loop_normals;
 
 		/* split vertices if normal is different
@@ -619,18 +617,18 @@ static void create_mesh(Scene *scene,
 			   is_zero(cross(mesh->verts[vi[2]] - mesh->verts[vi[0]], mesh->verts[vi[3]] - mesh->verts[vi[0]])))
 			{
 				// TODO(mai): order here is probably wrong
-				mesh->set_triangle(ti++, vi[0], vi[1], vi[3], shader, smooth, true);
-				mesh->set_triangle(ti++, vi[2], vi[3], vi[1], shader, smooth, true);
+				mesh->add_triangle(vi[0], vi[1], vi[3], shader, smooth, true);
+				mesh->add_triangle(vi[2], vi[3], vi[1], shader, smooth, true);
 				face_flags[fi] |= FACE_FLAG_DIVIDE_24;
 			}
 			else {
-				mesh->set_triangle(ti++, vi[0], vi[1], vi[2], shader, smooth, true);
-				mesh->set_triangle(ti++, vi[0], vi[2], vi[3], shader, smooth, true);
+				mesh->add_triangle(vi[0], vi[1], vi[2], shader, smooth, true);
+				mesh->add_triangle(vi[0], vi[2], vi[3], shader, smooth, true);
 				face_flags[fi] |= FACE_FLAG_DIVIDE_13;
 			}
 		}
 		else
-			mesh->set_triangle(ti++, vi[0], vi[1], vi[2], shader, smooth, false);
+			mesh->add_triangle(vi[0], vi[1], vi[2], shader, smooth, false);
 
 		nverts[fi] = n;
 	}
@@ -660,14 +658,14 @@ static void create_subd_mesh(Scene *scene,
                              BL::Object& b_ob,
                              BL::Mesh& b_mesh,
                              PointerRNA *cmesh,
-                             const vector<uint>& used_shaders,
+                             const vector<Shader*>& used_shaders,
                              float dicing_rate,
                              int max_subdivisions)
 {
 	Mesh basemesh;
 	create_mesh(scene, &basemesh, b_mesh, used_shaders);
 
-	SubdParams sdparams(mesh, used_shaders[0], true, false);
+	SubdParams sdparams(mesh, 0, true, false);
 	sdparams.dicing_rate = max(0.1f, RNA_float_get(cmesh, "dicing_rate") * dicing_rate);
 	sdparams.max_level = max_subdivisions;
 
@@ -700,7 +698,7 @@ Mesh *BlenderSync::sync_mesh(BL::Object& b_ob,
 	BL::Material material_override = render_layer.material_override;
 
 	/* find shader indices */
-	vector<uint> used_shaders;
+	vector<Shader*> used_shaders;
 
 	BL::Object::material_slots_iterator slot;
 	for(b_ob.material_slots.begin(slot); slot != b_ob.material_slots.end(); ++slot) {
@@ -742,8 +740,8 @@ Mesh *BlenderSync::sync_mesh(BL::Object& b_ob,
 			 * because the shader needs different mesh attributes */
 			bool attribute_recalc = false;
 
-			foreach(uint shader, mesh->used_shaders)
-				if(scene->shaders[shader]->need_update_attributes)
+			foreach(Shader *shader, mesh->used_shaders)
+				if(shader->need_update_attributes)
 					attribute_recalc = true;
 
 			if(!attribute_recalc)
@@ -760,11 +758,12 @@ Mesh *BlenderSync::sync_mesh(BL::Object& b_ob,
 	/* create derived mesh */
 	PointerRNA cmesh = RNA_pointer_get(&b_ob_data.ptr, "cycles");
 
-	vector<Mesh::Triangle> oldtriangle = mesh->triangles;
+	array<int> oldtriangle = mesh->triangles;
 	
 	/* compares curve_keys rather than strands in order to handle quick hair
 	 * adjustments in dynamic BVH - other methods could probably do this better*/
-	vector<float4> oldcurve_keys = mesh->curve_keys;
+	array<float3> oldcurve_keys = mesh->curve_keys;
+	array<float> oldcurve_radius = mesh->curve_radius;
 
 	mesh->clear();
 	mesh->used_shaders = used_shaders;
@@ -828,14 +827,21 @@ Mesh *BlenderSync::sync_mesh(BL::Object& b_ob,
 	if(oldtriangle.size() != mesh->triangles.size())
 		rebuild = true;
 	else if(oldtriangle.size()) {
-		if(memcmp(&oldtriangle[0], &mesh->triangles[0], sizeof(Mesh::Triangle)*oldtriangle.size()) != 0)
+		if(memcmp(&oldtriangle[0], &mesh->triangles[0], sizeof(int)*oldtriangle.size()) != 0)
 			rebuild = true;
 	}
 
 	if(oldcurve_keys.size() != mesh->curve_keys.size())
 		rebuild = true;
 	else if(oldcurve_keys.size()) {
-		if(memcmp(&oldcurve_keys[0], &mesh->curve_keys[0], sizeof(float4)*oldcurve_keys.size()) != 0)
+		if(memcmp(&oldcurve_keys[0], &mesh->curve_keys[0], sizeof(float3)*oldcurve_keys.size()) != 0)
+			rebuild = true;
+	}
+
+	if(oldcurve_radius.size() != mesh->curve_radius.size())
+		rebuild = true;
+	else if(oldcurve_radius.size()) {
+		if(memcmp(&oldcurve_radius[0], &mesh->curve_radius[0], sizeof(float)*oldcurve_radius.size()) != 0)
 			rebuild = true;
 	}
 	
@@ -932,8 +938,8 @@ void BlenderSync::sync_mesh_motion(BL::Object& b_ob,
 			Attribute *attr_mP = mesh->curve_attributes.find(ATTR_STD_MOTION_VERTEX_POSITION);
 
 			if(attr_mP) {
-				float4 *keys = &mesh->curve_keys[0];
-				memcpy(attr_mP->data_float4() + time_index*numkeys, keys, sizeof(float4)*numkeys);
+				float3 *keys = &mesh->curve_keys[0];
+				memcpy(attr_mP->data_float3() + time_index*numkeys, keys, sizeof(float3)*numkeys);
 			}
 		}
 

intern/cycles/blender/blender_object.cpp

@@ -155,13 +155,8 @@ void BlenderSync::sync_light(BL::Object& b_parent,
 	light->dir = -transform_get_column(&tfm, 2);
 
 	/* shader */
-	vector<uint> used_shaders;
-
+	vector<Shader*> used_shaders;
 	find_shader(b_lamp, used_shaders, scene->default_light);
-
-	if(used_shaders.size() == 0)
-		used_shaders.push_back(scene->default_light);
-
 	light->shader = used_shaders[0];
 
 	/* shadow */
@@ -370,13 +365,12 @@ Object *BlenderSync::sync_object(BL::Object& b_parent,
 	}
 
 	/* make holdout objects on excluded layer invisible for non-camera rays */
-	if(use_holdout && (layer_flag & render_layer.exclude_layer))
+	if(use_holdout && (layer_flag & render_layer.exclude_layer)) {
 		visibility &= ~(PATH_RAY_ALL_VISIBILITY - PATH_RAY_CAMERA);
+	}
 
-	/* camera flag is not actually used, instead is tested against render layer
-	 * flags */
-	if(visibility & PATH_RAY_CAMERA) {
-		visibility |= layer_flag << PATH_RAY_LAYER_SHIFT;
+	/* hide objects not on render layer from camera rays */
+	if(!(layer_flag & render_layer.layer)) {
 		visibility &= ~PATH_RAY_CAMERA;
 	}
 
@@ -577,7 +571,6 @@ void BlenderSync::sync_objects(BL::SpaceView3D& b_v3d, float motion_time)
 			bool hide = (render_layer.use_viewport_visibility)? b_ob.hide(): b_ob.hide_render();
 			uint ob_layer = get_layer(b_base->layers(),
 			                          b_base->layers_local_view(),
-			                          render_layer.use_localview,
 			                          object_is_light(b_ob),
 			                          scene_layers);
 			hide = hide || !(ob_layer & scene_layer);

intern/cycles/blender/blender_particles.cpp

@@ -76,7 +76,7 @@ bool BlenderSync::sync_dupli_particle(BL::Object& b_ob,
 	pa.velocity = get_float3(b_pa.velocity());
 	pa.angular_velocity = get_float3(b_pa.angular_velocity());
 
-	psys->particles.push_back(pa);
+	psys->particles.push_back_slow(pa);
 
 	if(object->particle_index != psys->particles.size() - 1)
 		scene->object_manager->tag_update(scene);

intern/cycles/blender/blender_python.cpp

@@ -70,6 +70,8 @@ bool debug_flags_sync_from_scene(BL::Scene b_scene)
 	flags.cpu.sse3 = get_boolean(cscene, "debug_use_cpu_sse3");
 	flags.cpu.sse2 = get_boolean(cscene, "debug_use_cpu_sse2");
 	flags.cpu.qbvh = get_boolean(cscene, "debug_use_qbvh");
+	/* Synchronize CUDA flags. */
+	flags.cuda.adaptive_compile = get_boolean(cscene, "debug_use_cuda_adaptive_compile");
 	/* Synchronize OpenCL kernel type. */
 	switch(get_enum(cscene, "debug_opencl_kernel_type")) {
 		case 0:

intern/cycles/blender/blender_session.cpp

@@ -473,7 +473,7 @@ void BlenderSession::render()
 		BL::RenderLayer b_rlay = *b_single_rlay;
 
 		/* add passes */
-		vector<Pass> passes;
+		array<Pass> passes;
 		Pass::add(PASS_COMBINED, passes);
 
 		if(session_params.device.advanced_shading) {

intern/cycles/blender/blender_shader.cpp

@@ -32,23 +32,18 @@ CCL_NAMESPACE_BEGIN
 
 typedef map<void*, ShaderInput*> PtrInputMap;
 typedef map<void*, ShaderOutput*> PtrOutputMap;
-typedef map<std::string, ProxyNode*> ProxyMap;
+typedef map<std::string, ConvertNode*> ProxyMap;
 
 /* Find */
 
 void BlenderSync::find_shader(BL::ID& id,
-                              vector<uint>& used_shaders,
-                              int default_shader)
+                              vector<Shader*>& used_shaders,
+                              Shader *default_shader)
 {
-	Shader *shader = (id)? shader_map.find(id): scene->shaders[default_shader];
+	Shader *shader = (id)? shader_map.find(id): default_shader;
 
-	for(size_t i = 0; i < scene->shaders.size(); i++) {
-		if(scene->shaders[i] == shader) {
-			used_shaders.push_back(i);
-			scene->shaders[i]->tag_used(scene);
-			break;
-		}
-	}
+	used_shaders.push_back(shader);
+	shader->tag_used(scene);
 }
 
 /* RNA translation utilities */
@@ -132,82 +127,60 @@ static float3 get_node_output_vector(BL::Node& b_node, const string& name)
 	return make_float3(value[0], value[1], value[2]);
 }
 
-static ShaderSocketType convert_socket_type(BL::NodeSocket& b_socket)
+static SocketType::Type convert_socket_type(BL::NodeSocket& b_socket)
 {
 	switch(b_socket.type()) {
 		case BL::NodeSocket::type_VALUE:
-			return SHADER_SOCKET_FLOAT;
+			return SocketType::FLOAT;
 		case BL::NodeSocket::type_INT:
-			return SHADER_SOCKET_INT;
+			return SocketType::INT;
 		case BL::NodeSocket::type_VECTOR:
-			return SHADER_SOCKET_VECTOR;
+			return SocketType::VECTOR;
 		case BL::NodeSocket::type_RGBA:
-			return SHADER_SOCKET_COLOR;
+			return SocketType::COLOR;
 		case BL::NodeSocket::type_STRING:
-			return SHADER_SOCKET_STRING;
+			return SocketType::STRING;
 		case BL::NodeSocket::type_SHADER:
-			return SHADER_SOCKET_CLOSURE;
+			return SocketType::CLOSURE;
 		
 		default:
-			return SHADER_SOCKET_UNDEFINED;
+			return SocketType::UNDEFINED;
 	}
 }
 
-#ifdef WITH_OSL
-static ShaderSocketType convert_osl_socket_type(OSL::OSLQuery& query,
-                                                BL::NodeSocket& b_socket)
-{
-	ShaderSocketType socket_type = convert_socket_type(b_socket);
-	if(socket_type == SHADER_SOCKET_VECTOR) {
-		/* TODO(sergey): Do we need compatible_name() here? */
-		const OSL::OSLQuery::Parameter *param = query.getparam(b_socket.name());
-		assert(param != NULL);
-		if(param != NULL) {
-			if(param->type.vecsemantics == TypeDesc::POINT) {
-				socket_type = SHADER_SOCKET_POINT;
-			}
-			else if(param->type.vecsemantics == TypeDesc::NORMAL) {
-				socket_type = SHADER_SOCKET_NORMAL;
-			}
-		}
-	}
-
-	return socket_type;
-}
-#endif  /* WITH_OSL */
-
 static void set_default_value(ShaderInput *input,
                               BL::NodeSocket& b_sock,
                               BL::BlendData& b_data,
                               BL::ID& b_id)
 {
-	/* copy values for non linked inputs */
-	switch(input->type) {
-		case SHADER_SOCKET_FLOAT: {
-			input->set(get_float(b_sock.ptr, "default_value"));
-			break;
-		}
-		case SHADER_SOCKET_INT: {
-			input->set((float)get_int(b_sock.ptr, "default_value"));
-			break;
-		}
-		case SHADER_SOCKET_COLOR: {
-			input->set(float4_to_float3(get_float4(b_sock.ptr, "default_value")));
-			break;
-		}
-		case SHADER_SOCKET_NORMAL:
-		case SHADER_SOCKET_POINT:
-		case SHADER_SOCKET_VECTOR: {
-			input->set(get_float3(b_sock.ptr, "default_value"));
-			break;
-		}
-		case SHADER_SOCKET_STRING: {
-			input->set((ustring)blender_absolute_path(b_data, b_id, get_string(b_sock.ptr, "default_value")));
-			break;
-		}
+	Node *node = input->parent;
+	const SocketType& socket = input->socket_type;
 
-		case SHADER_SOCKET_CLOSURE:
-		case SHADER_SOCKET_UNDEFINED:
+	/* copy values for non linked inputs */
+	switch(input->type()) {
+		case SocketType::FLOAT: {
+			node->set(socket, get_float(b_sock.ptr, "default_value"));
+			break;
+		}
+		case SocketType::INT: {
+			node->set(socket, get_int(b_sock.ptr, "default_value"));
+			break;
+		}
+		case SocketType::COLOR: {
+			node->set(socket, float4_to_float3(get_float4(b_sock.ptr, "default_value")));
+			break;
+		}
+		case SocketType::NORMAL:
+		case SocketType::POINT:
+		case SocketType::VECTOR: {
+			node->set(socket, get_float3(b_sock.ptr, "default_value"));
+			break;
+		}
+		case SocketType::STRING: {
+			node->set(socket, (ustring)blender_absolute_path(b_data, b_id, get_string(b_sock.ptr, "default_value")));
+			break;
+		}
+		default:
 			break;
 	}
 }
@@ -291,7 +264,7 @@ static ShaderNode *add_node(Scene *scene,
 		RGBRampNode *ramp = new RGBRampNode();
 		BL::ShaderNodeValToRGB b_ramp_node(b_node);
 		BL::ColorRamp b_color_ramp(b_ramp_node.color_ramp());
-		colorramp_to_array(b_color_ramp, ramp->ramp, RAMP_TABLE_SIZE);
+		colorramp_to_array(b_color_ramp, ramp->ramp, ramp->ramp_alpha, RAMP_TABLE_SIZE);
 		ramp->interpolate = b_color_ramp.interpolation() != BL::ColorRamp::interpolation_CONSTANT;
 		node = ramp;
 	}
@@ -320,11 +293,7 @@ static ShaderNode *add_node(Scene *scene,
 	else if(b_node.is_a(&RNA_ShaderNodeMixRGB)) {
 		BL::ShaderNodeMixRGB b_mix_node(b_node);
 		MixNode *mix = new MixNode();
-		mix->type = MixNode::type_enum[b_mix_node.blend_type()];
-		/* Tag if it's Mix */
-		if(b_mix_node.blend_type() == 0) 
-			mix->special_type = SHADER_SPECIAL_TYPE_MIX_RGB;
-
+		mix->type = (NodeMix)b_mix_node.blend_type();
 		mix->use_clamp = b_mix_node.use_clamp();
 		node = mix;
 	}
@@ -350,27 +319,27 @@ static ShaderNode *add_node(Scene *scene,
 		node = new HSVNode();
 	}
 	else if(b_node.is_a(&RNA_ShaderNodeRGBToBW)) {
-		node = new ConvertNode(SHADER_SOCKET_COLOR, SHADER_SOCKET_FLOAT);
+		node = new RGBToBWNode();
 	}
 	else if(b_node.is_a(&RNA_ShaderNodeMath)) {
 		BL::ShaderNodeMath b_math_node(b_node);
 		MathNode *math = new MathNode();
-		math->type = MathNode::type_enum[b_math_node.operation()];
+		math->type = (NodeMath)b_math_node.operation();
 		math->use_clamp = b_math_node.use_clamp();
 		node = math;
 	}
 	else if(b_node.is_a(&RNA_ShaderNodeVectorMath)) {
 		BL::ShaderNodeVectorMath b_vector_math_node(b_node);
 		VectorMathNode *vmath = new VectorMathNode();
-		vmath->type = VectorMathNode::type_enum[b_vector_math_node.operation()];
+		vmath->type = (NodeVectorMath)b_vector_math_node.operation();
 		node = vmath;
 	}
 	else if(b_node.is_a(&RNA_ShaderNodeVectorTransform)) {
 		BL::ShaderNodeVectorTransform b_vector_transform_node(b_node);
 		VectorTransformNode *vtransform = new VectorTransformNode();
-		vtransform->type = VectorTransformNode::type_enum[b_vector_transform_node.vector_type()];
-		vtransform->convert_from = VectorTransformNode::convert_space_enum[b_vector_transform_node.convert_from()];
-		vtransform->convert_to = VectorTransformNode::convert_space_enum[b_vector_transform_node.convert_to()];
+		vtransform->type = (NodeVectorTransformType)b_vector_transform_node.vector_type();
+		vtransform->convert_from = (NodeVectorTransformConvertSpace)b_vector_transform_node.convert_from();
+		vtransform->convert_to = (NodeVectorTransformConvertSpace)b_vector_transform_node.convert_to();
 		node = vtransform;
 	}
 	else if(b_node.is_a(&RNA_ShaderNodeNormal)) {
@@ -419,13 +388,13 @@ static ShaderNode *add_node(Scene *scene,
 
 		switch(b_aniso_node.distribution()) {
 			case BL::ShaderNodeBsdfAnisotropic::distribution_BECKMANN:
-				aniso->distribution = ustring("Beckmann");
+				aniso->distribution = CLOSURE_BSDF_MICROFACET_BECKMANN_ANISO_ID;
 				break;
 			case BL::ShaderNodeBsdfAnisotropic::distribution_GGX:
-				aniso->distribution = ustring("GGX");
+				aniso->distribution = CLOSURE_BSDF_MICROFACET_GGX_ANISO_ID;
 				break;
 			case BL::ShaderNodeBsdfAnisotropic::distribution_ASHIKHMIN_SHIRLEY:
-				aniso->distribution = ustring("Ashikhmin-Shirley");
+				aniso->distribution = CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ANISO_ID;
 				break;
 		}
 
@@ -441,13 +410,13 @@ static ShaderNode *add_node(Scene *scene,
 
 		switch(b_subsurface_node.falloff()) {
 			case BL::ShaderNodeSubsurfaceScattering::falloff_CUBIC:
-				subsurface->closure = CLOSURE_BSSRDF_CUBIC_ID;
+				subsurface->falloff = CLOSURE_BSSRDF_CUBIC_ID;
 				break;
 			case BL::ShaderNodeSubsurfaceScattering::falloff_GAUSSIAN:
-				subsurface->closure = CLOSURE_BSSRDF_GAUSSIAN_ID;
+				subsurface->falloff = CLOSURE_BSSRDF_GAUSSIAN_ID;
 				break;
 			case BL::ShaderNodeSubsurfaceScattering::falloff_BURLEY:
-				subsurface->closure = CLOSURE_BSSRDF_BURLEY_ID;
+				subsurface->falloff = CLOSURE_BSSRDF_BURLEY_ID;
 				break;
 		}
 
@@ -459,16 +428,16 @@ static ShaderNode *add_node(Scene *scene,
 		
 		switch(b_glossy_node.distribution()) {
 			case BL::ShaderNodeBsdfGlossy::distribution_SHARP:
-				glossy->distribution = ustring("Sharp");
+				glossy->distribution = CLOSURE_BSDF_REFLECTION_ID;
 				break;
 			case BL::ShaderNodeBsdfGlossy::distribution_BECKMANN:
-				glossy->distribution = ustring("Beckmann");
+				glossy->distribution = CLOSURE_BSDF_MICROFACET_BECKMANN_ID;
 				break;
 			case BL::ShaderNodeBsdfGlossy::distribution_GGX:
-				glossy->distribution = ustring("GGX");
+				glossy->distribution = CLOSURE_BSDF_MICROFACET_GGX_ID;
 				break;
 			case BL::ShaderNodeBsdfGlossy::distribution_ASHIKHMIN_SHIRLEY:
-				glossy->distribution = ustring("Ashikhmin-Shirley");
+				glossy->distribution = CLOSURE_BSDF_ASHIKHMIN_SHIRLEY_ANISO_ID;
 				break;
 		}
 		node = glossy;
@@ -478,13 +447,13 @@ static ShaderNode *add_node(Scene *scene,
 		GlassBsdfNode *glass = new GlassBsdfNode();
 		switch(b_glass_node.distribution()) {
 			case BL::ShaderNodeBsdfGlass::distribution_SHARP:
-				glass->distribution = ustring("Sharp");
+				glass->distribution = CLOSURE_BSDF_SHARP_GLASS_ID;
 				break;
 			case BL::ShaderNodeBsdfGlass::distribution_BECKMANN:
-				glass->distribution = ustring("Beckmann");
+				glass->distribution = CLOSURE_BSDF_MICROFACET_BECKMANN_GLASS_ID;
 				break;
 			case BL::ShaderNodeBsdfGlass::distribution_GGX:
-				glass->distribution = ustring("GGX");
+				glass->distribution = CLOSURE_BSDF_MICROFACET_GGX_GLASS_ID;
 				break;
 		}
 		node = glass;
@@ -494,13 +463,13 @@ static ShaderNode *add_node(Scene *scene,
 		RefractionBsdfNode *refraction = new RefractionBsdfNode();
 		switch(b_refraction_node.distribution()) {
 			case BL::ShaderNodeBsdfRefraction::distribution_SHARP:
-				refraction->distribution = ustring("Sharp");
+				refraction->distribution = CLOSURE_BSDF_REFRACTION_ID;
 				break;
 			case BL::ShaderNodeBsdfRefraction::distribution_BECKMANN:
-				refraction->distribution = ustring("Beckmann");
+				refraction->distribution = CLOSURE_BSDF_MICROFACET_BECKMANN_REFRACTION_ID;
 				break;
 			case BL::ShaderNodeBsdfRefraction::distribution_GGX:
-				refraction->distribution = ustring("GGX");
+				refraction->distribution = CLOSURE_BSDF_MICROFACET_GGX_REFRACTION_ID;
 				break;
 		}
 		node = refraction;
@@ -510,10 +479,10 @@ static ShaderNode *add_node(Scene *scene,
 		ToonBsdfNode *toon = new ToonBsdfNode();
 		switch(b_toon_node.component()) {
 			case BL::ShaderNodeBsdfToon::component_DIFFUSE:
-				toon->component = ustring("Diffuse");
+				toon->component = CLOSURE_BSDF_DIFFUSE_TOON_ID;
 				break;
 			case BL::ShaderNodeBsdfToon::component_GLOSSY:
-				toon->component = ustring("Glossy");
+				toon->component = CLOSURE_BSDF_GLOSSY_TOON_ID;
 				break;
 		}
 		node = toon;
@@ -523,10 +492,10 @@ static ShaderNode *add_node(Scene *scene,
 		HairBsdfNode *hair = new HairBsdfNode();
 		switch(b_hair_node.component()) {
 			case BL::ShaderNodeBsdfHair::component_Reflection:
-				hair->component = ustring("Reflection");
+				hair->component = CLOSURE_BSDF_HAIR_REFLECTION_ID;
 				break;
 			case BL::ShaderNodeBsdfHair::component_Transmission:
-				hair->component = ustring("Transmission");
+				hair->component = CLOSURE_BSDF_HAIR_TRANSMISSION_ID;
 				break;
 		}
 		node = hair;
@@ -593,62 +562,17 @@ static ShaderNode *add_node(Scene *scene,
 		if(scene->shader_manager->use_osl()) {
 			/* create script node */
 			BL::ShaderNodeScript b_script_node(b_node);
-			OSLScriptNode *script_node = new OSLScriptNode();
 
 			OSLShaderManager *manager = (OSLShaderManager*)scene->shader_manager;
 			string bytecode_hash = b_script_node.bytecode_hash();
 
-			/* Gather additional information from the shader, such as
-			 * input/output type info needed for proper node construction.
-			 */
-			OSL::OSLQuery query;
-
 			if(!bytecode_hash.empty()) {
-				query.open_bytecode(b_script_node.bytecode());
+				node = manager->osl_node("", bytecode_hash, b_script_node.bytecode());
 			}
 			else {
-				OSLShaderManager::osl_query(query, b_script_node.filepath());
+				string absolute_filepath = blender_absolute_path(b_data, b_ntree, b_script_node.filepath());
+				node = manager->osl_node(absolute_filepath, "");
 			}
-			/* TODO(sergey): Add proper query info error parsing. */
-
-			/* Generate inputs/outputs from node sockets
-			 *
-			 * Note: the node sockets are generated from OSL parameters,
-			 * so the names match those of the corresponding parameters exactly.
-			 *
-			 * Note 2: ShaderInput/ShaderOutput store shallow string copies only!
-			 * Socket names must be stored in the extra lists instead. */
-			BL::Node::inputs_iterator b_input;
-
-			for(b_script_node.inputs.begin(b_input); b_input != b_script_node.inputs.end(); ++b_input) {
-				script_node->input_names.push_back(ustring(b_input->name()));
-				ShaderInput *input = script_node->add_input(script_node->input_names.back().c_str(),
-				                                            convert_osl_socket_type(query, *b_input));
-				set_default_value(input, *b_input, b_data, b_ntree);
-			}
-
-			BL::Node::outputs_iterator b_output;
-
-			for(b_script_node.outputs.begin(b_output); b_output != b_script_node.outputs.end(); ++b_output) {
-				script_node->output_names.push_back(ustring(b_output->name()));
-				script_node->add_output(script_node->output_names.back().c_str(),
-				                        convert_osl_socket_type(query, *b_output));
-			}
-
-			/* load bytecode or filepath */
-			if(!bytecode_hash.empty()) {
-				/* loaded bytecode if not already done */
-				if(!manager->shader_test_loaded(bytecode_hash))
-					manager->shader_load_bytecode(bytecode_hash, b_script_node.bytecode());
-
-				script_node->bytecode_hash = bytecode_hash;
-			}
-			else {
-				/* set filepath */
-				script_node->filepath = blender_absolute_path(b_data, b_ntree, b_script_node.filepath());
-			}
-
-			node = script_node;
 		}
 #else
 		(void)b_data;
@@ -695,14 +619,14 @@ static ShaderNode *add_node(Scene *scene,
 			/* TODO(sergey): Does not work properly when we change builtin type. */
 			if(b_image.is_updated()) {
 				scene->image_manager->tag_reload_image(
-				        image->filename,
+				        image->filename.string(),
 				        image->builtin_data,
 				        get_image_interpolation(b_image_node),
 				        get_image_extension(b_image_node));
 			}
 		}
-		image->color_space = ImageTextureNode::color_space_enum[(int)b_image_node.color_space()];
-		image->projection = ImageTextureNode::projection_enum[(int)b_image_node.projection()];
+		image->color_space = (NodeImageColorSpace)b_image_node.color_space();
+		image->projection = (NodeImageProjection)b_image_node.projection();
 		image->interpolation = get_image_interpolation(b_image_node);
 		image->extension = get_image_extension(b_image_node);
 		image->projection_blend = b_image_node.projection_blend();
@@ -732,24 +656,24 @@ static ShaderNode *add_node(Scene *scene,
 				env->filename = image_user_file_path(b_image_user,
 				                                     b_image,
 				                                     b_scene.frame_current());
-				env->animated = b_env_node.image_user().use_auto_refresh();
 				env->builtin_data = NULL;
 			}
 
+			env->animated = b_env_node.image_user().use_auto_refresh();
 			env->use_alpha = b_image.use_alpha();
 
 			/* TODO(sergey): Does not work properly when we change builtin type. */
 			if(b_image.is_updated()) {
 				scene->image_manager->tag_reload_image(
-				        env->filename,
+				        env->filename.string(),
 				        env->builtin_data,
 				        get_image_interpolation(b_env_node),
 				        EXTENSION_REPEAT);
 			}
 		}
-		env->color_space = EnvironmentTextureNode::color_space_enum[(int)b_env_node.color_space()];
+		env->color_space = (NodeImageColorSpace)b_env_node.color_space();
 		env->interpolation = get_image_interpolation(b_env_node);
-		env->projection = EnvironmentTextureNode::projection_enum[(int)b_env_node.projection()];
+		env->projection = (NodeEnvironmentProjection)b_env_node.projection();
 		BL::TexMapping b_texture_mapping(b_env_node.texture_mapping());
 		get_tex_mapping(&env->tex_mapping, b_texture_mapping);
 		node = env;
@@ -757,7 +681,7 @@ static ShaderNode *add_node(Scene *scene,
 	else if(b_node.is_a(&RNA_ShaderNodeTexGradient)) {
 		BL::ShaderNodeTexGradient b_gradient_node(b_node);
 		GradientTextureNode *gradient = new GradientTextureNode();
-		gradient->type = GradientTextureNode::type_enum[(int)b_gradient_node.gradient_type()];
+		gradient->type = (NodeGradientType)b_gradient_node.gradient_type();
 		BL::TexMapping b_texture_mapping(b_gradient_node.texture_mapping());
 		get_tex_mapping(&gradient->tex_mapping, b_texture_mapping);
 		node = gradient;
@@ -765,7 +689,7 @@ static ShaderNode *add_node(Scene *scene,
 	else if(b_node.is_a(&RNA_ShaderNodeTexVoronoi)) {
 		BL::ShaderNodeTexVoronoi b_voronoi_node(b_node);
 		VoronoiTextureNode *voronoi = new VoronoiTextureNode();
-		voronoi->coloring = VoronoiTextureNode::coloring_enum[(int)b_voronoi_node.coloring()];
+		voronoi->coloring = (NodeVoronoiColoring)b_voronoi_node.coloring();
 		BL::TexMapping b_texture_mapping(b_voronoi_node.texture_mapping());
 		get_tex_mapping(&voronoi->tex_mapping, b_texture_mapping);
 		node = voronoi;
@@ -781,8 +705,8 @@ static ShaderNode *add_node(Scene *scene,
 	else if(b_node.is_a(&RNA_ShaderNodeTexWave)) {
 		BL::ShaderNodeTexWave b_wave_node(b_node);
 		WaveTextureNode *wave = new WaveTextureNode();
-		wave->type = WaveTextureNode::type_enum[(int)b_wave_node.wave_type()];
-		wave->profile = WaveTextureNode::profile_enum[(int)b_wave_node.wave_profile()];
+		wave->type = (NodeWaveType)b_wave_node.wave_type();
+		wave->profile = (NodeWaveProfile)b_wave_node.wave_profile();
 		BL::TexMapping b_texture_mapping(b_wave_node.texture_mapping());
 		get_tex_mapping(&wave->tex_mapping, b_texture_mapping);
 		node = wave;
@@ -815,7 +739,7 @@ static ShaderNode *add_node(Scene *scene,
 	else if(b_node.is_a(&RNA_ShaderNodeTexMusgrave)) {
 		BL::ShaderNodeTexMusgrave b_musgrave_node(b_node);
 		MusgraveTextureNode *musgrave = new MusgraveTextureNode();
-		musgrave->type = MusgraveTextureNode::type_enum[(int)b_musgrave_node.musgrave_type()];
+		musgrave->type = (NodeMusgraveType)b_musgrave_node.musgrave_type();
 		BL::TexMapping b_texture_mapping(b_musgrave_node.texture_mapping());
 		get_tex_mapping(&musgrave->tex_mapping, b_texture_mapping);
 		node = musgrave;
@@ -833,7 +757,7 @@ static ShaderNode *add_node(Scene *scene,
 	else if(b_node.is_a(&RNA_ShaderNodeTexSky)) {
 		BL::ShaderNodeTexSky b_sky_node(b_node);
 		SkyTextureNode *sky = new SkyTextureNode();
-		sky->type = SkyTextureNode::type_enum[(int)b_sky_node.sky_type()];
+		sky->type = (NodeSkyType)b_sky_node.sky_type();
 		sky->sun_direction = normalize(get_float3(b_sky_node.sun_direction()));
 		sky->turbidity = b_sky_node.turbidity();
 		sky->ground_albedo = b_sky_node.ground_albedo();
@@ -844,15 +768,15 @@ static ShaderNode *add_node(Scene *scene,
 	else if(b_node.is_a(&RNA_ShaderNodeNormalMap)) {
 		BL::ShaderNodeNormalMap b_normal_map_node(b_node);
 		NormalMapNode *nmap = new NormalMapNode();
-		nmap->space = NormalMapNode::space_enum[(int)b_normal_map_node.space()];
+		nmap->space = (NodeNormalMapSpace)b_normal_map_node.space();
 		nmap->attribute = b_normal_map_node.uv_map();
 		node = nmap;
 	}
 	else if(b_node.is_a(&RNA_ShaderNodeTangent)) {
 		BL::ShaderNodeTangent b_tangent_node(b_node);
 		TangentNode *tangent = new TangentNode();
-		tangent->direction_type = TangentNode::direction_type_enum[(int)b_tangent_node.direction_type()];
-		tangent->axis = TangentNode::axis_enum[(int)b_tangent_node.axis()];
+		tangent->direction_type = (NodeTangentDirectionType)b_tangent_node.direction_type();
+		tangent->axis = (NodeTangentAxis)b_tangent_node.axis();
 		tangent->attribute = b_tangent_node.uv_map();
 		node = tangent;
 	}
@@ -867,8 +791,7 @@ static ShaderNode *add_node(Scene *scene,
 		BL::ShaderNodeTexPointDensity b_point_density_node(b_node);
 		PointDensityTextureNode *point_density = new PointDensityTextureNode();
 		point_density->filename = b_point_density_node.name();
-		point_density->space =
-		        PointDensityTextureNode::space_enum[(int)b_point_density_node.space()];
+		point_density->space = (NodeTexVoxelSpace)b_point_density_node.space();
 		point_density->interpolation = get_image_interpolation(b_point_density_node);
 		point_density->builtin_data = b_point_density_node.ptr.data;
 
@@ -879,7 +802,7 @@ static ShaderNode *add_node(Scene *scene,
 		if(true) {
 			b_point_density_node.cache_point_density(b_scene, settings);
 			scene->image_manager->tag_reload_image(
-			        point_density->filename,
+			        point_density->filename.string(),
 			        point_density->builtin_data,
 			        point_density->interpolation,
 			        EXTENSION_CLIP);
@@ -1029,7 +952,8 @@ static void add_nodes(Scene *scene,
 			BL::Node::internal_links_iterator b_link;
 			for(b_node->internal_links.begin(b_link); b_link != b_node->internal_links.end(); ++b_link) {
 				BL::NodeSocket to_socket(b_link->to_socket());
-				ProxyNode *proxy = new ProxyNode(convert_socket_type(to_socket));
+				SocketType::Type to_socket_type = convert_socket_type(to_socket);
+				ConvertNode *proxy = new ConvertNode(to_socket_type, to_socket_type, true);
 
 				input_map[b_link->from_socket().ptr.data] = proxy->inputs[0];
 				output_map[b_link->to_socket().ptr.data] = proxy->outputs[0];
@@ -1051,7 +975,8 @@ static void add_nodes(Scene *scene,
 			 * so that links have something to connect to and assert won't fail.
 			 */
 			for(b_node->inputs.begin(b_input); b_input != b_node->inputs.end(); ++b_input) {
-				ProxyNode *proxy = new ProxyNode(convert_socket_type(*b_input));
+				SocketType::Type input_type = convert_socket_type(*b_input);
+				ConvertNode *proxy = new ConvertNode(input_type, input_type, true);
 				graph->add(proxy);
 
 				/* register the proxy node for internal binding */
@@ -1062,7 +987,8 @@ static void add_nodes(Scene *scene,
 				set_default_value(proxy->inputs[0], *b_input, b_data, b_ntree);
 			}
 			for(b_node->outputs.begin(b_output); b_output != b_node->outputs.end(); ++b_output) {
-				ProxyNode *proxy = new ProxyNode(convert_socket_type(*b_output));
+				SocketType::Type output_type = convert_socket_type(*b_output);
+				ConvertNode *proxy = new ConvertNode(output_type, output_type, true);
 				graph->add(proxy);
 
 				/* register the proxy node for internal binding */
@@ -1088,7 +1014,7 @@ static void add_nodes(Scene *scene,
 			for(b_node->outputs.begin(b_output); b_output != b_node->outputs.end(); ++b_output) {
 				ProxyMap::const_iterator proxy_it = proxy_input_map.find(b_output->identifier());
 				if(proxy_it != proxy_input_map.end()) {
-					ProxyNode *proxy = proxy_it->second;
+					ConvertNode *proxy = proxy_it->second;
 
 					output_map[b_output->ptr.data] = proxy->outputs[0];
 				}
@@ -1102,7 +1028,7 @@ static void add_nodes(Scene *scene,
 				for(b_node->inputs.begin(b_input); b_input != b_node->inputs.end(); ++b_input) {
 					ProxyMap::const_iterator proxy_it = proxy_output_map.find(b_input->identifier());
 					if(proxy_it != proxy_output_map.end()) {
-						ProxyNode *proxy = proxy_it->second;
+						ConvertNode *proxy = proxy_it->second;
 
 						input_map[b_input->ptr.data] = proxy->inputs[0];
 
@@ -1208,7 +1134,7 @@ static void add_nodes(Scene *scene,
 
 void BlenderSync::sync_materials(bool update_all)
 {
-	shader_map.set_default(scene->shaders[scene->default_surface]);
+	shader_map.set_default(scene->default_surface);
 
 	/* material loop */
 	BL::BlendData::materials_iterator b_mat;
@@ -1230,13 +1156,12 @@ void BlenderSync::sync_materials(bool update_all)
 				add_nodes(scene, b_engine, b_data, b_scene, !preview, graph, b_ntree);
 			}
 			else {
-				ShaderNode *closure, *out;
+				DiffuseBsdfNode *diffuse = new DiffuseBsdfNode();
+				diffuse->color = get_float3(b_mat->diffuse_color());
+				graph->add(diffuse);
 
-				closure = graph->add(new DiffuseBsdfNode());
-				closure->input("Color")->value = get_float3(b_mat->diffuse_color());
-				out = graph->output();
-
-				graph->connect(closure->output("BSDF"), out->input("Surface"));
+				ShaderNode *out = graph->output();
+				graph->connect(diffuse->output("BSDF"), out->input("Surface"));
 			}
 
 			/* settings */
@@ -1263,7 +1188,7 @@ void BlenderSync::sync_world(bool update_all)
 	BL::World b_world = b_scene.world();
 
 	if(world_recalc || update_all || b_world.ptr.data != world_map) {
-		Shader *shader = scene->shaders[scene->default_background];
+		Shader *shader = scene->default_background;
 		ShaderGraph *graph = new ShaderGraph();
 
 		/* create nodes */
@@ -1279,13 +1204,12 @@ void BlenderSync::sync_world(bool update_all)
 			shader->volume_interpolation_method = get_volume_interpolation(cworld);
 		}
 		else if(b_world) {
-			ShaderNode *closure, *out;
+			BackgroundNode *background = new BackgroundNode();
+			background->color = get_float3(b_world.horizon_color());
+			graph->add(background);
 
-			closure = graph->add(new BackgroundNode());
-			closure->input("Color")->value = get_float3(b_world.horizon_color());
-			out = graph->output();
-
-			graph->connect(closure->output("Background"), out->input("Surface"));
+			ShaderNode *out = graph->output();
+			graph->connect(background->output("Background"), out->input("Surface"));
 		}
 
 		if(b_world) {
@@ -1343,7 +1267,7 @@ void BlenderSync::sync_world(bool update_all)
 
 void BlenderSync::sync_lamps(bool update_all)
 {
-	shader_map.set_default(scene->shaders[scene->default_light]);
+	shader_map.set_default(scene->default_light);
 
 	/* lamp loop */
 	BL::BlendData::lamps_iterator b_lamp;
@@ -1364,7 +1288,6 @@ void BlenderSync::sync_lamps(bool update_all)
 				add_nodes(scene, b_engine, b_data, b_scene, !preview, graph, b_ntree);
 			}
 			else {
-				ShaderNode *closure, *out;
 				float strength = 1.0f;
 
 				if(b_lamp->type() == BL::Lamp::type_POINT ||
@@ -1374,12 +1297,13 @@ void BlenderSync::sync_lamps(bool update_all)
 					strength = 100.0f;
 				}
 
-				closure = graph->add(new EmissionNode());
-				closure->input("Color")->value = get_float3(b_lamp->color());
-				closure->input("Strength")->value.x = strength;
-				out = graph->output();
+				EmissionNode *emission = new EmissionNode();
+				emission->color = get_float3(b_lamp->color());
+				emission->strength = strength;
+				graph->add(emission);
 
-				graph->connect(closure->output("Emission"), out->input("Surface"));
+				ShaderNode *out = graph->output();
+				graph->connect(emission->output("Emission"), out->input("Surface"));
 			}
 
 			shader->set_graph(graph);

intern/cycles/blender/blender_sync.cpp

@@ -175,8 +175,8 @@ bool BlenderSync::sync_recalc()
 				world_recalc = true;
 			}
 			else if(b_world->node_tree() && b_world->use_nodes()) {
-				Shader *shader = scene->shaders[scene->default_background];
-				if(has_updated_objects && shader != NULL && shader->has_object_dependency) {
+				Shader *shader = scene->default_background;
+				if(has_updated_objects && shader->has_object_dependency) {
 					world_recalc = true;
 				}
 			}
@@ -269,8 +269,6 @@ void BlenderSync::sync_integrator()
 	        SAMPLING_NUM_PATTERNS,
 	        SAMPLING_PATTERN_SOBOL);
 
-	integrator->layer_flag = render_layer.layer;
-
 	integrator->sample_clamp_direct = get_float(cscene, "sample_clamp_direct");
 	integrator->sample_clamp_indirect = get_float(cscene, "sample_clamp_indirect");
 #ifdef __CAMERA_MOTION__
@@ -377,8 +375,7 @@ void BlenderSync::sync_render_layers(BL::SpaceView3D& b_v3d, const char *layer)
 			layer = layername.c_str();
 		}
 		else {
-			render_layer.use_localview = (b_v3d.local_view() ? true : false);
-			render_layer.scene_layer = get_layer(b_v3d.layers(), b_v3d.layers_local_view(), render_layer.use_localview);
+			render_layer.scene_layer = get_layer(b_v3d.layers(), b_v3d.layers_local_view());
 			render_layer.layer = render_layer.scene_layer;
 			render_layer.exclude_layer = 0;
 			render_layer.holdout_layer = 0;
@@ -421,7 +418,6 @@ void BlenderSync::sync_render_layers(BL::SpaceView3D& b_v3d, const char *layer)
 			render_layer.use_surfaces = b_rlay->use_solid();
 			render_layer.use_hair = b_rlay->use_strand();
 			render_layer.use_viewport_visibility = false;
-			render_layer.use_localview = false;
 
 			render_layer.bound_samples = (use_layer_samples == 1);
 			if(use_layer_samples != 2) {
@@ -631,9 +627,9 @@ SessionParams BlenderSync::get_session_params(BL::RenderEngine& b_engine,
 	else
 		params.threads = 0;
 
-	params.cancel_timeout = get_float(cscene, "debug_cancel_timeout");
-	params.reset_timeout = get_float(cscene, "debug_reset_timeout");
-	params.text_timeout = get_float(cscene, "debug_text_timeout");
+	params.cancel_timeout = (double)get_float(cscene, "debug_cancel_timeout");
+	params.reset_timeout = (double)get_float(cscene, "debug_reset_timeout");
+	params.text_timeout = (double)get_float(cscene, "debug_text_timeout");
 
 	params.progressive_refine = get_boolean(cscene, "use_progressive_refine");
 

intern/cycles/blender/blender_sync.h

@@ -146,7 +146,7 @@ private:
 	void sync_images();
 
 	/* util */
-	void find_shader(BL::ID& id, vector<uint>& used_shaders, int default_shader);
+	void find_shader(BL::ID& id, vector<Shader*>& used_shaders, Shader *default_shader);
 	bool BKE_object_is_modified(BL::Object& b_ob);
 	bool object_is_mesh(BL::Object& b_ob);
 	bool object_is_light(BL::Object& b_ob);
@@ -185,7 +185,6 @@ private:
 		  use_surfaces(true),
 		  use_hair(true),
 		  use_viewport_visibility(false),
-		  use_localview(false),
 		  samples(0), bound_samples(false)
 		{}
 
@@ -200,7 +199,6 @@ private:
 		bool use_surfaces;
 		bool use_hair;
 		bool use_viewport_visibility;
-		bool use_localview;
 		int samples;
 		bool bound_samples;
 	} render_layer;

intern/cycles/blender/blender_util.h

@@ -58,14 +58,19 @@ static inline BL::Mesh object_to_mesh(BL::BlendData& data,
 }
 
 static inline void colorramp_to_array(BL::ColorRamp& ramp,
-                                      float4 *data,
+                                      array<float3>& ramp_color,
+                                      array<float>& ramp_alpha,
                                       int size)
 {
+	ramp_color.resize(size);
+	ramp_alpha.resize(size);
+
 	for(int i = 0; i < size; i++) {
 		float color[4];
 
 		ramp.evaluate((float)i/(float)(size-1), color);
-		data[i] = make_float4(color[0], color[1], color[2], color[3]);
+		ramp_color[i] = make_float3(color[0], color[1], color[2]);
+		ramp_alpha[i] = color[3];
 	}
 }
 
@@ -93,11 +98,12 @@ static inline void curvemapping_minmax(/*const*/ BL::CurveMapping& cumap,
 }
 
 static inline void curvemapping_to_array(BL::CurveMapping& cumap,
-                                         float *data,
+                                         array<float>& data,
                                          int size)
 {
 	cumap.update();
 	BL::CurveMap curve = cumap.curves[0];
+	data.resize(size);
 	for(int i = 0; i < size; i++) {
 		float t = (float)i/(float)(size-1);
 		data[i] = curve.evaluate(t);
@@ -105,7 +111,7 @@ static inline void curvemapping_to_array(BL::CurveMapping& cumap,
 }
 
 static inline void curvemapping_color_to_array(BL::CurveMapping& cumap,
-                                               float4 *data,
+                                               array<float3>& data,
                                                int size,
                                                bool rgb_curve)
 {
@@ -132,6 +138,8 @@ static inline void curvemapping_color_to_array(BL::CurveMapping& cumap,
 	BL::CurveMap mapG = cumap.curves[1];
 	BL::CurveMap mapB = cumap.curves[2];
 
+	data.resize(size);
+
 	if(rgb_curve) {
 		BL::CurveMap mapI = cumap.curves[3];
 
@@ -268,7 +276,6 @@ static inline uint get_layer(const BL::Array<int, 20>& array)
 
 static inline uint get_layer(const BL::Array<int, 20>& array,
                              const BL::Array<int, 8>& local_array,
-                             bool use_local,
                              bool is_light = false,
                              uint scene_layers = (1 << 20) - 1)
 {
@@ -293,13 +300,6 @@ static inline uint get_layer(const BL::Array<int, 20>& array,
 				layer |= (1 << (20+i));
 	}
 
-	/* we don't have spare bits for localview (normally 20-28) because
-	 * PATH_RAY_LAYER_SHIFT uses 20-32. So - check if we have localview and if
-	 * so, shift local view bits down to 1-8, since this is done for the view
-	 * port only - it should be OK and not conflict with render layers. */
-	if(use_local)
-		layer >>= 20;
-
 	return layer;
 }
 

intern/cycles/bvh/CMakeLists.txt

@@ -1,6 +1,7 @@
 
 set(INC
 	.
+	../graph
 	../kernel
 	../kernel/svm
 	../render

intern/cycles/bvh/bvh.cpp

@@ -128,11 +128,11 @@ void BVH::pack_triangle(int idx, float4 storage[3])
 	const Mesh *mesh = objects[tob]->mesh;
 
 	int tidx = pack.prim_index[idx];
-	const int *vidx = mesh->triangles[tidx].v;
+	Mesh::Triangle t = mesh->get_triangle(tidx);
 	const float3* vpos = &mesh->verts[0];
-	float3 v0 = vpos[vidx[0]];
-	float3 v1 = vpos[vidx[1]];
-	float3 v2 = vpos[vidx[2]];
+	float3 v0 = vpos[t.v[0]];
+	float3 v1 = vpos[t.v[1]];
+	float3 v2 = vpos[t.v[2]];
 
 	storage[0] = float3_to_float4(v0);
 	storage[1] = float3_to_float4(v1);
@@ -506,10 +506,10 @@ void RegularBVH::refit_node(int idx, bool leaf, BoundBox& bbox, uint& visibility
 				if(pack.prim_type[prim] & PRIMITIVE_ALL_CURVE) {
 					/* curves */
 					int str_offset = (params.top_level)? mesh->curve_offset: 0;
-					const Mesh::Curve& curve = mesh->curves[pidx - str_offset];
+					Mesh::Curve curve = mesh->get_curve(pidx - str_offset);
 					int k = PRIMITIVE_UNPACK_SEGMENT(pack.prim_type[prim]);
 
-					curve.bounds_grow(k, &mesh->curve_keys[0], bbox);
+					curve.bounds_grow(k, &mesh->curve_keys[0], &mesh->curve_radius[0], bbox);
 
 					visibility |= PATH_RAY_CURVE;
 
@@ -520,17 +520,17 @@ void RegularBVH::refit_node(int idx, bool leaf, BoundBox& bbox, uint& visibility
 						if(attr) {
 							size_t mesh_size = mesh->curve_keys.size();
 							size_t steps = mesh->motion_steps - 1;
-							float4 *key_steps = attr->data_float4();
+							float3 *key_steps = attr->data_float3();
 
 							for(size_t i = 0; i < steps; i++)
-								curve.bounds_grow(k, key_steps + i*mesh_size, bbox);
+								curve.bounds_grow(k, key_steps + i*mesh_size, &mesh->curve_radius[0], bbox);
 						}
 					}
 				}
 				else {
 					/* triangles */
 					int tri_offset = (params.top_level)? mesh->tri_offset: 0;
-					const Mesh::Triangle& triangle = mesh->triangles[pidx - tri_offset];
+					Mesh::Triangle triangle = mesh->get_triangle(pidx - tri_offset);
 					const float3 *vpos = &mesh->verts[0];
 
 					triangle.bounds_grow(vpos, bbox);
@@ -770,10 +770,10 @@ void QBVH::refit_node(int idx, bool leaf, BoundBox& bbox, uint& visibility)
 				if(pack.prim_type[prim] & PRIMITIVE_ALL_CURVE) {
 					/* Curves. */
 					int str_offset = (params.top_level)? mesh->curve_offset: 0;
-					const Mesh::Curve& curve = mesh->curves[pidx - str_offset];
+					Mesh::Curve curve = mesh->get_curve(pidx - str_offset);
 					int k = PRIMITIVE_UNPACK_SEGMENT(pack.prim_type[prim]);
 
-					curve.bounds_grow(k, &mesh->curve_keys[0], bbox);
+					curve.bounds_grow(k, &mesh->curve_keys[0], &mesh->curve_radius[0], bbox);
 
 					visibility |= PATH_RAY_CURVE;
 
@@ -784,17 +784,17 @@ void QBVH::refit_node(int idx, bool leaf, BoundBox& bbox, uint& visibility)
 						if(attr) {
 							size_t mesh_size = mesh->curve_keys.size();
 							size_t steps = mesh->motion_steps - 1;
-							float4 *key_steps = attr->data_float4();
+							float3 *key_steps = attr->data_float3();
 
 							for(size_t i = 0; i < steps; i++)
-								curve.bounds_grow(k, key_steps + i*mesh_size, bbox);
+								curve.bounds_grow(k, key_steps + i*mesh_size, &mesh->curve_radius[0], bbox);
 						}
 					}
 				}
 				else {
 					/* Triangles. */
 					int tri_offset = (params.top_level)? mesh->tri_offset: 0;
-					const Mesh::Triangle& triangle = mesh->triangles[pidx - tri_offset];
+					Mesh::Triangle triangle = mesh->get_triangle(pidx - tri_offset);
 					const float3 *vpos = &mesh->verts[0];
 
 					triangle.bounds_grow(vpos, bbox);

intern/cycles/bvh/bvh_build.cpp

@@ -117,8 +117,9 @@ void BVHBuild::add_reference_mesh(BoundBox& root, BoundBox& center, Mesh *mesh,
 	if(mesh->has_motion_blur())
 		attr_mP = mesh->attributes.find(ATTR_STD_MOTION_VERTEX_POSITION);
 
-	for(uint j = 0; j < mesh->triangles.size(); j++) {
-		Mesh::Triangle t = mesh->triangles[j];
+	size_t num_triangles = mesh->num_triangles();
+	for(uint j = 0; j < num_triangles; j++) {
+		Mesh::Triangle t = mesh->get_triangle(j);
 		BoundBox bounds = BoundBox::empty;
 		PrimitiveType type = PRIMITIVE_TRIANGLE;
 
@@ -148,22 +149,23 @@ void BVHBuild::add_reference_mesh(BoundBox& root, BoundBox& center, Mesh *mesh,
 	if(mesh->has_motion_blur())
 		curve_attr_mP = mesh->curve_attributes.find(ATTR_STD_MOTION_VERTEX_POSITION);
 
-	for(uint j = 0; j < mesh->curves.size(); j++) {
-		Mesh::Curve curve = mesh->curves[j];
+	size_t num_curves = mesh->num_curves();
+	for(uint j = 0; j < num_curves; j++) {
+		Mesh::Curve curve = mesh->get_curve(j);
 		PrimitiveType type = PRIMITIVE_CURVE;
 
 		for(int k = 0; k < curve.num_keys - 1; k++) {
 			BoundBox bounds = BoundBox::empty;
-			curve.bounds_grow(k, &mesh->curve_keys[0], bounds);
+			curve.bounds_grow(k, &mesh->curve_keys[0], &mesh->curve_radius[0], bounds);
 
 			/* motion curve */
 			if(curve_attr_mP) {
 				size_t mesh_size = mesh->curve_keys.size();
 				size_t steps = mesh->motion_steps - 1;
-				float4 *key_steps = curve_attr_mP->data_float4();
+				float3 *key_steps = curve_attr_mP->data_float3();
 
 				for(size_t i = 0; i < steps; i++)
-					curve.bounds_grow(k, key_steps + i*mesh_size, bounds);
+					curve.bounds_grow(k, key_steps + i*mesh_size, &mesh->curve_radius[0], bounds);
 
 				type = PRIMITIVE_MOTION_CURVE;
 			}
@@ -188,10 +190,10 @@ void BVHBuild::add_reference_object(BoundBox& root, BoundBox& center, Object *ob
 
 static size_t count_curve_segments(Mesh *mesh)
 {
-	size_t num = 0, num_curves = mesh->curves.size();
+	size_t num = 0, num_curves = mesh->num_curves();
 
 	for(size_t i = 0; i < num_curves; i++)
-		num += mesh->curves[i].num_keys - 1;
+		num += mesh->get_curve(i).num_keys - 1;
 	
 	return num;
 }
@@ -203,15 +205,18 @@ void BVHBuild::add_references(BVHRange& root)
 
 	foreach(Object *ob, objects) {
 		if(params.top_level) {
+			if(!ob->is_traceable()) {
+				continue;
+			}
 			if(!ob->mesh->is_instanced()) {
-				num_alloc_references += ob->mesh->triangles.size();
+				num_alloc_references += ob->mesh->num_triangles();
 				num_alloc_references += count_curve_segments(ob->mesh);
 			}
 			else
 				num_alloc_references++;
 		}
 		else {
-			num_alloc_references += ob->mesh->triangles.size();
+			num_alloc_references += ob->mesh->num_triangles();
 			num_alloc_references += count_curve_segments(ob->mesh);
 		}
 	}
@@ -224,6 +229,10 @@ void BVHBuild::add_references(BVHRange& root)
 
 	foreach(Object *ob, objects) {
 		if(params.top_level) {
+			if(!ob->is_traceable()) {
+				++i;
+				continue;
+			}
 			if(!ob->mesh->is_instanced())
 				add_reference_mesh(bounds, center, ob->mesh, i);
 			else
@@ -326,11 +335,11 @@ BVHNode* BVHBuild::run()
 			VLOG(1) << "BVH build statistics:\n"
 			        << "  Build time: " << time_dt() - build_start_time << "\n"
 			        << "  Total number of nodes: "
-			        << rootnode->getSubtreeSize(BVH_STAT_NODE_COUNT) << "\n"
+			        << string_human_readable_number(rootnode->getSubtreeSize(BVH_STAT_NODE_COUNT)) << "\n"
 			        << "  Number of inner nodes: "
-			        << rootnode->getSubtreeSize(BVH_STAT_INNER_COUNT)  << "\n"
+			        << string_human_readable_number(rootnode->getSubtreeSize(BVH_STAT_INNER_COUNT)) << "\n"
 			        << "  Number of leaf nodes: "
-			        << rootnode->getSubtreeSize(BVH_STAT_LEAF_COUNT)  << "\n"
+			        << string_human_readable_number(rootnode->getSubtreeSize(BVH_STAT_LEAF_COUNT)) << "\n"
 			        << "  Allocation slop factor: "
 			               << ((prim_type.capacity() != 0)
 			                       ? (float)prim_type.size() / prim_type.capacity()
@@ -607,8 +616,10 @@ BVHNode* BVHBuild::create_leaf_node(const BVHRange& range,
 	vector<int, LeafStackAllocator> p_type[PRIMITIVE_NUM_TOTAL];
 	vector<int, LeafStackAllocator> p_index[PRIMITIVE_NUM_TOTAL];
 	vector<int, LeafStackAllocator> p_object[PRIMITIVE_NUM_TOTAL];
+
 	/* TODO(sergey): In theory we should be able to store references. */
-	vector<BVHReference, LeafStackAllocator> object_references;
+	typedef StackAllocator<256, BVHReference> LeafReferenceStackAllocator;
+	vector<BVHReference, LeafReferenceStackAllocator> object_references;
 
 	uint visibility[PRIMITIVE_NUM_TOTAL] = {0};
 	/* NOTE: Keep initializtion in sync with actual number of primitives. */
@@ -629,6 +640,9 @@ BVHNode* BVHBuild::create_leaf_node(const BVHRange& range,
 
 			bounds[type_index].grow(ref.bounds());
 			visibility[type_index] |= objects[ref.prim_object()]->visibility;
+			if(ref.prim_type() & PRIMITIVE_ALL_CURVE) {
+				visibility[type_index] |= PATH_RAY_CURVE;
+			}
 			++num_new_prims;
 		}
 		else {

intern/cycles/bvh/bvh_node.h

@@ -75,7 +75,7 @@ public:
 			m_visibility = 0; /* happens on build cancel */
 	}
 
-	InnerNode(const BoundBox& bounds)
+	explicit InnerNode(const BoundBox& bounds)
 	{
 		m_bounds = bounds;
 		m_visibility = 0;

intern/cycles/bvh/bvh_sort.cpp

@@ -40,7 +40,7 @@ struct BVHReferenceCompare {
 public:
 	int dim;
 
-	BVHReferenceCompare(int dim_)
+	explicit BVHReferenceCompare(int dim_)
 	{
 		dim = dim_;
 	}
@@ -125,7 +125,7 @@ static void bvh_reference_sort_threaded(TaskPool *task_pool,
 		if(compare.compare(data[left], data[right]) > 0) {
 			swap(data[left], data[right]);
 		}
-		if (compare.compare(data[center], data[right]) > 0) {
+		if(compare.compare(data[center], data[right]) > 0) {
 			swap(data[center], data[right]);
 		}
 		swap(data[center], data[right - 1]);
@@ -184,7 +184,7 @@ void bvh_reference_sort(int start, int end, BVHReference *data, int dim)
 	}
 	else {
 		TaskPool task_pool;
-		bvh_reference_sort_threaded(&task_pool, data, start, end - 1, dim);
+		bvh_reference_sort_threaded(&task_pool, data, start, end - 1, compare);
 		task_pool.wait_work();
 	}
 }

intern/cycles/bvh/bvh_split.cpp

@@ -292,13 +292,13 @@ void BVHSpatialSplit::split_triangle_primitive(const Mesh *mesh,
                                                BoundBox& left_bounds,
                                                BoundBox& right_bounds)
 {
-	const int *inds = mesh->triangles[prim_index].v;
+	Mesh::Triangle t = mesh->get_triangle(prim_index);
 	const float3 *verts = &mesh->verts[0];
-	float3 v1 = tfm ? transform_point(tfm, verts[inds[2]]) : verts[inds[2]];
+	float3 v1 = tfm ? transform_point(tfm, verts[t.v[2]]) : verts[t.v[2]];
 
 	for(int i = 0; i < 3; i++) {
 		float3 v0 = v1;
-		int vindex = inds[i];
+		int vindex = t.v[i];
 		v1 = tfm ? transform_point(tfm, verts[vindex]) : verts[vindex];
 		float v0p = v0[dim];
 		float v1p = v1[dim];
@@ -329,12 +329,11 @@ void BVHSpatialSplit::split_curve_primitive(const Mesh *mesh,
                                             BoundBox& right_bounds)
 {
 	/* curve split: NOTE - Currently ignores curve width and needs to be fixed.*/
-	const int k0 = mesh->curves[prim_index].first_key + segment_index;
+	Mesh::Curve curve = mesh->get_curve(prim_index);
+	const int k0 = curve.first_key + segment_index;
 	const int k1 = k0 + 1;
-	const float4& key0 = mesh->curve_keys[k0];
-	const float4& key1 = mesh->curve_keys[k1];
-	float3 v0 = float4_to_float3(key0);
-	float3 v1 = float4_to_float3(key1);
+	float3 v0 = mesh->curve_keys[k0];
+	float3 v1 = mesh->curve_keys[k1];
 
 	if(tfm != NULL) {
 		v0 = transform_point(tfm, v0);
@@ -405,7 +404,7 @@ void BVHSpatialSplit::split_object_reference(const Object *object,
                                              BoundBox& right_bounds)
 {
 	Mesh *mesh = object->mesh;
-	for(int tri_idx = 0; tri_idx < mesh->triangles.size(); ++tri_idx) {
+	for(int tri_idx = 0; tri_idx < mesh->num_triangles(); ++tri_idx) {
 		split_triangle_primitive(mesh,
 		                         &object->tfm,
 		                         tri_idx,
@@ -414,8 +413,8 @@ void BVHSpatialSplit::split_object_reference(const Object *object,
 		                         left_bounds,
 		                         right_bounds);
 	}
-	for(int curve_idx = 0; curve_idx < mesh->curves.size(); ++curve_idx) {
-		Mesh::Curve &curve = mesh->curves[curve_idx];
+	for(int curve_idx = 0; curve_idx < mesh->num_curves(); ++curve_idx) {
+		Mesh::Curve curve = mesh->get_curve(curve_idx);
 		for(int segment_idx = 0;
 		    segment_idx < curve.num_keys - 1;
 		    ++segment_idx)

intern/cycles/device/CMakeLists.txt

@@ -1,6 +1,7 @@
 
 set(INC
 	.
+	../graph
 	../kernel
 	../kernel/svm
 	../kernel/osl

intern/cycles/device/device.cpp

@@ -56,6 +56,8 @@ std::ostream& operator <<(std::ostream &os,
 	   << string_from_bool(requested_features.use_camera_motion)  << std::endl;
 	os << "Use Baking: "
 	   << string_from_bool(requested_features.use_baking)  << std::endl;
+	os << "Use Volume: "
+	   << string_from_bool(requested_features.use_volume)  << std::endl;
 	return os;
 }
 

intern/cycles/device/device.h

@@ -54,7 +54,7 @@ public:
 	bool display_device;
 	bool advanced_shading;
 	bool pack_images;
-	bool extended_images; /* flag for GPU and Multi device */
+	bool has_bindless_textures; /* flag for GPU and Multi device */
 	bool use_split_kernel; /* Denotes if the device is going to run cycles using split-kernel */
 	vector<DeviceInfo> multi_devices;
 
@@ -66,7 +66,7 @@ public:
 		display_device = false;
 		advanced_shading = true;
 		pack_images = false;
-		extended_images = false;
+		has_bindless_textures = false;
 		use_split_kernel = false;
 	}
 };
@@ -103,6 +103,9 @@ public:
 	/* Use subsurface scattering materials. */
 	bool use_subsurface;
 
+	/* Use volume materials. */
+	bool use_volume;
+
 	/* Use branched integrator. */
 	bool use_integrator_branched;
 
@@ -118,6 +121,7 @@ public:
 		use_camera_motion = false;
 		use_baking = false;
 		use_subsurface = false;
+		use_volume = false;
 		use_integrator_branched = false;
 	}
 
@@ -132,6 +136,7 @@ public:
 		         use_camera_motion == requested_features.use_camera_motion &&
 		         use_baking == requested_features.use_baking &&
 		         use_subsurface == requested_features.use_subsurface &&
+		         use_volume == requested_features.use_volume &&
 		         use_integrator_branched == requested_features.use_integrator_branched);
 	}
 
@@ -161,6 +166,9 @@ public:
 		if(!use_baking) {
 			build_options += " -D__NO_BAKING__";
 		}
+		if(!use_volume) {
+			build_options += " -D__NO_VOLUME__";
+		}
 		if(!use_subsurface) {
 			build_options += " -D__NO_SUBSURFACE__";
 		}
@@ -222,6 +230,7 @@ public:
 		(void)interpolation;  /* Ignored. */
 		(void)extension;  /* Ignored. */
 	};
+
 	virtual void tex_free(device_memory& /*mem*/) {};
 
 	/* pixel memory */

intern/cycles/device/device_cpu.cpp

@@ -155,7 +155,9 @@ public:
 	               InterpolationType interpolation,
 	               ExtensionType extension)
 	{
-		VLOG(1) << "Texture allocate: " << name << ", " << mem.memory_size() << " bytes.";
+		VLOG(1) << "Texture allocate: " << name << ", "
+		        << string_human_readable_number(mem.memory_size()) << " bytes. ("
+		        << string_human_readable_size(mem.memory_size()) << ")";
 		kernel_tex_copy(&kernel_globals,
 		                name,
 		                mem.data_pointer,
@@ -213,12 +215,7 @@ public:
 				return;
 		}
 
-		KernelGlobals kg = kernel_globals;
-
-#ifdef WITH_OSL
-		OSLShader::thread_init(&kg, &kernel_globals, &osl_globals);
-#endif
-
+		KernelGlobals kg = thread_kernel_globals_init();
 		RenderTile tile;
 
 		void(*path_trace_kernel)(KernelGlobals*, float*, unsigned int*, int, int, int, int, int);
@@ -289,9 +286,7 @@ public:
 			}
 		}
 
-#ifdef WITH_OSL
-		OSLShader::thread_free(&kg);
-#endif
+		thread_kernel_globals_free(&kg);
 	}
 
 	void thread_film_convert(DeviceTask& task)
@@ -481,6 +476,40 @@ public:
 	{
 		task_pool.cancel();
 	}
+
+protected:
+	inline KernelGlobals thread_kernel_globals_init()
+	{
+		KernelGlobals kg = kernel_globals;
+		kg.transparent_shadow_intersections = NULL;
+		const int decoupled_count = sizeof(kg.decoupled_volume_steps) /
+		                            sizeof(*kg.decoupled_volume_steps);
+		for(int i = 0; i < decoupled_count; ++i) {
+			kg.decoupled_volume_steps[i] = NULL;
+		}
+		kg.decoupled_volume_steps_index = 0;
+#ifdef WITH_OSL
+		OSLShader::thread_init(&kg, &kernel_globals, &osl_globals);
+#endif
+		return kg;
+	}
+
+	inline void thread_kernel_globals_free(KernelGlobals *kg)
+	{
+		if(kg->transparent_shadow_intersections != NULL) {
+			free(kg->transparent_shadow_intersections);
+		}
+		const int decoupled_count = sizeof(kg->decoupled_volume_steps) /
+		                            sizeof(*kg->decoupled_volume_steps);
+		for(int i = 0; i < decoupled_count; ++i) {
+			if(kg->decoupled_volume_steps[i] != NULL) {
+				free(kg->decoupled_volume_steps[i]);
+			}
+		}
+#ifdef WITH_OSL
+		OSLShader::thread_free(kg);
+#endif
+	}
 };
 
 Device *device_cpu_create(DeviceInfo& info, Stats &stats, bool background)

intern/cycles/device/device_cuda.cpp

@@ -41,11 +41,6 @@
 #include "util_types.h"
 #include "util_time.h"
 
-/* use feature-adaptive kernel compilation.
- * Requires CUDA toolkit to be installed and currently only works on Linux.
- */
-/* #define KERNEL_USE_ADAPTIVE */
-
 CCL_NAMESPACE_BEGIN
 
 #ifndef WITH_CUDA_DYNLOAD
@@ -90,10 +85,10 @@ public:
 	CUcontext cuContext;
 	CUmodule cuModule;
 	map<device_ptr, bool> tex_interp_map;
+	map<device_ptr, uint> tex_bindless_map;
 	int cuDevId;
 	int cuDevArchitecture;
 	bool first_error;
-	bool use_texture_storage;
 
 	struct PixelMem {
 		GLuint cuPBO;
@@ -104,6 +99,10 @@ public:
 
 	map<device_ptr, PixelMem> pixel_mem_map;
 
+	/* Bindless Textures */
+	device_vector<uint> bindless_mapping;
+	bool need_bindless_mapping;
+
 	CUdeviceptr cuda_device_ptr(device_ptr mem)
 	{
 		return (CUdeviceptr)mem;
@@ -181,12 +180,13 @@ public:
 	{
 		first_error = true;
 		background = background_;
-		use_texture_storage = true;
 
 		cuDevId = info.num;
 		cuDevice = 0;
 		cuContext = 0;
 
+		need_bindless_mapping = false;
+
 		/* intialize */
 		if(cuda_error(cuInit(0)))
 			return;
@@ -216,11 +216,6 @@ public:
 		cuDeviceComputeCapability(&major, &minor, cuDevId);
 		cuDevArchitecture = major*100 + minor*10;
 
-		/* In order to use full 6GB of memory on Titan cards, use arrays instead
-		 * of textures. On earlier cards this seems slower, but on Titan it is
-		 * actually slightly faster in tests. */
-		use_texture_storage = (cuDevArchitecture < 300);
-
 		cuda_pop_context();
 	}
 
@@ -228,6 +223,10 @@ public:
 	{
 		task_pool.stop();
 
+		if(info.has_bindless_textures) {
+			tex_free(bindless_mapping);
+		}
+
 		cuda_assert(cuCtxDestroy(cuContext));
 	}
 
@@ -245,40 +244,52 @@ public:
 		return true;
 	}
 
+	bool use_adaptive_compilation()
+	{
+		return DebugFlags().cuda.adaptive_compile;
+	}
+
 	string compile_kernel(const DeviceRequestedFeatures& requested_features)
 	{
-		/* compute cubin name */
+		/* Compute cubin name. */
 		int major, minor;
 		cuDeviceComputeCapability(&major, &minor, cuDevId);
 		string cubin;
 
-		/* attempt to use kernel provided with blender */
-		cubin = path_get(string_printf("lib/kernel_sm_%d%d.cubin", major, minor));
-		VLOG(1) << "Testing for pre-compiled kernel " << cubin;
-		if(path_exists(cubin)) {
-			VLOG(1) << "Using precompiled kernel";
-			return cubin;
+		/* Adaptive Compile.
+		 * If enabled, always use that */
+		bool use_adaptive_compile = use_adaptive_compilation();
+
+		/* Attempt to use kernel provided with Blender. */
+		if(!use_adaptive_compile) {
+			cubin = path_get(string_printf("lib/kernel_sm_%d%d.cubin", major, minor));
+			VLOG(1) << "Testing for pre-compiled kernel " << cubin;
+			if(path_exists(cubin)) {
+				VLOG(1) << "Using precompiled kernel";
+				return cubin;
+			}
 		}
 
-		/* not found, try to use locally compiled kernel */
+		/* Try to use locally compiled kernel. */
 		string kernel_path = path_get("kernel");
 		string md5 = path_files_md5_hash(kernel_path);
 
-#ifdef KERNEL_USE_ADAPTIVE
-		string feature_build_options = requested_features.get_build_options();
-		string device_md5 = util_md5_string(feature_build_options);
-		cubin = string_printf("cycles_kernel_%s_sm%d%d_%s.cubin",
-		                      device_md5.c_str(),
-		                      major, minor,
-		                      md5.c_str());
-#else
-		(void)requested_features;
-		cubin = string_printf("cycles_kernel_sm%d%d_%s.cubin", major, minor, md5.c_str());
-#endif
+		string feature_build_options;
+		if(use_adaptive_compile) {
+			feature_build_options = requested_features.get_build_options();
+			string device_md5 = util_md5_string(feature_build_options);
+			cubin = string_printf("cycles_kernel_%s_sm%d%d_%s.cubin",
+		                          device_md5.c_str(),
+		                          major, minor,
+		                          md5.c_str());
+		}
+		else {
+			cubin = string_printf("cycles_kernel_sm%d%d_%s.cubin", major, minor, md5.c_str());
+		}
 
 		cubin = path_user_get(path_join("cache", cubin));
 		VLOG(1) << "Testing for locally compiled kernel " << cubin;
-		/* if exists already, use it */
+		/* If exists already, use it. */
 		if(path_exists(cubin)) {
 			VLOG(1) << "Using locally compiled kernel";
 			return cubin;
@@ -294,7 +305,7 @@ public:
 		}
 #endif
 
-		/* if not, find CUDA compiler */
+		/* If not, find CUDA compiler. */
 		const char *nvcc = cuewCompilerPath();
 
 		if(nvcc == NULL) {
@@ -316,7 +327,7 @@ public:
 		else if(cuda_version != 75)
 			printf("CUDA version %d.%d detected, build may succeed but only CUDA 7.5 is officially supported.\n", cuda_version/10, cuda_version%10);
 
-		/* compile */
+		/* Compile. */
 		string kernel = path_join(kernel_path, path_join("kernels", path_join("cuda", "kernel.cu")));
 		string include = kernel_path;
 		const int machine = system_cpu_bits();
@@ -331,9 +342,8 @@ public:
 			"-DNVCC -D__KERNEL_CUDA_VERSION__=%d",
 			nvcc, major, minor, machine, kernel.c_str(), cubin.c_str(), include.c_str(), cuda_version);
 
-#ifdef KERNEL_USE_ADAPTIVE
-		command += " " + feature_build_options;
-#endif
+		if(use_adaptive_compile)
+			command += " " + feature_build_options;
 
 		const char* extra_cflags = getenv("CYCLES_CUDA_EXTRA_CFLAGS");
 		if(extra_cflags) {
@@ -351,7 +361,7 @@ public:
 			return "";
 		}
 
-		/* verify if compilation succeeded */
+		/* Verify if compilation succeeded */
 		if(!path_exists(cubin)) {
 			cuda_error_message("CUDA kernel compilation failed, see console for details.");
 			return "";
@@ -397,6 +407,15 @@ public:
 		return (result == CUDA_SUCCESS);
 	}
 
+	void load_bindless_mapping()
+	{
+		if(info.has_bindless_textures && need_bindless_mapping) {
+			tex_free(bindless_mapping);
+			tex_alloc("__bindless_mapping", bindless_mapping, INTERPOLATION_NONE, EXTENSION_REPEAT);
+			need_bindless_mapping = false;
+		}
+	}
+
 	void mem_alloc(device_memory& mem, MemoryType /*type*/)
 	{
 		cuda_push_context();
@@ -474,128 +493,103 @@ public:
 	               InterpolationType interpolation,
 	               ExtensionType extension)
 	{
-		VLOG(1) << "Texture allocate: " << name << ", " << mem.memory_size() << " bytes.";
+		VLOG(1) << "Texture allocate: " << name << ", "
+		        << string_human_readable_number(mem.memory_size()) << " bytes. ("
+		        << string_human_readable_size(mem.memory_size()) << ")";
 
+		/* Check if we are on sm_30 or above.
+		 * We use arrays and bindles textures for storage there */
+		bool has_bindless_textures = info.has_bindless_textures;
+
+		/* General variables for both architectures */
 		string bind_name = name;
-		if(mem.data_depth > 1) {
-			/* Kernel uses different bind names for 2d and 3d float textures,
-			 * so we have to adjust couple of things here.
-			 */
-			vector<string> tokens;
-			string_split(tokens, name, "_");
-			bind_name = string_printf("__tex_image_%s3d_%s",
-			                          tokens[2].c_str(),
-			                          tokens[3].c_str());
-		}
-
-		/* determine format */
-		CUarray_format_enum format;
 		size_t dsize = datatype_size(mem.data_type);
 		size_t size = mem.memory_size();
-		bool use_texture = (interpolation != INTERPOLATION_NONE) || use_texture_storage;
 
-		if(use_texture) {
+		CUaddress_mode address_mode = CU_TR_ADDRESS_MODE_WRAP;
+		switch(extension) {
+			case EXTENSION_REPEAT:
+				address_mode = CU_TR_ADDRESS_MODE_WRAP;
+				break;
+			case EXTENSION_EXTEND:
+				address_mode = CU_TR_ADDRESS_MODE_CLAMP;
+				break;
+			case EXTENSION_CLIP:
+				address_mode = CU_TR_ADDRESS_MODE_BORDER;
+				break;
+			default:
+				assert(0);
+				break;
+		}
 
-			switch(mem.data_type) {
-				case TYPE_UCHAR: format = CU_AD_FORMAT_UNSIGNED_INT8; break;
-				case TYPE_UINT: format = CU_AD_FORMAT_UNSIGNED_INT32; break;
-				case TYPE_INT: format = CU_AD_FORMAT_SIGNED_INT32; break;
-				case TYPE_FLOAT: format = CU_AD_FORMAT_FLOAT; break;
-				default: assert(0); return;
+		CUfilter_mode filter_mode;
+		if(interpolation == INTERPOLATION_CLOSEST) {
+			filter_mode = CU_TR_FILTER_MODE_POINT;
+		}
+		else {
+			filter_mode = CU_TR_FILTER_MODE_LINEAR;
+		}
+
+		CUarray_format_enum format;
+		switch(mem.data_type) {
+			case TYPE_UCHAR: format = CU_AD_FORMAT_UNSIGNED_INT8; break;
+			case TYPE_UINT: format = CU_AD_FORMAT_UNSIGNED_INT32; break;
+			case TYPE_INT: format = CU_AD_FORMAT_SIGNED_INT32; break;
+			case TYPE_FLOAT: format = CU_AD_FORMAT_FLOAT; break;
+			default: assert(0); return;
+		}
+
+		/* General variables for Fermi */
+		CUtexref texref = NULL;
+
+		if(!has_bindless_textures) {
+			if(mem.data_depth > 1) {
+				/* Kernel uses different bind names for 2d and 3d float textures,
+				 * so we have to adjust couple of things here.
+				 */
+				vector<string> tokens;
+				string_split(tokens, name, "_");
+				bind_name = string_printf("__tex_image_%s_3d_%s",
+				                          tokens[2].c_str(),
+				                          tokens[3].c_str());
 			}
 
-			CUtexref texref = NULL;
-
 			cuda_push_context();
 			cuda_assert(cuModuleGetTexRef(&texref, cuModule, bind_name.c_str()));
+			cuda_pop_context();
 
 			if(!texref) {
-				cuda_pop_context();
 				return;
 			}
+		}
 
-			if(interpolation != INTERPOLATION_NONE) {
-				CUarray handle = NULL;
+		/* Data Storage */
+		if(interpolation == INTERPOLATION_NONE) {
+			if(has_bindless_textures) {
+				mem_alloc(mem, MEM_READ_ONLY);
+				mem_copy_to(mem);
 
-				if(mem.data_depth > 1) {
-					CUDA_ARRAY3D_DESCRIPTOR desc;
+				cuda_push_context();
 
-					desc.Width = mem.data_width;
-					desc.Height = mem.data_height;
-					desc.Depth = mem.data_depth;
-					desc.Format = format;
-					desc.NumChannels = mem.data_elements;
-					desc.Flags = 0;
+				CUdeviceptr cumem;
+				size_t cubytes;
 
-					cuda_assert(cuArray3DCreate(&handle, &desc));
+				cuda_assert(cuModuleGetGlobal(&cumem, &cubytes, cuModule, bind_name.c_str()));
+
+				if(cubytes == 8) {
+					/* 64 bit device pointer */
+					uint64_t ptr = mem.device_pointer;
+					cuda_assert(cuMemcpyHtoD(cumem, (void*)&ptr, cubytes));
 				}
 				else {
-					CUDA_ARRAY_DESCRIPTOR desc;
-
-					desc.Width = mem.data_width;
-					desc.Height = mem.data_height;
-					desc.Format = format;
-					desc.NumChannels = mem.data_elements;
-
-					cuda_assert(cuArrayCreate(&handle, &desc));
+					/* 32 bit device pointer */
+					uint32_t ptr = (uint32_t)mem.device_pointer;
+					cuda_assert(cuMemcpyHtoD(cumem, (void*)&ptr, cubytes));
 				}
 
-				if(!handle) {
-					cuda_pop_context();
-					return;
-				}
-
-				if(mem.data_depth > 1) {
-					CUDA_MEMCPY3D param;
-					memset(&param, 0, sizeof(param));
-					param.dstMemoryType = CU_MEMORYTYPE_ARRAY;
-					param.dstArray = handle;
-					param.srcMemoryType = CU_MEMORYTYPE_HOST;
-					param.srcHost = (void*)mem.data_pointer;
-					param.srcPitch = mem.data_width*dsize*mem.data_elements;
-					param.WidthInBytes = param.srcPitch;
-					param.Height = mem.data_height;
-					param.Depth = mem.data_depth;
-
-					cuda_assert(cuMemcpy3D(&param));
-				}
-				if(mem.data_height > 1) {
-					CUDA_MEMCPY2D param;
-					memset(&param, 0, sizeof(param));
-					param.dstMemoryType = CU_MEMORYTYPE_ARRAY;
-					param.dstArray = handle;
-					param.srcMemoryType = CU_MEMORYTYPE_HOST;
-					param.srcHost = (void*)mem.data_pointer;
-					param.srcPitch = mem.data_width*dsize*mem.data_elements;
-					param.WidthInBytes = param.srcPitch;
-					param.Height = mem.data_height;
-
-					cuda_assert(cuMemcpy2D(&param));
-				}
-				else
-					cuda_assert(cuMemcpyHtoA(handle, 0, (void*)mem.data_pointer, size));
-
-				cuda_assert(cuTexRefSetArray(texref, handle, CU_TRSA_OVERRIDE_FORMAT));
-
-				if(interpolation == INTERPOLATION_CLOSEST) {
-					cuda_assert(cuTexRefSetFilterMode(texref, CU_TR_FILTER_MODE_POINT));
-				}
-				else if(interpolation == INTERPOLATION_LINEAR) {
-					cuda_assert(cuTexRefSetFilterMode(texref, CU_TR_FILTER_MODE_LINEAR));
-				}
-				else {/* CUBIC and SMART are unsupported for CUDA */
-					cuda_assert(cuTexRefSetFilterMode(texref, CU_TR_FILTER_MODE_LINEAR));
-				}
-				cuda_assert(cuTexRefSetFlags(texref, CU_TRSF_NORMALIZED_COORDINATES));
-
-				mem.device_pointer = (device_ptr)handle;
-				mem.device_size = size;
-
-				stats.mem_alloc(size);
+				cuda_pop_context();
 			}
 			else {
-				cuda_pop_context();
-
 				mem_alloc(mem, MEM_READ_ONLY);
 				mem_copy_to(mem);
 
@@ -604,23 +598,137 @@ public:
 				cuda_assert(cuTexRefSetAddress(NULL, texref, cuda_device_ptr(mem.device_pointer), size));
 				cuda_assert(cuTexRefSetFilterMode(texref, CU_TR_FILTER_MODE_POINT));
 				cuda_assert(cuTexRefSetFlags(texref, CU_TRSF_READ_AS_INTEGER));
+
+				cuda_pop_context();
+			}
+		}
+		/* Texture Storage */
+		else {
+			CUarray handle = NULL;
+
+			cuda_push_context();
+
+			if(mem.data_depth > 1) {
+				CUDA_ARRAY3D_DESCRIPTOR desc;
+
+				desc.Width = mem.data_width;
+				desc.Height = mem.data_height;
+				desc.Depth = mem.data_depth;
+				desc.Format = format;
+				desc.NumChannels = mem.data_elements;
+				desc.Flags = 0;
+
+				cuda_assert(cuArray3DCreate(&handle, &desc));
+			}
+			else {
+				CUDA_ARRAY_DESCRIPTOR desc;
+
+				desc.Width = mem.data_width;
+				desc.Height = mem.data_height;
+				desc.Format = format;
+				desc.NumChannels = mem.data_elements;
+
+				cuda_assert(cuArrayCreate(&handle, &desc));
 			}
 
-			CUaddress_mode address_mode = CU_TR_ADDRESS_MODE_WRAP;
-			switch(extension) {
-				case EXTENSION_REPEAT:
-					address_mode = CU_TR_ADDRESS_MODE_WRAP;
-					break;
-				case EXTENSION_EXTEND:
-					address_mode = CU_TR_ADDRESS_MODE_CLAMP;
-					break;
-				case EXTENSION_CLIP:
-					address_mode = CU_TR_ADDRESS_MODE_BORDER;
-					break;
-				default:
-					assert(0);
-					break;
+			if(!handle) {
+				cuda_pop_context();
+				return;
 			}
+
+			/* Allocate 3D, 2D or 1D memory */
+			if(mem.data_depth > 1) {
+				CUDA_MEMCPY3D param;
+				memset(&param, 0, sizeof(param));
+				param.dstMemoryType = CU_MEMORYTYPE_ARRAY;
+				param.dstArray = handle;
+				param.srcMemoryType = CU_MEMORYTYPE_HOST;
+				param.srcHost = (void*)mem.data_pointer;
+				param.srcPitch = mem.data_width*dsize*mem.data_elements;
+				param.WidthInBytes = param.srcPitch;
+				param.Height = mem.data_height;
+				param.Depth = mem.data_depth;
+
+				cuda_assert(cuMemcpy3D(&param));
+			}
+			else if(mem.data_height > 1) {
+				CUDA_MEMCPY2D param;
+				memset(&param, 0, sizeof(param));
+				param.dstMemoryType = CU_MEMORYTYPE_ARRAY;
+				param.dstArray = handle;
+				param.srcMemoryType = CU_MEMORYTYPE_HOST;
+				param.srcHost = (void*)mem.data_pointer;
+				param.srcPitch = mem.data_width*dsize*mem.data_elements;
+				param.WidthInBytes = param.srcPitch;
+				param.Height = mem.data_height;
+
+				cuda_assert(cuMemcpy2D(&param));
+			}
+			else
+				cuda_assert(cuMemcpyHtoA(handle, 0, (void*)mem.data_pointer, size));
+
+			/* Fermi and Kepler */
+			mem.device_pointer = (device_ptr)handle;
+			mem.device_size = size;
+
+			stats.mem_alloc(size);
+
+			/* Bindless Textures - Kepler */
+			if(has_bindless_textures) {
+				int flat_slot = 0;
+				if(string_startswith(name, "__tex_image")) {
+					int pos =  string(name).rfind("_");
+					flat_slot = atoi(name + pos + 1);
+				}
+				else {
+					assert(0);
+				}
+
+				CUDA_RESOURCE_DESC resDesc;
+				memset(&resDesc, 0, sizeof(resDesc));
+				resDesc.resType = CU_RESOURCE_TYPE_ARRAY;
+				resDesc.res.array.hArray = handle;
+				resDesc.flags = 0;
+
+				CUDA_TEXTURE_DESC texDesc;
+				memset(&texDesc, 0, sizeof(texDesc));
+				texDesc.addressMode[0] = address_mode;
+				texDesc.addressMode[1] = address_mode;
+				texDesc.addressMode[2] = address_mode;
+				texDesc.filterMode = filter_mode;
+				texDesc.flags = CU_TRSF_NORMALIZED_COORDINATES;
+
+				CUtexObject tex = 0;
+				cuda_assert(cuTexObjectCreate(&tex, &resDesc, &texDesc, NULL));
+
+				/* Safety check */
+				if((uint)tex > UINT_MAX) {
+					assert(0);
+				}
+
+				/* Resize once */
+				if(flat_slot >= bindless_mapping.size())
+					bindless_mapping.resize(4096); /*TODO(dingto): Make this a variable */
+
+				/* Set Mapping and tag that we need to (re-)upload to device */
+				bindless_mapping.get_data()[flat_slot] = (uint)tex;
+				tex_bindless_map[mem.device_pointer] = (uint)tex;
+				need_bindless_mapping = true;
+			}
+			/* Regular Textures - Fermi */
+			else {
+				cuda_assert(cuTexRefSetArray(texref, handle, CU_TRSA_OVERRIDE_FORMAT));
+				cuda_assert(cuTexRefSetFilterMode(texref, filter_mode));
+				cuda_assert(cuTexRefSetFlags(texref, CU_TRSF_NORMALIZED_COORDINATES));
+			}
+
+			cuda_pop_context();
+		}
+
+		/* Fermi, Data and Image Textures */
+		if(!has_bindless_textures) {
+			cuda_push_context();
+
 			cuda_assert(cuTexRefSetAddressMode(texref, 0, address_mode));
 			cuda_assert(cuTexRefSetAddressMode(texref, 1, address_mode));
 			if(mem.data_depth > 1) {
@@ -631,31 +739,8 @@ public:
 
 			cuda_pop_context();
 		}
-		else {
-			mem_alloc(mem, MEM_READ_ONLY);
-			mem_copy_to(mem);
-
-			cuda_push_context();
-
-			CUdeviceptr cumem;
-			size_t cubytes;
-
-			cuda_assert(cuModuleGetGlobal(&cumem, &cubytes, cuModule, bind_name.c_str()));
-
-			if(cubytes == 8) {
-				/* 64 bit device pointer */
-				uint64_t ptr = mem.device_pointer;
-				cuda_assert(cuMemcpyHtoD(cumem, (void*)&ptr, cubytes));
-			}
-			else {
-				/* 32 bit device pointer */
-				uint32_t ptr = (uint32_t)mem.device_pointer;
-				cuda_assert(cuMemcpyHtoD(cumem, (void*)&ptr, cubytes));
-			}
-
-			cuda_pop_context();
-		}
 
+		/* Fermi and Kepler */
 		tex_interp_map[mem.device_pointer] = (interpolation != INTERPOLATION_NONE);
 	}
 
@@ -667,6 +752,12 @@ public:
 				cuArrayDestroy((CUarray)mem.device_pointer);
 				cuda_pop_context();
 
+				/* Free CUtexObject (Bindless Textures) */
+				if(info.has_bindless_textures && tex_bindless_map[mem.device_pointer]) {
+					uint flat_slot = tex_bindless_map[mem.device_pointer];
+					cuTexObjectDestroy(flat_slot);
+				}
+
 				tex_interp_map.erase(tex_interp_map.find(mem.device_pointer));
 				mem.device_pointer = 0;
 
@@ -723,8 +814,8 @@ public:
 		printf("threads_per_block %d\n", threads_per_block);
 		printf("num_registers %d\n", num_registers);*/
 
-		int xthreads = (int)sqrt((float)threads_per_block);
-		int ythreads = (int)sqrt((float)threads_per_block);
+		int xthreads = (int)sqrt(threads_per_block);
+		int ythreads = (int)sqrt(threads_per_block);
 		int xblocks = (rtile.w + xthreads - 1)/xthreads;
 		int yblocks = (rtile.h + ythreads - 1)/ythreads;
 
@@ -777,8 +868,8 @@ public:
 		int threads_per_block;
 		cuda_assert(cuFuncGetAttribute(&threads_per_block, CU_FUNC_ATTRIBUTE_MAX_THREADS_PER_BLOCK, cuFilmConvert));
 
-		int xthreads = (int)sqrt((float)threads_per_block);
-		int ythreads = (int)sqrt((float)threads_per_block);
+		int xthreads = (int)sqrt(threads_per_block);
+		int ythreads = (int)sqrt(threads_per_block);
 		int xblocks = (task.w + xthreads - 1)/xthreads;
 		int yblocks = (task.h + ythreads - 1)/ythreads;
 
@@ -1108,6 +1199,9 @@ public:
 			RenderTile tile;
 			
 			bool branched = task->integrator_branched;
+
+			/* Upload Bindless Mapping */
+			load_bindless_mapping();
 			
 			/* keep rendering tiles until done */
 			while(task->acquire_tile(this, tile)) {
@@ -1131,6 +1225,9 @@ public:
 			}
 		}
 		else if(task->type == DeviceTask::SHADER) {
+			/* Upload Bindless Mapping */
+			load_bindless_mapping();
+
 			shader(*task);
 
 			cuda_push_context();
@@ -1266,11 +1363,12 @@ void device_cuda_info(vector<DeviceInfo>& devices)
 		info.num = num;
 
 		info.advanced_shading = (major >= 2);
-		info.extended_images = (major >= 3);
+		info.has_bindless_textures = (major >= 3);
 		info.pack_images = false;
 
 		/* if device has a kernel timeout, assume it is used for display */
 		if(cuDeviceGetAttribute(&attr, CU_DEVICE_ATTRIBUTE_KERNEL_EXEC_TIMEOUT, num) == CUDA_SUCCESS && attr == 1) {
+			info.description += " (Display)";
 			info.display_device = true;
 			display_devices.push_back(info);
 		}

intern/cycles/device/device_multi.cpp

@@ -35,7 +35,7 @@ class MultiDevice : public Device
 {
 public:
 	struct SubDevice {
-		SubDevice(Device *device_)
+		explicit SubDevice(Device *device_)
 		: device(device_) {}
 
 		Device *device;
@@ -175,7 +175,9 @@ public:
 	               interpolation,
 	               ExtensionType extension)
 	{
-		VLOG(1) << "Texture allocate: " << name << ", " << mem.memory_size() << " bytes.";
+		VLOG(1) << "Texture allocate: " << name << ", "
+		        << string_human_readable_number(mem.memory_size()) << " bytes. ("
+		        << string_human_readable_size(mem.memory_size()) << ")";
 
 		foreach(SubDevice& sub, devices) {
 			mem.device_pointer = 0;
@@ -352,7 +354,7 @@ static bool device_multi_add(vector<DeviceInfo>& devices, DeviceType type, bool
 
 	info.advanced_shading = with_advanced_shading;
 	info.pack_images = false;
-	info.extended_images = true;
+	info.has_bindless_textures = true;
 
 	foreach(DeviceInfo& subinfo, devices) {
 		if(subinfo.type == type) {
@@ -376,7 +378,7 @@ static bool device_multi_add(vector<DeviceInfo>& devices, DeviceType type, bool
 			if(subinfo.display_device)
 				info.display_device = true;
 			info.pack_images = info.pack_images || subinfo.pack_images;
-			info.extended_images = info.extended_images && subinfo.extended_images;
+			info.has_bindless_textures = info.has_bindless_textures && subinfo.has_bindless_textures;
 			num_added++;
 		}
 	}

intern/cycles/device/device_network.cpp

@@ -168,7 +168,9 @@ public:
 	               InterpolationType interpolation,
 	               ExtensionType extension)
 	{
-		VLOG(1) << "Texture allocate: " << name << ", " << mem.memory_size() << " bytes.";
+		VLOG(1) << "Texture allocate: " << name << ", "
+		        << string_human_readable_number(mem.memory_size()) << " bytes. ("
+		        << string_human_readable_size(mem.memory_size()) << ")";
 
 		thread_scoped_lock lock(rpc_lock);
 

intern/cycles/device/device_network.h

@@ -322,7 +322,7 @@ protected:
 
 class ServerDiscovery {
 public:
-	ServerDiscovery(bool discover = false)
+	explicit ServerDiscovery(bool discover = false)
 	: listen_socket(io_service), collect_servers(false)
 	{
 		/* setup listen socket */

intern/cycles/device/device_opencl.cpp

@@ -1187,7 +1187,9 @@ public:
 	               InterpolationType /*interpolation*/,
 	               ExtensionType /*extension*/)
 	{
-		VLOG(1) << "Texture allocate: " << name << ", " << mem.memory_size() << " bytes.";
+		VLOG(1) << "Texture allocate: " << name << ", "
+		        << string_human_readable_number(mem.memory_size()) << " bytes. ("
+		        << string_human_readable_size(mem.memory_size()) << ")";
 		mem_alloc(mem, MEM_READ_ONLY);
 		mem_copy_to(mem);
 		assert(mem_map.find(name) == mem_map.end());
@@ -1222,18 +1224,28 @@ public:
 			CL_KERNEL_WORK_GROUP_SIZE, sizeof(size_t), &workgroup_size, NULL);
 		clGetDeviceInfo(cdDevice,
 			CL_DEVICE_MAX_WORK_ITEM_SIZES, sizeof(size_t)*3, max_work_items, NULL);
-	
-		/* try to divide evenly over 2 dimensions */
+
+		/* Try to divide evenly over 2 dimensions. */
 		size_t sqrt_workgroup_size = max((size_t)sqrt((double)workgroup_size), 1);
 		size_t local_size[2] = {sqrt_workgroup_size, sqrt_workgroup_size};
 
-		/* some implementations have max size 1 on 2nd dimension */
+		/* Some implementations have max size 1 on 2nd dimension. */
 		if(local_size[1] > max_work_items[1]) {
 			local_size[0] = workgroup_size/max_work_items[1];
 			local_size[1] = max_work_items[1];
 		}
 
-		size_t global_size[2] = {global_size_round_up(local_size[0], w), global_size_round_up(local_size[1], h)};
+		size_t global_size[2] = {global_size_round_up(local_size[0], w),
+		                         global_size_round_up(local_size[1], h)};
+
+		/* Vertical size of 1 is coming from bake/shade kernels where we should
+		 * not round anything up because otherwise we'll either be doing too
+		 * much work per pixel (if we don't check global ID on Y axis) or will
+		 * be checking for global ID to always have Y of 0.
+		 */
+		if (h == 1) {
+			global_size[h] = 1;
+		}
 
 		/* run kernel */
 		opencl_assert(clEnqueueNDRangeKernel(cqCommandQueue, kernel, 2, NULL, global_size, NULL, 0, NULL, NULL));
@@ -1318,48 +1330,49 @@ public:
 		else
 			kernel = ckShaderKernel;
 
+		cl_uint start_arg_index =
+			kernel_set_args(kernel,
+			                0,
+			                d_data,
+			                d_input,
+			                d_output);
+
+		if(task.shader_eval_type < SHADER_EVAL_BAKE) {
+			start_arg_index += kernel_set_args(kernel,
+			                                   start_arg_index,
+			                                   d_output_luma);
+		}
+
+#define KERNEL_TEX(type, ttype, name) \
+		set_kernel_arg_mem(kernel, &start_arg_index, #name);
+#include "kernel_textures.h"
+#undef KERNEL_TEX
+
+		start_arg_index += kernel_set_args(kernel,
+		                                   start_arg_index,
+		                                   d_shader_eval_type);
+		if(task.shader_eval_type >= SHADER_EVAL_BAKE) {
+			start_arg_index += kernel_set_args(kernel,
+			                                   start_arg_index,
+			                                   d_shader_filter);
+		}
+		start_arg_index += kernel_set_args(kernel,
+		                                   start_arg_index,
+		                                   d_shader_x,
+		                                   d_shader_w,
+		                                   d_offset);
+
 		for(int sample = 0; sample < task.num_samples; sample++) {
 
 			if(task.get_cancel())
 				break;
 
-			cl_int d_sample = sample;
-
-			cl_uint start_arg_index =
-				kernel_set_args(kernel,
-				                0,
-				                d_data,
-				                d_input,
-				                d_output);
-
-			if(task.shader_eval_type < SHADER_EVAL_BAKE) {
-				start_arg_index += kernel_set_args(kernel,
-				                                   start_arg_index,
-				                                   d_output_luma);
-			}
-
-#define KERNEL_TEX(type, ttype, name) \
-			set_kernel_arg_mem(kernel, &start_arg_index, #name);
-#include "kernel_textures.h"
-#undef KERNEL_TEX
-
-			start_arg_index += kernel_set_args(kernel,
-			                                   start_arg_index,
-			                                   d_shader_eval_type);
-			if(task.shader_eval_type >= SHADER_EVAL_BAKE) {
-				start_arg_index += kernel_set_args(kernel,
-				                                   start_arg_index,
-				                                   d_shader_filter);
-			}
-			start_arg_index += kernel_set_args(kernel,
-			                                   start_arg_index,
-			                                   d_shader_x,
-			                                   d_shader_w,
-			                                   d_offset,
-			                                   d_sample);
+			kernel_set_args(kernel, start_arg_index, sample);
 
 			enqueue_kernel(kernel, task.shader_w, 1);
 
+			clFinish(cqCommandQueue);
+
 			task.update_progress(NULL);
 		}
 	}

intern/cycles/device/device_task.h

@@ -51,7 +51,7 @@ public:
 	int shader_filter;
 	int shader_x, shader_w;
 
-	DeviceTask(Type type = PATH_TRACE);
+	explicit DeviceTask(Type type = PATH_TRACE);
 
 	int get_subtask_count(int num, int max_size = 0);
 	void split(list<DeviceTask>& tasks, int num, int max_size = 0);

intern/cycles/graph/CMakeLists.txt

@@ -0,0 +1,24 @@
+
+set(INC
+	.
+	../util
+)
+
+set(SRC
+	node.cpp
+	node_type.cpp
+	node_xml.cpp
+)
+
+set(SRC_HEADERS
+	node.h
+	node_enum.h
+	node_type.h
+	node_xml.h
+)
+
+include_directories(${INC})
+include_directories(SYSTEM ${INC_SYS})
+
+add_library(cycles_graph ${SRC} ${SRC_HEADERS})
+

intern/cycles/graph/node.cpp

@@ -0,0 +1,407 @@
+/*
+ * Copyright 2011-2016 Blender Foundation
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "node.h"
+#include "node_type.h"
+
+#include "util_foreach.h"
+#include "util_param.h"
+#include "util_transform.h"
+
+CCL_NAMESPACE_BEGIN
+
+/* Node Type */
+
+Node::Node(const NodeType *type_, ustring name_)
+: name(name_), type(type_)
+{
+	assert(type);
+
+	/* assign non-empty name, convenient for debugging */
+	if(name.empty()) {
+		name = type->name;
+	}
+
+	/* initialize default values */
+	foreach(const SocketType& socket, type->inputs) {
+		set_default_value(socket);
+	}
+}
+
+Node::~Node()
+{
+}
+
+template<typename T>
+static T& get_socket_value(const Node *node, const SocketType& socket)
+{
+	return (T&)*(((char*)node) + socket.struct_offset);
+}
+
+#ifndef NDEBUG
+static bool is_socket_float3(const SocketType& socket)
+{
+	return socket.type == SocketType::COLOR ||
+	       socket.type == SocketType::POINT ||
+		   socket.type == SocketType::VECTOR ||
+		   socket.type == SocketType::NORMAL;
+}
+
+static bool is_socket_array_float3(const SocketType& socket)
+{
+	return socket.type == SocketType::COLOR_ARRAY ||
+	       socket.type == SocketType::POINT_ARRAY ||
+		   socket.type == SocketType::VECTOR_ARRAY ||
+		   socket.type == SocketType::NORMAL_ARRAY;
+}
+#endif
+
+/* set values */
+void Node::set(const SocketType& input, bool value)
+{
+	assert(input.type == SocketType::BOOLEAN);
+	get_socket_value<bool>(this, input) = value;
+}
+
+void Node::set(const SocketType& input, int value)
+{
+	assert((input.type == SocketType::INT || input.type == SocketType::ENUM));
+	get_socket_value<int>(this, input) = value;
+}
+
+void Node::set(const SocketType& input, uint value)
+{
+	assert(input.type == SocketType::UINT);
+	get_socket_value<uint>(this, input) = value;
+}
+
+void Node::set(const SocketType& input, float value)
+{
+	assert(input.type == SocketType::FLOAT);
+	get_socket_value<float>(this, input) = value;
+}
+
+void Node::set(const SocketType& input, float2 value)
+{
+	assert(input.type == SocketType::FLOAT);
+	get_socket_value<float2>(this, input) = value;
+}
+
+void Node::set(const SocketType& input, float3 value)
+{
+	assert(is_socket_float3(input));
+	get_socket_value<float3>(this, input) = value;
+}
+
+void Node::set(const SocketType& input, const char *value)
+{
+	set(input, ustring(value));
+}
+
+void Node::set(const SocketType& input, ustring value)
+{
+	if(input.type == SocketType::STRING) {
+		get_socket_value<ustring>(this, input) = value;
+	}
+	else if(input.type == SocketType::ENUM) {
+		const NodeEnum& enm = *input.enum_values;
+		if(enm.exists(value)) {
+			get_socket_value<int>(this, input) = enm[value];
+		}
+		else {
+			assert(0);
+		}
+	}
+	else {
+		assert(0);
+	}
+}
+
+void Node::set(const SocketType& input, const Transform& value)
+{
+	assert(input.type == SocketType::TRANSFORM);
+	get_socket_value<Transform>(this, input) = value;
+}
+
+void Node::set(const SocketType& input, Node *value)
+{
+	assert(input.type == SocketType::TRANSFORM);
+	get_socket_value<Node*>(this, input) = value;
+}
+
+/* set array values */
+void Node::set(const SocketType& input, array<bool>& value)
+{
+	assert(input.type == SocketType::BOOLEAN_ARRAY);
+	get_socket_value<array<bool> >(this, input).steal_data(value);
+}
+
+void Node::set(const SocketType& input, array<int>& value)
+{
+	assert(input.type == SocketType::INT_ARRAY);
+	get_socket_value<array<int> >(this, input).steal_data(value);
+}
+
+void Node::set(const SocketType& input, array<float>& value)
+{
+	assert(input.type == SocketType::FLOAT_ARRAY);
+	get_socket_value<array<float> >(this, input).steal_data(value);
+}
+
+void Node::set(const SocketType& input, array<float2>& value)
+{
+	assert(input.type == SocketType::FLOAT_ARRAY);
+	get_socket_value<array<float2> >(this, input).steal_data(value);
+}
+
+void Node::set(const SocketType& input, array<float3>& value)
+{
+	assert(is_socket_array_float3(input));
+	get_socket_value<array<float3> >(this, input).steal_data(value);
+}
+
+void Node::set(const SocketType& input, array<ustring>& value)
+{
+	assert(input.type == SocketType::STRING_ARRAY);
+	get_socket_value<array<ustring> >(this, input).steal_data(value);
+}
+
+void Node::set(const SocketType& input, array<Transform>& value)
+{
+	assert(input.type == SocketType::TRANSFORM_ARRAY);
+	get_socket_value<array<Transform> >(this, input).steal_data(value);
+}
+
+void Node::set(const SocketType& input, array<Node*>& value)
+{
+	assert(input.type == SocketType::TRANSFORM_ARRAY);
+	get_socket_value<array<Node*> >(this, input).steal_data(value);
+}
+
+/* get values */
+bool Node::get_bool(const SocketType& input) const
+{
+	assert(input.type == SocketType::BOOLEAN);
+	return get_socket_value<bool>(this, input);
+}
+
+int Node::get_int(const SocketType& input) const
+{
+	assert(input.type == SocketType::INT || input.type == SocketType::ENUM);
+	return get_socket_value<int>(this, input);
+}
+
+uint Node::get_uint(const SocketType& input) const
+{
+	assert(input.type == SocketType::UINT);
+	return get_socket_value<uint>(this, input);
+}
+
+float Node::get_float(const SocketType& input) const
+{
+	assert(input.type == SocketType::FLOAT);
+	return get_socket_value<float>(this, input);
+}
+
+float2 Node::get_float2(const SocketType& input) const
+{
+	assert(input.type == SocketType::FLOAT);
+	return get_socket_value<float2>(this, input);
+}
+
+float3 Node::get_float3(const SocketType& input) const
+{
+	assert(is_socket_float3(input));
+	return get_socket_value<float3>(this, input);
+}
+
+ustring Node::get_string(const SocketType& input) const
+{
+	if(input.type == SocketType::STRING) {
+		return get_socket_value<ustring>(this, input);
+	}
+	else if(input.type == SocketType::ENUM) {
+		const NodeEnum& enm = *input.enum_values;
+		int intvalue = get_socket_value<int>(this, input);
+		return (enm.exists(intvalue)) ? enm[intvalue] : ustring();
+	}
+	else {
+		assert(0);
+		return ustring();
+	}
+}
+
+Transform Node::get_transform(const SocketType& input) const
+{
+	assert(input.type == SocketType::TRANSFORM);
+	return get_socket_value<Transform>(this, input);
+}
+
+Node *Node::get_node(const SocketType& input) const
+{
+	assert(input.type == SocketType::NODE);
+	return get_socket_value<Node*>(this, input);
+}
+
+/* get array values */
+const array<bool>& Node::get_bool_array(const SocketType& input) const
+{
+	assert(input.type == SocketType::BOOLEAN_ARRAY);
+	return get_socket_value<array<bool> >(this, input);
+}
+
+const array<int>& Node::get_int_array(const SocketType& input) const
+{
+	assert(input.type == SocketType::INT_ARRAY);
+	return get_socket_value<array<int> >(this, input);
+}
+
+const array<float>& Node::get_float_array(const SocketType& input) const
+{
+	assert(input.type == SocketType::FLOAT_ARRAY);
+	return get_socket_value<array<float> >(this, input);
+}
+
+const array<float2>& Node::get_float2_array(const SocketType& input) const
+{
+	assert(input.type == SocketType::FLOAT_ARRAY);
+	return get_socket_value<array<float2> >(this, input);
+}
+
+const array<float3>& Node::get_float3_array(const SocketType& input) const
+{
+	assert(is_socket_array_float3(input));
+	return get_socket_value<array<float3> >(this, input);
+}
+
+const array<ustring>& Node::get_string_array(const SocketType& input) const
+{
+	assert(input.type == SocketType::STRING_ARRAY);
+	return get_socket_value<array<ustring> >(this, input);
+}
+
+const array<Transform>& Node::get_transform_array(const SocketType& input) const
+{
+	assert(input.type == SocketType::TRANSFORM_ARRAY);
+	return get_socket_value<array<Transform> >(this, input);
+}
+
+const array<Node*>& Node::get_node_array(const SocketType& input) const
+{
+	assert(input.type == SocketType::NODE_ARRAY);
+	return get_socket_value<array<Node*> >(this, input);
+}
+
+/* generic value operations */
+
+bool Node::has_default_value(const SocketType& input) const
+{
+	const void *src = input.default_value;
+	void *dst = &get_socket_value<char>(this, input);
+	return memcmp(dst, src, input.size()) == 0;
+}
+
+void Node::set_default_value(const SocketType& socket)
+{
+	const void *src = socket.default_value;
+	void *dst = ((char*)this) + socket.struct_offset;
+	memcpy(dst, src, socket.size());
+}
+
+template<typename T>
+static void copy_array(const Node *node, const SocketType& socket, const Node *other, const SocketType& other_socket)
+{
+	const array<T>* src = (const array<T>*)(((char*)other) + other_socket.struct_offset);
+	array<T>* dst = (array<T>*)(((char*)node) + socket.struct_offset);
+	*dst = *src;
+}
+
+void Node::copy_value(const SocketType& socket, const Node& other, const SocketType& other_socket)
+{
+	assert(socket.type == other_socket.type);
+
+	if(socket.is_array()) {
+		switch(socket.type) {
+			case SocketType::BOOLEAN_ARRAY: copy_array<bool>(this, socket, &other, other_socket); break;
+			case SocketType::FLOAT_ARRAY: copy_array<float>(this, socket, &other, other_socket); break;
+			case SocketType::INT_ARRAY: copy_array<int>(this, socket, &other, other_socket); break;
+			case SocketType::COLOR_ARRAY: copy_array<float3>(this, socket, &other, other_socket); break;
+			case SocketType::VECTOR_ARRAY: copy_array<float3>(this, socket, &other, other_socket); break;
+			case SocketType::POINT_ARRAY: copy_array<float3>(this, socket, &other, other_socket); break;
+			case SocketType::NORMAL_ARRAY: copy_array<float3>(this, socket, &other, other_socket); break;
+			case SocketType::POINT2_ARRAY: copy_array<float2>(this, socket, &other, other_socket); break;
+			case SocketType::STRING_ARRAY: copy_array<ustring>(this, socket, &other, other_socket); break;
+			case SocketType::TRANSFORM_ARRAY: copy_array<Transform>(this, socket, &other, other_socket); break;
+			case SocketType::NODE_ARRAY: copy_array<void*>(this, socket, &other, other_socket); break;
+			default: assert(0); break;
+		}
+	}
+	else {
+		const void *src = ((char*)&other) + other_socket.struct_offset;
+		void *dst = ((char*)this) + socket.struct_offset;
+		memcpy(dst, src, socket.size());
+	}
+}
+
+template<typename T>
+static bool is_array_equal(const Node *node, const Node *other, const SocketType& socket)
+{
+	const array<T>* a = (const array<T>*)(((char*)node) + socket.struct_offset);
+	const array<T>* b = (const array<T>*)(((char*)other) + socket.struct_offset);
+	return *a == *b;
+}
+
+bool Node::equals_value(const Node& other, const SocketType& socket) const
+{
+	if(socket.is_array()) {
+		switch(socket.type) {
+			case SocketType::BOOLEAN_ARRAY: return is_array_equal<bool>(this, &other, socket);
+			case SocketType::FLOAT_ARRAY: return is_array_equal<float>(this, &other, socket);
+			case SocketType::INT_ARRAY: return is_array_equal<int>(this, &other, socket);
+			case SocketType::COLOR_ARRAY: return is_array_equal<float3>(this, &other, socket);
+			case SocketType::VECTOR_ARRAY: return is_array_equal<float3>(this, &other, socket);
+			case SocketType::POINT_ARRAY: return is_array_equal<float3>(this, &other, socket);
+			case SocketType::NORMAL_ARRAY: return is_array_equal<float3>(this, &other, socket);
+			case SocketType::POINT2_ARRAY: return is_array_equal<float2>(this, &other, socket);
+			case SocketType::STRING_ARRAY: return is_array_equal<ustring>(this, &other, socket);
+			case SocketType::TRANSFORM_ARRAY: return is_array_equal<Transform>(this, &other, socket);
+			case SocketType::NODE_ARRAY: return is_array_equal<void*>(this, &other, socket);
+			default: assert(0); return true;
+		}
+	}
+	else {
+		const void *a = ((char*)this) + socket.struct_offset;
+		const void *b = ((char*)&other) + socket.struct_offset;
+		return (memcmp(a, b, socket.size()) == 0);
+	}
+}
+
+/* equals */
+
+bool Node::equals(const Node& other) const
+{
+	assert(type == other.type);
+
+	foreach(const SocketType& socket, type->inputs) {
+		if(!equals_value(other, socket))
+			return false;
+	}
+
+	return true;
+}
+
+CCL_NAMESPACE_END
+

intern/cycles/graph/node.h

@@ -0,0 +1,96 @@
+/*
+ * Copyright 2011-2016 Blender Foundation
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#pragma once
+
+#include "node_type.h"
+
+#include "util_map.h"
+#include "util_param.h"
+#include "util_vector.h"
+
+CCL_NAMESPACE_BEGIN
+
+struct Node;
+struct NodeType;
+struct Transform;
+
+/* Node */
+
+struct Node
+{
+	explicit Node(const NodeType *type, ustring name = ustring());
+	virtual ~Node();
+
+	/* set values */
+	void set(const SocketType& input, bool value);
+	void set(const SocketType& input, int value);
+	void set(const SocketType& input, uint value);
+	void set(const SocketType& input, float value);
+	void set(const SocketType& input, float2 value);
+	void set(const SocketType& input, float3 value);
+	void set(const SocketType& input, const char *value);
+	void set(const SocketType& input, ustring value);
+	void set(const SocketType& input, const Transform& value);
+	void set(const SocketType& input, Node *value);
+
+	/* set array values. the memory from the input array will taken over
+	 * by the node and the input array will be empty after return */
+	void set(const SocketType& input, array<bool>& value);
+	void set(const SocketType& input, array<int>& value);
+	void set(const SocketType& input, array<float>& value);
+	void set(const SocketType& input, array<float2>& value);
+	void set(const SocketType& input, array<float3>& value);
+	void set(const SocketType& input, array<ustring>& value);
+	void set(const SocketType& input, array<Transform>& value);
+	void set(const SocketType& input, array<Node*>& value);
+
+	/* get values */
+	bool get_bool(const SocketType& input) const;
+	int get_int(const SocketType& input) const;
+	uint get_uint(const SocketType& input) const;
+	float get_float(const SocketType& input) const;
+	float2 get_float2(const SocketType& input) const;
+	float3 get_float3(const SocketType& input) const;
+	ustring get_string(const SocketType& input) const;
+	Transform get_transform(const SocketType& input) const;
+	Node *get_node(const SocketType& input) const;
+
+	/* get array values */
+	const array<bool>& get_bool_array(const SocketType& input) const;
+	const array<int>& get_int_array(const SocketType& input) const;
+	const array<float>& get_float_array(const SocketType& input) const;
+	const array<float2>& get_float2_array(const SocketType& input) const;
+	const array<float3>& get_float3_array(const SocketType& input) const;
+	const array<ustring>& get_string_array(const SocketType& input) const;
+	const array<Transform>& get_transform_array(const SocketType& input) const;
+	const array<Node*>& get_node_array(const SocketType& input) const;
+
+	/* generic values operations */
+	bool has_default_value(const SocketType& input) const;
+	void set_default_value(const SocketType& input);
+	bool equals_value(const Node& other, const SocketType& input) const;
+	void copy_value(const SocketType& input, const Node& other, const SocketType& other_input);
+
+	/* equals */
+	bool equals(const Node& other) const;
+
+	ustring name;
+	const NodeType *type;
+};
+
+CCL_NAMESPACE_END
+

intern/cycles/graph/node_enum.h

@@ -0,0 +1,48 @@
+/*
+ * Copyright 2011-2016 Blender Foundation
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#pragma once
+
+#include "util_map.h"
+#include "util_param.h"
+
+CCL_NAMESPACE_BEGIN
+
+/* Enum
+ *
+ * Utility class for enum values. */
+
+struct NodeEnum {
+	bool empty() const { return left.empty(); }
+	void insert(const char *x, int y) {
+		left[ustring(x)] = y;
+		right[y] = ustring(x);
+	}
+
+	bool exists(ustring x) const { return left.find(x) != left.end(); }
+	bool exists(int y) const { return right.find(y) != right.end(); }
+
+	int operator[](const char *x) const { return left.find(ustring(x))->second; }
+	int operator[](ustring x) const { return left.find(x)->second; }
+	ustring operator[](int y) const { return right.find(y)->second; }
+
+private:
+	unordered_map<ustring, int, ustringHash> left;
+	unordered_map<int, ustring> right;
+};
+
+CCL_NAMESPACE_END
+

intern/cycles/graph/node_type.cpp

@@ -0,0 +1,219 @@
+/*
+ * Copyright 2011-2016 Blender Foundation
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "node_type.h"
+#include "util_foreach.h"
+#include "util_transform.h"
+
+CCL_NAMESPACE_BEGIN
+
+/* Node Socket Type */
+
+size_t SocketType::size() const
+{
+	return size(type);
+}
+
+bool SocketType::is_array() const
+{
+	return (type >= BOOLEAN_ARRAY);
+}
+
+size_t SocketType::size(Type type)
+{
+	switch(type)
+	{
+		case UNDEFINED: return 0;
+
+		case BOOLEAN: return sizeof(bool);
+		case FLOAT: return sizeof(float);
+		case INT: return sizeof(int);
+		case UINT: return sizeof(uint);
+		case COLOR: return sizeof(float3);
+		case VECTOR: return sizeof(float3);
+		case POINT: return sizeof(float3);
+		case NORMAL: return sizeof(float3);
+		case POINT2: return sizeof(float2);
+		case CLOSURE: return 0;
+		case STRING: return sizeof(ustring);
+		case ENUM: return sizeof(int);
+		case TRANSFORM: return sizeof(Transform);
+		case NODE: return sizeof(void*);
+
+		case BOOLEAN_ARRAY: return sizeof(array<bool>);
+		case FLOAT_ARRAY: return sizeof(array<float>);
+		case INT_ARRAY: return sizeof(array<int>);
+		case COLOR_ARRAY: return sizeof(array<float3>);
+		case VECTOR_ARRAY: return sizeof(array<float3>);
+		case POINT_ARRAY: return sizeof(array<float3>);
+		case NORMAL_ARRAY: return sizeof(array<float3>);
+		case POINT2_ARRAY: return sizeof(array<float2>);
+		case STRING_ARRAY: return sizeof(array<ustring>);
+		case TRANSFORM_ARRAY: return sizeof(array<Transform>);
+		case NODE_ARRAY: return sizeof(array<void*>);
+	}
+
+	assert(0);
+	return 0;
+}
+
+size_t SocketType::max_size()
+{
+	return sizeof(Transform);
+}
+
+void *SocketType::zero_default_value()
+{
+	static Transform zero_transform = {{0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0}};
+	return &zero_transform;
+}
+
+ustring SocketType::type_name(Type type)
+{
+	static ustring names[] = {
+		ustring("undefined"),
+
+		ustring("boolean"),
+		ustring("float"),
+		ustring("int"),
+		ustring("color"),
+		ustring("vector"),
+		ustring("point"),
+		ustring("normal"),
+		ustring("point2"),
+		ustring("closure"),
+		ustring("string"),
+		ustring("enum"),
+		ustring("transform"),
+		ustring("node"),
+
+		ustring("array_boolean"),
+		ustring("array_float"),
+		ustring("array_int"),
+		ustring("array_color"),
+		ustring("array_vector"),
+		ustring("array_point"),
+		ustring("array_normal"),
+		ustring("array_point2"),
+		ustring("array_string"),
+		ustring("array_transform"),
+		ustring("array_node")};
+
+	return names[(int)type];
+}
+
+bool SocketType::is_float3(Type type)
+{
+	return (type == COLOR || type == VECTOR || type == POINT || type == NORMAL);
+}
+
+/* Node Type */
+
+NodeType::NodeType(Type type_)
+: type(type_)
+{
+}
+
+NodeType::~NodeType()
+{
+}
+
+void NodeType::register_input(ustring name, ustring ui_name, SocketType::Type type, int struct_offset,
+                              const void *default_value, const NodeEnum *enum_values,
+							  const NodeType **node_type, int flags, int extra_flags)
+{
+	SocketType socket;
+	socket.name = name;
+	socket.ui_name = ui_name;
+	socket.type = type;
+	socket.struct_offset = struct_offset;
+	socket.default_value = default_value;
+	socket.enum_values = enum_values;
+	socket.node_type = node_type;
+	socket.flags = flags | extra_flags;
+	inputs.push_back(socket);
+}
+
+void NodeType::register_output(ustring name, ustring ui_name, SocketType::Type type)
+{
+	SocketType socket;
+	socket.name = name;
+	socket.ui_name = ui_name;
+	socket.type = type;
+	socket.struct_offset = 0;
+	socket.default_value = NULL;
+	socket.enum_values = NULL;
+	socket.node_type = NULL;
+	socket.flags = SocketType::LINKABLE;
+	outputs.push_back(socket);
+}
+
+const SocketType *NodeType::find_input(ustring name) const
+{
+	foreach(const SocketType& socket, inputs) {
+		if(socket.name == name) {
+			return &socket;
+		}
+	}
+
+	return NULL;
+}
+
+const SocketType *NodeType::find_output(ustring name) const
+{
+	foreach(const SocketType& socket, outputs) {
+		if(socket.name == name) {
+			return &socket;
+		}
+	}
+
+	return NULL;
+}
+
+/* Node Type Registry */
+
+unordered_map<ustring, NodeType, ustringHash>& NodeType::types()
+{
+	static unordered_map<ustring, NodeType, ustringHash> _types;
+	return _types;
+}
+
+NodeType *NodeType::add(const char *name_, CreateFunc create_, Type type_)
+{
+	ustring name(name_);
+
+	if(types().find(name) != types().end()) {
+		fprintf(stderr, "Node type %s registered twice!\n", name_);
+		assert(0);
+		return NULL;
+	}
+
+	types()[name] = NodeType(type_);
+
+	NodeType *type = &types()[name];
+	type->name = name;
+	type->create = create_;
+	return type;
+}
+
+const NodeType *NodeType::find(ustring name)
+{
+	unordered_map<ustring, NodeType, ustringHash>::iterator it = types().find(name);
+	return (it == types().end()) ? NULL : &it->second;
+}
+
+CCL_NAMESPACE_END
+

intern/cycles/graph/node_type.h

@@ -0,0 +1,265 @@
+/*
+ * Copyright 2011-2016 Blender Foundation
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#pragma once
+
+#include "node_enum.h"
+
+#include "util_map.h"
+#include "util_param.h"
+#include "util_string.h"
+#include "util_vector.h"
+
+CCL_NAMESPACE_BEGIN
+
+struct Node;
+struct NodeType;
+
+/* Socket Type */
+
+struct SocketType
+{
+	enum Type
+	{
+		UNDEFINED,
+
+		BOOLEAN,
+		FLOAT,
+		INT,
+		UINT,
+		COLOR,
+		VECTOR,
+		POINT,
+		NORMAL,
+		POINT2,
+		CLOSURE,
+		STRING,
+		ENUM,
+		TRANSFORM,
+		NODE,
+
+		BOOLEAN_ARRAY,
+		FLOAT_ARRAY,
+		INT_ARRAY,
+		COLOR_ARRAY,
+		VECTOR_ARRAY,
+		POINT_ARRAY,
+		NORMAL_ARRAY,
+		POINT2_ARRAY,
+		STRING_ARRAY,
+		TRANSFORM_ARRAY,
+		NODE_ARRAY,
+	};
+
+	enum Flags {
+		LINKABLE               = (1 << 0),
+		ANIMATABLE             = (1 << 1),
+
+		SVM_INTERNAL           = (1 << 2),
+		OSL_INTERNAL           = (1 << 3),
+		INTERNAL               = (1 << 2) | (1 << 3),
+
+		LINK_TEXTURE_GENERATED = (1 << 4),
+		LINK_TEXTURE_UV        = (1 << 5),
+		LINK_INCOMING          = (1 << 6),
+		LINK_NORMAL            = (1 << 7),
+		LINK_POSITION          = (1 << 8),
+		LINK_TANGENT           = (1 << 9),
+		DEFAULT_LINK_MASK      = (1 << 4) | (1 << 5) | (1 << 6) | (1 << 7) | (1 << 8) | (1 << 9)
+	};
+
+	ustring name;
+	Type type;
+	int struct_offset;
+	const void *default_value;
+	const NodeEnum *enum_values;
+	const NodeType **node_type;
+	int flags;
+	ustring ui_name;
+
+	size_t size() const;
+	bool is_array() const;
+	static size_t size(Type type);
+	static size_t max_size();
+	static ustring type_name(Type type);
+	static void *zero_default_value();
+	static bool is_float3(Type type);
+};
+
+/* Node Type */
+
+struct NodeType
+{
+	enum Type {
+		NONE,
+		SHADER
+	};
+
+	explicit NodeType(Type type = NONE);
+	~NodeType();
+
+	void register_input(ustring name, ustring ui_name, SocketType::Type type,
+	                    int struct_offset, const void *default_value,
+						const NodeEnum *enum_values = NULL,
+						const NodeType **node_type = NULL,
+						int flags = 0, int extra_flags = 0);
+	void register_output(ustring name, ustring ui_name, SocketType::Type type);
+
+	const SocketType *find_input(ustring name) const;
+	const SocketType *find_output(ustring name) const;
+
+	typedef Node *(*CreateFunc)(const NodeType *type);
+
+	ustring name;
+	Type type;
+	std::vector<SocketType> inputs;
+	std::vector<SocketType> outputs;
+	CreateFunc create;
+
+	static NodeType *add(const char *name, CreateFunc create, Type type = NONE);
+	static const NodeType *find(ustring name);
+	static unordered_map<ustring, NodeType, ustringHash>& types();
+};
+
+/* Node Definition Macros */
+
+#define NODE_DECLARE                       \
+template<typename T>                       \
+static const NodeType *register_type();    \
+static Node *create(const NodeType *type); \
+static const NodeType *node_type;
+
+#define NODE_DEFINE(structname)                                                  \
+const NodeType *structname::node_type = structname::register_type<structname>(); \
+Node *structname::create(const NodeType*) { return new structname(); }           \
+template<typename T>                                                             \
+const NodeType *structname::register_type()
+
+/* Sock Definition Macros */
+
+#define SOCKET_OFFSETOF(T, name) (((char *)&(((T *)1)->name)) - (char *)1)
+#define SOCKET_SIZEOF(T, name) (sizeof(((T *)1)->name))
+#define SOCKET_DEFINE(name, ui_name, default_value, datatype, TYPE, flags, ...) \
+	{ \
+		static datatype defval = default_value; \
+		CHECK_TYPE(((T *)1)->name, datatype); \
+		type->register_input(ustring(#name), ustring(ui_name), TYPE, SOCKET_OFFSETOF(T, name), &defval, NULL, NULL, flags, ##__VA_ARGS__); \
+	}
+
+#define SOCKET_BOOLEAN(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, bool, SocketType::BOOLEAN, 0, ##__VA_ARGS__)
+#define SOCKET_INT(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, int, SocketType::INT, 0, ##__VA_ARGS__)
+#define SOCKET_UINT(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, uint, SocketType::UINT, 0, ##__VA_ARGS__)
+#define SOCKET_FLOAT(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, float, SocketType::FLOAT, 0, ##__VA_ARGS__)
+#define SOCKET_COLOR(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, float3, SocketType::COLOR, 0, ##__VA_ARGS__)
+#define SOCKET_VECTOR(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, float3, SocketType::VECTOR, 0, ##__VA_ARGS__)
+#define SOCKET_POINT(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, float3, SocketType::POINT, 0, ##__VA_ARGS__)
+#define SOCKET_NORMAL(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, float3, SocketType::NORMAL, 0, ##__VA_ARGS__)
+#define SOCKET_POINT2(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, float2, SocketType::POINT2, 0, ##__VA_ARGS__)
+#define SOCKET_STRING(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, ustring, SocketType::STRING, 0, ##__VA_ARGS__)
+#define SOCKET_TRANSFORM(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, Transform, SocketType::TRANSFORM, 0, ##__VA_ARGS__)
+#define SOCKET_ENUM(name, ui_name, values, default_value, ...) \
+	{ \
+		static int defval = default_value; \
+		assert(SOCKET_SIZEOF(T, name) == sizeof(int)); \
+		type->register_input(ustring(#name), ustring(ui_name), SocketType::ENUM, SOCKET_OFFSETOF(T, name), &defval, &values, NULL, ##__VA_ARGS__); \
+	}
+#define SOCKET_NODE(name, ui_name, node_type, ...) \
+	{ \
+	    static Node *defval = NULL; \
+		assert(SOCKET_SIZEOF(T, name) == sizeof(Node*)); \
+		type->register_input(ustring(#name), ustring(ui_name), SocketType::NODE, SOCKET_OFFSETOF(T, name), &defval, NULL, node_type, ##__VA_ARGS__); \
+	}
+
+#define SOCKET_BOOLEAN_ARRAY(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, array<bool>, SocketType::BOOLEAN_ARRAY, 0, ##__VA_ARGS__)
+#define SOCKET_INT_ARRAY(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, array<int>, SocketType::INT_ARRAY, 0, ##__VA_ARGS__)
+#define SOCKET_FLOAT_ARRAY(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, array<float>, SocketType::FLOAT_ARRAY, 0, ##__VA_ARGS__)
+#define SOCKET_COLOR_ARRAY(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, array<float3>, SocketType::COLOR_ARRAY, 0, ##__VA_ARGS__)
+#define SOCKET_VECTOR_ARRAY(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, array<float3>, SocketType::VECTOR_ARRAY, 0, ##__VA_ARGS__)
+#define SOCKET_POINT_ARRAY(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, array<float3>, SocketType::POINT_ARRAY, 0, ##__VA_ARGS__)
+#define SOCKET_NORMAL_ARRAY(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, array<float3>, SocketType::NORMAL_ARRAY, 0, ##__VA_ARGS__)
+#define SOCKET_POINT2_ARRAY(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, array<float2>, SocketType::POINT2_ARRAY, 0, ##__VA_ARGS__)
+#define SOCKET_STRING_ARRAY(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, array<ustring>, SocketType::STRING_ARRAY, 0, ##__VA_ARGS__)
+#define SOCKET_TRANSFORM_ARRAY(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, array<Transform>, SocketType::TRANSFORM_ARRAY, 0, ##__VA_ARGS__)
+#define SOCKET_NODE_ARRAY(name, ui_name, node_type, ...) \
+	{ \
+	    static Node *defval = NULL; \
+		assert(SOCKET_SIZEOF(T, name) == sizeof(Node*)); \
+		type->register_input(ustring(#name), ustring(ui_name), SocketType::NODE_ARRAY, SOCKET_OFFSETOF(T, name), &defval, NULL, node_type, ##__VA_ARGS__); \
+	}
+
+#define SOCKET_IN_BOOLEAN(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, bool, SocketType::BOOLEAN, SocketType::LINKABLE, ##__VA_ARGS__)
+#define SOCKET_IN_INT(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, int, SocketType::INT, SocketType::LINKABLE, ##__VA_ARGS__)
+#define SOCKET_IN_FLOAT(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, float, SocketType::FLOAT, SocketType::LINKABLE, ##__VA_ARGS__)
+#define SOCKET_IN_COLOR(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, float3, SocketType::COLOR, SocketType::LINKABLE, ##__VA_ARGS__)
+#define SOCKET_IN_VECTOR(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, float3, SocketType::VECTOR, SocketType::LINKABLE, ##__VA_ARGS__)
+#define SOCKET_IN_POINT(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, float3, SocketType::POINT, SocketType::LINKABLE, ##__VA_ARGS__)
+#define SOCKET_IN_NORMAL(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, float3, SocketType::NORMAL, SocketType::LINKABLE, ##__VA_ARGS__)
+#define SOCKET_IN_STRING(name, ui_name, default_value, ...) \
+	SOCKET_DEFINE(name, ui_name, default_value, ustring, SocketType::STRING, SocketType::LINKABLE, ##__VA_ARGS__)
+#define SOCKET_IN_CLOSURE(name, ui_name, ...) \
+	type->register_input(ustring(#name), ustring(ui_name), SocketType::CLOSURE, 0, NULL, NULL, NULL, SocketType::LINKABLE, ##__VA_ARGS__)
+
+#define SOCKET_OUT_BOOLEAN(name, ui_name) \
+	{ type->register_output(ustring(#name), ustring(ui_name), SocketType::BOOLEAN); }
+#define SOCKET_OUT_INT(name, ui_name) \
+	{ type->register_output(ustring(#name), ustring(ui_name), SocketType::INT); }
+#define SOCKET_OUT_FLOAT(name, ui_name) \
+	{ type->register_output(ustring(#name), ustring(ui_name), SocketType::FLOAT); }
+#define SOCKET_OUT_COLOR(name, ui_name) \
+	{ type->register_output(ustring(#name), ustring(ui_name), SocketType::COLOR); }
+#define SOCKET_OUT_VECTOR(name, ui_name) \
+	{ type->register_output(ustring(#name), ustring(ui_name), SocketType::VECTOR); }
+#define SOCKET_OUT_POINT(name, ui_name) \
+	{ type->register_output(ustring(#name), ustring(ui_name), SocketType::POINT); }
+#define SOCKET_OUT_NORMAL(name, ui_name) \
+	{ type->register_output(ustring(#name), ustring(ui_name), SocketType::NORMAL); }
+#define SOCKET_OUT_CLOSURE(name, ui_name) \
+	{ type->register_output(ustring(#name), ustring(ui_name), SocketType::CLOSURE); }
+#define SOCKET_OUT_STRING(name, ui_name) \
+	{ type->register_output(ustring(#name), ustring(ui_name), SocketType::STRING); }
+#define SOCKET_OUT_ENUM(name, ui_name) \
+	{ type->register_output(ustring(#name), ustring(ui_name), SocketType::ENUM); }
+
+CCL_NAMESPACE_END
+

intern/cycles/graph/node_xml.cpp

@@ -0,0 +1,462 @@
+/*
+ * Copyright 2011-2016 Blender Foundation
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "node_xml.h"
+
+#include "util_foreach.h"
+#include "util_string.h"
+#include "util_transform.h"
+
+CCL_NAMESPACE_BEGIN
+
+static bool xml_read_boolean(const char *value)
+{
+	return string_iequals(value, "true") || (atoi(value) != 0);
+}
+
+static const char *xml_write_boolean(bool value)
+{
+	return (value) ? "true" : "false";
+}
+
+template<int VECTOR_SIZE, typename T>
+static void xml_read_float_array(T& value, pugi::xml_attribute attr)
+{
+	vector<string> tokens;
+	string_split(tokens, attr.value());
+
+	if(tokens.size() % VECTOR_SIZE != 0) {
+		return;
+	}
+
+	value.resize(tokens.size() / VECTOR_SIZE);
+	for(size_t i = 0; i < value.size(); i++) {
+		float *value_float = (float*)&value[i];
+
+		for(size_t j = 0; j < VECTOR_SIZE; j++)
+			value_float[j] = (float)atof(tokens[i * VECTOR_SIZE + j].c_str());
+	}
+}
+
+void xml_read_node(XMLReader& reader, Node *node, pugi::xml_node xml_node)
+{
+	pugi::xml_attribute name_attr = xml_node.attribute("name");
+	if(name_attr) {
+		node->name = ustring(name_attr.value());
+	}
+
+	foreach(const SocketType& socket, node->type->inputs) {
+		if(socket.type == SocketType::CLOSURE || socket.type == SocketType::UNDEFINED) {
+			continue;
+		}
+		if(socket.flags & SocketType::INTERNAL) {
+			continue;
+		}
+
+		pugi::xml_attribute attr = xml_node.attribute(socket.name.c_str());
+
+		if(!attr) {
+			continue;
+		}
+
+		switch(socket.type)
+		{
+			case SocketType::BOOLEAN:
+			{
+				node->set(socket, xml_read_boolean(attr.value()));
+				break;
+			}
+			case SocketType::BOOLEAN_ARRAY:
+			{
+				vector<string> tokens;
+				string_split(tokens, attr.value());
+
+				array<bool> value;
+				value.resize(tokens.size());
+				for(size_t i = 0; i < value.size(); i++)
+					value[i] = xml_read_boolean(tokens[i].c_str());
+				node->set(socket, value);
+				break;
+			}
+			case SocketType::FLOAT:
+			{
+				node->set(socket, (float)atof(attr.value()));
+				break;
+			}
+			case SocketType::FLOAT_ARRAY:
+			{
+				array<float> value;
+				xml_read_float_array<1>(value, attr);
+				node->set(socket, value);
+				break;
+			}
+			case SocketType::INT:
+			{
+				node->set(socket, (int)atoi(attr.value()));
+				break;
+			}
+			case SocketType::UINT:
+			{
+				node->set(socket, (uint)atoi(attr.value()));
+				break;
+			}
+			case SocketType::INT_ARRAY:
+			{
+				vector<string> tokens;
+				string_split(tokens, attr.value());
+
+				array<int> value;
+				value.resize(tokens.size());
+				for(size_t i = 0; i < value.size(); i++) {
+					value[i] = (int)atoi(attr.value());
+				}
+				node->set(socket, value);
+				break;
+			}
+			case SocketType::COLOR:
+			case SocketType::VECTOR:
+			case SocketType::POINT:
+			case SocketType::NORMAL:
+			{
+				array<float3> value;
+				xml_read_float_array<3>(value, attr);
+				if(value.size() == 1) {
+					node->set(socket, value[0]);
+				}
+				break;
+			}
+			case SocketType::COLOR_ARRAY:
+			case SocketType::VECTOR_ARRAY:
+			case SocketType::POINT_ARRAY:
+			case SocketType::NORMAL_ARRAY:
+			{
+				array<float3> value;
+				xml_read_float_array<3>(value, attr);
+				node->set(socket, value);
+				break;
+			}
+			case SocketType::POINT2:
+			{
+				array<float2> value;
+				xml_read_float_array<2>(value, attr);
+				if(value.size() == 1) {
+					node->set(socket, value[0]);
+				}
+				break;
+			}
+			case SocketType::POINT2_ARRAY:
+			{
+				array<float2> value;
+				xml_read_float_array<2>(value, attr);
+				node->set(socket, value);
+				break;
+			}
+			case SocketType::STRING:
+			{
+				node->set(socket, attr.value());
+				break;
+			}
+			case SocketType::ENUM:
+			{
+				ustring value(attr.value());
+				if(socket.enum_values->exists(value)) {
+					node->set(socket, value);
+				}
+				else {
+					fprintf(stderr, "Unknown value \"%s\" for attribute \"%s\".\n", value.c_str(), socket.name.c_str());
+				}
+				break;
+			}
+			case SocketType::STRING_ARRAY:
+			{
+				vector<string> tokens;
+				string_split(tokens, attr.value());
+
+				array<ustring> value;
+				value.resize(tokens.size());
+				for(size_t i = 0; i < value.size(); i++) {
+					value[i] = ustring(tokens[i]);
+				}
+				node->set(socket, value);
+				break;
+			}
+			case SocketType::TRANSFORM:
+			{
+				array<Transform> value;
+				xml_read_float_array<16>(value, attr);
+				if(value.size() == 1) {
+					node->set(socket, value[0]);
+				}
+				break;
+			}
+			case SocketType::TRANSFORM_ARRAY:
+			{
+				array<Transform> value;
+				xml_read_float_array<16>(value, attr);
+				node->set(socket, value);
+				break;
+			}
+			case SocketType::NODE:
+			{
+				ustring value(attr.value());
+				map<ustring, Node*>::iterator it = reader.node_map.find(value);
+				if(it != reader.node_map.end())
+				{
+					Node *value_node = it->second;
+					if(value_node->type == *(socket.node_type))
+						node->set(socket, it->second);
+				}
+				break;
+			}
+			case SocketType::NODE_ARRAY:
+			{
+				vector<string> tokens;
+				string_split(tokens, attr.value());
+
+				array<Node*> value;
+				value.resize(tokens.size());
+				for(size_t i = 0; i < value.size(); i++)
+				{
+					map<ustring, Node*>::iterator it = reader.node_map.find(ustring(tokens[i]));
+					if(it != reader.node_map.end())
+					{
+						Node *value_node = it->second;
+						value[i] = (value_node->type == *(socket.node_type)) ? value_node : NULL;
+					}
+					else
+					{
+						value[i] = NULL;
+					}
+				}
+				node->set(socket, value);
+				break;
+			}
+			case SocketType::CLOSURE:
+			case SocketType::UNDEFINED:
+				break;
+		}
+	}
+
+	if(node->name)
+		reader.node_map[node->name] = node;
+}
+
+pugi::xml_node xml_write_node(Node *node, pugi::xml_node xml_root)
+{
+	pugi::xml_node xml_node = xml_root.append_child(node->type->name.c_str());
+
+	xml_node.append_attribute("name") = node->name.c_str();
+
+	foreach(const SocketType& socket, node->type->inputs) {
+		if(socket.type == SocketType::CLOSURE || socket.type == SocketType::UNDEFINED) {
+			continue;
+		}
+		if(socket.flags & SocketType::INTERNAL) {
+			continue;
+		}
+		if(node->has_default_value(socket)) {
+			continue;
+		}
+
+		pugi::xml_attribute attr = xml_node.append_attribute(socket.name.c_str());
+
+		switch(socket.type)
+		{
+			case SocketType::BOOLEAN:
+			{
+				attr = xml_write_boolean(node->get_bool(socket));
+				break;
+			}
+			case SocketType::BOOLEAN_ARRAY:
+			{
+				std::stringstream ss;
+				const array<bool>& value = node->get_bool_array(socket);
+				for(size_t i = 0; i < value.size(); i++) {
+					ss << xml_write_boolean(value[i]);
+					if(i != value.size() - 1)
+						ss << " ";
+				}
+				attr = ss.str().c_str();
+				break;
+			}
+			case SocketType::FLOAT:
+			{
+				attr = (double)node->get_float(socket);
+				break;
+			}
+			case SocketType::FLOAT_ARRAY:
+			{
+				std::stringstream ss;
+				const array<float>& value = node->get_float_array(socket);
+				for(size_t i = 0; i < value.size(); i++) {
+					ss << value[i];
+					if(i != value.size() - 1) {
+						ss << " ";
+					}
+				}
+				attr = ss.str().c_str();
+				break;
+			}
+			case SocketType::INT:
+			{
+				attr = node->get_int(socket);
+				break;
+			}
+			case SocketType::UINT:
+			{
+				attr = node->get_uint(socket);
+				break;
+			}
+			case SocketType::INT_ARRAY:
+			{
+				std::stringstream ss;
+				const array<int>& value = node->get_int_array(socket);
+				for(size_t i = 0; i < value.size(); i++) {
+					ss << value[i];
+					if(i != value.size() - 1) {
+						ss << " ";
+					}
+				}
+				attr = ss.str().c_str();
+				break;
+			}
+			case SocketType::COLOR:
+			case SocketType::VECTOR:
+			case SocketType::POINT:
+			case SocketType::NORMAL:
+			{
+				float3 value = node->get_float3(socket);
+				attr = string_printf("%g %g %g", (double)value.x, (double)value.y, (double)value.z).c_str();
+				break;
+			}
+			case SocketType::COLOR_ARRAY:
+			case SocketType::VECTOR_ARRAY:
+			case SocketType::POINT_ARRAY:
+			case SocketType::NORMAL_ARRAY:
+			{
+				std::stringstream ss;
+				const array<float3>& value = node->get_float3_array(socket);
+				for(size_t i = 0; i < value.size(); i++) {
+					ss << string_printf("%g %g %g", (double)value[i].x, (double)value[i].y, (double)value[i].z);
+					if(i != value.size() - 1) {
+						ss << " ";
+					}
+				}
+				attr = ss.str().c_str();
+				break;
+			}
+			case SocketType::POINT2:
+			{
+				float2 value = node->get_float2(socket);
+				attr = string_printf("%g %g", (double)value.x, (double)value.y).c_str();
+				break;
+			}
+			case SocketType::POINT2_ARRAY:
+			{
+				std::stringstream ss;
+				const array<float2>& value = node->get_float2_array(socket);
+				for(size_t i = 0; i < value.size(); i++) {
+					ss << string_printf("%g %g", (double)value[i].x, (double)value[i].y);
+					if(i != value.size() - 1) {
+						ss << " ";
+					}
+				}
+				attr = ss.str().c_str();
+				break;
+			}
+			case SocketType::STRING:
+			case SocketType::ENUM:
+			{
+				attr = node->get_string(socket).c_str();
+				break;
+			}
+			case SocketType::STRING_ARRAY:
+			{
+				std::stringstream ss;
+				const array<ustring>& value = node->get_string_array(socket);
+				for(size_t i = 0; i < value.size(); i++) {
+					ss << value[i];
+					if(i != value.size() - 1) {
+						ss << " ";
+					}
+				}
+				attr = ss.str().c_str();
+				break;
+			}
+			case SocketType::TRANSFORM:
+			{
+				Transform tfm = node->get_transform(socket);
+				std::stringstream ss;
+				for(int i = 0; i < 4; i++) {
+					ss << string_printf("%g %g %g %g", (double)tfm[i][0], (double)tfm[i][1], (double)tfm[i][2], (double)tfm[i][3]);
+					if(i != 3) {
+						ss << " ";
+					}
+				}
+				attr = ss.str().c_str();
+				break;
+			}
+			case SocketType::TRANSFORM_ARRAY:
+			{
+				std::stringstream ss;
+				const array<Transform>& value = node->get_transform_array(socket);
+				for(size_t j = 0; j < value.size(); j++) {
+					const Transform& tfm = value[j];
+
+					for(int i = 0; i < 4; i++) {
+						ss << string_printf("%g %g %g %g", (double)tfm[i][0], (double)tfm[i][1], (double)tfm[i][2], (double)tfm[i][3]);
+						if(j != value.size() - 1 || i != 3) {
+							ss << " ";
+						}
+					}
+				}
+				attr = ss.str().c_str();
+				break;
+			}
+			case SocketType::NODE:
+			{
+				Node *value = node->get_node(socket);
+				if(value) {
+					attr = value->name.c_str();
+				}
+				break;
+			}
+			case SocketType::NODE_ARRAY:
+			{
+				std::stringstream ss;
+				const array<Node*>& value = node->get_node_array(socket);
+				for(size_t i = 0; i < value.size(); i++) {
+					if(value[i]) {
+						ss << value[i]->name.c_str();
+					}
+					if(i != value.size() - 1) {
+						ss << " ";
+					}
+				}
+				attr = ss.str().c_str();
+				break;
+			}
+			case SocketType::CLOSURE:
+			case SocketType::UNDEFINED:
+				break;
+		}
+	}
+
+	return xml_node;
+}
+
+CCL_NAMESPACE_END
+

intern/cycles/graph/node_xml.h

@@ -0,0 +1,35 @@
+/*
+ * Copyright 2011-2016 Blender Foundation
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#pragma once
+
+#include "node.h"
+
+#include "util_map.h"
+#include "util_string.h"
+#include "util_xml.h"
+
+CCL_NAMESPACE_BEGIN
+
+struct XMLReader {
+	map<ustring, Node*> node_map;
+};
+
+void xml_read_node(XMLReader& reader, Node *node, pugi::xml_node xml_node);
+pugi::xml_node xml_write_node(Node *node, pugi::xml_node xml_root);
+
+CCL_NAMESPACE_END
+

intern/cycles/kernel/CMakeLists.txt

@@ -67,6 +67,7 @@ set(SRC_KERNELS_CPU_HEADERS
 	kernel.h
 	kernels/cpu/kernel_cpu.h
 	kernels/cpu/kernel_cpu_impl.h
+	kernels/cpu/kernel_cpu_image.h
 )
 
 set(SRC_CLOSURE_HEADERS
@@ -195,8 +196,8 @@ if(WITH_CYCLES_CUDA_BINARIES)
 
 	# CUDA version
 	execute_process(COMMAND ${CUDA_NVCC_EXECUTABLE} "--version" OUTPUT_VARIABLE NVCC_OUT)
-	string(REGEX REPLACE ".*release ([0-9]+)\\.([0-9]+).*" "\\1" CUDA_VERSION_MAJOR ${NVCC_OUT})
-	string(REGEX REPLACE ".*release ([0-9]+)\\.([0-9]+).*" "\\2" CUDA_VERSION_MINOR ${NVCC_OUT})
+	string(REGEX REPLACE ".*release ([0-9]+)\\.([0-9]+).*" "\\1" CUDA_VERSION_MAJOR "${NVCC_OUT}")
+	string(REGEX REPLACE ".*release ([0-9]+)\\.([0-9]+).*" "\\2" CUDA_VERSION_MINOR "${NVCC_OUT}")
 	set(CUDA_VERSION "${CUDA_VERSION_MAJOR}${CUDA_VERSION_MINOR}")
 
 	# warn for other versions
@@ -233,6 +234,7 @@ if(WITH_CYCLES_CUDA_BINARIES)
 			OUTPUT ${cuda_cubin}
 			COMMAND ${CUDA_NVCC_EXECUTABLE}
 					-arch=${arch}
+					${CUDA_NVCC_FLAGS}
 					-m${CUDA_BITS}
 					--cubin ${CMAKE_CURRENT_SOURCE_DIR}/kernels/cuda/kernel.cu
 					-o ${CMAKE_CURRENT_BINARY_DIR}/${cuda_cubin}

intern/cycles/kernel/closure/bsdf_microfacet.h

@@ -276,7 +276,7 @@ ccl_device float3 bsdf_microfacet_ggx_eval_reflect(const ShaderClosure *sc, cons
 	bool m_refractive = sc->type == CLOSURE_BSDF_MICROFACET_GGX_REFRACTION_ID;
 	float3 N = sc->N;
 
-	if(m_refractive || fmaxf(alpha_x, alpha_y) <= 1e-4f)
+	if(m_refractive || alpha_x*alpha_y <= 1e-7f)
 		return make_float3(0.0f, 0.0f, 0.0f);
 
 	float cosNO = dot(N, I);
@@ -367,7 +367,7 @@ ccl_device float3 bsdf_microfacet_ggx_eval_transmit(const ShaderClosure *sc, con
 	bool m_refractive = sc->type == CLOSURE_BSDF_MICROFACET_GGX_REFRACTION_ID;
 	float3 N = sc->N;
 
-	if(!m_refractive || fmaxf(alpha_x, alpha_y) <= 1e-4f)
+	if(!m_refractive || alpha_x*alpha_y <= 1e-7f)
 		return make_float3(0.0f, 0.0f, 0.0f);
 
 	float cosNO = dot(N, I);
@@ -450,7 +450,7 @@ ccl_device int bsdf_microfacet_ggx_sample(KernelGlobals *kg, const ShaderClosure
 				*omega_in = 2 * cosMO * m - I;
 
 				if(dot(Ng, *omega_in) > 0) {
-					if(fmaxf(alpha_x, alpha_y) <= 1e-4f) {
+					if(alpha_x*alpha_y <= 1e-7f) {
 						/* some high number for MIS */
 						*pdf = 1e6f;
 						*eval = make_float3(1e6f, 1e6f, 1e6f);
@@ -539,7 +539,7 @@ ccl_device int bsdf_microfacet_ggx_sample(KernelGlobals *kg, const ShaderClosure
 				*domega_in_dy = dTdy;
 #endif
 
-				if(fmaxf(alpha_x, alpha_y) <= 1e-4f || fabsf(m_eta - 1.0f) < 1e-4f) {
+				if(alpha_x*alpha_y <= 1e-7f || fabsf(m_eta - 1.0f) < 1e-4f) {
 					/* some high number for MIS */
 					*pdf = 1e6f;
 					*eval = make_float3(1e6f, 1e6f, 1e6f);
@@ -622,7 +622,7 @@ ccl_device float3 bsdf_microfacet_beckmann_eval_reflect(const ShaderClosure *sc,
 	bool m_refractive = sc->type == CLOSURE_BSDF_MICROFACET_BECKMANN_REFRACTION_ID;
 	float3 N = sc->N;
 
-	if(m_refractive || fmaxf(alpha_x, alpha_y) <= 1e-4f)
+	if(m_refractive || alpha_x*alpha_y <= 1e-7f)
 		return make_float3(0.0f, 0.0f, 0.0f);
 
 	float cosNO = dot(N, I);
@@ -716,7 +716,7 @@ ccl_device float3 bsdf_microfacet_beckmann_eval_transmit(const ShaderClosure *sc
 	bool m_refractive = sc->type == CLOSURE_BSDF_MICROFACET_BECKMANN_REFRACTION_ID;
 	float3 N = sc->N;
 
-	if(!m_refractive || fmaxf(alpha_x, alpha_y) <= 1e-4f)
+	if(!m_refractive || alpha_x*alpha_y <= 1e-7f)
 		return make_float3(0.0f, 0.0f, 0.0f);
 
 	float cosNO = dot(N, I);
@@ -798,7 +798,7 @@ ccl_device int bsdf_microfacet_beckmann_sample(KernelGlobals *kg, const ShaderCl
 				*omega_in = 2 * cosMO * m - I;
 
 				if(dot(Ng, *omega_in) > 0) {
-					if(fmaxf(alpha_x, alpha_y) <= 1e-4f) {
+					if(alpha_x*alpha_y <= 1e-7f) {
 						/* some high number for MIS */
 						*pdf = 1e6f;
 						*eval = make_float3(1e6f, 1e6f, 1e6f);
@@ -889,7 +889,7 @@ ccl_device int bsdf_microfacet_beckmann_sample(KernelGlobals *kg, const ShaderCl
 				*domega_in_dy = dTdy;
 #endif
 
-				if(fmaxf(alpha_x, alpha_y) <= 1e-4f || fabsf(m_eta - 1.0f) < 1e-4f) {
+				if(alpha_x*alpha_y <= 1e-7f || fabsf(m_eta - 1.0f) < 1e-4f) {
 					/* some high number for MIS */
 					*pdf = 1e6f;
 					*eval = make_float3(1e6f, 1e6f, 1e6f);

intern/cycles/kernel/geom/geom_bvh.h

@@ -48,6 +48,28 @@ CCL_NAMESPACE_BEGIN
 
 #define BVH_FEATURE(f) (((BVH_FUNCTION_FEATURES) & (f)) != 0)
 
+/* Debugging heleprs */
+#ifdef __KERNEL_DEBUG__
+#  define BVH_DEBUG_INIT() \
+	do { \
+		isect->num_traversal_steps = 0; \
+		isect->num_traversed_instances = 0; \
+	} while(0)
+#  define BVH_DEBUG_NEXT_STEP() \
+	do { \
+		++isect->num_traversal_steps; \
+	} while(0)
+#  define BVH_DEBUG_NEXT_INSTANCE() \
+	do { \
+		++isect->num_traversed_instances; \
+	} while(0)
+#else  /* __KERNEL_DEBUG__ */
+#  define BVH_DEBUG_INIT()
+#  define BVH_DEBUG_NEXT_STEP()
+#  define BVH_DEBUG_NEXT_INSTANCE()
+#endif  /* __KERNEL_DEBUG__ */
+
+
 /* Common QBVH functions. */
 #ifdef __QBVH__
 #  include "geom_qbvh.h"