Merge branch 'master' into gpencil-new-data-proposalgpencil-new-data-proposal

18 files changed, 102 insertions, 33 deletions

diff --git a/intern/cycles/device/CMakeLists.txt b/intern/cycles/device/CMakeLists.txt
index 5296d819e42..bfca3ab6aea 100644
--- a/intern/cycles/device/CMakeLists.txt
+++ b/intern/cycles/device/CMakeLists.txt
@@ -224,7 +224,8 @@ include_directories(SYSTEM ${INC_SYS})
 cycles_add_library(cycles_device "${LIB}" ${SRC})
 
 if(WITH_CYCLES_DEVICE_ONEAPI)
-  # Need to have proper rebuilding in case of changes in cycles_kernel_oneapi due external project behaviour
+  # Need to have proper rebuilding in case of changes
+  # in cycles_kernel_oneapi due external project behavior.
   add_dependencies(cycles_device cycles_kernel_oneapi)
 endif()
 
diff --git a/intern/cycles/device/metal/kernel.mm b/intern/cycles/device/metal/kernel.mm
index 55938d1a03a..dc8af9a5358 100644
--- a/intern/cycles/device/metal/kernel.mm
+++ b/intern/cycles/device/metal/kernel.mm
@@ -45,6 +45,36 @@ bool kernel_has_intersection(DeviceKernel device_kernel)
 struct ShaderCache {
   ShaderCache(id<MTLDevice> _mtlDevice) : mtlDevice(_mtlDevice)
   {
+    /* Initialize occupancy tuning LUT. */
+    if (MetalInfo::get_device_vendor(mtlDevice) == METAL_GPU_APPLE) {
+      switch (MetalInfo::get_apple_gpu_architecture(mtlDevice)) {
+        default:
+        case APPLE_M2:
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_COMPACT_SHADOW_STATES] = {32, 32};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_INIT_FROM_CAMERA] = {832, 32};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_INTERSECT_CLOSEST] = {64, 64};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_INTERSECT_SHADOW] = {64, 64};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_INTERSECT_SUBSURFACE] = {704, 32};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_QUEUED_PATHS_ARRAY] = {1024, 256};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_SHADE_BACKGROUND] = {64, 32};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_SHADE_SHADOW] = {256, 256};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE] = {448, 384};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_SORTED_PATHS_ARRAY] = {1024, 1024};
+          break;
+        case APPLE_M1:
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_COMPACT_SHADOW_STATES] = {256, 128};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_INIT_FROM_CAMERA] = {768, 32};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_INTERSECT_CLOSEST] = {512, 128};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_INTERSECT_SHADOW] = {384, 128};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_INTERSECT_SUBSURFACE] = {512, 64};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_QUEUED_PATHS_ARRAY] = {512, 256};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_SHADE_BACKGROUND] = {512, 128};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_SHADE_SHADOW] = {384, 32};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE] = {576, 384};
+          occupancy_tuning[DEVICE_KERNEL_INTEGRATOR_SORTED_PATHS_ARRAY] = {832, 832};
+          break;
+      }
+    }
   }
   ~ShaderCache();
 
@@ -73,6 +103,11 @@ struct ShaderCache {
     std::function<void(MetalKernelPipeline *)> completionHandler;
   };
 
+  struct OccupancyTuningParameters {
+    int threads_per_threadgroup = 0;
+    int num_threads_per_block = 0;
+  } occupancy_tuning[DEVICE_KERNEL_NUM];
+
   std::mutex cache_mutex;
 
   PipelineCollection pipelines[DEVICE_KERNEL_NUM];
@@ -230,6 +265,13 @@ void ShaderCache::load_kernel(DeviceKernel device_kernel,
   request.pipeline->device_kernel = device_kernel;
   request.pipeline->threads_per_threadgroup = device->max_threads_per_threadgroup;
 
+  if (occupancy_tuning[device_kernel].threads_per_threadgroup) {
+    request.pipeline->threads_per_threadgroup =
+        occupancy_tuning[device_kernel].threads_per_threadgroup;
+    request.pipeline->num_threads_per_block =
+        occupancy_tuning[device_kernel].num_threads_per_block;
+  }
+
   /* metalrt options */
   request.pipeline->use_metalrt = device->use_metalrt;
   request.pipeline->metalrt_hair = device->use_metalrt &&
@@ -374,13 +416,6 @@ void MetalKernelPipeline::compile()
   const std::string function_name = std::string("cycles_metal_") +
                                     device_kernel_as_string(device_kernel);
 
-  int threads_per_threadgroup = this->threads_per_threadgroup;
-  if (device_kernel > DEVICE_KERNEL_INTEGRATOR_MEGAKERNEL &&
-      device_kernel < DEVICE_KERNEL_INTEGRATOR_RESET) {
-    /* Always use 512 for the sorting kernels */
-    threads_per_threadgroup = 512;
-  }
-
   NSString *entryPoint = [@(function_name.c_str()) copy];
 
   NSError *error = NULL;
@@ -644,12 +679,14 @@ void MetalKernelPipeline::compile()
       return;
     }
 
-    int num_threads_per_block = round_down(computePipelineState.maxTotalThreadsPerThreadgroup,
-                                           computePipelineState.threadExecutionWidth);
-    num_threads_per_block = std::max(num_threads_per_block,
-                                     (int)computePipelineState.threadExecutionWidth);
+    if (!num_threads_per_block) {
+      num_threads_per_block = round_down(computePipelineState.maxTotalThreadsPerThreadgroup,
+                                         computePipelineState.threadExecutionWidth);
+      num_threads_per_block = std::max(num_threads_per_block,
+                                       (int)computePipelineState.threadExecutionWidth);
+    }
+
     this->pipeline = computePipelineState;
-    this->num_threads_per_block = num_threads_per_block;
 
     if (@available(macOS 11.0, *)) {
       if (creating_new_archive || recreate_archive) {
diff --git a/intern/cycles/kernel/CMakeLists.txt b/intern/cycles/kernel/CMakeLists.txt
index 81c5f593974..3779fdc697a 100644
--- a/intern/cycles/kernel/CMakeLists.txt
+++ b/intern/cycles/kernel/CMakeLists.txt
@@ -866,8 +866,8 @@ if(WITH_CYCLES_DEVICE_ONEAPI)
   else()
     list(APPEND sycl_compiler_flags -fPIC)
 
-    # We avoid getting __FAST_MATH__ to be defined when building on CentOS 7 until the compilation crash
-    # it triggers at either AoT or JIT stages gets fixed.
+    # We avoid getting __FAST_MATH__ to be defined when building on CentOS 7 until the compilation
+    # crash it triggers at either AoT or JIT stages gets fixed.
     list(APPEND sycl_compiler_flags -fhonor-nans)
 
     # add $ORIGIN to cycles_kernel_oneapi.so rpath so libsycl.so and
@@ -881,7 +881,8 @@ if(WITH_CYCLES_DEVICE_ONEAPI)
       OUTPUT ${cycles_kernel_oneapi_lib}
       COMMAND ${CMAKE_COMMAND} -E env
               "LD_LIBRARY_PATH=${sycl_compiler_root}/../lib:${OCLOC_INSTALL_DIR}/lib:${IGC_INSTALL_DIR}/lib"
-              "PATH=${OCLOC_INSTALL_DIR}/bin:${sycl_compiler_root}:$ENV{PATH}" # env PATH is for compiler to find ld
+              # `$ENV{PATH}` is for compiler to find `ld`.
+              "PATH=${OCLOC_INSTALL_DIR}/bin:${sycl_compiler_root}:$ENV{PATH}"
               ${SYCL_COMPILER} $<$<CONFIG:Debug>:-g>$<$<CONFIG:RelWithDebInfo>:-g> ${sycl_compiler_flags}
       DEPENDS ${cycles_oneapi_kernel_sources})
   endif()
diff --git a/intern/cycles/test/CMakeLists.txt b/intern/cycles/test/CMakeLists.txt
index b126247de5f..c3ae81ed1db 100644
--- a/intern/cycles/test/CMakeLists.txt
+++ b/intern/cycles/test/CMakeLists.txt
@@ -34,6 +34,7 @@ set(SRC
   render_graph_finalize_test.cpp
   util_aligned_malloc_test.cpp
   util_math_test.cpp
+  util_md5_test.cpp
   util_path_test.cpp
   util_string_test.cpp
   util_task_test.cpp
diff --git a/intern/cycles/test/util_md5_test.cpp b/intern/cycles/test/util_md5_test.cpp
new file mode 100644
index 00000000000..abc147b70a1
--- /dev/null
+++ b/intern/cycles/test/util_md5_test.cpp
@@ -0,0 +1,16 @@
+/* SPDX-License-Identifier: Apache-2.0
+ * Copyright 2011-2022 Blender Foundation */
+
+#include "testing/testing.h"
+
+#include "util/md5.h"
+
+CCL_NAMESPACE_BEGIN
+
+TEST(util, util_md5_string)
+{
+  /* The hash is calculated using `echo -n "Hello, World\!" | md5 | tr '[:lower:]' '[:upper:]'`. */
+  EXPECT_EQ(util_md5_string("Hello, World!"), "65A8E27D8879283831B664BD8B7F0AD4");
+}
+
+CCL_NAMESPACE_END
diff --git a/intern/cycles/util/md5.cpp b/intern/cycles/util/md5.cpp
index 1c7e6b9bf3e..3342d7a509a 100644
--- a/intern/cycles/util/md5.cpp
+++ b/intern/cycles/util/md5.cpp
@@ -347,13 +347,18 @@ void MD5Hash::finish(uint8_t digest[16])
 
 string MD5Hash::get_hex()
 {
+  constexpr char kHexDigits[] = {
+      '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B', 'C', 'D', 'E', 'F'};
+
   uint8_t digest[16];
   char buf[16 * 2 + 1];
 
   finish(digest);
 
-  for (int i = 0; i < 16; i++)
-    sprintf(buf + i * 2, "%02X", (unsigned int)digest[i]);
+  for (int i = 0; i < 16; i++) {
+    buf[i * 2 + 0] = kHexDigits[digest[i] / 0x10];
+    buf[i * 2 + 1] = kHexDigits[digest[i] % 0x10];
+  }
   buf[sizeof(buf) - 1] = '\0';
 
   return string(buf);
diff --git a/intern/ffmpeg/tests/ffmpeg_codecs.cc b/intern/ffmpeg/tests/ffmpeg_codecs.cc
index e5c33202417..10cbe4b938b 100644
--- a/intern/ffmpeg/tests/ffmpeg_codecs.cc
+++ b/intern/ffmpeg/tests/ffmpeg_codecs.cc
@@ -40,7 +40,7 @@ bool test_acodec(const AVCodec *codec, AVSampleFormat fmt)
     if (ctx) {
       ctx->sample_fmt = fmt;
       ctx->sample_rate = 48000;
-      ctx->channel_layout = AV_CH_LAYOUT_MONO;
+      av_channel_layout_from_mask(&ctx->ch_layout, AV_CH_LAYOUT_MONO);
       ctx->bit_rate = 128000;
       int open = avcodec_open2(ctx, codec, NULL);
       if (open >= 0) {
diff --git a/intern/ghost/intern/GHOST_ContextGLX.cpp b/intern/ghost/intern/GHOST_ContextGLX.cpp
index 93708983f37..d9f2df21ee0 100644
--- a/intern/ghost/intern/GHOST_ContextGLX.cpp
+++ b/intern/ghost/intern/GHOST_ContextGLX.cpp
@@ -140,7 +140,7 @@ GHOST_TSuccess GHOST_ContextGLX::initializeDrawingContext()
   /* End Inline GLEW. */
   /* -------------------------------------------------------------------- */
 #else
-  /* Important to initialize only glxew (_not_ GLEW),
+  /* Important to initialize only GLXEW (_not_ GLEW),
    * since this breaks w/ Mesa's `swrast`, see: T46431. */
   glxewInit();
 #endif /* USE_GLXEW_INIT_WORKAROUND */
diff --git a/intern/ghost/intern/GHOST_ImeWin32.h b/intern/ghost/intern/GHOST_ImeWin32.h
index 85c8ed7b4bd..cb6d8a770cf 100644
--- a/intern/ghost/intern/GHOST_ImeWin32.h
+++ b/intern/ghost/intern/GHOST_ImeWin32.h
@@ -266,7 +266,7 @@ class GHOST_ImeWin32 {
    * Parameters
    *   * window_handle [in] (HWND)
    *     Represents the window handle of the caller.
-   *   * caret_rect [in] (const gfx::Rect&)
+   *   * caret_rect [in] (`const gfx::Rect&`)
    *     Represent the rectangle of the input caret.
    *     This rectangle is used for controlling the positions of IME windows.
    *   * complete [in] (bool)
diff --git a/intern/ghost/intern/GHOST_SystemWayland.cpp b/intern/ghost/intern/GHOST_SystemWayland.cpp
index 5afbf5b0b5f..3a0ba5cd21a 100644
--- a/intern/ghost/intern/GHOST_SystemWayland.cpp
+++ b/intern/ghost/intern/GHOST_SystemWayland.cpp
@@ -101,7 +101,7 @@ static const struct GWL_RegistryHandler *gwl_registry_handler_from_interface_slo
     int interface_slot);
 
 /* -------------------------------------------------------------------- */
-/** \name Workaround Compositor Sprsific Bugs
+/** \name Workaround Compositor Specific Bugs
  * \{ */
 
 /**
@@ -4013,7 +4013,8 @@ static void gwl_seat_capability_touch_disable(GWL_Seat *seat)
 }
 
 static void seat_handle_capabilities(void *data,
-                                     struct wl_seat *wl_seat,
+                                     /* Only used in an assert. */
+                                     [[maybe_unused]] struct wl_seat *wl_seat,
                                      const uint32_t capabilities)
 {
   CLOG_INFO(LOG,
diff --git a/intern/guardedalloc/MEM_guardedalloc.h b/intern/guardedalloc/MEM_guardedalloc.h
index fdd77fb9eef..5ae33343949 100644
--- a/intern/guardedalloc/MEM_guardedalloc.h
+++ b/intern/guardedalloc/MEM_guardedalloc.h
@@ -271,7 +271,7 @@ void MEM_use_guarded_allocator(void);
 template<typename T, typename... Args>
 inline T *MEM_new(const char *allocation_name, Args &&...args)
 {
-  void *buffer = MEM_mallocN(sizeof(T), allocation_name);
+  void *buffer = MEM_mallocN_aligned(sizeof(T), alignof(T), allocation_name);
   return new (buffer) T(std::forward<Args>(args)...);
 }
 
diff --git a/intern/guardedalloc/intern/leak_detector.cc b/intern/guardedalloc/intern/leak_detector.cc
index 288d78fd206..5b565b15920 100644
--- a/intern/guardedalloc/intern/leak_detector.cc
+++ b/intern/guardedalloc/intern/leak_detector.cc
@@ -41,7 +41,7 @@ class MemLeakPrinter {
 
     if (fail_on_memleak) {
       /* There are many other ways to change the exit code to failure here:
-       * - Make the destructor noexcept(false) and throw an exception.
+       * - Make the destructor `noexcept(false)` and throw an exception.
        * - Call exit(EXIT_FAILURE).
        * - Call terminate().
        */
diff --git a/intern/wayland_dynload/intern/wayland_dynload_client.c b/intern/wayland_dynload/intern/wayland_dynload_client.c
index 68ba5374aba..617a8878199 100644
--- a/intern/wayland_dynload/intern/wayland_dynload_client.c
+++ b/intern/wayland_dynload/intern/wayland_dynload_client.c
@@ -44,7 +44,8 @@ bool wayland_dynload_client_init(const bool verbose)
 #define WAYLAND_DYNLOAD_IFACE(symbol) \
   { \
     const void *symbol_val; \
-    if (!(symbol_val = dynamic_library_find_with_error(lib, #symbol, paths[path_found]))) { \
+    if (!(symbol_val = dynamic_library_find_with_error( \
+              lib, #symbol, paths[path_found], verbose))) { \
       return false; \
     } \
     memcpy(&symbol, symbol_val, sizeof(symbol)); \
@@ -54,7 +55,7 @@ bool wayland_dynload_client_init(const bool verbose)
 
 #define WAYLAND_DYNLOAD_FN(symbol) \
   if (!(wayland_dynload_client.symbol = dynamic_library_find_with_error( \
-            lib, #symbol, paths[path_found]))) { \
+            lib, #symbol, paths[path_found], verbose))) { \
     return false; \
   }
 #include "wayland_dynload_client.h"
diff --git a/intern/wayland_dynload/intern/wayland_dynload_cursor.c b/intern/wayland_dynload/intern/wayland_dynload_cursor.c
index 3d0526c7ba6..cc62a43bc01 100644
--- a/intern/wayland_dynload/intern/wayland_dynload_cursor.c
+++ b/intern/wayland_dynload/intern/wayland_dynload_cursor.c
@@ -36,7 +36,7 @@ bool wayland_dynload_cursor_init(const bool verbose)
 
 #define WAYLAND_DYNLOAD_FN(symbol) \
   if (!(wayland_dynload_cursor.symbol = dynamic_library_find_with_error( \
-            lib, #symbol, paths[path_index]))) { \
+            lib, #symbol, paths[path_index], verbose))) { \
     return false; \
   }
 #include "wayland_dynload_cursor.h"
diff --git a/intern/wayland_dynload/intern/wayland_dynload_egl.c b/intern/wayland_dynload/intern/wayland_dynload_egl.c
index cfc195c0408..d8e4dfe0dad 100644
--- a/intern/wayland_dynload/intern/wayland_dynload_egl.c
+++ b/intern/wayland_dynload/intern/wayland_dynload_egl.c
@@ -36,7 +36,7 @@ bool wayland_dynload_egl_init(const bool verbose)
 
 #define WAYLAND_DYNLOAD_FN(symbol) \
   if (!(wayland_dynload_egl.symbol = dynamic_library_find_with_error( \
-            lib, #symbol, paths[path_found]))) { \
+            lib, #symbol, paths[path_found], verbose))) { \
     return false; \
   }
 #include "wayland_dynload_egl.h"
diff --git a/intern/wayland_dynload/intern/wayland_dynload_libdecor.c b/intern/wayland_dynload/intern/wayland_dynload_libdecor.c
index d8bdd27bb27..dafd1badfda 100644
--- a/intern/wayland_dynload/intern/wayland_dynload_libdecor.c
+++ b/intern/wayland_dynload/intern/wayland_dynload_libdecor.c
@@ -36,7 +36,7 @@ bool wayland_dynload_libdecor_init(const bool verbose)
 
 #define WAYLAND_DYNLOAD_FN(symbol) \
   if (!(wayland_dynload_libdecor.symbol = dynamic_library_find_with_error( \
-            lib, #symbol, paths[path_index]))) { \
+            lib, #symbol, paths[path_index], verbose))) { \
     return false; \
   }
 #include "wayland_dynload_libdecor.h"
diff --git a/intern/wayland_dynload/intern/wayland_dynload_utils.c b/intern/wayland_dynload/intern/wayland_dynload_utils.c
index 743dac14eec..666de20c5d3 100644
--- a/intern/wayland_dynload/intern/wayland_dynload_utils.c
+++ b/intern/wayland_dynload/intern/wayland_dynload_utils.c
@@ -30,11 +30,16 @@ DynamicLibrary dynamic_library_open_array_with_error(const char **paths,
   return lib;
 }
 
-void *dynamic_library_find_with_error(DynamicLibrary lib, const char *symbol, const char *path_lib)
+void *dynamic_library_find_with_error(DynamicLibrary lib,
+                                      const char *symbol,
+                                      const char *path_lib,
+                                      const bool verbose)
 {
   void *symbol_var = dynamic_library_find(lib, symbol);
   if (symbol_var == NULL) {
-    fprintf(stderr, "Unable to find '%s' in '%s'.\n", symbol, path_lib);
+    if (verbose) {
+      fprintf(stderr, "Unable to find '%s' in '%s'.\n", symbol, path_lib);
+    }
   }
   return symbol_var;
 }
diff --git a/intern/wayland_dynload/intern/wayland_dynload_utils.h b/intern/wayland_dynload/intern/wayland_dynload_utils.h
index 785f32521e4..1814879615b 100644
--- a/intern/wayland_dynload/intern/wayland_dynload_utils.h
+++ b/intern/wayland_dynload/intern/wayland_dynload_utils.h
@@ -26,4 +26,5 @@ DynamicLibrary dynamic_library_open_array_with_error(const char **paths,
 /** Find a symbol, printing an error when the symbol isn't found. */
 void *dynamic_library_find_with_error(DynamicLibrary lib,
                                       const char *symbol,
-                                      const char *path_lib);
+                                      const char *path_lib,
+                                      bool verbose);