diff options
author | Brecht Van Lommel <brecht@blender.org> | 2021-09-20 18:59:20 +0300 |
---|---|---|
committer | Brecht Van Lommel <brecht@blender.org> | 2021-09-21 15:55:54 +0300 |
commit | 08031197250aeecbaca3803254e6f25b8c7b7b37 (patch) | |
tree | 6fe7ab045f0dc0a423d6557c4073f34309ef4740 /intern/cycles/device/device.h | |
parent | fa6b1007bad065440950cd67deb16a04f368856f (diff) |
Cycles: merge of cycles-x branch, a major update to the renderer
This includes much improved GPU rendering performance, viewport interactivity,
new shadow catcher, revamped sampling settings, subsurface scattering anisotropy,
new GPU volume sampling, improved PMJ sampling pattern, and more.
Some features have also been removed or changed, breaking backwards compatibility.
Including the removal of the OpenCL backend, for which alternatives are under
development.
Release notes and code docs:
https://wiki.blender.org/wiki/Reference/Release_Notes/3.0/Cycles
https://wiki.blender.org/wiki/Source/Render/Cycles
Credits:
* Sergey Sharybin
* Brecht Van Lommel
* Patrick Mours (OptiX backend)
* Christophe Hery (subsurface scattering anisotropy)
* William Leeson (PMJ sampling pattern)
* Alaska (various fixes and tweaks)
* Thomas Dinges (various fixes)
For the full commit history, see the cycles-x branch. This squashes together
all the changes since intermediate changes would often fail building or tests.
Ref T87839, T87837, T87836
Fixes T90734, T89353, T80267, T80267, T77185, T69800
Diffstat (limited to 'intern/cycles/device/device.h')
-rw-r--r-- | intern/cycles/device/device.h | 366 |
1 files changed, 74 insertions, 292 deletions
diff --git a/intern/cycles/device/device.h b/intern/cycles/device/device.h index ecf79bcdfa6..02b6edb56d0 100644 --- a/intern/cycles/device/device.h +++ b/intern/cycles/device/device.h @@ -21,31 +21,34 @@ #include "bvh/bvh_params.h" +#include "device/device_denoise.h" #include "device/device_memory.h" -#include "device/device_task.h" +#include "util/util_function.h" #include "util/util_list.h" +#include "util/util_logging.h" #include "util/util_stats.h" #include "util/util_string.h" #include "util/util_texture.h" #include "util/util_thread.h" #include "util/util_types.h" +#include "util/util_unique_ptr.h" #include "util/util_vector.h" CCL_NAMESPACE_BEGIN class BVH; +class DeviceQueue; class Progress; -class RenderTile; +class CPUKernels; +class CPUKernelThreadGlobals; /* Device Types */ enum DeviceType { DEVICE_NONE = 0, DEVICE_CPU, - DEVICE_OPENCL, DEVICE_CUDA, - DEVICE_NETWORK, DEVICE_MULTI, DEVICE_OPTIX, DEVICE_DUMMY, @@ -53,20 +56,11 @@ enum DeviceType { enum DeviceTypeMask { DEVICE_MASK_CPU = (1 << DEVICE_CPU), - DEVICE_MASK_OPENCL = (1 << DEVICE_OPENCL), DEVICE_MASK_CUDA = (1 << DEVICE_CUDA), DEVICE_MASK_OPTIX = (1 << DEVICE_OPTIX), - DEVICE_MASK_NETWORK = (1 << DEVICE_NETWORK), DEVICE_MASK_ALL = ~0 }; -enum DeviceKernelStatus { - DEVICE_KERNEL_FEATURE_KERNEL_AVAILABLE, - DEVICE_KERNEL_USING_FEATURE_KERNEL, - DEVICE_KERNEL_FEATURE_KERNEL_INVALID, - DEVICE_KERNEL_UNKNOWN, -}; - #define DEVICE_MASK(type) (DeviceTypeMask)(1 << type) class DeviceInfo { @@ -75,20 +69,16 @@ class DeviceInfo { string description; string id; /* used for user preferences, should stay fixed with changing hardware config */ int num; - bool display_device; /* GPU is used as a display device. */ - bool has_half_images; /* Support half-float textures. */ - bool has_nanovdb; /* Support NanoVDB volumes. */ - bool has_volume_decoupled; /* Decoupled volume shading. */ - bool has_branched_path; /* Supports branched path tracing. */ - bool has_adaptive_stop_per_sample; /* Per-sample adaptive sampling stopping. */ - bool has_osl; /* Support Open Shading Language. */ - bool use_split_kernel; /* Use split or mega kernel. */ - bool has_profiling; /* Supports runtime collection of profiling info. */ - bool has_peer_memory; /* GPU has P2P access to memory of another GPU. */ - DenoiserTypeMask denoisers; /* Supported denoiser types. */ + bool display_device; /* GPU is used as a display device. */ + bool has_nanovdb; /* Support NanoVDB volumes. */ + bool has_half_images; /* Support half-float textures. */ + bool has_osl; /* Support Open Shading Language. */ + bool has_profiling; /* Supports runtime collection of profiling info. */ + bool has_peer_memory; /* GPU has P2P access to memory of another GPU. */ + bool has_gpu_queue; /* Device supports GPU queue. */ + DenoiserTypeMask denoisers; /* Supported denoiser types. */ int cpu_threads; vector<DeviceInfo> multi_devices; - vector<DeviceInfo> denoising_devices; string error_msg; DeviceInfo() @@ -100,227 +90,35 @@ class DeviceInfo { display_device = false; has_half_images = false; has_nanovdb = false; - has_volume_decoupled = false; - has_branched_path = true; - has_adaptive_stop_per_sample = false; has_osl = false; - use_split_kernel = false; has_profiling = false; has_peer_memory = false; + has_gpu_queue = false; denoisers = DENOISER_NONE; } - bool operator==(const DeviceInfo &info) + bool operator==(const DeviceInfo &info) const { /* Multiple Devices with the same ID would be very bad. */ assert(id != info.id || (type == info.type && num == info.num && description == info.description)); return id == info.id; } - - /* Add additional devices needed for the specified denoiser. */ - void add_denoising_devices(DenoiserType denoiser_type); -}; - -class DeviceRequestedFeatures { - public: - /* Use experimental feature set. */ - bool experimental; - - /* Selective nodes compilation. */ - - /* Identifier of a node group up to which all the nodes needs to be - * compiled in. Nodes from higher group indices will be ignores. - */ - int max_nodes_group; - - /* Features bitfield indicating which features from the requested group - * will be compiled in. Nodes which corresponds to features which are not - * in this bitfield will be ignored even if they're in the requested group. - */ - int nodes_features; - - /* BVH/sampling kernel features. */ - bool use_hair; - bool use_hair_thick; - bool use_object_motion; - bool use_camera_motion; - - /* Denotes whether baking functionality is needed. */ - bool use_baking; - - /* Use subsurface scattering materials. */ - bool use_subsurface; - - /* Use volume materials. */ - bool use_volume; - - /* Use branched integrator. */ - bool use_integrator_branched; - - /* Use OpenSubdiv patch evaluation */ - bool use_patch_evaluation; - - /* Use Transparent shadows */ - bool use_transparent; - - /* Use various shadow tricks, such as shadow catcher. */ - bool use_shadow_tricks; - - /* Per-uber shader usage flags. */ - bool use_principled; - - /* Denoising features. */ - bool use_denoising; - - /* Use raytracing in shaders. */ - bool use_shader_raytrace; - - /* Use true displacement */ - bool use_true_displacement; - - /* Use background lights */ - bool use_background_light; - - DeviceRequestedFeatures() - { - /* TODO(sergey): Find more meaningful defaults. */ - max_nodes_group = 0; - nodes_features = 0; - use_hair = false; - use_hair_thick = false; - use_object_motion = false; - use_camera_motion = false; - use_baking = false; - use_subsurface = false; - use_volume = false; - use_integrator_branched = false; - use_patch_evaluation = false; - use_transparent = false; - use_shadow_tricks = false; - use_principled = false; - use_denoising = false; - use_shader_raytrace = false; - use_true_displacement = false; - use_background_light = false; - } - - bool modified(const DeviceRequestedFeatures &requested_features) - { - return !(max_nodes_group == requested_features.max_nodes_group && - nodes_features == requested_features.nodes_features && - use_hair == requested_features.use_hair && - use_hair_thick == requested_features.use_hair_thick && - use_object_motion == requested_features.use_object_motion && - 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 && - use_patch_evaluation == requested_features.use_patch_evaluation && - use_transparent == requested_features.use_transparent && - use_shadow_tricks == requested_features.use_shadow_tricks && - use_principled == requested_features.use_principled && - use_denoising == requested_features.use_denoising && - use_shader_raytrace == requested_features.use_shader_raytrace && - use_true_displacement == requested_features.use_true_displacement && - use_background_light == requested_features.use_background_light); - } - - /* Convert the requested features structure to a build options, - * which could then be passed to compilers. - */ - string get_build_options() const - { - string build_options = ""; - if (experimental) { - build_options += "-D__KERNEL_EXPERIMENTAL__ "; - } - build_options += "-D__NODES_MAX_GROUP__=" + string_printf("%d", max_nodes_group); - build_options += " -D__NODES_FEATURES__=" + string_printf("%d", nodes_features); - if (!use_hair) { - build_options += " -D__NO_HAIR__"; - } - if (!use_object_motion) { - build_options += " -D__NO_OBJECT_MOTION__"; - } - if (!use_camera_motion) { - build_options += " -D__NO_CAMERA_MOTION__"; - } - if (!use_baking) { - build_options += " -D__NO_BAKING__"; - } - if (!use_volume) { - build_options += " -D__NO_VOLUME__"; - } - if (!use_subsurface) { - build_options += " -D__NO_SUBSURFACE__"; - } - if (!use_integrator_branched) { - build_options += " -D__NO_BRANCHED_PATH__"; - } - if (!use_patch_evaluation) { - build_options += " -D__NO_PATCH_EVAL__"; - } - if (!use_transparent && !use_volume) { - build_options += " -D__NO_TRANSPARENT__"; - } - if (!use_shadow_tricks) { - build_options += " -D__NO_SHADOW_TRICKS__"; - } - if (!use_principled) { - build_options += " -D__NO_PRINCIPLED__"; - } - if (!use_denoising) { - build_options += " -D__NO_DENOISING__"; - } - if (!use_shader_raytrace) { - build_options += " -D__NO_SHADER_RAYTRACE__"; - } - return build_options; - } }; -std::ostream &operator<<(std::ostream &os, const DeviceRequestedFeatures &requested_features); - /* Device */ -struct DeviceDrawParams { - function<void()> bind_display_space_shader_cb; - function<void()> unbind_display_space_shader_cb; -}; - class Device { friend class device_sub_ptr; protected: - enum { - FALLBACK_SHADER_STATUS_NONE = 0, - FALLBACK_SHADER_STATUS_ERROR, - FALLBACK_SHADER_STATUS_SUCCESS, - }; - - Device(DeviceInfo &info_, Stats &stats_, Profiler &profiler_, bool background) - : background(background), - vertex_buffer(0), - fallback_status(FALLBACK_SHADER_STATUS_NONE), - fallback_shader_program(0), - info(info_), - stats(stats_), - profiler(profiler_) + Device(const DeviceInfo &info_, Stats &stats_, Profiler &profiler_) + : info(info_), stats(stats_), profiler(profiler_) { } - bool background; string error_msg; - /* used for real time display */ - unsigned int vertex_buffer; - int fallback_status, fallback_shader_program; - int image_texture_location, fullscreen_location; - - bool bind_fallback_display_space_shader(const float width, const float height); - virtual device_ptr mem_alloc_sub_ptr(device_memory & /*mem*/, int /*offset*/, int /*size*/) { /* Only required for devices that implement denoising. */ @@ -361,67 +159,31 @@ class Device { Stats &stats; Profiler &profiler; - /* memory alignment */ - virtual int mem_sub_ptr_alignment() - { - return MIN_ALIGNMENT_CPU_DATA_TYPES; - } - /* constant memory */ virtual void const_copy_to(const char *name, void *host, size_t size) = 0; - /* open shading language, only for CPU device */ - virtual void *osl_memory() - { - return NULL; - } - /* load/compile kernels, must be called before adding tasks */ - virtual bool load_kernels(const DeviceRequestedFeatures & /*requested_features*/) + virtual bool load_kernels(uint /*kernel_features*/) { return true; } - /* Wait for device to become available to upload data and receive tasks - * This method is used by the OpenCL device to load the - * optimized kernels or when not (yet) available load the - * generic kernels (only during foreground rendering) */ - virtual bool wait_for_availability(const DeviceRequestedFeatures & /*requested_features*/) - { - return true; - } - /* Check if there are 'better' kernels available to be used - * We can switch over to these kernels - * This method is used to determine if we can switch the preview kernels - * to regular kernels */ - virtual DeviceKernelStatus get_active_kernel_switch_state() - { - return DEVICE_KERNEL_USING_FEATURE_KERNEL; - } + /* GPU device only functions. + * These may not be used on CPU or multi-devices. */ - /* tasks */ - virtual int get_split_task_count(DeviceTask &) - { - return 1; - } + /* Create new queue for executing kernels in. */ + virtual unique_ptr<DeviceQueue> gpu_queue_create(); + + /* CPU device only functions. + * These may not be used on GPU or multi-devices. */ - virtual void task_add(DeviceTask &task) = 0; - virtual void task_wait() = 0; - virtual void task_cancel() = 0; - - /* opengl drawing */ - virtual void draw_pixels(device_memory &mem, - int y, - int w, - int h, - int width, - int height, - int dx, - int dy, - int dw, - int dh, - bool transparent, - const DeviceDrawParams &draw_params); + /* Get CPU kernel functions for native instruction set. */ + virtual const CPUKernels *get_cpu_kernels() const; + /* Get kernel globals to pass to kernels. */ + virtual void get_cpu_kernel_thread_globals( + vector<CPUKernelThreadGlobals> & /*kernel_thread_globals*/); + /* Get OpenShadingLanguage memory buffer. */ + virtual void *get_cpu_osl_memory(); /* acceleration structure building */ virtual void build_bvh(BVH *bvh, Progress &progress, bool refit); @@ -429,25 +191,11 @@ class Device { /* OptiX specific destructor. */ virtual void release_optix_bvh(BVH * /*bvh*/){}; -#ifdef WITH_NETWORK - /* networking */ - void server_run(); -#endif - /* multi device */ - virtual void map_tile(Device * /*sub_device*/, RenderTile & /*tile*/) - { - } virtual int device_number(Device * /*sub_device*/) { return 0; } - virtual void map_neighbor_tiles(Device * /*sub_device*/, RenderTileNeighbors & /*neighbors*/) - { - } - virtual void unmap_neighbor_tiles(Device * /*sub_device*/, RenderTileNeighbors & /*neighbors*/) - { - } virtual bool is_resident(device_ptr /*key*/, Device *sub_device) { @@ -460,11 +208,47 @@ class Device { return false; } + /* Graphics resources interoperability. + * + * The interoperability comes here by the meaning that the device is capable of computing result + * directly into an OpenGL (or other graphics library) buffer. */ + + /* Check display si to be updated using graphics interoperability. + * The interoperability can not be used is it is not supported by the device. But the device + * might also force disable the interoperability if it detects that it will be slower than + * copying pixels from the render buffer. */ + virtual bool should_use_graphics_interop() + { + return false; + } + + /* Buffer denoising. */ + + /* Returns true if task is fully handled. */ + virtual bool denoise_buffer(const DeviceDenoiseTask & /*task*/) + { + LOG(ERROR) << "Request buffer denoising from a device which does not support it."; + return false; + } + + virtual DeviceQueue *get_denoise_queue() + { + LOG(ERROR) << "Request denoising queue from a device which does not support it."; + return nullptr; + } + + /* Sub-devices */ + + /* Run given callback for every individual device which will be handling rendering. + * For the single device the callback is called for the device itself. For the multi-device the + * callback is only called for the sub-devices. */ + virtual void foreach_device(const function<void(Device *)> &callback) + { + callback(this); + } + /* static */ - static Device *create(DeviceInfo &info, - Stats &stats, - Profiler &profiler, - bool background = true); + static Device *create(const DeviceInfo &info, Stats &stats, Profiler &profiler); static DeviceType type_from_string(const char *name); static string string_from_type(DeviceType type); @@ -499,9 +283,7 @@ class Device { static thread_mutex device_mutex; static vector<DeviceInfo> cuda_devices; static vector<DeviceInfo> optix_devices; - static vector<DeviceInfo> opencl_devices; static vector<DeviceInfo> cpu_devices; - static vector<DeviceInfo> network_devices; static uint devices_initialized_mask; }; |