diff options
Diffstat (limited to 'drivers')
| -rw-r--r-- | drivers/dummy/rasterizer_dummy.h | 445 | ||||
| -rw-r--r-- | drivers/dummy/texture_loader_dummy.cpp | 2 | ||||
| -rw-r--r-- | drivers/dummy/texture_loader_dummy.h | 2 | ||||
| -rw-r--r-- | drivers/vulkan/rendering_device_vulkan.cpp | 308 | ||||
| -rw-r--r-- | drivers/vulkan/rendering_device_vulkan.h | 17 | ||||
| -rw-r--r-- | drivers/vulkan/vulkan_context.cpp | 62 | ||||
| -rw-r--r-- | drivers/vulkan/vulkan_context.h | 9 |
7 files changed, 416 insertions, 429 deletions
diff --git a/drivers/dummy/rasterizer_dummy.h b/drivers/dummy/rasterizer_dummy.h index 72ab18d115..082f247476 100644 --- a/drivers/dummy/rasterizer_dummy.h +++ b/drivers/dummy/rasterizer_dummy.h @@ -80,11 +80,11 @@ public: int sdfgi_get_pending_region_count(RID p_render_buffers) const override { return 0; } AABB sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const override { return AABB(); } uint32_t sdfgi_get_pending_region_cascade(RID p_render_buffers, int p_region) const override { return 0; } - void sdfgi_update_probes(RID p_render_buffers, RID p_environment, const Vector<RID> &p_directional_lights, const RID *p_positional_light_instances, uint32_t p_positional_light_count) override {} /* SKY API */ - RID sky_create() override { return RID(); } + RID sky_allocate() override { return RID(); } + void sky_initialize(RID p_rid) override {} void sky_set_radiance_size(RID p_sky, int p_radiance_size) override {} void sky_set_mode(RID p_sky, RS::SkyMode p_samples) override {} void sky_set_material(RID p_sky, RID p_material) override {} @@ -92,8 +92,8 @@ public: /* ENVIRONMENT API */ - RID environment_create() override { return RID(); } - + RID environment_allocate() override { return RID(); } + void environment_initialize(RID p_rid) override {} void environment_set_background(RID p_env, RS::EnvironmentBG p_bg) override {} void environment_set_sky(RID p_env, RID p_sky) override {} void environment_set_sky_custom_fov(RID p_env, float p_scale) override {} @@ -112,7 +112,7 @@ public: void environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_power, float p_detail, float p_horizon, float p_sharpness, float p_light_affect, float p_ao_channel_affect) override {} void environment_set_ssao_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) override {} - void environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, bool p_use_multibounce, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias) override {} + void environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, float p_bounce_feedback, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias) override {} void environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count) override {} void environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames) override {} @@ -123,11 +123,9 @@ public: void environment_set_adjustment(RID p_env, bool p_enable, float p_brightness, float p_contrast, float p_saturation, bool p_use_1d_color_correction, RID p_color_correction) override {} void environment_set_fog(RID p_env, bool p_enable, const Color &p_light_color, float p_light_energy, float p_sun_scatter, float p_density, float p_height, float p_height_density, float p_aerial_perspective) override {} - void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, RS::EnvVolumetricFogShadowFilter p_shadow_filter) override {} + void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount) override {} void environment_set_volumetric_fog_volume_size(int p_size, int p_depth) override {} void environment_set_volumetric_fog_filter_active(bool p_enable) override {} - void environment_set_volumetric_fog_directional_shadow_shrink_size(int p_shrink_size) override {} - void environment_set_volumetric_fog_positional_shadow_shrink_size(int p_shrink_size) override {} Ref<Image> environment_bake_panorama(RID p_env, bool p_bake_irradiance, const Size2i &p_size) override { return Ref<Image>(); } @@ -135,8 +133,8 @@ public: RS::EnvironmentBG environment_get_background(RID p_env) const override { return RS::ENV_BG_KEEP; } int environment_get_canvas_max_layer(RID p_env) const override { return 0; } - RID camera_effects_create() override { return RID(); } - + RID camera_effects_allocate() override { return RID(); } + void camera_effects_initialize(RID p_rid) override {} void camera_effects_set_dof_blur_quality(RS::DOFBlurQuality p_quality, bool p_use_jitter) override {} void camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape p_shape) override {} @@ -177,11 +175,8 @@ public: void gi_probe_set_quality(RS::GIProbeQuality) override {} - void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold) override {} - void render_shadow(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0) override {} + void render_scene(RID p_render_buffers, const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_gi_probes, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data = nullptr) override {} void render_material(const Transform &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override {} - void render_sdfgi(RID p_render_buffers, int p_region, const PagedArray<GeometryInstance *> &p_instances) override {} - void render_sdfgi_static_lights(RID p_render_buffers, uint32_t p_cascade_count, const uint32_t *p_cascade_indices, const PagedArray<RID> *p_positional_lights) override {} void render_particle_collider_heightfield(RID p_collider, const Transform &p_transform, const PagedArray<GeometryInstance *> &p_instances) override {} void set_scene_pass(uint64_t p_pass) override {} @@ -212,50 +207,22 @@ public: class RasterizerStorageDummy : public RendererStorage { public: - /* TEXTURE API */ - struct DummyTexture { - int width = 0; - int height = 0; - uint32_t flags = 0; - Image::Format format = Image::Format::FORMAT_MAX; - Ref<Image> image; - String path; - }; - - struct DummySurface { - uint32_t format = 0; - RS::PrimitiveType primitive = RS::PrimitiveType::PRIMITIVE_MAX; - Vector<uint8_t> array; - int vertex_count = 0; - Vector<uint8_t> index_array; - int index_count = 0; - AABB aabb; - Vector<Vector<uint8_t>> blend_shapes; - Vector<AABB> bone_aabbs; - }; - - struct DummyMesh { - Vector<DummySurface> surfaces; - int blend_shape_count = 0; - RS::BlendShapeMode blend_shape_mode = RS::BlendShapeMode::BLEND_SHAPE_MODE_NORMALIZED; - }; - - mutable RID_PtrOwner<DummyTexture> texture_owner; - mutable RID_PtrOwner<DummyMesh> mesh_owner; - - RID texture_2d_create(const Ref<Image> &p_image) override { return RID(); } - RID texture_2d_layered_create(const Vector<Ref<Image>> &p_layers, RS::TextureLayeredType p_layered_type) override { return RID(); } - RID texture_3d_create(Image::Format, int p_width, int p_height, int p_depth, bool p_mipmaps, const Vector<Ref<Image>> &p_data) override { return RID(); } - RID texture_proxy_create(RID p_base) override { return RID(); } + bool can_create_resources_async() const override { return false; } + /* TEXTURE API */ + RID texture_allocate() override { return RID(); } + void texture_2d_initialize(RID p_texture, const Ref<Image> &p_image) override {} + void texture_2d_layered_initialize(RID p_texture, const Vector<Ref<Image>> &p_layers, RS::TextureLayeredType p_layered_type) override {} void texture_2d_update_immediate(RID p_texture, const Ref<Image> &p_image, int p_layer = 0) override {} void texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer = 0) override {} + void texture_3d_initialize(RID p_texture, Image::Format, int p_width, int p_height, int p_depth, bool p_mipmaps, const Vector<Ref<Image>> &p_data) override {} void texture_3d_update(RID p_texture, const Vector<Ref<Image>> &p_data) override {} + void texture_proxy_initialize(RID p_texture, RID p_base) override {} void texture_proxy_update(RID p_proxy, RID p_base) override {} - RID texture_2d_placeholder_create() override { return RID(); } - RID texture_2d_layered_placeholder_create(RenderingServer::TextureLayeredType p_layered_type) override { return RID(); } - RID texture_3d_placeholder_create() override { return RID(); } + void texture_2d_placeholder_initialize(RID p_texture) override {} + void texture_2d_layered_placeholder_initialize(RID p_texture, RenderingServer::TextureLayeredType p_layered_type) override {} + void texture_3d_placeholder_initialize(RID p_texture) override {} Ref<Image> texture_2d_get(RID p_texture) const override { return Ref<Image>(); } Ref<Image> texture_2d_layer_get(RID p_texture, int p_layer) const override { return Ref<Image>(); } @@ -263,10 +230,6 @@ public: void texture_replace(RID p_texture, RID p_by_texture) override {} void texture_set_size_override(RID p_texture, int p_width, int p_height) override {} -// FIXME: Disabled during Vulkan refactoring, should be ported. -#if 0 - void texture_bind(RID p_texture, uint32_t p_texture_no) = 0; -#endif void texture_set_path(RID p_texture, const String &p_path) override {} String texture_get_path(RID p_texture) const override { return String(); } @@ -284,113 +247,18 @@ public: /* CANVAS TEXTURE API */ - RID canvas_texture_create() override { return RID(); } + RID canvas_texture_allocate() override { return RID(); } + void canvas_texture_initialize(RID p_rid) override {} void canvas_texture_set_channel(RID p_canvas_texture, RS::CanvasTextureChannel p_channel, RID p_texture) override {} void canvas_texture_set_shading_parameters(RID p_canvas_texture, const Color &p_base_color, float p_shininess) override {} void canvas_texture_set_texture_filter(RID p_item, RS::CanvasItemTextureFilter p_filter) override {} void canvas_texture_set_texture_repeat(RID p_item, RS::CanvasItemTextureRepeat p_repeat) override {} -#if 0 - RID texture_create() override { - DummyTexture *texture = memnew(DummyTexture); - ERR_FAIL_COND_V(!texture, RID()); - return texture_owner.make_rid(texture); - } - - void texture_allocate(RID p_texture, int p_width, int p_height, int p_depth_3d, Image::Format p_format, RenderingServer::TextureType p_type = RS::TEXTURE_TYPE_2D, uint32_t p_flags = RS::TEXTURE_FLAGS_DEFAULT) override { - DummyTexture *t = texture_owner.getornull(p_texture); - ERR_FAIL_COND(!t); - t->width = p_width; - t->height = p_height; - t->flags = p_flags; - t->format = p_format; - t->image = Ref<Image>(memnew(Image)); - t->image->create(p_width, p_height, false, p_format); - } - void texture_set_data(RID p_texture, const Ref<Image> &p_image, int p_level) override { - DummyTexture *t = texture_owner.getornull(p_texture); - ERR_FAIL_COND(!t); - t->width = p_image->get_width(); - t->height = p_image->get_height(); - t->format = p_image->get_format(); - t->image->create(t->width, t->height, false, t->format, p_image->get_data()); - } - - void texture_set_data_partial(RID p_texture, const Ref<Image> &p_image, int src_x, int src_y, int src_w, int src_h, int dst_x, int dst_y, int p_dst_mip, int p_level) override { - DummyTexture *t = texture_owner.getornull(p_texture); - - ERR_FAIL_COND(!t); - ERR_FAIL_COND_MSG(p_image.is_null(), "It's not a reference to a valid Image object."); - ERR_FAIL_COND(t->format != p_image->get_format()); - ERR_FAIL_COND(src_w <= 0 || src_h <= 0); - ERR_FAIL_COND(src_x < 0 || src_y < 0 || src_x + src_w > p_image->get_width() || src_y + src_h > p_image->get_height()); - ERR_FAIL_COND(dst_x < 0 || dst_y < 0 || dst_x + src_w > t->width || dst_y + src_h > t->height); - - t->image->blit_rect(p_image, Rect2(src_x, src_y, src_w, src_h), Vector2(dst_x, dst_y)); - } - - Ref<Image> texture_get_data(RID p_texture, int p_level) const override { - DummyTexture *t = texture_owner.getornull(p_texture); - ERR_FAIL_COND_V(!t, Ref<Image>()); - return t->image; - } - void texture_set_flags(RID p_texture, uint32_t p_flags) override { - DummyTexture *t = texture_owner.getornull(p_texture); - ERR_FAIL_COND(!t); - t->flags = p_flags; - } - uint32_t texture_get_flags(RID p_texture) const override { - DummyTexture *t = texture_owner.getornull(p_texture); - ERR_FAIL_COND_V(!t, 0); - return t->flags; - } - Image::Format texture_get_format(RID p_texture) const override { - DummyTexture *t = texture_owner.getornull(p_texture); - ERR_FAIL_COND_V(!t, Image::FORMAT_RGB8); - return t->format; - } - - RenderingServer::TextureType texture_get_type(RID p_texture) const override { return RS::TEXTURE_TYPE_2D; } - uint32_t texture_get_texid(RID p_texture) const override { return 0; } - uint32_t texture_get_width(RID p_texture) const override { return 0; } - uint32_t texture_get_height(RID p_texture) const override { return 0; } - uint32_t texture_get_depth(RID p_texture) const override { return 0; } - void texture_set_size_override(RID p_texture, int p_width, int p_height, int p_depth_3d) override {} - void texture_bind(RID p_texture, uint32_t p_texture_no) override {} - - void texture_set_path(RID p_texture, const String &p_path) override { - DummyTexture *t = texture_owner.getornull(p_texture); - ERR_FAIL_COND(!t); - t->path = p_path; - } - String texture_get_path(RID p_texture) const override { - DummyTexture *t = texture_owner.getornull(p_texture); - ERR_FAIL_COND_V(!t, String()); - return t->path; - } - - void texture_set_shrink_all_x2_on_set_data(bool p_enable) override {} - - void texture_debug_usage(List<RS::TextureInfo> *r_info) override {} - - RID texture_create_radiance_cubemap(RID p_source, int p_resolution = -1) const override { return RID(); } - - void texture_set_detect_3d_callback(RID p_texture, RenderingServer::TextureDetectCallback p_callback, void *p_userdata) override {} - void texture_set_detect_srgb_callback(RID p_texture, RenderingServer::TextureDetectCallback p_callback, void *p_userdata) override {} - void texture_set_detect_normal_callback(RID p_texture, RenderingServer::TextureDetectCallback p_callback, void *p_userdata) override {} - - void textures_keep_original(bool p_enable) override {} - - void texture_set_proxy(RID p_proxy, RID p_base) override {} - Size2 texture_size_with_proxy(RID p_texture) const override { return Size2(); } - void texture_set_force_redraw_if_visible(RID p_texture, bool p_enable) override {} -#endif - /* SHADER API */ - RID shader_create() override { return RID(); } - + RID shader_allocate() override { return RID(); } + void shader_initialize(RID p_rid) override {} void shader_set_code(RID p_shader, const String &p_code) override {} String shader_get_code(RID p_shader) const override { return ""; } void shader_get_param_list(RID p_shader, List<PropertyInfo> *p_param_list) const override {} @@ -403,8 +271,8 @@ public: /* COMMON MATERIAL API */ - RID material_create() override { return RID(); } - + RID material_allocate() override { return RID(); } + void material_initialize(RID p_rid) override {} void material_set_render_priority(RID p_material, int priority) override {} void material_set_shader(RID p_shader_material, RID p_shader) override {} @@ -420,14 +288,8 @@ public: /* MESH API */ - RID mesh_create() override { - DummyMesh *mesh = memnew(DummyMesh); - ERR_FAIL_COND_V(!mesh, RID()); - mesh->blend_shape_count = 0; - mesh->blend_shape_mode = RS::BLEND_SHAPE_MODE_NORMALIZED; - return mesh_owner.make_rid(mesh); - } - + RID mesh_allocate() override { return RID(); } + void mesh_initialize(RID p_rid) override {} void mesh_set_blend_shape_count(RID p_mesh, int p_blend_shape_count) override {} bool mesh_needs_instance(RID p_mesh, bool p_has_skeleton) override { return false; } RID mesh_instance_create(RID p_base) override { return RID(); } @@ -440,139 +302,31 @@ public: void mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_surface) override {} -#if 0 - void mesh_add_surface(RID p_mesh, uint32_t p_format, RS::PrimitiveType p_primitive, const Vector<uint8_t> &p_array, int p_vertex_count, const Vector<uint8_t> &p_index_array, int p_index_count, const AABB &p_aabb, const Vector<Vector<uint8_t> > &p_blend_shapes = Vector<Vector<uint8_t> >(), const Vector<AABB> &p_bone_aabbs = Vector<AABB>()) override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND(!m); - - m->surfaces.push_back(DummySurface()); - DummySurface *s = &m->surfaces.write[m->surfaces.size() - 1]; - s->format = p_format; - s->primitive = p_primitive; - s->array = p_array; - s->vertex_count = p_vertex_count; - s->index_array = p_index_array; - s->index_count = p_index_count; - s->aabb = p_aabb; - s->blend_shapes = p_blend_shapes; - s->bone_aabbs = p_bone_aabbs; - } - - void mesh_set_blend_shape_count(RID p_mesh, int p_amount) override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND(!m); - m->blend_shape_count = p_amount; - } -#endif - - int mesh_get_blend_shape_count(RID p_mesh) const override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND_V(!m, 0); - return m->blend_shape_count; - } + int mesh_get_blend_shape_count(RID p_mesh) const override { return 0; } - void mesh_set_blend_shape_mode(RID p_mesh, RS::BlendShapeMode p_mode) override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND(!m); - m->blend_shape_mode = p_mode; - } - RS::BlendShapeMode mesh_get_blend_shape_mode(RID p_mesh) const override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND_V(!m, RS::BLEND_SHAPE_MODE_NORMALIZED); - return m->blend_shape_mode; - } + void mesh_set_blend_shape_mode(RID p_mesh, RS::BlendShapeMode p_mode) override {} + RS::BlendShapeMode mesh_get_blend_shape_mode(RID p_mesh) const override { return RS::BLEND_SHAPE_MODE_NORMALIZED; } void mesh_surface_update_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) override {} void mesh_surface_set_material(RID p_mesh, int p_surface, RID p_material) override {} RID mesh_surface_get_material(RID p_mesh, int p_surface) const override { return RID(); } -#if 0 - int mesh_surface_get_array_len(RID p_mesh, int p_surface) const override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND_V(!m, 0); - - return m->surfaces[p_surface].vertex_count; - } - int mesh_surface_get_array_index_len(RID p_mesh, int p_surface) const override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND_V(!m, 0); - - return m->surfaces[p_surface].index_count; - } - - Vector<uint8_t> mesh_surface_get_array(RID p_mesh, int p_surface) const override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND_V(!m, Vector<uint8_t>()); - - return m->surfaces[p_surface].array; - } - Vector<uint8_t> mesh_surface_get_index_array(RID p_mesh, int p_surface) const override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND_V(!m, Vector<uint8_t>()); - - return m->surfaces[p_surface].index_array; - } - - uint32_t mesh_surface_get_format(RID p_mesh, int p_surface) const override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND_V(!m, 0); - - return m->surfaces[p_surface].format; - } - RS::PrimitiveType mesh_surface_get_primitive_type(RID p_mesh, int p_surface) const override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND_V(!m, RS::PRIMITIVE_POINTS); - - return m->surfaces[p_surface].primitive; - } - - AABB mesh_surface_get_aabb(RID p_mesh, int p_surface) const override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND_V(!m, AABB()); - - return m->surfaces[p_surface].aabb; - } - Vector<Vector<uint8_t> > mesh_surface_get_blend_shapes(RID p_mesh, int p_surface) const override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND_V(!m, Vector<Vector<uint8_t> >()); - - return m->surfaces[p_surface].blend_shapes; - } - Vector<AABB> mesh_surface_get_skeleton_aabb(RID p_mesh, int p_surface) const override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND_V(!m, Vector<AABB>()); - - return m->surfaces[p_surface].bone_aabbs; - } - - void mesh_remove_surface(RID p_mesh, int p_index) override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND(!m); - ERR_FAIL_COND(p_index >= m->surfaces.size()); - - m->surfaces.remove(p_index); - } -#endif - RS::SurfaceData mesh_get_surface(RID p_mesh, int p_surface) const override { return RS::SurfaceData(); } - int mesh_get_surface_count(RID p_mesh) const override { - DummyMesh *m = mesh_owner.getornull(p_mesh); - ERR_FAIL_COND_V(!m, 0); - return m->surfaces.size(); - } + int mesh_get_surface_count(RID p_mesh) const override { return 0; } void mesh_set_custom_aabb(RID p_mesh, const AABB &p_aabb) override {} AABB mesh_get_custom_aabb(RID p_mesh) const override { return AABB(); } AABB mesh_get_aabb(RID p_mesh, RID p_skeleton = RID()) override { return AABB(); } + void mesh_set_shadow_mesh(RID p_mesh, RID p_shadow_mesh) override {} void mesh_clear(RID p_mesh) override {} /* MULTIMESH API */ - RID multimesh_create() override { return RID(); } - - void multimesh_allocate(RID p_multimesh, int p_instances, RS::MultimeshTransformFormat p_transform_format, bool p_use_colors = false, bool p_use_custom_data = false) override {} + RID multimesh_allocate() override { return RID(); } + void multimesh_initialize(RID p_rid) override {} + void multimesh_allocate_data(RID p_multimesh, int p_instances, RS::MultimeshTransformFormat p_transform_format, bool p_use_colors = false, bool p_use_custom_data = false) override {} int multimesh_get_instance_count(RID p_multimesh) const override { return 0; } void multimesh_set_mesh(RID p_multimesh, RID p_mesh) override {} @@ -596,7 +350,8 @@ public: /* IMMEDIATE API */ - RID immediate_create() override { return RID(); } + RID immediate_allocate() override { return RID(); } + void immediate_initialize(RID p_rid) override {} void immediate_begin(RID p_immediate, RS::PrimitiveType p_rimitive, RID p_texture = RID()) override {} void immediate_vertex(RID p_immediate, const Vector3 &p_vertex) override {} void immediate_normal(RID p_immediate, const Vector3 &p_normal) override {} @@ -612,8 +367,9 @@ public: /* SKELETON API */ - RID skeleton_create() override { return RID(); } - void skeleton_allocate(RID p_skeleton, int p_bones, bool p_2d_skeleton = false) override {} + RID skeleton_allocate() override { return RID(); } + void skeleton_initialize(RID p_rid) override {} + void skeleton_allocate_data(RID p_skeleton, int p_bones, bool p_2d_skeleton = false) override {} void skeleton_set_base_transform_2d(RID p_skeleton, const Transform2D &p_base_transform) override {} int skeleton_get_bone_count(RID p_skeleton) const override { return 0; } void skeleton_bone_set_transform(RID p_skeleton, int p_bone, const Transform &p_transform) override {} @@ -623,7 +379,14 @@ public: /* Light API */ - RID light_create(RS::LightType p_type) override { return RID(); } + RID directional_light_allocate() override { return RID(); } + void directional_light_initialize(RID p_rid) override {} + RID omni_light_allocate() override { return RID(); } + void omni_light_initialize(RID p_rid) override {} + RID spot_light_allocate() override { return RID(); } + void spot_light_initialize(RID p_rid) override {} + RID reflection_probe_allocate() override { return RID(); } + void reflection_probe_initialize(RID p_rid) override {} void light_set_color(RID p_light, const Color &p_color) override {} void light_set_param(RID p_light, RS::LightParam p_param, float p_value) override {} @@ -661,8 +424,6 @@ public: /* PROBE API */ - RID reflection_probe_create() override { return RID(); } - void reflection_probe_set_update_mode(RID p_probe, RS::ReflectionProbeUpdateMode p_mode) override {} void reflection_probe_set_intensity(RID p_probe, float p_intensity) override {} void reflection_probe_set_ambient_mode(RID p_probe, RS::ReflectionProbeAmbientMode p_mode) override {} @@ -690,7 +451,8 @@ public: /* DECAL API */ - RID decal_create() override { return RID(); } + RID decal_allocate() override { return RID(); } + void decal_initialize(RID p_rid) override {} void decal_set_extents(RID p_decal, const Vector3 &p_extents) override {} void decal_set_texture(RID p_decal, RS::DecalTexture p_type, RID p_texture) override {} void decal_set_emission_energy(RID p_decal, float p_energy) override {} @@ -705,9 +467,9 @@ public: /* GI PROBE API */ - RID gi_probe_create() override { return RID(); } - - void gi_probe_allocate(RID p_gi_probe, const Transform &p_to_cell_xform, const AABB &p_aabb, const Vector3i &p_octree_size, const Vector<uint8_t> &p_octree_cells, const Vector<uint8_t> &p_data_cells, const Vector<uint8_t> &p_distance_field, const Vector<int> &p_level_counts) override {} + RID gi_probe_allocate() override { return RID(); } + void gi_probe_initialize(RID p_rid) override {} + void gi_probe_allocate_data(RID p_gi_probe, const Transform &p_to_cell_xform, const AABB &p_aabb, const Vector3i &p_octree_size, const Vector<uint8_t> &p_octree_cells, const Vector<uint8_t> &p_data_cells, const Vector<uint8_t> &p_distance_field, const Vector<int> &p_level_counts) override {} AABB gi_probe_get_bounds(RID p_gi_probe) const override { return AABB(); } Vector3i gi_probe_get_octree_size(RID p_gi_probe) const override { return Vector3i(); } @@ -751,72 +513,8 @@ public: uint32_t gi_probe_get_version(RID p_gi_probe) override { return 0; } /* LIGHTMAP CAPTURE */ -#if 0 - struct Instantiable { - SelfList<RendererSceneRender::GeometryInstance>::List instance_list; - - _FORCE_INLINE_ void instance_change_notify(bool p_aabb = true, bool p_materials = true) override { - SelfList<RendererSceneRender::GeometryInstance> *instances = instance_list.first(); - while (instances) override { - //instances->self()->base_changed(p_aabb, p_materials); - instances = instances->next(); - } - } - - _FORCE_INLINE_ void instance_remove_deps() override { - SelfList<RendererSceneRender::GeometryInstance> *instances = instance_list.first(); - while (instances) override { - SelfList<RendererSceneRender::GeometryInstance> *next = instances->next(); - //instances->self()->base_removed(); - instances = next; - } - } - - Instantiable() override {} - ~Instantiable() override { - } - }; - - struct LightmapCapture : public Instantiable { - Vector<LightmapCaptureOctree> octree; - AABB bounds; - Transform cell_xform; - int cell_subdiv; - float energy; - LightmapCapture() override { - energy = 1.0; - cell_subdiv = 1; - } - }; - - mutable RID_PtrOwner<LightmapCapture> lightmap_capture_data_owner; - void lightmap_capture_set_bounds(RID p_capture, const AABB &p_bounds) override {} - AABB lightmap_capture_get_bounds(RID p_capture) const override { return AABB(); } - void lightmap_capture_set_octree(RID p_capture, const Vector<uint8_t> &p_octree) override {} - RID lightmap_capture_create() override { - LightmapCapture *capture = memnew(LightmapCapture); - return lightmap_capture_data_owner.make_rid(capture); - } - Vector<uint8_t> lightmap_capture_get_octree(RID p_capture) const override { - const LightmapCapture *capture = lightmap_capture_data_owner.getornull(p_capture); - ERR_FAIL_COND_V(!capture, Vector<uint8_t>()); - return Vector<uint8_t>(); - } - void lightmap_capture_set_octree_cell_transform(RID p_capture, const Transform &p_xform) override {} - Transform lightmap_capture_get_octree_cell_transform(RID p_capture) const override { return Transform(); } - void lightmap_capture_set_octree_cell_subdiv(RID p_capture, int p_subdiv) override {} - int lightmap_capture_get_octree_cell_subdiv(RID p_capture) const override { return 0; } - void lightmap_capture_set_energy(RID p_capture, float p_energy) override {} - float lightmap_capture_get_energy(RID p_capture) const override { return 0.0; } - const Vector<LightmapCaptureOctree> *lightmap_capture_get_octree_ptr(RID p_capture) const override { - const LightmapCapture *capture = lightmap_capture_data_owner.getornull(p_capture); - ERR_FAIL_COND_V(!capture, nullptr); - return &capture->octree; - } -#endif - - RID lightmap_create() override { return RID(); } - + RID lightmap_allocate() override { return RID(); } + void lightmap_initialize(RID p_rid) override {} void lightmap_set_textures(RID p_lightmap, RID p_light, bool p_uses_spherical_haromics) override {} void lightmap_set_probe_bounds(RID p_lightmap, const AABB &p_bounds) override {} void lightmap_set_probe_interior(RID p_lightmap, bool p_interior) override {} @@ -833,8 +531,8 @@ public: /* PARTICLES */ - RID particles_create() override { return RID(); } - + RID particles_allocate() override { return RID(); } + void particles_initialize(RID p_rid) override {} void particles_emit(RID p_particles, const Transform &p_transform, const Vector3 &p_velocity, const Color &p_color, const Color &p_custom, uint32_t p_emit_flags) override {} void particles_set_emitting(RID p_particles, bool p_emitting) override {} void particles_set_amount(RID p_particles, int p_amount) override {} @@ -876,7 +574,8 @@ public: /* PARTICLES COLLISION */ - RID particles_collision_create() override { return RID(); } + RID particles_collision_allocate() override { return RID(); } + void particles_collision_initialize(RID p_rid) override {} void particles_collision_set_collision_type(RID p_particles_collision, RS::ParticlesCollisionType p_type) override {} void particles_collision_set_cull_mask(RID p_particles_collision, uint32_t p_cull_mask) override {} void particles_collision_set_sphere_radius(RID p_particles_collision, float p_radius) override {} @@ -935,30 +634,8 @@ public: void render_target_set_sdf_size_and_scale(RID p_render_target, RS::ViewportSDFOversize p_size, RS::ViewportSDFScale p_scale) override {} Rect2i render_target_get_sdf_rect(RID p_render_target) const override { return Rect2i(); } - RS::InstanceType get_base_type(RID p_rid) const override { - if (mesh_owner.owns(p_rid)) { - return RS::INSTANCE_MESH; - } - - return RS::INSTANCE_NONE; - } - - bool free(RID p_rid) override { - if (texture_owner.owns(p_rid)) { - // delete the texture - DummyTexture *texture = texture_owner.getornull(p_rid); - texture_owner.free(p_rid); - memdelete(texture); - } - - if (mesh_owner.owns(p_rid)) { - // delete the mesh - DummyMesh *mesh = mesh_owner.getornull(p_rid); - mesh_owner.free(p_rid); - memdelete(mesh); - } - return true; - } + RS::InstanceType get_base_type(RID p_rid) const override { return RS::INSTANCE_NONE; } + bool free(RID p_rid) override { return true; } bool has_os_feature(const String &p_feature) const override { return false; } diff --git a/drivers/dummy/texture_loader_dummy.cpp b/drivers/dummy/texture_loader_dummy.cpp index 2484e6d140..f148e42845 100644 --- a/drivers/dummy/texture_loader_dummy.cpp +++ b/drivers/dummy/texture_loader_dummy.cpp @@ -35,7 +35,7 @@ #include <string.h> -RES ResourceFormatDummyTexture::load(const String &p_path, const String &p_original_path, Error *r_error, bool p_use_sub_threads, float *r_progress, bool p_no_cache) { +RES ResourceFormatDummyTexture::load(const String &p_path, const String &p_original_path, Error *r_error, bool p_use_sub_threads, float *r_progress, CacheMode p_cache_mode) { unsigned int width = 8; unsigned int height = 8; diff --git a/drivers/dummy/texture_loader_dummy.h b/drivers/dummy/texture_loader_dummy.h index 3b71b5824f..00e6b9cc53 100644 --- a/drivers/dummy/texture_loader_dummy.h +++ b/drivers/dummy/texture_loader_dummy.h @@ -36,7 +36,7 @@ class ResourceFormatDummyTexture : public ResourceFormatLoader { public: - virtual RES load(const String &p_path, const String &p_original_path = "", Error *r_error = nullptr, bool p_use_sub_threads = false, float *r_progress = nullptr, bool p_no_cache = false); + virtual RES load(const String &p_path, const String &p_original_path = "", Error *r_error = nullptr, bool p_use_sub_threads = false, float *r_progress = nullptr, CacheMode p_cache_mode = CACHE_MODE_REUSE); virtual void get_recognized_extensions(List<String> *p_extensions) const; virtual bool handles_type(const String &p_type) const; virtual String get_resource_type(const String &p_path) const; diff --git a/drivers/vulkan/rendering_device_vulkan.cpp b/drivers/vulkan/rendering_device_vulkan.cpp index ef331ec4b6..9584dd3f67 100644 --- a/drivers/vulkan/rendering_device_vulkan.cpp +++ b/drivers/vulkan/rendering_device_vulkan.cpp @@ -74,11 +74,13 @@ RenderingDeviceVulkan::Buffer *RenderingDeviceVulkan::_get_buffer_from_owner(RID } else if (texture_buffer_owner.owns(p_buffer)) { if (p_post_barrier & BARRIER_MASK_RASTER) { r_stage_mask |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT; + r_access_mask |= VK_ACCESS_SHADER_READ_BIT; } if (p_post_barrier & BARRIER_MASK_COMPUTE) { r_stage_mask |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; + r_access_mask |= VK_ACCESS_SHADER_READ_BIT; } - r_access_mask |= VK_ACCESS_SHADER_READ_BIT; + buffer = &texture_buffer_owner.getornull(p_buffer)->buffer; } else if (storage_buffer_owner.owns(p_buffer)) { buffer = storage_buffer_owner.getornull(p_buffer); @@ -1627,6 +1629,9 @@ void RenderingDeviceVulkan::_memory_barrier(VkPipelineStageFlags p_src_stage_mas mem_barrier.srcAccessMask = p_src_access; mem_barrier.dstAccessMask = p_dst_sccess; + if (p_src_stage_mask == 0 || p_dst_stage_mask == 0) { + return; //no barrier, since this is invalid + } vkCmdPipelineBarrier(p_sync_with_draw ? frames[frame].draw_command_buffer : frames[frame].setup_command_buffer, p_src_stage_mask, p_dst_stage_mask, 0, 1, &mem_barrier, 0, nullptr, 0, nullptr); } @@ -2477,6 +2482,10 @@ Error RenderingDeviceVulkan::texture_update(RID p_texture, uint32_t p_layer, con access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT; } + if (barrier_flags == 0) { + barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT; + } + VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; @@ -2496,6 +2505,13 @@ Error RenderingDeviceVulkan::texture_update(RID p_texture, uint32_t p_layer, con vkCmdPipelineBarrier(command_buffer, VK_ACCESS_TRANSFER_WRITE_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier); } + if (texture->used_in_frame != frames_drawn) { + texture->used_in_raster = false; + texture->used_in_compute = false; + texture->used_in_frame = frames_drawn; + } + texture->used_in_transfer = true; + return OK; } @@ -2844,6 +2860,10 @@ Error RenderingDeviceVulkan::texture_copy(RID p_from_texture, RID p_to_texture, access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT; } + if (barrier_flags == 0) { + barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT; + } + { //restore src VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; @@ -3011,6 +3031,10 @@ Error RenderingDeviceVulkan::texture_resolve_multisample(RID p_from_texture, RID access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT; } + if (barrier_flags == 0) { + barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT; + } + { //restore src VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; @@ -3143,6 +3167,10 @@ Error RenderingDeviceVulkan::texture_clear(RID p_texture, const Color &p_color, access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT; } + if (barrier_flags == 0) { + barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT; + } + VkImageMemoryBarrier image_memory_barrier; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; @@ -3163,6 +3191,13 @@ Error RenderingDeviceVulkan::texture_clear(RID p_texture, const Color &p_color, vkCmdPipelineBarrier(command_buffer, VK_PIPELINE_STAGE_TRANSFER_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier); } + if (src_tex->used_in_frame != frames_drawn) { + src_tex->used_in_raster = false; + src_tex->used_in_compute = false; + src_tex->used_in_frame = frames_drawn; + } + src_tex->used_in_transfer = true; + return OK; } @@ -3289,6 +3324,7 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF dependency_from_external.srcStageMask |= reading_stages; } } break; + case INITIAL_ACTION_CLEAR_REGION_CONTINUE: case INITIAL_ACTION_CONTINUE: { if (p_format[i].usage_flags & TEXTURE_USAGE_COLOR_ATTACHMENT_BIT) { description.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD; @@ -3296,7 +3332,7 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; } else if (p_format[i].usage_flags & TEXTURE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT) { description.loadOp = VK_ATTACHMENT_LOAD_OP_LOAD; - description.initialLayout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL; //don't care what is there + description.initialLayout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL; description.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD; } else { description.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; @@ -3425,8 +3461,13 @@ VkRenderPass RenderingDeviceVulkan::_render_pass_create(const Vector<AttachmentF render_pass_create_info.pAttachments = attachments.ptr(); render_pass_create_info.subpassCount = 1; render_pass_create_info.pSubpasses = &subpass; - render_pass_create_info.dependencyCount = 2; - render_pass_create_info.pDependencies = dependencies; + // Commenting this because it seems it just avoids raster and compute to work at the same time. + // Other barriers seem to be protecting the render pass fine. + // render_pass_create_info.dependencyCount = 2; + // render_pass_create_info.pDependencies = dependencies; + + render_pass_create_info.dependencyCount = 0; + render_pass_create_info.pDependencies = nullptr; VkRenderPass render_pass; VkResult res = vkCreateRenderPass(device, &render_pass_create_info, nullptr, &render_pass); @@ -4108,6 +4149,8 @@ RID RenderingDeviceVulkan::shader_create(const Vector<ShaderStageData> &p_stages bool is_compute = false; + uint32_t compute_local_size[3] = { 0, 0, 0 }; + for (int i = 0; i < p_stages.size(); i++) { if (p_stages[i].shader_stage == SHADER_STAGE_COMPUTE) { is_compute = true; @@ -4124,6 +4167,11 @@ RID RenderingDeviceVulkan::shader_create(const Vector<ShaderStageData> &p_stages ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, RID(), "Reflection of SPIR-V shader stage '" + String(shader_stage_names[p_stages[i].shader_stage]) + "' failed parsing shader."); + if (is_compute) { + compute_local_size[0] = module.entry_points->local_size.x; + compute_local_size[1] = module.entry_points->local_size.y; + compute_local_size[2] = module.entry_points->local_size.z; + } uint32_t binding_count = 0; result = spvReflectEnumerateDescriptorBindings(&module, &binding_count, nullptr); ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, RID(), @@ -4328,6 +4376,7 @@ RID RenderingDeviceVulkan::shader_create(const Vector<ShaderStageData> &p_stages } } } + uint32_t pc_count = 0; result = spvReflectEnumeratePushConstantBlocks(&module, &pc_count, nullptr); ERR_FAIL_COND_V_MSG(result != SPV_REFLECT_RESULT_SUCCESS, RID(), @@ -4376,6 +4425,9 @@ RID RenderingDeviceVulkan::shader_create(const Vector<ShaderStageData> &p_stages shader.fragment_outputs = fragment_outputs; shader.push_constant = push_constant; shader.is_compute = is_compute; + shader.compute_local_size[0] = compute_local_size[0]; + shader.compute_local_size[1] = compute_local_size[1]; + shader.compute_local_size[2] = compute_local_size[2]; String error_text; @@ -5216,7 +5268,14 @@ Error RenderingDeviceVulkan::buffer_update(RID p_buffer, uint32_t p_offset, uint #ifdef FORCE_FULL_BARRIER _full_barrier(true); #else - _buffer_memory_barrier(buffer->buffer, p_offset, p_size, VK_PIPELINE_STAGE_TRANSFER_BIT, dst_stage_mask, VK_ACCESS_TRANSFER_WRITE_BIT, dst_access, true); + if (dst_stage_mask == 0) { + dst_stage_mask = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT; + } + + if (p_post_barrier != RD::BARRIER_MASK_NO_BARRIER) { + _buffer_memory_barrier(buffer->buffer, p_offset, p_size, VK_PIPELINE_STAGE_TRANSFER_BIT, dst_stage_mask, VK_ACCESS_TRANSFER_WRITE_BIT, dst_access, true); + } + #endif return err; } @@ -5255,7 +5314,12 @@ Error RenderingDeviceVulkan::buffer_clear(RID p_buffer, uint32_t p_offset, uint3 #ifdef FORCE_FULL_BARRIER _full_barrier(true); #else - _buffer_memory_barrier(buffer->buffer, p_offset, p_size, VK_PIPELINE_STAGE_TRANSFER_BIT, dst_stage_mask, VK_ACCESS_TRANSFER_WRITE_BIT, dst_access, p_post_barrier); + if (dst_stage_mask == 0) { + dst_stage_mask = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT; + } + + _buffer_memory_barrier(buffer->buffer, p_offset, p_size, VK_PIPELINE_STAGE_TRANSFER_BIT, dst_stage_mask, VK_ACCESS_TRANSFER_WRITE_BIT, dst_access, dst_stage_mask); + #endif return OK; } @@ -5710,6 +5774,9 @@ RID RenderingDeviceVulkan::compute_pipeline_create(RID p_shader) { pipeline.pipeline_layout = shader->pipeline_layout; pipeline.shader = p_shader; pipeline.push_constant_size = shader->push_constant.push_constant_size; + pipeline.local_group_size[0] = shader->compute_local_size[0]; + pipeline.local_group_size[1] = shader->compute_local_size[1]; + pipeline.local_group_size[2] = shader->compute_local_size[2]; //create ID to associate with this pipeline RID id = compute_pipeline_owner.make_rid(pipeline); @@ -6019,7 +6086,7 @@ RenderingDevice::DrawListID RenderingDeviceVulkan::draw_list_begin(RID p_framebu _THREAD_SAFE_METHOD_ ERR_FAIL_COND_V_MSG(draw_list != nullptr, INVALID_ID, "Only one draw list can be active at the same time."); - ERR_FAIL_COND_V_MSG(compute_list != nullptr, INVALID_ID, "Only one draw/compute list can be active at the same time."); + ERR_FAIL_COND_V_MSG(compute_list != nullptr && !compute_list->state.allow_draw_overlap, INVALID_ID, "Only one draw/compute list can be active at the same time."); Framebuffer *framebuffer = framebuffer_owner.getornull(p_framebuffer); ERR_FAIL_COND_V(!framebuffer, INVALID_ID); @@ -6040,7 +6107,14 @@ RenderingDevice::DrawListID RenderingDeviceVulkan::draw_list_begin(RID p_framebu viewport_offset = regioni.position; viewport_size = regioni.size; - + if (p_initial_color_action == INITIAL_ACTION_CLEAR_REGION_CONTINUE) { + needs_clear_color = true; + p_initial_color_action = INITIAL_ACTION_CONTINUE; + } + if (p_initial_depth_action == INITIAL_ACTION_CLEAR_REGION_CONTINUE) { + needs_clear_depth = true; + p_initial_depth_action = INITIAL_ACTION_CONTINUE; + } if (p_initial_color_action == INITIAL_ACTION_CLEAR_REGION) { needs_clear_color = true; p_initial_color_action = INITIAL_ACTION_KEEP; @@ -6388,6 +6462,19 @@ void RenderingDeviceVulkan::draw_list_bind_uniform_set(DrawListID p_list, RID p_ dl->state.sets[p_index].uniform_set_format = uniform_set->format; dl->state.sets[p_index].uniform_set = p_uniform_set; + uint32_t mst_count = uniform_set->mutable_storage_textures.size(); + if (mst_count) { + Texture **mst_textures = const_cast<UniformSet *>(uniform_set)->mutable_storage_textures.ptrw(); + for (uint32_t i = 0; i < mst_count; i++) { + if (mst_textures[i]->used_in_frame != frames_drawn) { + mst_textures[i]->used_in_frame = frames_drawn; + mst_textures[i]->used_in_transfer = false; + mst_textures[i]->used_in_compute = false; + } + mst_textures[i]->used_in_raster = true; + } + } + #ifdef DEBUG_ENABLED { //validate that textures bound are not attached as framebuffer bindings uint32_t attachable_count = uniform_set->attachable_textures.size(); @@ -6673,23 +6760,43 @@ void RenderingDeviceVulkan::draw_list_end(uint32_t p_post_barrier) { access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } if (p_post_barrier & BARRIER_MASK_RASTER) { - barrier_flags |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT; - access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_INDEX_READ_BIT | VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | VK_ACCESS_INDIRECT_COMMAND_READ_BIT; + barrier_flags |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT /*| VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT*/; + access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_INDEX_READ_BIT | VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT /*| VK_ACCESS_INDIRECT_COMMAND_READ_BIT*/; } if (p_post_barrier & BARRIER_MASK_TRANSFER) { barrier_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT; access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT; } + if (barrier_flags == 0) { + barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT; + } + draw_list_bound_textures.clear(); - for (int i = 0; i < draw_list_storage_textures.size(); i++) { + VkImageMemoryBarrier *image_barriers = nullptr; + + uint32_t image_barrier_count = draw_list_storage_textures.size(); + + if (image_barrier_count) { + image_barriers = (VkImageMemoryBarrier *)alloca(sizeof(VkImageMemoryBarrier) * draw_list_storage_textures.size()); + } + + uint32_t src_stage = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT; + uint32_t src_access = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT; + + if (image_barrier_count) { + src_stage |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT; + src_access |= VK_ACCESS_SHADER_WRITE_BIT; + } + + for (uint32_t i = 0; i < image_barrier_count; i++) { Texture *texture = texture_owner.getornull(draw_list_storage_textures[i]); - VkImageMemoryBarrier image_memory_barrier; + VkImageMemoryBarrier &image_memory_barrier = image_barriers[i]; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; - image_memory_barrier.srcAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; + image_memory_barrier.srcAccessMask = src_access; image_memory_barrier.dstAccessMask = access_flags; image_memory_barrier.oldLayout = texture->layout; image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; @@ -6703,8 +6810,6 @@ void RenderingDeviceVulkan::draw_list_end(uint32_t p_post_barrier) { image_memory_barrier.subresourceRange.baseArrayLayer = texture->base_layer; image_memory_barrier.subresourceRange.layerCount = texture->layers; - vkCmdPipelineBarrier(frames[frame].draw_command_buffer, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier); - texture->layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; } @@ -6717,7 +6822,17 @@ void RenderingDeviceVulkan::draw_list_end(uint32_t p_post_barrier) { #ifdef FORCE_FULL_BARRIER _full_barrier(true); #else - _memory_barrier(VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT, barrier_flags, VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT, access_flags, true); + + VkMemoryBarrier mem_barrier; + mem_barrier.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER; + mem_barrier.pNext = nullptr; + mem_barrier.srcAccessMask = src_access; + mem_barrier.dstAccessMask = access_flags; + + if (image_barrier_count > 0 || p_post_barrier != BARRIER_MASK_NO_BARRIER) { + vkCmdPipelineBarrier(frames[frame].draw_command_buffer, src_stage, barrier_flags, 0, 1, &mem_barrier, 0, nullptr, image_barrier_count, image_barriers); + } + #endif } @@ -6725,12 +6840,13 @@ void RenderingDeviceVulkan::draw_list_end(uint32_t p_post_barrier) { /**** COMPUTE LISTS ****/ /***********************/ -RenderingDevice::ComputeListID RenderingDeviceVulkan::compute_list_begin() { - ERR_FAIL_COND_V_MSG(draw_list != nullptr, INVALID_ID, "Only one draw list can be active at the same time."); +RenderingDevice::ComputeListID RenderingDeviceVulkan::compute_list_begin(bool p_allow_draw_overlap) { + ERR_FAIL_COND_V_MSG(!p_allow_draw_overlap && draw_list != nullptr, INVALID_ID, "Only one draw list can be active at the same time."); ERR_FAIL_COND_V_MSG(compute_list != nullptr, INVALID_ID, "Only one draw/compute list can be active at the same time."); compute_list = memnew(ComputeList); compute_list->command_buffer = frames[frame].draw_command_buffer; + compute_list->state.allow_draw_overlap = p_allow_draw_overlap; return ID_TYPE_COMPUTE_LIST; } @@ -6787,6 +6903,9 @@ void RenderingDeviceVulkan::compute_list_bind_compute_pipeline(ComputeListID p_l } cl->state.pipeline_shader = pipeline->shader; + cl->state.local_group_size[0] = pipeline->local_group_size[0]; + cl->state.local_group_size[1] = pipeline->local_group_size[1]; + cl->state.local_group_size[2] = pipeline->local_group_size[2]; } #ifdef DEBUG_ENABLED @@ -6824,11 +6943,24 @@ void RenderingDeviceVulkan::compute_list_bind_uniform_set(ComputeListID p_list, cl->state.sets[p_index].uniform_set = p_uniform_set; uint32_t textures_to_sampled_count = uniform_set->mutable_sampled_textures.size(); + uint32_t textures_to_storage_count = uniform_set->mutable_storage_textures.size(); + Texture **textures_to_sampled = uniform_set->mutable_sampled_textures.ptrw(); + VkImageMemoryBarrier *texture_barriers = nullptr; + + if (textures_to_sampled_count + textures_to_storage_count) { + texture_barriers = (VkImageMemoryBarrier *)alloca(sizeof(VkImageMemoryBarrier) * (textures_to_sampled_count + textures_to_storage_count)); + } + uint32_t texture_barrier_count = 0; + + uint32_t src_stage_flags = 0; + for (uint32_t i = 0; i < textures_to_sampled_count; i++) { if (textures_to_sampled[i]->layout != VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL) { - VkImageMemoryBarrier image_memory_barrier; + src_stage_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; + + VkImageMemoryBarrier &image_memory_barrier = texture_barriers[texture_barrier_count++]; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; image_memory_barrier.srcAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; @@ -6845,23 +6977,55 @@ void RenderingDeviceVulkan::compute_list_bind_uniform_set(ComputeListID p_list, image_memory_barrier.subresourceRange.baseArrayLayer = textures_to_sampled[i]->base_layer; image_memory_barrier.subresourceRange.layerCount = textures_to_sampled[i]->layers; - vkCmdPipelineBarrier(cl->command_buffer, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier); - textures_to_sampled[i]->layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; cl->state.textures_to_sampled_layout.erase(textures_to_sampled[i]); } + + if (textures_to_sampled[i]->used_in_frame != frames_drawn) { + textures_to_sampled[i]->used_in_frame = frames_drawn; + textures_to_sampled[i]->used_in_transfer = false; + textures_to_sampled[i]->used_in_raster = false; + } + textures_to_sampled[i]->used_in_compute = true; } - uint32_t textures_to_storage_count = uniform_set->mutable_storage_textures.size(); Texture **textures_to_storage = uniform_set->mutable_storage_textures.ptrw(); for (uint32_t i = 0; i < textures_to_storage_count; i++) { if (textures_to_storage[i]->layout != VK_IMAGE_LAYOUT_GENERAL) { - VkImageMemoryBarrier image_memory_barrier; + uint32_t src_access_flags = 0; + + if (textures_to_storage[i]->used_in_frame == frames_drawn) { + if (textures_to_storage[i]->used_in_compute) { + src_stage_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; + src_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; + } + if (textures_to_storage[i]->used_in_raster) { + src_stage_flags |= VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT; + src_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; + } + if (textures_to_storage[i]->used_in_transfer) { + src_stage_flags |= VK_PIPELINE_STAGE_TRANSFER_BIT; + src_access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT; + } + + textures_to_storage[i]->used_in_compute = false; + textures_to_storage[i]->used_in_raster = false; + textures_to_storage[i]->used_in_compute = false; + + } else { + src_access_flags = 0; + textures_to_storage[i]->used_in_compute = false; + textures_to_storage[i]->used_in_raster = false; + textures_to_storage[i]->used_in_compute = false; + textures_to_storage[i]->used_in_frame = frames_drawn; + } + + VkImageMemoryBarrier &image_memory_barrier = texture_barriers[texture_barrier_count++]; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; - image_memory_barrier.srcAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; + image_memory_barrier.srcAccessMask = src_access_flags; image_memory_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; image_memory_barrier.oldLayout = textures_to_storage[i]->layout; image_memory_barrier.newLayout = VK_IMAGE_LAYOUT_GENERAL; @@ -6875,14 +7039,20 @@ void RenderingDeviceVulkan::compute_list_bind_uniform_set(ComputeListID p_list, image_memory_barrier.subresourceRange.baseArrayLayer = textures_to_storage[i]->base_layer; image_memory_barrier.subresourceRange.layerCount = textures_to_storage[i]->layers; - vkCmdPipelineBarrier(cl->command_buffer, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier); - textures_to_storage[i]->layout = VK_IMAGE_LAYOUT_GENERAL; cl->state.textures_to_sampled_layout.insert(textures_to_storage[i]); //needs to go back to sampled layout afterwards } } + if (texture_barrier_count) { + if (src_stage_flags == 0) { + src_stage_flags = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT; + } + + vkCmdPipelineBarrier(cl->command_buffer, src_stage_flags, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, 0, nullptr, 0, nullptr, texture_barrier_count, texture_barriers); + } + #if 0 { //validate that textures bound are not attached as framebuffer bindings uint32_t attachable_count = uniform_set->attachable_textures.size(); @@ -6976,6 +7146,27 @@ void RenderingDeviceVulkan::compute_list_dispatch(ComputeListID p_list, uint32_t vkCmdDispatch(cl->command_buffer, p_x_groups, p_y_groups, p_z_groups); } +void RenderingDeviceVulkan::compute_list_dispatch_threads(ComputeListID p_list, uint32_t p_x_threads, uint32_t p_y_threads, uint32_t p_z_threads) { + ERR_FAIL_COND(p_list != ID_TYPE_COMPUTE_LIST); + ERR_FAIL_COND(!compute_list); + + ComputeList *cl = compute_list; + +#ifdef DEBUG_ENABLED + + ERR_FAIL_COND_MSG(!cl->validation.pipeline_active, "No compute pipeline was set before attempting to draw."); + + if (cl->validation.pipeline_push_constant_size > 0) { + //using push constants, check that they were supplied + ERR_FAIL_COND_MSG(!cl->validation.pipeline_push_constant_supplied, + "The shader in this pipeline requires a push constant to be set before drawing, but it's not present."); + } + +#endif + + compute_list_dispatch(p_list, (p_x_threads - 1) / cl->state.local_group_size[0] + 1, (p_y_threads - 1) / cl->state.local_group_size[1] + 1, (p_z_threads - 1) / cl->state.local_group_size[2] + 1); +} + void RenderingDeviceVulkan::compute_list_dispatch_indirect(ComputeListID p_list, RID p_buffer, uint32_t p_offset) { ERR_FAIL_COND(p_list != ID_TYPE_COMPUTE_LIST); ERR_FAIL_COND(!compute_list); @@ -7047,7 +7238,7 @@ void RenderingDeviceVulkan::compute_list_end(uint32_t p_post_barrier) { uint32_t access_flags = 0; if (p_post_barrier & BARRIER_MASK_COMPUTE) { barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; - access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_INDIRECT_COMMAND_READ_BIT; + access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } if (p_post_barrier & BARRIER_MASK_RASTER) { barrier_flags |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT; @@ -7058,8 +7249,22 @@ void RenderingDeviceVulkan::compute_list_end(uint32_t p_post_barrier) { access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT; } + if (barrier_flags == 0) { + barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT; + } + + VkImageMemoryBarrier *image_barriers = nullptr; + + uint32_t image_barrier_count = compute_list->state.textures_to_sampled_layout.size(); + + if (image_barrier_count) { + image_barriers = (VkImageMemoryBarrier *)alloca(sizeof(VkImageMemoryBarrier) * image_barrier_count); + } + + uint32_t barrier_idx = 0; + for (Set<Texture *>::Element *E = compute_list->state.textures_to_sampled_layout.front(); E; E = E->next()) { - VkImageMemoryBarrier image_memory_barrier; + VkImageMemoryBarrier &image_memory_barrier = image_barriers[barrier_idx++]; image_memory_barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; image_memory_barrier.pNext = nullptr; image_memory_barrier.srcAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; @@ -7076,19 +7281,33 @@ void RenderingDeviceVulkan::compute_list_end(uint32_t p_post_barrier) { image_memory_barrier.subresourceRange.baseArrayLayer = E->get()->base_layer; image_memory_barrier.subresourceRange.layerCount = E->get()->layers; - // TODO: Look at the usages in the compute list and determine tighter dst stage and access masks based on some "final" usage equivalent - vkCmdPipelineBarrier(compute_list->command_buffer, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, barrier_flags, 0, 0, nullptr, 0, nullptr, 1, &image_memory_barrier); - E->get()->layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; + + if (E->get()->used_in_frame != frames_drawn) { + E->get()->used_in_transfer = false; + E->get()->used_in_raster = false; + E->get()->used_in_compute = false; + E->get()->used_in_frame = frames_drawn; + } } - memdelete(compute_list); - compute_list = nullptr; #ifdef FORCE_FULL_BARRIER _full_barrier(true); #else - _memory_barrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, barrier_flags, VK_ACCESS_SHADER_WRITE_BIT, VK_ACCESS_INDEX_READ_BIT | VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | VK_ACCESS_TRANSFER_READ_BIT | VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_INDIRECT_COMMAND_READ_BIT, true); + VkMemoryBarrier mem_barrier; + mem_barrier.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER; + mem_barrier.pNext = nullptr; + mem_barrier.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT; + mem_barrier.dstAccessMask = access_flags; + + if (image_barrier_count > 0 || p_post_barrier != BARRIER_MASK_NO_BARRIER) { + vkCmdPipelineBarrier(compute_list->command_buffer, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, barrier_flags, 0, 1, &mem_barrier, 0, nullptr, image_barrier_count, image_barriers); + } + #endif + + memdelete(compute_list); + compute_list = nullptr; } void RenderingDeviceVulkan::barrier(uint32_t p_from, uint32_t p_to) { @@ -7107,11 +7326,15 @@ void RenderingDeviceVulkan::barrier(uint32_t p_from, uint32_t p_to) { src_access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT; } + if (p_from == 0) { + src_barrier_flags = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT; + } + uint32_t dst_barrier_flags = 0; uint32_t dst_access_flags = 0; if (p_to & BARRIER_MASK_COMPUTE) { dst_barrier_flags |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; - dst_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_INDIRECT_COMMAND_READ_BIT; + dst_access_flags |= VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT; } if (p_to & BARRIER_MASK_RASTER) { dst_barrier_flags |= VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT; @@ -7122,6 +7345,10 @@ void RenderingDeviceVulkan::barrier(uint32_t p_from, uint32_t p_to) { dst_access_flags |= VK_ACCESS_TRANSFER_WRITE_BIT | VK_ACCESS_TRANSFER_READ_BIT; } + if (p_to == 0) { + dst_barrier_flags = VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT; + } + _memory_barrier(src_barrier_flags, dst_barrier_flags, src_access_flags, dst_access_flags, true); } @@ -7325,6 +7552,16 @@ void RenderingDeviceVulkan::draw_command_end_label() { context->command_end_label(frames[frame].draw_command_buffer); } +String RenderingDeviceVulkan::get_device_vendor_name() const { + return context->get_device_vendor_name(); +} +String RenderingDeviceVulkan::get_device_name() const { + return context->get_device_name(); +} +String RenderingDeviceVulkan::get_device_pipeline_cache_uuid() const { + return context->get_device_pipeline_cache_uuid(); +} + void RenderingDeviceVulkan::_finalize_command_bufers() { if (draw_list) { ERR_PRINT("Found open draw list at the end of the frame, this should never happen (further drawing will likely not work)."); @@ -7377,6 +7614,7 @@ void RenderingDeviceVulkan::_begin_frame() { if (frames[frame].timestamp_count) { vkGetQueryPoolResults(device, frames[frame].timestamp_pool, 0, frames[frame].timestamp_count, sizeof(uint64_t) * max_timestamp_query_elements, frames[frame].timestamp_result_values, sizeof(uint64_t), VK_QUERY_RESULT_64_BIT); + vkCmdResetQueryPool(frames[frame].setup_command_buffer, frames[frame].timestamp_pool, 0, frames[frame].timestamp_count); SWAP(frames[frame].timestamp_names, frames[frame].timestamp_result_names); SWAP(frames[frame].timestamp_cpu_values, frames[frame].timestamp_cpu_result_values); } diff --git a/drivers/vulkan/rendering_device_vulkan.h b/drivers/vulkan/rendering_device_vulkan.h index 4bea17e4a1..a2527d5c33 100644 --- a/drivers/vulkan/rendering_device_vulkan.h +++ b/drivers/vulkan/rendering_device_vulkan.h @@ -141,6 +141,11 @@ class RenderingDeviceVulkan : public RenderingDevice { VkImageLayout layout; + uint64_t used_in_frame = 0; + bool used_in_transfer = false; + bool used_in_raster = false; + bool used_in_compute = false; + uint32_t read_aspect_mask = 0; uint32_t barrier_aspect_mask = 0; bool bound = false; //bound to framebffer @@ -528,6 +533,8 @@ class RenderingDeviceVulkan : public RenderingDevice { PushConstant push_constant; + uint32_t compute_local_size[3] = { 0, 0, 0 }; + bool is_compute = false; int max_output = 0; Vector<Set> sets; @@ -686,6 +693,7 @@ class RenderingDeviceVulkan : public RenderingDevice { VkPipeline pipeline = VK_NULL_HANDLE; uint32_t push_constant_size = 0; uint32_t push_constant_stages = 0; + uint32_t local_group_size[3] = { 0, 0, 0 }; }; RID_Owner<ComputePipeline, true> compute_pipeline_owner; @@ -808,8 +816,10 @@ class RenderingDeviceVulkan : public RenderingDevice { uint32_t set_count = 0; RID pipeline; RID pipeline_shader; + uint32_t local_group_size[3] = { 0, 0, 0 }; VkPipelineLayout pipeline_layout = VK_NULL_HANDLE; uint32_t pipeline_push_constant_stages = 0; + bool allow_draw_overlap; } state; #ifdef DEBUG_ENABLED @@ -1028,13 +1038,14 @@ public: /**** COMPUTE LISTS ****/ /***********************/ - virtual ComputeListID compute_list_begin(); + virtual ComputeListID compute_list_begin(bool p_allow_draw_overlap = false); virtual void compute_list_bind_compute_pipeline(ComputeListID p_list, RID p_compute_pipeline); virtual void compute_list_bind_uniform_set(ComputeListID p_list, RID p_uniform_set, uint32_t p_index); virtual void compute_list_set_push_constant(ComputeListID p_list, const void *p_data, uint32_t p_data_size); virtual void compute_list_add_barrier(ComputeListID p_list); virtual void compute_list_dispatch(ComputeListID p_list, uint32_t p_x_groups, uint32_t p_y_groups, uint32_t p_z_groups); + virtual void compute_list_dispatch_threads(ComputeListID p_list, uint32_t p_x_threads, uint32_t p_y_threads, uint32_t p_z_threads); virtual void compute_list_dispatch_indirect(ComputeListID p_list, RID p_buffer, uint32_t p_offset); virtual void compute_list_end(uint32_t p_post_barrier = BARRIER_MASK_ALL); @@ -1085,6 +1096,10 @@ public: virtual void draw_command_insert_label(String p_label_name, const Color p_color = Color(1, 1, 1, 1)); virtual void draw_command_end_label(); + virtual String get_device_vendor_name() const; + virtual String get_device_name() const; + virtual String get_device_pipeline_cache_uuid() const; + RenderingDeviceVulkan(); ~RenderingDeviceVulkan(); }; diff --git a/drivers/vulkan/vulkan_context.cpp b/drivers/vulkan/vulkan_context.cpp index 1a631f4a2c..c564cee757 100644 --- a/drivers/vulkan/vulkan_context.cpp +++ b/drivers/vulkan/vulkan_context.cpp @@ -331,7 +331,7 @@ Error VulkanContext::_create_physical_device() { * function to register the final callback. */ VkDebugUtilsMessengerCreateInfoEXT dbg_messenger_create_info; - if (use_validation_layers) { + if (enabled_debug_utils) { // VK_EXT_debug_utils style dbg_messenger_create_info.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_MESSENGER_CREATE_INFO_EXT; dbg_messenger_create_info.pNext = nullptr; @@ -380,7 +380,8 @@ Error VulkanContext::_create_physical_device() { ERR_FAIL_V(ERR_CANT_CREATE); } /* for now, just grab the first physical device */ - gpu = physical_devices[0]; + uint32_t device_index = 0; + gpu = physical_devices[device_index]; free(physical_devices); /* Look for device extensions */ @@ -389,6 +390,40 @@ Error VulkanContext::_create_physical_device() { enabled_extension_count = 0; memset(extension_names, 0, sizeof(extension_names)); + /* Get identifier properties */ + vkGetPhysicalDeviceProperties(gpu, &gpu_props); + + static const struct { + uint32_t id; + const char *name; + } vendor_names[] = { + { 0x1002, "AMD" }, + { 0x1010, "ImgTec" }, + { 0x10DE, "NVIDIA" }, + { 0x13B5, "ARM" }, + { 0x5143, "Qualcomm" }, + { 0x8086, "INTEL" }, + { 0, nullptr }, + }; + device_name = gpu_props.deviceName; + pipeline_cache_id = String::hex_encode_buffer(gpu_props.pipelineCacheUUID, VK_UUID_SIZE); + pipeline_cache_id += "-driver-" + itos(gpu_props.driverVersion); + { + device_vendor = "Unknown"; + uint32_t vendor_idx = 0; + while (vendor_names[vendor_idx].name != nullptr) { + if (gpu_props.vendorID == vendor_names[vendor_idx].id) { + device_vendor = vendor_names[vendor_idx].name; + break; + } + vendor_idx++; + } + } +#ifdef DEBUG_ENABLED + print_line("Using Vulkan Device #" + itos(device_index) + ": " + device_vendor + " - " + device_name); +#endif + device_api_version = gpu_props.apiVersion; + err = vkEnumerateDeviceExtensionProperties(gpu, nullptr, &device_extension_count, nullptr); ERR_FAIL_COND_V(err, ERR_CANT_CREATE); @@ -498,7 +533,6 @@ Error VulkanContext::_create_physical_device() { break; } } - vkGetPhysicalDeviceProperties(gpu, &gpu_props); /* Call with NULL data to get count */ vkGetPhysicalDeviceQueueFamilyProperties(gpu, &queue_family_count, nullptr); @@ -565,6 +599,7 @@ Error VulkanContext::_create_device() { } err = vkCreateDevice(gpu, &sdevice, nullptr, &device); ERR_FAIL_COND_V(err, ERR_CANT_CREATE); + return OK; } @@ -1590,10 +1625,12 @@ void VulkanContext::command_begin_label(VkCommandBuffer p_command_buffer, String if (!enabled_debug_utils) { return; } + + CharString cs = p_label_name.utf8().get_data(); VkDebugUtilsLabelEXT label; label.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_LABEL_EXT; label.pNext = nullptr; - label.pLabelName = p_label_name.utf8().get_data(); + label.pLabelName = cs.get_data(); label.color[0] = p_color[0]; label.color[1] = p_color[1]; label.color[2] = p_color[2]; @@ -1605,10 +1642,11 @@ void VulkanContext::command_insert_label(VkCommandBuffer p_command_buffer, Strin if (!enabled_debug_utils) { return; } + CharString cs = p_label_name.utf8().get_data(); VkDebugUtilsLabelEXT label; label.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_LABEL_EXT; label.pNext = nullptr; - label.pLabelName = p_label_name.utf8().get_data(); + label.pLabelName = cs.get_data(); label.color[0] = p_color[0]; label.color[1] = p_color[1]; label.color[2] = p_color[2]; @@ -1627,16 +1665,26 @@ void VulkanContext::set_object_name(VkObjectType p_object_type, uint64_t p_objec if (!enabled_debug_utils) { return; } + CharString obj_data = p_object_name.utf8(); VkDebugUtilsObjectNameInfoEXT name_info; name_info.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_OBJECT_NAME_INFO_EXT; name_info.pNext = nullptr; name_info.objectType = p_object_type; name_info.objectHandle = p_object_handle; - CharString object_name = p_object_name.utf8(); - name_info.pObjectName = object_name.get_data(); + name_info.pObjectName = obj_data.get_data(); SetDebugUtilsObjectNameEXT(device, &name_info); } +String VulkanContext::get_device_vendor_name() const { + return device_vendor; +} +String VulkanContext::get_device_name() const { + return device_name; +} +String VulkanContext::get_device_pipeline_cache_uuid() const { + return pipeline_cache_id; +} + VulkanContext::VulkanContext() { use_validation_layers = Engine::get_singleton()->is_validation_layers_enabled(); diff --git a/drivers/vulkan/vulkan_context.h b/drivers/vulkan/vulkan_context.h index 5cb762aca8..dc6b0410bc 100644 --- a/drivers/vulkan/vulkan_context.h +++ b/drivers/vulkan/vulkan_context.h @@ -57,6 +57,11 @@ class VulkanContext { bool device_initialized = false; bool inst_initialized = false; + String device_vendor; + String device_name; + String pipeline_cache_id; + uint32_t device_api_version = 0; + bool buffers_prepared = false; // Present queue. @@ -215,6 +220,10 @@ public: void command_end_label(VkCommandBuffer p_command_buffer); void set_object_name(VkObjectType p_object_type, uint64_t p_object_handle, String p_object_name); + String get_device_vendor_name() const; + String get_device_name() const; + String get_device_pipeline_cache_uuid() const; + VulkanContext(); virtual ~VulkanContext(); }; |