fyrox_graphics/

server.rs

// Copyright (c) 2019-present Dmitry Stepanov and Fyrox Engine contributors.
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in all
// copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
// SOFTWARE.

#![warn(missing_docs)]

//! Graphics server is an abstraction layer over various graphics APIs used on different platforms
//! supported by the engine.

use crate::{
    buffer::{GpuBuffer, GpuBufferDescriptor},
    error::FrameworkError,
    framebuffer::{Attachment, GpuFrameBuffer},
    geometry_buffer::{GpuGeometryBuffer, GpuGeometryBufferDescriptor},
    gpu_program::{GpuProgram, GpuShader, ShaderKind, ShaderResourceDefinition},
    gpu_texture::{GpuTexture, GpuTextureDescriptor, GpuTextureKind, PixelKind},
    query::GpuQuery,
    read_buffer::GpuAsyncReadBuffer,
    sampler::{GpuSampler, GpuSamplerDescriptor},
    stats::PipelineStatistics,
    PolygonFace, PolygonFillMode,
};
use fyrox_core::define_as_any_trait;
use std::fmt::{Display, Formatter};
use std::rc::{Rc, Weak};

/// Graphics server capabilities.
#[derive(Debug)]
pub struct ServerCapabilities {
    /// The maximum size in basic machine units of a uniform block, which must be at least 16384.
    pub max_uniform_block_size: usize,
    /// The minimum required alignment for uniform buffer sizes and offset. The initial value is 1.
    pub uniform_buffer_offset_alignment: usize,
    /// The maximum, absolute value of the texture level-of-detail bias. The value must be at least
    /// 2.0.
    pub max_lod_bias: f32,
}

/// Contains information about used memory per each category of GPU resource. This is not precise
/// data; it only calculates total requested memory by the user of a graphics server and does not
/// include additional memory overhead in the video driver. Yet this information could still be
/// useful.
#[derive(Default, Debug, Clone)]
pub struct ServerMemoryUsage {
    /// Total number of bytes used by all textures (including render targets).
    pub textures: usize,
    /// Total number of bytes used by all buffers (vertex, index, uniform, etc.)
    pub buffers: usize,
}

impl Display for ServerMemoryUsage {
    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
        const MEGABYTE: f32 = 1024.0 * 1024.0;
        write!(
            f,
            "Textures: {:.3} Mb\nBuffers: {:.3} Mb",
            self.textures as f32 / MEGABYTE,
            self.buffers as f32 / MEGABYTE
        )
    }
}

/// A shared reference to a graphics server.
pub type SharedGraphicsServer = Rc<dyn GraphicsServer>;

define_as_any_trait!(GraphicsServerAsAny => GraphicsServer);

/// A named debug scope RAII object that automatically exits the scope on drop.
pub struct RenderingScope {
    server: Weak<dyn GraphicsServer>,
}

impl Drop for RenderingScope {
    fn drop(&mut self) {
        if let Some(server) = self.server.upgrade() {
            server.pop_debug_group();
        }
    }
}

/// Graphics server is an abstraction layer over various graphics APIs used on different platforms
/// supported by the engine. Such abstraction layer tries to provide more or less high-level and
/// unified interface, that can be used to build graphics pipelines quickly and more or less efficiently.
///
/// Low-level GAPI-specific optimizations could be performed using direct access to the underlying API,
/// by downcasting to a specific type.
pub trait GraphicsServer: GraphicsServerAsAny {
    /// Creates a GPU buffer with the given size and kind. Usage is a hint to the video driver
    /// that allows to perform some potential performance optimizations.
    fn create_buffer(&self, desc: GpuBufferDescriptor) -> Result<GpuBuffer, FrameworkError>;

    /// Creates a new GPU texture using the given descriptor.
    fn create_texture(&self, desc: GpuTextureDescriptor) -> Result<GpuTexture, FrameworkError>;

    /// Creates a new GPU sampler that can be used to sample texels from a texture.
    fn create_sampler(&self, desc: GpuSamplerDescriptor) -> Result<GpuSampler, FrameworkError>;

    /// Creates a new frame buffer using the given depth and color attachments. Depth attachment
    /// not exist, but there must be at least one color attachment of a format that supports rendering.
    fn create_frame_buffer(
        &self,
        depth_attachment: Option<Attachment>,
        color_attachments: Vec<Attachment>,
    ) -> Result<GpuFrameBuffer, FrameworkError>;

    /// Creates a frame buffer that "connected" to the final image that will be displayed to the
    /// screen.
    fn back_buffer(&self) -> GpuFrameBuffer;

    /// Creates a new GPU query, that can perform asynchronous data fetching from GPU. Usually it
    /// is used to create occlusion queries.
    fn create_query(&self) -> Result<GpuQuery, FrameworkError>;

    /// Creates a new named GPU shader. The name could be used for debugging purposes. The
    /// implementation of graphics server will generate proper resource bindings in the shader code
    /// for you.
    fn create_shader(
        &self,
        name: String,
        kind: ShaderKind,
        source: String,
        resources: &[ShaderResourceDefinition],
        line_offset: isize,
    ) -> Result<GpuShader, FrameworkError>;

    /// Creates a new named GPU program using source code of both vertex and fragment shaders. The
    /// name could be used for debugging purposes. The implementation of graphics server will generate
    /// proper resource bindings in the shader code for you.
    fn create_program(
        &self,
        name: &str,
        vertex_source: String,
        vertex_source_line_offset: isize,
        fragment_source: String,
        fragment_source_line_offset: isize,
        resources: &[ShaderResourceDefinition],
    ) -> Result<GpuProgram, FrameworkError>;

    /// Creates a new named GPU program using a pair of vertex and fragment shaders. The name could
    /// be used for debugging purposes. The implementation of graphics server will generate proper
    /// resource bindings in the shader code for you.
    fn create_program_from_shaders(
        &self,
        name: &str,
        vertex_shader: &GpuShader,
        fragment_shader: &GpuShader,
        resources: &[ShaderResourceDefinition],
    ) -> Result<GpuProgram, FrameworkError>;

    /// Creates a new read-back buffer, that can be used to obtain texture data from GPU. It can be
    /// used to read rendering result from GPU to CPU memory and save the result to disk.
    fn create_async_read_buffer(
        &self,
        name: &str,
        pixel_size: usize,
        pixel_count: usize,
    ) -> Result<GpuAsyncReadBuffer, FrameworkError>;

    /// Creates a new geometry buffer, which consists of one or more vertex buffers and only one
    /// element buffer. Geometry buffer could be considered as a complex mesh storage allocated on
    /// GPU.
    fn create_geometry_buffer(
        &self,
        desc: GpuGeometryBufferDescriptor,
    ) -> Result<GpuGeometryBuffer, FrameworkError>;

    /// Creates a weak reference to the shared graphics server.
    fn weak(&self) -> Weak<dyn GraphicsServer>;

    /// Sends all scheduled GPU command buffers for execution on GPU without waiting for a certain
    /// threshold.
    fn flush(&self);

    /// Waits until all the scheduled GPU commands are fully executed. This is blocking operation, and
    /// it blocks the current thread until all the commands are fully executed.
    fn finish(&self);

    /// Unbinds the all bound resources from the graphics pipeline.
    fn invalidate_resource_bindings_cache(&self);

    /// Returns GPU pipeline statistics. See [`PipelineStatistics`] for more info.
    fn pipeline_statistics(&self) -> PipelineStatistics;

    /// Swaps the front and back buffers and thus presenting the final image on screen. There could
    /// be more than two buffers, and it is up to the graphics server implementation to choose the
    /// right amount, but it can't be less than two.
    fn swap_buffers(&self) -> Result<(), FrameworkError>;

    /// Notifies the graphics server that the size of the back buffer has changed. It has very limited
    /// use and there are very few platforms (Linux with Wayland mostly) that needs this function to
    /// be called.
    fn set_frame_size(&self, new_size: (u32, u32));

    /// Returns current capabilities of the graphics server. See [`ServerCapabilities`] for more info.
    fn capabilities(&self) -> ServerCapabilities;

    /// Sets current polygon fill mode for front faces, back faces, or both.
    /// The mode of front faces is controlled separately from the mode of back faces,
    /// and `polygon_face` determines which mode is set by this method.
    /// See [`PolygonFace`] and [`PolygonFillMode`] docs for more info.
    fn set_polygon_fill_mode(&self, polygon_face: PolygonFace, polygon_fill_mode: PolygonFillMode);

    /// Generates mipmaps for the given texture. Graphics server implementation can pick any desired
    /// way of mipmaps generation, depending on the underlying GAPI capabilities.
    fn generate_mipmap(&self, texture: &GpuTexture);

    /// Fetches the total amount of memory used by the graphics server.
    fn memory_usage(&self) -> ServerMemoryUsage;

    /// Begins a new named debug group. It is recommended to use [`Self::begin_scope`] instead,
    /// so that the compiler will manage the scope lifetime for your correctly. Otherwise, a forgotten
    /// call to [`Self::pop_debug_group`] may cause stack overflow or underflow errors.
    fn push_debug_group(&self, name: &str);

    /// Ends the current debug group.
    fn pop_debug_group(&self);

    /// Begins a new debug scope by creating a temporary object that automatically exits the
    /// scope on drop. The scope could be created like so: `let _debug_scope = server.begin_scope("ScopeName");`
    /// Note the `let _debug_scope = ...` part - it is important to keep the produced object alive
    /// until the end of the current semantic scope. Do not call this method like so:
    /// `server.begin_scope("VisibilityTest");` because it will enter and leave the scope instantly.
    fn begin_scope(&self, name: &str) -> RenderingScope {
        self.push_debug_group(name);

        RenderingScope {
            server: self.weak(),
        }
    }

    /// A shortcut for [`Self::create_texture`], that creates a rectangular texture with the given
    /// size and pixel kind.
    fn create_2d_render_target(
        &self,
        name: &str,
        pixel_kind: PixelKind,
        width: usize,
        height: usize,
    ) -> Result<GpuTexture, FrameworkError> {
        self.create_texture(GpuTextureDescriptor {
            name,
            kind: GpuTextureKind::Rectangle { width, height },
            pixel_kind,
            ..Default::default()
        })
    }

    /// A shortcut for [`Self::create_texture`], that creates a cube texture with the given
    /// size and pixel kind.
    fn create_cube_render_target(
        &self,
        name: &str,
        pixel_kind: PixelKind,
        size: usize,
    ) -> Result<GpuTexture, FrameworkError> {
        self.create_texture(GpuTextureDescriptor {
            name,
            kind: GpuTextureKind::Cube { size },
            pixel_kind,
            ..Default::default()
        })
    }
}