kas-wgpu 0.17.1

// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License in the LICENSE-APACHE file or at:
//     https://www.apache.org/licenses/LICENSE-2.0

//! Drawing API for `kas_wgpu`

use futures_lite::future::block_on;
use std::f32::consts::FRAC_PI_2;
use wgpu::util::DeviceExt;

use super::*;
use crate::DrawShadedImpl;
use crate::Options;
use kas::cast::traits::*;
use kas::config::RasterConfig;
use kas::draw::color::Rgba;
use kas::draw::*;
use kas::geom::{Quad, Size, Vec2};
use kas::runner::{GraphicsFeatures, RunError};
use kas::text::{Effect, TextDisplay};

impl<C: CustomPipe> DrawPipe<C> {
    /// Construct
    pub fn new<CB: CustomPipeBuilder<Pipe = C>>(
        instance: &wgpu::Instance,
        custom: &mut CB,
        options: &Options,
        surface: Option<&wgpu::Surface>,
        features: GraphicsFeatures,
    ) -> Result<Self, RunError> {
        let mut adapter_options = options.adapter_options();
        adapter_options.compatible_surface = surface;
        let req = instance.request_adapter(&adapter_options);
        let adapter = match block_on(req) {
            Ok(a) => a,
            Err(e) => return Err(RunError::Graphics(Box::new(e))),
        };
        log::info!("Using graphics adapter: {}", adapter.get_info().name);

        // Use adapter texture size limits to support the largest window surface possible
        let mut desc = CB::device_descriptor(&adapter);
        if features.subpixel_rendering
            && adapter
                .features()
                .contains(wgpu::Features::DUAL_SOURCE_BLENDING)
        {
            desc.required_features |= wgpu::Features::DUAL_SOURCE_BLENDING;
        }
        desc.required_limits = desc.required_limits.using_resolution(adapter.limits());

        let req = adapter.request_device(&desc);
        let (device, queue) = block_on(req).map_err(|e| RunError::Graphics(Box::new(e)))?;

        let shaders = ShaderManager::new(&device);

        // Create staging belt and a local pool
        let staging_belt = wgpu::util::StagingBelt::new(device.clone(), 1024);

        let bgl_common = device.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
            label: Some("common bind group layout"),
            entries: &[
                wgpu::BindGroupLayoutEntry {
                    binding: 0,
                    visibility: wgpu::ShaderStages::VERTEX,
                    ty: wgpu::BindingType::Buffer {
                        ty: wgpu::BufferBindingType::Uniform,
                        has_dynamic_offset: false,
                        min_binding_size: wgpu::BufferSize::new(16),
                    },
                    count: None,
                },
                wgpu::BindGroupLayoutEntry {
                    binding: 1,
                    visibility: wgpu::ShaderStages::FRAGMENT,
                    ty: wgpu::BindingType::Buffer {
                        ty: wgpu::BufferBindingType::Uniform,
                        has_dynamic_offset: false,
                        min_binding_size: wgpu::BufferSize::new(16),
                    },
                    count: None,
                },
            ],
        });

        // Light dir: `(a, b)` where `0 ≤ a < pi/2` is the angle to the screen
        // normal (i.e. `a = 0` is straight at the screen) and `b` is the bearing
        // (from UP, clockwise), both in radians.
        let dir: (f32, f32) = (0.3, 0.4);
        assert!(0.0 <= dir.0 && dir.0 < FRAC_PI_2);
        let a = (dir.0.sin(), dir.0.cos());
        // We normalise intensity:
        let f = a.0 / a.1;
        let light_norm = [dir.1.sin() * f, -dir.1.cos() * f, 1.0, 0.0];

        let light_norm_buf = device.create_buffer_init(&wgpu::util::BufferInitDescriptor {
            label: Some("light_norm_buf"),
            contents: bytemuck::cast_slice(&light_norm),
            usage: wgpu::BufferUsages::UNIFORM | wgpu::BufferUsages::COPY_DST,
        });

        let images = images::Images::new(&device, &shaders, &bgl_common);
        let shaded_square = shaded_square::Pipeline::new(&device, &shaders, &bgl_common);
        let shaded_round = shaded_round::Pipeline::new(&device, &shaders, &bgl_common);
        let flat_round = flat_round::Pipeline::new(&device, &shaders, &bgl_common);
        let round_2col = round_2col::Pipeline::new(&device, &shaders, &bgl_common);
        let custom = custom.build(&device, &bgl_common, RENDER_TEX_FORMAT);

        Ok(DrawPipe {
            adapter,
            device,
            queue,
            staging_belt,
            bgl_common,
            light_norm_buf,
            bg_common: vec![],
            images,
            text: kas::text::raster::State::default(),
            shaded_square,
            shaded_round,
            flat_round,
            round_2col,
            custom,
        })
    }

    /// Process window resize
    pub fn resize(&self, window: &mut DrawWindow<C::Window>, size: Size) {
        window.clip_regions[0].rect.size = size;

        let vsize = Vec2::conv(size);
        let off = vsize * -0.5;
        let scale = Vec2::splat(2.0) / vsize;
        window.scale = [off.0, off.1, scale.0, -scale.1];

        self.custom
            .resize(&mut window.custom, &self.device, &self.queue, size);

        self.queue.submit(std::iter::empty());
    }

    /// Render batched draw instructions via `rpass`
    pub fn render(
        &mut self,
        window: &mut DrawWindow<C::Window>,
        frame_view: &wgpu::TextureView,
        clear_color: wgpu::Color,
    ) {
        // Update all bind groups. We use a separate bind group for each clip
        // region and update on each render, although they don't always change.
        // NOTE: we could use push constants instead.
        let mut scale = window.scale;
        let base_offset = (scale[0], scale[1]);
        for (region, bg) in window.clip_regions.iter().zip(self.bg_common.iter()) {
            let offset = Vec2::conv(region.offset);
            scale[0] = base_offset.0 - offset.0;
            scale[1] = base_offset.1 - offset.1;
            self.queue
                .write_buffer(&bg.0, 0, bytemuck::cast_slice(&scale));
        }
        let device = &self.device;
        let bg_len = self.bg_common.len();
        if window.clip_regions.len() > bg_len {
            let (bgl_common, light_norm_buf) = (&self.bgl_common, &self.light_norm_buf);
            self.bg_common
                .extend(window.clip_regions[bg_len..].iter().map(|region| {
                    let offset = Vec2::conv(region.offset);
                    scale[0] = base_offset.0 - offset.0;
                    scale[1] = base_offset.1 - offset.1;
                    let scale_buf = device.create_buffer_init(&wgpu::util::BufferInitDescriptor {
                        label: Some("scale_buf"),
                        contents: bytemuck::cast_slice(&scale),
                        usage: wgpu::BufferUsages::UNIFORM | wgpu::BufferUsages::COPY_DST,
                    });
                    let bg_common = device.create_bind_group(&wgpu::BindGroupDescriptor {
                        label: Some("common bind group"),
                        layout: bgl_common,
                        entries: &[
                            wgpu::BindGroupEntry {
                                binding: 0,
                                resource: wgpu::BindingResource::Buffer(wgpu::BufferBinding {
                                    buffer: &scale_buf,
                                    offset: 0,
                                    size: None,
                                }),
                            },
                            wgpu::BindGroupEntry {
                                binding: 1,
                                resource: wgpu::BindingResource::Buffer(wgpu::BufferBinding {
                                    buffer: light_norm_buf,
                                    offset: 0,
                                    size: None,
                                }),
                            },
                        ],
                    });
                    (scale_buf, bg_common)
                }));
        }
        self.queue.submit(std::iter::empty());

        let mut encoder = self
            .device
            .create_command_encoder(&wgpu::CommandEncoderDescriptor {
                label: Some("render"),
            });

        self.images.prepare(
            &mut window.images,
            &self.device,
            &self.queue,
            &mut self.staging_belt,
            &mut encoder,
            &mut self.text,
        );
        window
            .shaded_square
            .write_buffers(&self.device, &mut self.staging_belt, &mut encoder);
        window
            .shaded_round
            .write_buffers(&self.device, &mut self.staging_belt, &mut encoder);
        window
            .flat_round
            .write_buffers(&self.device, &mut self.staging_belt, &mut encoder);
        window
            .round_2col
            .write_buffers(&self.device, &mut self.staging_belt, &mut encoder);
        self.custom.prepare(
            &mut window.custom,
            &self.device,
            &mut self.staging_belt,
            &mut encoder,
        );

        let mut color_attachments = [Some(wgpu::RenderPassColorAttachment {
            view: frame_view,
            depth_slice: None,
            resolve_target: None,
            ops: wgpu::Operations {
                load: wgpu::LoadOp::Clear(clear_color),
                store: wgpu::StoreOp::Store,
            },
        })];

        // Order passes to ensure overlays are drawn after other content
        let mut passes: Vec<_> = window
            .clip_regions
            .iter()
            .map(|pass| pass.order)
            .enumerate()
            .collect();
        // Note that sorting is stable (does not re-order equal elements):
        passes.sort_by_key(|pass| pass.1);

        // We use a separate render pass for each clipped region.
        for (pass, _) in passes.drain(..) {
            let rect = window.clip_regions[pass].rect;
            if rect.size.0 == 0 || rect.size.1 == 0 {
                continue;
            }
            let bg_common = &self.bg_common[pass].1;

            {
                let mut rpass = encoder.begin_render_pass(&wgpu::RenderPassDescriptor {
                    label: Some("kas-wgpu render pass"),
                    color_attachments: &color_attachments,
                    depth_stencil_attachment: None,
                    timestamp_writes: None,
                    occlusion_query_set: None,
                    multiview_mask: None,
                });
                rpass.set_scissor_rect(
                    rect.pos.0.cast(),
                    rect.pos.1.cast(),
                    rect.size.0.cast(),
                    rect.size.1.cast(),
                );

                self.round_2col
                    .render(&window.round_2col, pass, &mut rpass, bg_common);
                self.shaded_square
                    .render(&window.shaded_square, pass, &mut rpass, bg_common);
                self.images
                    .render(&window.images, pass, &mut rpass, bg_common);
                self.shaded_round
                    .render(&window.shaded_round, pass, &mut rpass, bg_common);
                self.flat_round
                    .render(&window.flat_round, pass, &mut rpass, bg_common);
                self.custom.render_pass(
                    &mut window.custom,
                    &self.device,
                    pass,
                    &mut rpass,
                    bg_common,
                );
            }

            color_attachments[0].as_mut().unwrap().ops.load = wgpu::LoadOp::Load;
        }

        let size = window.clip_regions[0].rect.size;

        self.custom.render_final(
            &mut window.custom,
            &self.device,
            &mut encoder,
            frame_view,
            size,
        );

        // Keep only first clip region (which is the entire window)
        window.clip_regions.truncate(1);

        self.staging_belt.finish();
        self.queue.submit(std::iter::once(encoder.finish()));

        self.staging_belt.recall();
    }
}

impl<C: CustomPipe> DrawSharedImpl for DrawPipe<C> {
    type Draw = DrawWindow<C::Window>;

    fn max_texture_dimension_2d(&self) -> u32 {
        self.device.limits().max_texture_dimension_2d
    }

    fn set_raster_config(&mut self, config: &RasterConfig) {
        self.text.set_raster_config(config);
    }

    #[inline]
    fn image_alloc(&mut self, format: ImageFormat, size: Size) -> Result<ImageId, AllocError> {
        self.images.alloc(format, size)
    }

    #[inline]
    fn image_upload(&mut self, id: ImageId, data: &[u8]) -> Result<(), UploadError> {
        self.images.upload(&self.device, &self.queue, id, data)
    }

    #[inline]
    fn image_free(&mut self, id: ImageId) {
        self.images.free(id);
    }

    #[inline]
    fn image_size(&self, id: ImageId) -> Option<Size> {
        self.images.image_size(id)
    }

    #[inline]
    fn draw_image(&self, draw: &mut Self::Draw, pass: PassId, id: ImageId, rect: Quad) {
        if let Some((atlas, tex)) = self.images.get_im_atlas_coords(id) {
            draw.images.rect(pass, atlas, tex, rect);
        };
    }

    #[inline]
    fn draw_text(
        &mut self,
        draw: &mut Self::Draw,
        pass: PassId,
        pos: Vec2,
        bb: Quad,
        text: &TextDisplay,
        col: Rgba,
    ) {
        let time = std::time::Instant::now();
        self.text
            .text(&mut self.images, &mut draw.images, pass, pos, bb, text, col);
        draw.common.report_dur_text(time.elapsed());
    }

    fn draw_text_effects(
        &mut self,
        draw: &mut Self::Draw,
        pass: PassId,
        pos: Vec2,
        bb: Quad,
        text: &TextDisplay,
        colors: &[Rgba],
        effects: &[Effect],
    ) {
        let time = std::time::Instant::now();
        self.text.text_effects(
            &mut self.images,
            &mut draw.images,
            pass,
            pos,
            bb,
            text,
            colors,
            effects,
            |quad, col| {
                draw.shaded_square.rect(pass, quad, col);
            },
        );
        draw.common.report_dur_text(time.elapsed());
    }
}

impl<CW: CustomWindow> DrawImpl for DrawWindow<CW> {
    fn common_mut(&mut self) -> &mut WindowCommon {
        &mut self.common
    }

    fn new_pass(
        &mut self,
        parent_pass: PassId,
        rect: Rect,
        offset: Offset,
        class: PassType,
    ) -> PassId {
        let parent = match class {
            PassType::Clip => &self.clip_regions[parent_pass.pass()],
            PassType::Overlay => {
                // NOTE: parent_pass.pass() is always zero in this case since
                // this is only invoked from the Window (root).
                &self.clip_regions[0]
            }
        };
        let order = match class {
            PassType::Clip => (parent.order << 4) + 1,
            PassType::Overlay => (parent.order << 16) + 1,
        };
        let rect = rect - parent.offset;
        let offset = offset + parent.offset;
        let rect = rect.intersection(&parent.rect).unwrap_or(Rect::ZERO);
        let pass = self.clip_regions.len().cast();
        self.clip_regions.push(ClipRegion {
            rect,
            offset,
            order,
        });
        PassId::new(pass)
    }

    #[inline]
    fn get_clip_rect(&self, pass: PassId) -> Rect {
        let region = &self.clip_regions[pass.pass()];
        region.rect + region.offset
    }

    #[inline]
    fn rect(&mut self, pass: PassId, rect: Quad, col: Rgba) {
        self.shaded_square.rect(pass, rect, col);
    }

    #[inline]
    fn frame(&mut self, pass: PassId, outer: Quad, inner: Quad, col: Rgba) {
        self.shaded_square.frame(pass, outer, inner, col);
    }

    #[inline]
    fn line(&mut self, pass: PassId, p1: Vec2, p2: Vec2, width: f32, col: Rgba) {
        self.flat_round.line(pass, p1, p2, 0.5 * width, col);
    }
}

impl<CW: CustomWindow> DrawRoundedImpl for DrawWindow<CW> {
    #[inline]
    fn rounded_line(&mut self, pass: PassId, p1: Vec2, p2: Vec2, radius: f32, col: Rgba) {
        self.flat_round.line(pass, p1, p2, radius, col);
    }

    #[inline]
    fn circle(&mut self, pass: PassId, rect: Quad, inner_radius: f32, col: Rgba) {
        self.flat_round.circle(pass, rect, inner_radius, col);
    }

    #[inline]
    fn circle_2col(&mut self, pass: PassId, rect: Quad, col1: Rgba, col2: Rgba) {
        self.round_2col.circle(pass, rect, col1, col2);
    }

    #[inline]
    fn rounded_frame(&mut self, pass: PassId, outer: Quad, inner: Quad, r1: f32, col: Rgba) {
        self.flat_round.rounded_frame(pass, outer, inner, r1, col);
    }

    #[inline]
    fn rounded_frame_2col(&mut self, pass: PassId, outer: Quad, inner: Quad, c1: Rgba, c2: Rgba) {
        self.round_2col.frame(pass, outer, inner, c1, c2);
    }
}

impl<CW: CustomWindow> DrawShadedImpl for DrawWindow<CW> {
    #[inline]
    fn shaded_square(&mut self, pass: PassId, rect: Quad, norm: (f32, f32), col: Rgba) {
        self.shaded_square
            .shaded_rect(pass, rect, Vec2::from(norm), col);
    }

    #[inline]
    fn shaded_circle(&mut self, pass: PassId, rect: Quad, norm: (f32, f32), col: Rgba) {
        self.shaded_round.circle(pass, rect, Vec2::from(norm), col);
    }

    #[inline]
    fn shaded_square_frame(
        &mut self,
        pass: PassId,
        outer: Quad,
        inner: Quad,
        norm: (f32, f32),
        outer_col: Rgba,
        inner_col: Rgba,
    ) {
        self.shaded_square
            .shaded_frame(pass, outer, inner, Vec2::from(norm), outer_col, inner_col);
    }

    #[inline]
    fn shaded_round_frame(
        &mut self,
        pass: PassId,
        outer: Quad,
        inner: Quad,
        norm: (f32, f32),
        col: Rgba,
    ) {
        self.shaded_round
            .shaded_frame(pass, outer, inner, Vec2::from(norm), col);
    }
}