Add fxaa compute shader, improves visual quality of renders

crates/render/build.rs

@@ -27,8 +27,9 @@ fn visit_files(dir: &Path, f: &mut dyn FnMut(&Path) -> io::Result<()>) -> io::Re
 
 fn main() -> io::Result<()> {
     let mut compiler = shaderc::Compiler::new().expect("failed to initialize glsl compiler");
-    let mut compile_options = shaderc::CompileOptions::new().expect("failed to initialize compiler options");
-
+    let mut compile_options =
+        shaderc::CompileOptions::new().expect("failed to initialize compiler options");
+
     if env::var("TARGET").unwrap() == "wasm32-unknown-unknown" {
         compile_options.add_macro_definition("TARGET_WASM", None);
     }

crates/render/shaders/fxaa.comp

@@ -21,13 +21,11 @@ void main(void) {
     const vec2 global_coord = vec2(gl_GlobalInvocationID.xy);
     const vec2 inverse_size = 1.0 / vec2(gl_NumWorkGroups.xy * gl_WorkGroupSize.xy);
 
-    // Grab the pixel we're sitting directly on
-    // this fills in the center of the `loaded_texels` matrix.
     const vec4 color_center = texture(sampler2D(color_input, linear_sampler), global_coord * inverse_size);
-    const vec3 color_nw = texture(sampler2D(color_input, linear_sampler), (global_coord + vec2(1.0, 1.0)) * inverse_size).rgb;
+    const vec3 color_nw = texture(sampler2D(color_input, linear_sampler), (global_coord + vec2(-1.0, -1.0)) * inverse_size).rgb;
     const vec3 color_ne = texture(sampler2D(color_input, linear_sampler), (global_coord + vec2(1.0, -1.0)) * inverse_size).rgb;
     const vec3 color_sw = texture(sampler2D(color_input, linear_sampler), (global_coord + vec2(-1.0, 1.0)) * inverse_size).rgb;
-    const vec3 color_se = texture(sampler2D(color_input, linear_sampler), (global_coord + vec2(-1.0, -1.0)) * inverse_size).rgb;
+    const vec3 color_se = texture(sampler2D(color_input, linear_sampler), (global_coord + vec2(1.0, 1.0)) * inverse_size).rgb;
 
     const vec3 luma = vec3(0.299, 0.587, 0.114);
     const float luma_nw = dot(color_nw, luma);
@@ -39,7 +37,7 @@ void main(void) {
     float luma_min = min(luma_center, min(min(luma_nw, luma_ne), min(luma_sw, luma_se)));
     float luma_max = max(luma_center, max(max(luma_nw, luma_ne), max(luma_sw, luma_se)));
 
-    vec2 dir = vec2(
+    mediump vec2 dir = vec2(
         -((luma_nw + luma_ne) - (luma_sw + luma_se)),
         (luma_nw + luma_sw) - (luma_ne + luma_se)
     );

crates/render/src/atoms.rs

@@ -101,12 +101,25 @@ impl Atoms {
         }
     }
 
-    pub fn copy_new(&self, gpu_resources: &GlobalGpuResources, fragment_id: FragmentId, encoder: &mut wgpu::CommandEncoder) -> Self {
-        let buffer = self.buffer.copy_new(gpu_resources, |len, src_buffer, dest_buffer| {
-            let header_size = mem::size_of::<AtomBufferHeader>() as u64;
-            encoder.copy_buffer_to_buffer(src_buffer, header_size, dest_buffer, header_size, self.buffer.len());
-            len
-        });
+    pub fn copy_new(
+        &self,
+        gpu_resources: &GlobalGpuResources,
+        fragment_id: FragmentId,
+        encoder: &mut wgpu::CommandEncoder,
+    ) -> Self {
+        let buffer = self
+            .buffer
+            .copy_new(gpu_resources, |len, src_buffer, dest_buffer| {
+                let header_size = mem::size_of::<AtomBufferHeader>() as u64;
+                encoder.copy_buffer_to_buffer(
+                    src_buffer,
+                    header_size,
+                    dest_buffer,
+                    header_size,
+                    self.buffer.len(),
+                );
+                len
+            });
 
         let bind_group = gpu_resources
             .device

crates/render/src/bind_groups.rs

@@ -6,6 +6,7 @@ pub struct BindGroupLayouts {
     pub global: wgpu::BindGroupLayout,
     pub atoms: wgpu::BindGroupLayout,
     pub blit: wgpu::BindGroupLayout,
+    pub fxaa: wgpu::BindGroupLayout,
 }
 
 impl BindGroupLayouts {
@@ -56,9 +57,7 @@ impl BindGroupLayouts {
                     wgpu::BindGroupLayoutEntry {
                         binding: 0,
                         visibility: wgpu::ShaderStage::FRAGMENT,
-                        ty: wgpu::BindingType::Sampler {
-                            comparison: false,
-                        },
+                        ty: wgpu::BindingType::Sampler { comparison: false },
                         count: None,
                     },
                     wgpu::BindGroupLayoutEntry {
@@ -71,8 +70,39 @@ impl BindGroupLayouts {
                         },
                         count: None,
                     },
-                ]
-            })
+                ],
+            }),
+            fxaa: device.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
+                label: None,
+                entries: &[
+                    wgpu::BindGroupLayoutEntry {
+                        binding: 0,
+                        visibility: wgpu::ShaderStage::COMPUTE,
+                        ty: wgpu::BindingType::Sampler { comparison: false },
+                        count: None,
+                    },
+                    wgpu::BindGroupLayoutEntry {
+                        binding: 1,
+                        visibility: wgpu::ShaderStage::COMPUTE,
+                        ty: wgpu::BindingType::SampledTexture {
+                            dimension: wgpu::TextureViewDimension::D2,
+                            component_type: wgpu::TextureComponentType::Float,
+                            multisampled: false,
+                        },
+                        count: None,
+                    },
+                    wgpu::BindGroupLayoutEntry {
+                        binding: 2,
+                        visibility: wgpu::ShaderStage::COMPUTE,
+                        ty: wgpu::BindingType::StorageTexture {
+                            dimension: wgpu::TextureViewDimension::D2,
+                            format: crate::STORAGE_TEXTURE_FORMAT,
+                            readonly: false,
+                        },
+                        count: None,
+                    },
+                ],
+            }),
         }
     }
 }

crates/render/src/buffer_vec.rs

@@ -34,7 +34,6 @@ impl BufferVec {
     where
         F: FnOnce(&mut [u8]),
     {
-
         #[cfg(not(target_arch = "wasm32"))]
         let buffer = {
             let buffer = device.create_buffer(&wgpu::BufferDescriptor {
@@ -53,7 +52,7 @@ impl BufferVec {
         };
         #[cfg(target_arch = "wasm32")]
         let buffer = {
-            use wgpu::util::{DeviceExt as _, BufferInitDescriptor};
+            use wgpu::util::{BufferInitDescriptor, DeviceExt as _};
             let mut vec = vec![0; size as usize];
             fill(&mut vec);
             device.create_buffer_init(&BufferInitDescriptor {
@@ -136,7 +135,9 @@ impl BufferVec {
                         usage: self.usage,
                         mapped_at_creation: false,
                     });
-                    gpu_resources.queue.write_buffer(&new_buffer, inner.len, data);
+                    gpu_resources
+                        .queue
+                        .write_buffer(&new_buffer, inner.len, data);
                     new_buffer
                 };
 
@@ -217,7 +218,7 @@ impl BufferVec {
 
     pub fn copy_new<F>(&self, gpu_resources: &GlobalGpuResources, f: F) -> Self
     where
-        F: FnOnce(u64, &wgpu::Buffer /* from */, &wgpu::Buffer /* to */) -> u64
+        F: FnOnce(u64, &wgpu::Buffer /* from */, &wgpu::Buffer /* to */) -> u64,
     {
         if let Some(inner) = self.inner.as_ref() {
             let copied_buffer = gpu_resources.device.create_buffer(&wgpu::BufferDescriptor {