f16 wip

Cargo.toml

@@ -50,7 +50,8 @@ regex = "1.10.6"
 bytemuck = { version = "1.17.0", features = [ "derive" ] }
 swash = "0.1.18"
 serde = { version = "1.0.208", features = ["derive"] }
-wgpu = "22.1.0"
+wgpu = { git = "https://github.com/raphamorim/wgpu-simd" }
+# 22.1.0"
 libc = "0.2.158"
 smol_str = "0.3.1"
 futures = "0.3.30"

sugarloaf/src/components/rich_text/rich_text.wgsl

@@ -1,3 +1,5 @@
+enable f16;
+
 struct Globals {
     transform: mat4x4<f32>,
 }
@@ -8,16 +10,16 @@ struct Globals {
 
 struct VertexInput {
     @builtin(vertex_index) vertex_index: u32,
-    @location(0) v_pos: vec4<f32>,
-    @location(1) v_color: vec4<f32>,
-    @location(2) v_uv: vec2<f32>,
+    @location(0) v_pos: vec4<f16>,
+    @location(1) v_color: vec4<f16>,
+    @location(2) v_uv: vec2<f16>,
     @location(3) layers: vec2<i32>,
 }
 
 struct VertexOutput {
-    @builtin(position) position: vec4<f32>,
-    @location(0) f_color: vec4<f32>,
-    @location(1) f_uv: vec2<f32>,
+    @builtin(position) position: vec4<f16>,
+    @location(0) f_color: vec4<f16>,
+    @location(1) f_uv: vec2<f16>,
     @location(2) color_layer: i32,
     @location(3) mask_layer: i32,
 }
@@ -36,14 +38,14 @@ fn vs_main(input: VertexInput) -> VertexOutput {
 
 @fragment
 fn fs_main(input: VertexOutput) -> @location(0) vec4<f32> {
-    var out: vec4<f32> = input.f_color;
+    var out: vec4<f16> = input.f_color;
 
     if input.color_layer > 0 {
         out = textureSampleLevel(font_texture, font_sampler, input.f_uv, 0.0);
     }
 
     if input.mask_layer > 0 {
-        out = vec4<f32>(out.xyz, textureSampleLevel(font_texture, font_sampler, input.f_uv, 0.0).x);
+        out = vec4<f16>(out.xyz, textureSampleLevel(font_texture, font_sampler, input.f_uv, 0.0).x);
     }
 
     return out;

sugarloaf/src/context/mod.rs

@@ -112,14 +112,18 @@ impl Context<'_> {
         let (device, queue) = {
             {
                 if let Ok(result) = futures::executor::block_on(
-                    adapter.request_device(&wgpu::DeviceDescriptor::default(), None),
+                    adapter.request_device(&wgpu::DeviceDescriptor {
+                        label: None,
+                        required_features: wgpu::Features::SHADER_F16,
+                        required_limits: wgpu::Limits::downlevel_defaults(),
+                    }, None),
                 ) {
                     result
                 } else {
                     // These downlevel limits will allow the code to run on all possible hardware
                     futures::executor::block_on(adapter.request_device(
                         &wgpu::DeviceDescriptor {
-                            memory_hints: wgpu::MemoryHints::Performance,
+                            // memory_hints: wgpu::MemoryHints::Performance,
                             label: None,
                             required_features: wgpu::Features::empty(),
                             required_limits: wgpu::Limits::downlevel_webgl2_defaults(),