webgpu: Use wgpu's instead of string errors and update limits handling (servo#32925)

* Use wgpu specific errors

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

* fixup expect

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

* WIP

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

* Fix records erasing enforcerange

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

* page can already be destroyed

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

* Support more limits

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

* Set good results

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

* Set OK (not PASS) expect CRASH

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

* fixup expectation

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

* bad expectations

gfx-rs/wgpu#6075
Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

* set bad expectation

render bundleencoder needs to be rewritten

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

---------

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

Author: sagudev

components/script/dom/bindings/codegen/CodegenRust.py

@@ -718,7 +718,7 @@ def wrapObjectTemplate(templateBody, nullValue, isDefinitelyObject, type,
                                                 isMember="Sequence",
                                                 isAutoRooted=isAutoRooted)
         declType = wrapInNativeContainerType(type, innerInfo.declType)
-        config = getConversionConfigForType(type, isEnforceRange, isClamp, treatNullAs)
+        config = getConversionConfigForType(type, innerContainerType(type).hasEnforceRange(), isClamp, treatNullAs)
 
         if type.nullable():
             declType = CGWrapper(declType, pre="Option<", post=" >")

components/script/dom/gpuadapter.rs

@@ -2,15 +2,17 @@
  * License, v. 2.0. If a copy of the MPL was not distributed with this
  * file, You can obtain one at https://mozilla.org/MPL/2.0/. */
 
-use std::convert::TryFrom;
 use std::rc::Rc;
 
 use dom_struct::dom_struct;
 use js::jsapi::{Heap, JSObject};
+use webgpu::wgc::instance::RequestDeviceError;
 use webgpu::wgt::MemoryHints;
 use webgpu::{wgt, WebGPU, WebGPUAdapter, WebGPURequest, WebGPUResponse};
 
+use super::bindings::codegen::Bindings::WebGPUBinding::GPUDeviceLostReason;
 use super::gpusupportedfeatures::GPUSupportedFeatures;
+use super::gpusupportedlimits::set_limit;
 use super::types::{GPUAdapterInfo, GPUSupportedLimits};
 use crate::dom::bindings::codegen::Bindings::WebGPUBinding::{
     GPUAdapterMethods, GPUDeviceDescriptor,
@@ -129,84 +131,10 @@ impl GPUAdapterMethods for GPUAdapter {
         };
         if let Some(limits) = &descriptor.requiredLimits {
             for (limit, value) in (*limits).iter() {
-                let v = u32::try_from(*value).unwrap_or(u32::MAX);
-                match limit.as_ref() {
-                    "maxTextureDimension1D" => desc.required_limits.max_texture_dimension_1d = v,
-                    "maxTextureDimension2D" => desc.required_limits.max_texture_dimension_2d = v,
-                    "maxTextureDimension3D" => desc.required_limits.max_texture_dimension_3d = v,
-                    "maxTextureArrayLayers" => desc.required_limits.max_texture_array_layers = v,
-                    "maxBindGroups" => desc.required_limits.max_bind_groups = v,
-                    "maxBindingsPerBindGroup" => {
-                        desc.required_limits.max_bindings_per_bind_group = v
-                    },
-                    "maxDynamicUniformBuffersPerPipelineLayout" => {
-                        desc.required_limits
-                            .max_dynamic_uniform_buffers_per_pipeline_layout = v
-                    },
-                    "maxDynamicStorageBuffersPerPipelineLayout" => {
-                        desc.required_limits
-                            .max_dynamic_storage_buffers_per_pipeline_layout = v
-                    },
-                    "maxSampledTexturesPerShaderStage" => {
-                        desc.required_limits.max_sampled_textures_per_shader_stage = v
-                    },
-                    "maxSamplersPerShaderStage" => {
-                        desc.required_limits.max_samplers_per_shader_stage = v
-                    },
-                    "maxStorageBuffersPerShaderStage" => {
-                        desc.required_limits.max_storage_buffers_per_shader_stage = v
-                    },
-                    "maxStorageTexturesPerShaderStage" => {
-                        desc.required_limits.max_storage_textures_per_shader_stage = v
-                    },
-                    "maxUniformBuffersPerShaderStage" => {
-                        desc.required_limits.max_uniform_buffers_per_shader_stage = v
-                    },
-                    "maxUniformBufferBindingSize" => {
-                        desc.required_limits.max_uniform_buffer_binding_size = v
-                    },
-                    "maxStorageBufferBindingSize" => {
-                        desc.required_limits.max_storage_buffer_binding_size = v
-                    },
-                    "minUniformBufferOffsetAlignment" => {
-                        desc.required_limits.min_uniform_buffer_offset_alignment = v
-                    },
-                    "minStorageBufferOffsetAlignment" => {
-                        desc.required_limits.min_storage_buffer_offset_alignment = v
-                    },
-                    "maxVertexBuffers" => desc.required_limits.max_vertex_buffers = v,
-                    "maxBufferSize" => desc.required_limits.max_buffer_size = *value,
-                    "maxVertexAttributes" => desc.required_limits.max_vertex_attributes = v,
-                    "maxVertexBufferArrayStride" => {
-                        desc.required_limits.max_vertex_buffer_array_stride = v
-                    },
-                    "maxInterStageShaderComponents" => {
-                        desc.required_limits.max_inter_stage_shader_components = v
-                    },
-                    "maxComputeWorkgroupStorageSize" => {
-                        desc.required_limits.max_compute_workgroup_storage_size = v
-                    },
-                    "maxComputeInvocationsPerWorkgroup" => {
-                        desc.required_limits.max_compute_invocations_per_workgroup = v
-                    },
-                    "maxComputeWorkgroupSizeX" => {
-                        desc.required_limits.max_compute_workgroup_size_x = v
-                    },
-                    "maxComputeWorkgroupSizeY" => {
-                        desc.required_limits.max_compute_workgroup_size_y = v
-                    },
-                    "maxComputeWorkgroupSizeZ" => {
-                        desc.required_limits.max_compute_workgroup_size_z = v
-                    },
-                    "maxComputeWorkgroupsPerDimension" => {
-                        desc.required_limits.max_compute_workgroups_per_dimension = v
-                    },
-                    _ => {
-                        error!("Unknown required limit: {limit} with value {value}");
-                        // we should reject but spec is still evolving
-                        // promise.reject_error(Error::Operation);
-                        // return promise;
-                    },
+                if !set_limit(&mut desc.required_limits, limit.as_ref(), *value) {
+                    warn!("Unknown GPUDevice limit: {limit}");
+                    promise.reject_error(Error::Operation);
+                    return promise;
                 }
             }
         }
@@ -267,25 +195,45 @@ impl GPUAdapterMethods for GPUAdapter {
 impl AsyncWGPUListener for GPUAdapter {
     fn handle_response(&self, response: WebGPUResponse, promise: &Rc<Promise>) {
         match response {
-            WebGPUResponse::Device(Ok(device)) => {
-                let descriptor = device.descriptor;
+            WebGPUResponse::Device((device_id, queue_id, Ok(descriptor))) => {
                 let device = GPUDevice::new(
                     &self.global(),
                     self.channel.clone(),
                     self,
                     Heap::default(),
                     descriptor.required_features,
                     descriptor.required_limits,
-                    device.device_id,
-                    device.queue_id,
+                    device_id,
+                    queue_id,
                     descriptor.label.unwrap_or_default(),
                 );
                 self.global().add_gpu_device(&device);
                 promise.resolve_native(&device);
             },
-            WebGPUResponse::Device(Err(e)) => {
-                warn!("Could not get GPUDevice({:?})", e);
-                promise.reject_error(Error::Operation);
+            WebGPUResponse::Device((_, _, Err(RequestDeviceError::UnsupportedFeature(f)))) => {
+                promise.reject_error(Error::Type(
+                    RequestDeviceError::UnsupportedFeature(f).to_string(),
+                ))
+            },
+            WebGPUResponse::Device((
+                _,
+                _,
+                Err(RequestDeviceError::LimitsExceeded(_) | RequestDeviceError::InvalidAdapter),
+            )) => promise.reject_error(Error::Operation),
+            WebGPUResponse::Device((device_id, queue_id, Err(e))) => {
+                let device = GPUDevice::new(
+                    &self.global(),
+                    self.channel.clone(),
+                    self,
+                    Heap::default(),
+                    wgt::Features::default(),
+                    wgt::Limits::default(),
+                    device_id,
+                    queue_id,
+                    String::new(),
+                );
+                device.lose(GPUDeviceLostReason::Unknown, e.to_string());
+                promise.resolve_native(&device);
             },
             WebGPUResponse::None => unreachable!("Failed to get a response for RequestDevice"),
             _ => unreachable!("GPUAdapter received wrong WebGPUResponse"),

components/script/dom/gpusupportedlimits.rs

@@ -3,6 +3,7 @@
  * file, You can obtain one at https://mozilla.org/MPL/2.0/. */
 
 use dom_struct::dom_struct;
+use num_traits::bounds::UpperBounded;
 use webgpu::wgt::Limits;
 use GPUSupportedLimits_Binding::GPUSupportedLimitsMethods;
 
@@ -173,4 +174,144 @@ impl GPUSupportedLimitsMethods for GPUSupportedLimits {
     fn MaxComputeWorkgroupsPerDimension(&self) -> u32 {
         self.limits.max_compute_workgroups_per_dimension
     }
+
+    /// <https://gpuweb.github.io/gpuweb/#dom-gpusupportedlimits-maxbindgroupsplusvertexbuffers>
+    fn MaxBindGroupsPlusVertexBuffers(&self) -> u32 {
+        // Not on wgpu yet, so we craft it manually
+        self.limits.max_bind_groups + self.limits.max_vertex_buffers
+    }
+
+    /// <https://gpuweb.github.io/gpuweb/#dom-gpusupportedlimits-maxinterstageshadervariables>
+    fn MaxInterStageShaderVariables(&self) -> u32 {
+        // Not in wgpu yet, so we use default value from spec
+        16
+    }
+
+    /// <https://gpuweb.github.io/gpuweb/#dom-gpusupportedlimits-maxcolorattachments>
+    fn MaxColorAttachments(&self) -> u32 {
+        self.limits.max_color_attachments
+    }
+
+    /// <https://gpuweb.github.io/gpuweb/#dom-gpusupportedlimits-maxcolorattachmentbytespersample>
+    fn MaxColorAttachmentBytesPerSample(&self) -> u32 {
+        self.limits.max_color_attachment_bytes_per_sample
+    }
+}
+
+/// Returns false if unknown limit or other value error
+pub fn set_limit(limits: &mut Limits, limit: &str, value: u64) -> bool {
+    /// per spec defaults are lower bounds for values
+    ///
+    /// https://www.w3.org/TR/webgpu/#limit-class-maximum
+    fn set_maximum<T>(limit: &mut T, value: u64) -> bool
+    where
+        T: Ord + Copy + TryFrom<u64> + UpperBounded,
+    {
+        if let Ok(value) = T::try_from(value) {
+            *limit = value.max(*limit);
+            true
+        } else {
+            false
+        }
+    }
+
+    /// per spec defaults are higher bounds for values
+    ///
+    /// <https://www.w3.org/TR/webgpu/#limit-class-alignment>
+    fn set_alignment<T>(limit: &mut T, value: u64) -> bool
+    where
+        T: Ord + Copy + TryFrom<u64> + UpperBounded,
+    {
+        if !value.is_power_of_two() {
+            return false;
+        }
+        if let Ok(value) = T::try_from(value) {
+            *limit = value.min(*limit);
+            true
+        } else {
+            false
+        }
+    }
+
+    match limit {
+        "maxTextureDimension1D" => set_maximum(&mut limits.max_texture_dimension_1d, value),
+        "maxTextureDimension2D" => set_maximum(&mut limits.max_texture_dimension_2d, value),
+        "maxTextureDimension3D" => set_maximum(&mut limits.max_texture_dimension_3d, value),
+        "maxTextureArrayLayers" => set_maximum(&mut limits.max_texture_array_layers, value),
+        "maxBindGroups" => set_maximum(&mut limits.max_bind_groups, value),
+        "maxBindGroupsPlusVertexBuffers" => {
+            // not in wgpu but we're allowed to give back better limits than requested.
+            // we use dummy value to still produce value verification
+            let mut v: u32 = 0;
+            set_maximum(&mut v, value)
+        },
+        "maxBindingsPerBindGroup" => set_maximum(&mut limits.max_bindings_per_bind_group, value),
+        "maxDynamicUniformBuffersPerPipelineLayout" => set_maximum(
+            &mut limits.max_dynamic_uniform_buffers_per_pipeline_layout,
+            value,
+        ),
+        "maxDynamicStorageBuffersPerPipelineLayout" => set_maximum(
+            &mut limits.max_dynamic_storage_buffers_per_pipeline_layout,
+            value,
+        ),
+        "maxSampledTexturesPerShaderStage" => {
+            set_maximum(&mut limits.max_sampled_textures_per_shader_stage, value)
+        },
+        "maxSamplersPerShaderStage" => {
+            set_maximum(&mut limits.max_samplers_per_shader_stage, value)
+        },
+        "maxStorageBuffersPerShaderStage" => {
+            set_maximum(&mut limits.max_storage_buffers_per_shader_stage, value)
+        },
+        "maxStorageTexturesPerShaderStage" => {
+            set_maximum(&mut limits.max_storage_textures_per_shader_stage, value)
+        },
+        "maxUniformBuffersPerShaderStage" => {
+            set_maximum(&mut limits.max_uniform_buffers_per_shader_stage, value)
+        },
+        "maxUniformBufferBindingSize" => {
+            set_maximum(&mut limits.max_uniform_buffer_binding_size, value)
+        },
+        "maxStorageBufferBindingSize" => {
+            set_maximum(&mut limits.max_storage_buffer_binding_size, value)
+        },
+        "minUniformBufferOffsetAlignment" => {
+            set_alignment(&mut limits.min_uniform_buffer_offset_alignment, value)
+        },
+        "minStorageBufferOffsetAlignment" => {
+            set_alignment(&mut limits.min_storage_buffer_offset_alignment, value)
+        },
+        "maxVertexBuffers" => set_maximum(&mut limits.max_vertex_buffers, value),
+        "maxBufferSize" => set_maximum(&mut limits.max_buffer_size, value),
+        "maxVertexAttributes" => set_maximum(&mut limits.max_vertex_attributes, value),
+        "maxVertexBufferArrayStride" => {
+            set_maximum(&mut limits.max_vertex_buffer_array_stride, value)
+        },
+        "maxInterStageShaderComponents" => {
+            set_maximum(&mut limits.max_inter_stage_shader_components, value)
+        },
+        "maxInterStageShaderVariables" => {
+            // not in wgpu but we're allowed to give back better limits than requested.
+            // we use dummy value to still produce value verification
+            let mut v: u32 = 0;
+            set_maximum(&mut v, value)
+        },
+        "maxColorAttachments" => set_maximum(&mut limits.max_color_attachments, value),
+        "maxColorAttachmentBytesPerSample" => {
+            set_maximum(&mut limits.max_color_attachment_bytes_per_sample, value)
+        },
+        "maxComputeWorkgroupStorageSize" => {
+            set_maximum(&mut limits.max_compute_workgroup_storage_size, value)
+        },
+        "maxComputeInvocationsPerWorkgroup" => {
+            set_maximum(&mut limits.max_compute_invocations_per_workgroup, value)
+        },
+        "maxComputeWorkgroupSizeX" => set_maximum(&mut limits.max_compute_workgroup_size_x, value),
+        "maxComputeWorkgroupSizeY" => set_maximum(&mut limits.max_compute_workgroup_size_y, value),
+        "maxComputeWorkgroupSizeZ" => set_maximum(&mut limits.max_compute_workgroup_size_z, value),
+        "maxComputeWorkgroupsPerDimension" => {
+            set_maximum(&mut limits.max_compute_workgroups_per_dimension, value)
+        },
+        _ => false,
+    }
 }

components/script/dom/webidls/WebGPU.webidl

@@ -21,7 +21,7 @@ interface GPUSupportedLimits {
     readonly attribute unsigned long maxTextureDimension3D;
     readonly attribute unsigned long maxTextureArrayLayers;
     readonly attribute unsigned long maxBindGroups;
-    //readonly attribute unsigned long maxBindGroupsPlusVertexBuffers;
+    readonly attribute unsigned long maxBindGroupsPlusVertexBuffers;
     readonly attribute unsigned long maxBindingsPerBindGroup;
     readonly attribute unsigned long maxDynamicUniformBuffersPerPipelineLayout;
     readonly attribute unsigned long maxDynamicStorageBuffersPerPipelineLayout;
@@ -39,9 +39,9 @@ interface GPUSupportedLimits {
     readonly attribute unsigned long maxVertexAttributes;
     readonly attribute unsigned long maxVertexBufferArrayStride;
     readonly attribute unsigned long maxInterStageShaderComponents;
-    //readonly attribute unsigned long maxInterStageShaderVariables;
-    //readonly attribute unsigned long maxColorAttachments;
-    //readonly attribute unsigned long maxColorAttachmentBytesPerSample;
+    readonly attribute unsigned long maxInterStageShaderVariables;
+    readonly attribute unsigned long maxColorAttachments;
+    readonly attribute unsigned long maxColorAttachmentBytesPerSample;
     readonly attribute unsigned long maxComputeWorkgroupStorageSize;
     readonly attribute unsigned long maxComputeInvocationsPerWorkgroup;
     readonly attribute unsigned long maxComputeWorkgroupSizeX;

components/script/script_thread.rs

@@ -2427,8 +2427,9 @@ impl ScriptThread {
                 pipeline_id,
             } => {
                 self.gpu_id_hub.free_device_id(device_id);
-                let global = self.documents.borrow().find_global(pipeline_id).unwrap();
-                global.remove_gpu_device(WebGPUDevice(device_id));
+                if let Some(global) = self.documents.borrow().find_global(pipeline_id) {
+                    global.remove_gpu_device(WebGPUDevice(device_id));
+                } // page can already be destroyed
             },
             WebGPUMsg::FreeBuffer(id) => self.gpu_id_hub.free_buffer_id(id),
             WebGPUMsg::FreePipelineLayout(id) => self.gpu_id_hub.free_pipeline_layout_id(id),

components/webgpu/ipc_messages/to_dom.rs

@@ -7,6 +7,8 @@
 use ipc_channel::ipc::IpcSharedMemory;
 use serde::{Deserialize, Serialize};
 use wgc::pipeline::CreateShaderModuleError;
+use wgpu_core::instance::{RequestAdapterError, RequestDeviceError};
+use wgpu_core::resource::BufferAccessError;
 pub use {wgpu_core as wgc, wgpu_types as wgt};
 
 use crate::identity::*;
@@ -63,22 +65,20 @@ pub struct Adapter {
     pub channel: WebGPU,
 }
 
-#[derive(Debug, Deserialize, Serialize)]
-pub struct Device {
-    pub device_id: WebGPUDevice,
-    pub queue_id: WebGPUQueue,
-    pub descriptor: wgt::DeviceDescriptor<Option<String>>,
-}
-
 #[derive(Debug, Deserialize, Serialize)]
 #[allow(clippy::large_enum_variant)]
 pub enum WebGPUResponse {
     /// WebGPU is disabled
     None,
-    // TODO: use wgpu errors
-    Adapter(Result<Adapter, String>),
-    Device(Result<Device, String>),
-    BufferMapAsync(Result<IpcSharedMemory, String>),
+    Adapter(Result<Adapter, RequestAdapterError>),
+    Device(
+        (
+            WebGPUDevice,
+            WebGPUQueue,
+            Result<wgt::DeviceDescriptor<Option<String>>, RequestDeviceError>,
+        ),
+    ),
+    BufferMapAsync(Result<IpcSharedMemory, BufferAccessError>),
     SubmittedWorkDone,
     PoppedErrorScope(Result<Option<Error>, PopError>),
     CompilationInfo(Option<ShaderCompilationInfo>),