docs: remove all WebGPU/Metal/WGSL references from documentation

v1.0.0 docs cleanup:

Deleted:
- docs/14-wgpu-codegen.md (WGSL transpiler doc — crate removed)
- docs/PRODUCTION_READINESS_ROADMAP.md (superseded by v1.0.0 plan)

Updated 12 docs:
- 02-crate-structure: remove 4 crate entries, dependency sections
- 05-gpu-backends: remove Metal (~70 lines) and WebGPU (~65 lines) sections
- 09-testing: remove wgpu-codegen and wavesim3d test counts
- 15-showcase-applications: remove WaveSim3D, update to 4 apps
- 16-cli-tool: CUDA-only codegen examples
- README index: remove WebGPU/Metal from backend table
- 03/04/07/08/10/13: targeted removal of backend references

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: mivertowski

docs/02-crate-structure.md

@@ -60,25 +60,17 @@ RustCompute/
 │   │   └── tests/
 │   │       └── gpu_execution_verify.rs  # GPU execution verification
 │   │
-│   ├── ringkernel-metal/         # Metal backend (scaffolded)
-│   │   └── src/lib.rs
-│   │
-│   ├── ringkernel-wgpu/          # WebGPU backend (scaffolded)
-│   │   └── src/lib.rs
-│   │
 │   ├── ringkernel-codegen/       # Kernel code generation
 │   │   └── src/lib.rs
 │   │
-│   ├── ringkernel-ir/            # Unified IR for multi-backend codegen
+│   ├── ringkernel-ir/            # Unified IR for codegen
 │   │   └── src/
 │   │       ├── lib.rs            # IrModule, IrBuilder, ValueId, BlockId
 │   │       ├── builder.rs        # Fluent IR construction
 │   │       ├── nodes.rs          # IR node types (BinaryOp, Load, Store, etc.)
 │   │       ├── types.rs          # IrType, ScalarType, VectorType
 │   │       ├── optimize.rs       # Optimization passes (DCE, constant folding)
 │   │       ├── lower_cuda.rs     # CUDA backend lowering
-│   │       ├── lower_wgsl.rs     # WGSL backend lowering
-│   │       ├── lower_msl.rs      # MSL backend lowering
 │   │       └── validation.rs     # IR validation
 │   │
 │   ├── ringkernel-cli/           # CLI tool for scaffolding and codegen
@@ -136,14 +128,6 @@ RustCompute/
 │   │       ├── simulation/       # Grid, kernels, backends
 │   │       └── gui/              # Interactive visualization
 │   │
-│   ├── ringkernel-wavesim3d/     # Showcase: 3D wave simulation
-│   │   └── src/
-│   │       ├── lib.rs
-│   │       ├── simulation/       # 3D FDTD, actor backend, physics
-│   │       ├── audio/            # Binaural audio, sources, virtual head
-│   │       ├── visualization/    # Volume renderer, slices, camera
-│   │       └── gui/              # Controls panel
-│   │
 │   ├── ringkernel-txmon/         # Showcase: Transaction monitoring
 │   │   └── src/
 │   │       ├── lib.rs
@@ -169,8 +153,7 @@ RustCompute/
 ├── examples/                     # 20+ working examples
 │   ├── basic/
 │   │   ├── hello_kernel.rs       # Runtime, lifecycle, suspend/resume
-│   │   ├── kernel_states.rs      # State machine, multi-kernel
-│   │   └── wgpu_hello.rs         # WebGPU backend
+│   │   └── kernel_states.rs      # State machine, multi-kernel
 │   ├── messaging/
 │   │   ├── request_response.rs   # Correlation IDs, priorities
 │   │   ├── pub_sub.rs            # Topic wildcards, QoS
@@ -262,16 +245,12 @@ RustCompute/
 | `ringkernel-core` | Working | Core traits: RingKernel, KernelHandle, HLC, PubSub, K2K |
 | `ringkernel-cpu` | Working | CPU backend for development/testing |
 | `ringkernel-cuda` | Working | NVIDIA CUDA backend with PTX kernels |
-| `ringkernel-metal` | Scaffolded | Apple Metal backend (API defined) |
-| `ringkernel-wgpu` | Working | WebGPU cross-platform backend |
 | `ringkernel-derive` | Working | Proc macros for message/kernel definitions |
 | `ringkernel-codegen` | In Development | GPU kernel code generation |
 | `ringkernel-cuda-codegen` | Working | Rust-to-CUDA transpiler for GPU kernels |
-| `ringkernel-wgpu-codegen` | Working | Rust-to-WGSL transpiler for GPU kernels |
 | `ringkernel-ecosystem` | Working | Integration utilities |
 | `ringkernel-audio-fft` | Working | Example: GPU audio FFT processing |
 | `ringkernel-wavesim` | Working | Example: 2D wave simulation with FDTD |
-| `ringkernel-wavesim3d` | Working | Showcase: 3D wave simulation with binaural audio |
 | `ringkernel-txmon` | Working | Showcase: GPU-accelerated transaction monitoring |
 | `ringkernel-accnet` | Working | Showcase: Accounting network analytics |
 | `ringkernel-procint` | Working | Showcase: Process intelligence with DFG mining |
@@ -288,8 +267,6 @@ members = [
     "crates/ringkernel-core",
     "crates/ringkernel-derive",
     "crates/ringkernel-cuda",
-    "crates/ringkernel-metal",
-    "crates/ringkernel-wgpu",
     "crates/ringkernel-cpu",
     "crates/ringkernel-codegen",
 ]
@@ -325,8 +302,6 @@ tracing-subscriber = "0.3"
 
 # GPU backends
 cudarc = { version = "0.18.2", optional = true }       # CUDA (updated API)
-metal = { version = "0.31", optional = true }          # Metal
-wgpu = { version = "27.0", optional = true }           # WebGPU (Arc-based)
 
 # Web frameworks
 axum = { version = "0.8", optional = true }
@@ -390,20 +365,6 @@ Requires:
 
 ---
 
-## Platform-Specific Compilation
-
-```toml
-# crates/ringkernel-cuda/Cargo.toml
-[target.'cfg(any(target_os = "linux", target_os = "windows"))'.dependencies]
-cudarc = { workspace = true }
-
-# crates/ringkernel-metal/Cargo.toml
-[target.'cfg(any(target_os = "macos", target_os = "ios"))'.dependencies]
-metal = { workspace = true }
-```
-
----
-
 ## Build Scripts
 
 ### CUDA Backend (build.rs)

docs/03-core-abstractions.md

@@ -159,7 +159,7 @@ use async_trait::async_trait;
 /// Runtime for managing ring kernel lifecycles.
 #[async_trait]
 pub trait RingKernelRuntime: Send + Sync {
-    /// Backend identifier (e.g., "cuda", "metal", "wgpu").
+    /// Backend identifier (e.g., "cuda", "cpu").
     fn backend_name(&self) -> &'static str;
 
     /// Launch a kernel (initially inactive).

docs/04-memory-management.md

@@ -400,37 +400,6 @@ println!("Hit rate: {:.1}%", stats.hit_rate() * 100.0);
 
 ---
 
-## WebGPU Staging Buffer Pool (v0.3.0)
-
-Efficient staging buffer reuse for GPU-to-host transfers:
-
-```rust
-use ringkernel_wgpu::memory::{StagingBufferPool, StagingPoolStats};
-
-// Create staging pool
-let pool = StagingBufferPool::new(&device, 16);  // Max 16 buffers
-
-// Acquire staging buffer for readback
-let staging = pool.acquire(data_size)?;
-
-// Use for GPU → host copy
-encoder.copy_buffer_to_buffer(&gpu_buffer, 0, staging.buffer(), 0, data_size);
-
-// Map and read
-staging.map_async(wgpu::MapMode::Read);
-device.poll(wgpu::Maintain::Wait);
-let data = staging.get_mapped_range();
-
-// Buffer returned to pool on drop
-drop(staging);
-
-// Check statistics
-let stats = pool.stats();
-println!("Staging buffer reuse rate: {:.1}%", stats.hit_rate() * 100.0);
-```
-
----
-
 ## GPU Stratified Memory Pool (v0.4.0)
 
 For GPU-side buffer reuse with O(1) allocation from free lists:
@@ -563,11 +532,6 @@ impl MemoryVisibility {
         {
             Self::SystemScope // Native Windows CUDA works
         }
-
-        #[cfg(target_os = "macos")]
-        {
-            Self::DeviceScope // Metal doesn't have system atomics
-        }
     }
 }