Crate rlx_wgpu 

RLX wgpu backend — cross-platform GPU execution via the wgpu Rust crate (Metal on macOS, Vulkan on Linux, DX12 on Windows, WebGPU in browsers).

Compared to rlx-metal: same overall shape (device singleton, buffer arena, per-op compute pipelines, command-buffer-per-forward-pass) but with WGSL kernels and the wgpu Rust API instead of MSL + the metal crate. Pure Rust deps — no FFI / submodules to manage.

Layout:

• device — wgpu instance/adapter/device singleton (sync wrapper)
• buffer — typed GPU buffer + arena
• kernels — WGSL source strings + per-kernel pipeline cache
• backend — Backend trait impl + per-op dispatch

Re-exports

pub use device::is_vulkan_available;

pub use device::select_vulkan_backend;

Modules

backend

WgpuExecutable — compiles an rlx-ir Graph into a sequence of kernel dispatches against a pre-allocated arena buffer.

buffer

Buffer arena for the wgpu backend. Mirrors the rlx-metal arena shape: pre-plan one big storage buffer at compile time, sub-allocate per-node offsets at known positions, treat I/O as write_buffer / read_buffer against those offsets.

calibrate

On-disk wgpu calibration for cost-model ranking.

coop_f16_vk

device

wgpu device discovery + capabilities.

fft_dispatch

fft_host

gdn_host

Host-side Op::GatedDeltaNet for wgpu arenas (readback → CPU → writeback).

gguf_host

Host-side GGUF K-quant Op::DequantMatMul for wgpu arenas.

im2col_host

kernels

WGSL kernel sources + per-kernel pipeline cache.

llada2_gate_host

Host-side Op::Custom("llada2.group_limited_gate") for wgpu arenas.

log_mel_host

training_bwd_host

Host-side training backward ops for wgpu arenas (readback → CPU → writeback).

umap_knn_host

Host-side Op::Custom("umap.knn") for wgpu arenas (small n only).

unfuse

IR-level “unfusion” pass for the wgpu backend.

welch_peaks_dispatch

welch_peaks_host

Functions

is_available

True if a wgpu adapter is reachable on this system. Always available at the crate level; the runtime registry only registers the backend when this returns true so tests on weird CI machines without a GPU don’t trip up.