Struct SimdGroupOperations 

pub struct SimdGroupOperations;

SIMD group operations for Metal.

Metal simd_shuffle* / simd_sum intrinsics compile to MSL and are gated on real Metal hardware. The WGSL emitted here maps each onto the portable WGSL subgroup built-ins when the device exposes wgpu::Features::SUBGROUP (native == true); otherwise it emits a workgroupBarrier()-synchronised workgroup-shared-memory emulation with matching semantics within a workgroup. Every helper takes trailing (lid: u32, n: u32) — the invocation’s local_invocation_index and the active-invocation count — which the native path ignores.

Implementations

impl SimdGroupOperations

pub fn simd_shuffle_shader(native: bool) -> String

Generate shader code for SIMD group shuffle operations.

pub fn simd_reduce_shader(native: bool) -> String

Generate shader code for SIMD group reductions.

Native reductions broadcast the reduced value to every lane of the subgroup (matching Metal’s simd_sum family). The emulation reduces across the whole workgroup (n active invocations, n <= 256) and likewise returns the result on every invocation.

pub fn native_subgroups(context: &GpuContext) -> bool

Return true when the device exposes native subgroup intrinsics.

Convenience wrapper to pick the native argument for Self::simd_shuffle_shader / Self::simd_reduce_shader from a live GpuContext.