Skip to main content

PACKED_KERNELS_HEADER

Constant PACKED_KERNELS_HEADER 

Source
pub const PACKED_KERNELS_HEADER: &str = r#"// CUDA Kernels for Packed Tile-Based FDTD Wave Simulation
// Generated by ringkernel-cuda-codegen from Rust DSL
//
// All tiles packed contiguously: [Tile(0,0)][Tile(1,0)]...[Tile(n,m)]
// Each tile is 18x18 floats (16x16 interior + 1-cell halo)
//
// Benefits:
// - Zero host<->GPU transfers during simulation
// - All tiles computed in parallel
// - Halo exchange is just GPU memory copies
"#;
Expand description

Header comment for generated packed kernels.