Struct KernelLaunch 

pub struct KernelLaunch<'a> { /* private fields */ }

2026-09-25: Builder for a kernel launch: grid, block, dynamic shared memory and typed arguments. launch() passes the arguments as &[KernelArg] to GpuBackend::launch_typed. Grid and block default to [1, 1, 1], shared memory to 0.

Implementations

impl<'a> KernelLaunch<'a>

pub fn new(gpu: &'a dyn GpuBackend, kernel: KernelHandle) -> Self

pub fn grid(self, grid: [u32; 3]) -> Self

pub fn block(self, block: [u32; 3]) -> Self

pub fn shared_mem(self, bytes: u32) -> Self

pub fn arg_ptr(self, p: DevicePtr) -> Self

2026-09-25: Add a device buffer argument.

pub fn arg_tensormap(self, map: &[u8; 128]) -> Self

2026-09-25: Add a 128-byte CUtensorMap by value, for a kernel parameter declared __grid_constant__ const CUtensorMap. The bytes fill 16 consecutive slots and form one argument.

pub fn arg_u32(self, v: u32) -> Self

pub fn arg_u64(self, v: u64) -> Self

pub fn arg_i32(self, v: i32) -> Self

pub fn arg_f32(self, v: f32) -> Self

pub fn launch(self, stream: u64) -> Result<()>

2026-09-25: Launch on stream through GpuBackend::launch_typed.

With debug_sync_kernels on, the stream is synchronised after a successful launch and a fault is returned with the launch geometry and a backtrace.