Function tvm::ffi::cuda_api::ConstructLaunchConfig#
Defined in File unified_api.h
Function Documentation#
-
inline ResultType tvm::ffi::cuda_api::ConstructLaunchConfig(KernelHandle kernel, StreamHandle stream, uint32_t smem_size, tvm::ffi::dim3 grid, tvm::ffi::dim3 block, int cluster_dim, LaunchConfig &config, LaunchAttrType &attr)#
Construct a launch configuration with optional cluster dimensions.
- Parameters:
kernel – The kernel handle.
stream – The CUDA stream.
smem_size – Dynamic shared memory size in bytes.
grid – Grid dimensions.
block – Block dimensions.
cluster_dim – Cluster dimension (1 = no clustering, >1 enables cluster launch).
config – [out] The launch configuration to populate.
attr – [out] Storage for a launch attribute (must outlive the launch call).
- Returns:
kSuccess.