Launcher

cuda kernel function launcher with runtime numbers of blocks/threads

version(grain_cuda)

struct Launcher (

Args...

) {

CUfunction cuFunction;

auto kernelParams(T args);

void launch(uint[3] grid, uint[3] block, uint sharedMemBytes, CUstream stream);

enum CUDA_NUM_THREADS;

static getBlocks(uint n);

void launch(uint n, uint sharedMemBytes, CUstream stream);

}

Members

Functions

kernelParams auto kernelParams(T args): create kernel function as voidArgs.length
launch void launch(uint[3] grid, uint[3] block, uint sharedMemBytes, CUstream stream): detailed launch function
launch void launch(uint n, uint sharedMemBytes, CUstream stream): convinient launch function

Manifest constants

CUDA_NUM_THREADS enum CUDA_NUM_THREADS;: Undocumented in source.

Static functions

getBlocks getBlocks(uint n): Undocumented in source. Be warned that the author may not have intended to support it.

Variables

args Args args;: Undocumented in source.
cuFunction CUfunction cuFunction;: Undocumented in source.

Meta

Source

See Implementation

grain cuda aliases

CudaElementType

classes

Global

functions

static variables

structs

variables

isDeviceMemory