CUDA C

Enjoy this cheat sheet at its fullest within Dash, the macOS documentation browser.

Function Type Qualifiers

Function type qualifiers are used in function declarations and definitions.

They specify where the functions can be called from and where they can be executed.

Host is the CPU side, and device is the GPU side.

__device__

Executed on the device. Callable from the device only.

__global__

Executed on the device. Callable from the host or from the device for devices of compute capability 3.x or higher. Must have void return type.

__host__

Executed on the host. Callable from the host only (equivalent to declaring the function without any qualifiers).

Built-in Vector Types

Types

charX, ucharX, shortX, intX, uintX, longX, ulongX, floatX, where X = 1, 2, 3, or 4.

doubleX, longlongX, ulonglongX, where X = 1, or 2.

Note: dim3 is a uint3 with default components initalized to 1.

Constructor Function

make_<type name> constructs the built-in vector type with type specified by replacing <type name> with one of the types above.

Component Access

The 1st, 2nd, 3rd, and 4th components are accessible through the fields x, y, z, and w.

Example

int4 intVec = make_int4(0, 42, 3, 5) creates an int4 vector typed variable named intVec with the given int elements. intVec.z accesses its third element, 3.

Built-in Variables

Inside functions executed on the device (GPU), grid and block dimensions, and block and thread indices can be accessed using built-in variables listed below.

gridDim

Dimensions of the grid (dim3).

blockIdx

Block index within the grid (uint3).

blockDim

Dimensions of the block (dim3).

threadIdx

Thread index within the block (uint3).

warpSize

Warp size in threads (int).

Device Memory Management

Allocating memory

cudaError_t cudaMalloc(void **devPtr, size_t size)

Allocates size bytes of linear memory on the device and points devPtr to the allocated memory.

Freeing memory

cudaError_t cudaFree(void *devPtr)

Frees the memory space pointed to by devPtr.

Transferring data

cudaError_t cudaMemcpy(void *dst, const void *src, size_t count, cudaMemcpyKind kind)

Copies count bytes of data from the memory area pointed to by src to the memory area pointed to by dst. The direction of copy is specified by kind, and is one of cudaMemcpyHostToHost, cudaMemcpyHostToDevice, cudaMemcpyDeviceToHost, or cudaMemcpyDeviceToDevice.

Kernel Launch

A kernel function declared as __global__ void Func(float *parameter) can be called without the optional arguments as Func<<<numBlocks, threadsPerBlock>>>(parameter) or with the optional arguments as Func<<<numBlocks, threadsPerBlock, Ns, S>>>(parameter).

numBlocks is of type dim3 and specifies the number of blocks,

threadsPerBlock is of type dim3 and specifies the number of threads per block,

Ns is of type size_t and specifies bytes in shared memory (optional),

S is of type cudaStream_t and specifies associated stream (optional).

Notes