README.md

Function Execution Space Specifiers

Source: NVIDIA CUDA Documentation

• Function Execution Space Specifiers denote whether a function executes on host or on the device and whether it is callable from the host or the device.

• Source Code

Types:

__device__ :

• Declares a function that is:

  • Executed on device.

  • Callable from device only.

  • Compiler inlines any device function when deemed appropriate.

__global__ :

• Declares a function as being a kernel:

  • Executed on the device.

  • Callable from host.

  • Callable from device for devices of compute capability 3.2 or higher.

  • Must have a void return type, and cannot be member of a class.

  • A call to this type of function is asynchronous, meaning it returns before the device has completed its execution.

  • Any call to this function must specify its Execution Configuration.

__host__ :

• Declares a function as:

  • Executed on host.

  • Callable from host only.

  • A function declared without any of the Specifiers is treated as of this type.

---

__noinline__ :

• This qualifier can be used as a hint for compiler not to inline the function.

__forceinline__ :

• This qualifier can be used to force compiler to inline the function.