Skip to content

Latest commit

 

History

History
83 lines (62 loc) · 3.56 KB

how_does_opencl_work.md

File metadata and controls

83 lines (62 loc) · 3.56 KB
Raw

How Does OpenCL Work?

OpenCL is a programming framework and runtime that enables a programmer to create small programs, called kernel programs (or kernels), that can be compiled and executed, in parallel, across any processors in a system. The processors can be any mix of different types, including CPUs, GPUs, DSPs, FPGAs or Tensor Processors - which is why OpenCL is often called a solution for heterogeneous parallel programming.

The OpenCL framework contains two APIs. The Platform Layer API is run on the host CPU and is used first to enable a program to discover what parallel processors or compute devices are available in a system. By querying for what compute devices are available an application can run portably across diverse systems - adapting to different combinations of accelerator hardware. Once the compute devices are discovered, the Platform API enables the application to select and initialize the devices it wants to use.

The second API is the Runtime API which enables the application's kernel programs to be compiled for the compute devices on which they are going to run, loaded in parallel onto those processors and executed. Once the kernel programs finish execution the Runtime API is used to gather the results.



Compiling and Executing OpenCL Kernels

The most commonly used language for programming the kernels that are compiled and executed across the available parallel processors is called OpenCL C. OpenCL C is based on C99 and is defined as part of the OpenCL specification. Kernels written in other programming languages may be executed using OpenCL by compiling to an intermediate program representation, such as SPIR-V.

OpenCL is a low-level programming framework so the programmer has direct, explicit control over where and when kernels are run, how the memory they use is allocated and how the compute devices and host CPU synchronize their operations to ensure that data and computed results flow correctly - even when the host and compute kernels are running in parallel.

Executing an OpenCL Program

OpenCL regards a kernel program as the basic unit of executable code (similar to a C function). Kernels can execute with data or task-parallelism. An OpenCL program is a collection of kernels and functions (similar to dynamic library with run-time linking).

An OpenCL command queue is used by the host application to send kernels and data transfer functions to a device for execution. By enqueueing commands into a command queue, kernels and data transfer functions may execute asynchronously and in parallel with application host code.

The kernels and functions in a command queue can be executed in-order or out-of-order. A compute device may have multiple command queues.



Sequence for Executing OpenCL Kernels

A complete sequence for executing an OpenCL program is:

  1. Query for available OpenCL platforms and devices

  2. Create a context for one or more OpenCL devices in a platform

  3. Create and build programs for OpenCL devices in the context

  4. Select kernels to execute from the programs

  5. Create memory objects for kernels to operate on

  6. Create command queues to execute commands on an OpenCL device

  7. Enqueue data transfer commands into the memory objects, if needed

  8. Enqueue kernels into the command queue for execution

  9. Enqueue commands to transfer data back to the host, if needed