Although ArrayFire is quite extensive, there remain many cases in which you may want to write custom kernels in OpenCL or CUDA. For example, you may wish to add ArrayFire to an existing code base to increase your productivity, or you may need to supplement ArrayFire's functionality with your own custom implementation of specific algorithms.

ArrayFire manages its own context, queue, memory, and creates custom IDs for devices. As such, most of the interoperability functions focus on reducing potential synchronization conflicts between ArrayFire and OpenCL.

Basics

It is fairly straightforward to interface ArrayFire with your own custom OpenCL code. ArrayFire provides several functions to ease this process including:

Additionally, the OpenCL backend permits the programmer to add and remove custom devices from the ArrayFire device manager. These permit you to attach ArrayFire directly to the OpenCL queue used by other portions of your application.

Below we provide two worked examples on how ArrayFire can be integrated into new and existing projects.

Adding custom OpenCL kernels to an existing ArrayFire application

By default, ArrayFire manages its own context, queue, memory, and creates custom IDs for devices. Thus there is some bookkeeping that needs to be done to integrate your custom OpenCL kernel.

If your kernels can share operate in the same queue as ArrayFire, you should:

1. Add an include for af/opencl.h to your project
2. Obtain the OpenCL context, device, and queue used by ArrayFire
3. Obtain cl_mem references to af::array objects
4. Load, build, and use your kernels
5. Return control of af::array memory to ArrayFire

Note, ArrayFire uses an in-order queue, thus when ArrayFire and your kernels are operating in the same queue, there is no need to perform any synchronization operations.

This process is best illustrated with a fully worked example:

If your kernels needs to operate in their own OpenCL queue, the process is essentially identical, except you need to instruct ArrayFire to complete its computations using the af::sync() function prior to launching your own kernel and ensure your kernels are complete using clFinish (or similar) commands prior to returning control of the memory to ArrayFire:

1. Add an include for af/opencl.h to your project
2. Obtain the OpenCL context, device, and queue used by ArrayFire
3. Obtain cl_mem references to af::array objects
4. Instruct ArrayFire to finish operations using af::sync()
5. Load, build, and use your kernels
6. Instruct OpenCL to finish operations using clFinish() or similar commands.

1. Return control of af::array memory to ArrayFire

Adding ArrayFire to an existing OpenCL application

Adding ArrayFire to an existing OpenCL application is slightly more involved and can be somewhat tricky due to several optimizations we implement. The most important are as follows:

• ArrayFire assumes control of all memory provided to it.
• ArrayFire does not (in general) support in-place memory transactions.

We will discuss the implications of these items below. To add ArrayFire to existing code you need to:

1. Add includes
2. Instruct OpenCL to complete its operations using clFinish (or similar)
3. Instruct ArrayFire to use the user-created OpenCL Context
4. Create ArrayFire arrays from OpenCL memory objects
5. Perform ArrayFire operations on the Arrays
6. Instruct ArrayFire to finish operations using af::sync()
7. Obtain cl_mem references for important memory
8. Continue your OpenCL application

To create the af::array objects, you should use one of the following constructors:

NOTE: With all of these constructors, ArrayFire's memory manager automatically assumes responsibility for any memory provided to it. If you are creating an array from a cl::Buffer, you should specify retain=true to ensure your memory is not deallocated if your cl::Buffer were to go out of scope. We use this technique in the example below. If you do not wish for ArrayFire to manage your memory, you may call the array::unlock() function and manage the memory yourself; however, if you do so, please be cautious not to call clReleaseMemObj on a cl_mem when ArrayFire might be using it!

The eight steps above are best illustrated using a fully-worked example. Below we use the OpenCL C++ API and omit error checking to keep the code readable.

Using multiple devices

If you are using ArrayFire and OpenCL with multiple devices be sure to use afcl::addDevice to add your custom context + device + queue to ArrayFire's device manager. This will let you switch ArrayFire devices using your current cl_device_id and cl_context.