OpenCL Context

This document is auto-generated for Owl’s APIs. #23 entries have been extracted.

Github: {Signature} {Implementation}

Type definition

type num =
  | F of float
  | F32 of (float, float32_elt) Owl_dense_ndarray_generic.t
  | F64 of (float, float64_elt) Owl_dense_ndarray_generic.t

Type of supported number types in the param of eval function.

type t = {
  mutable device : cl_device_id array;
  mutable context : cl_context;
  mutable program : cl_program;
  mutable progsrc : string array;
  mutable command_queue : (cl_device_id, cl_command_queue) Hashtbl.t;
  }

Type of context. Note this is different from OpenCL’s context object.

val default : t

Default context, with all GPU devices included and pre-compiled core kernels.

source code

Query platform

val platforms : unit -> cl_platform_id array

List all the platforms on this computer.

source code

val devices : unit -> cl_device_id array

List all the devices (including CPUs, GPUs, accelerators, etc.) on

source code

val cpu_devices : unit -> cl_device_id array

List all the CPU devices on this computer.

source code

val gpu_devices : unit -> cl_device_id array

List all the GPU devices on this computer.

source code

val accelerators : unit -> cl_device_id array

List all the accelerators on this computer.

source code

Manipulate context

val create : cl_device_id array -> string array -> t

Create a context with the given devices. The corresponding command queues are created and core kernels are compiled.

source code

val get_opencl_ctx : t -> cl_context

Return the OpenCL context object. Note this is different from Owl’s context.

source code

val get_program : t -> cl_program

Return the program associated with the context.

source code

val get_dev : t -> int -> cl_device_id

Return the ith device object associated with the context.

source code

val get_cmdq : t -> cl_device_id -> cl_command_queue

Return the corresponding command queue object of the given device object and its associated with the context.

source code

Manipulate kernels

val kernels : t -> string array

List all the installed kernels in the given context.

source code

val add_kernels : t -> string array -> unit

add_kernels ctx code adds list of kernels to the existing context. src contains the source code of all the kernels. Note this function call also causes all the existing kernels in the current context to be recompiled with the passed in ones.

source code

val make_kernel : t -> string -> cl_kernel

make_kernel ctx fun_name makes a kernel object from passed in context with the given function name.

source code

val ba_kernel : ('a, 'b) kind -> string -> cl_program -> cl_kernel

This function is similar to make_kernel but specifically for making Bigarray function.

source code

Evaluate kernels

val eval : ?param:num array -> ?ctx:t -> ?dev_id:int -> ?work_dim:int -> ?work_size:int array -> string -> unit

eval fun_name evaluates a kernel function in the given context, by calling Kernel.enqueue_ndrange function.

Parameters:
  • param: an arrray of num type elements which are passed into kernel function. Note the order of array elements is the same as the order of parameters of kernel function.
  • ctx: the context for running the kernel function, default is used if this parameter is not specified.
  • dev_id: the device that the kernel function will run on. The default value is 0.
  • work_dim: The number of dimensions used to specify the global work-items and work-items in the work-group. work_dim must be greater than zero and less than or equal to three.
  • work_size: Global work size, if not specified the number of elements of the first ndarray in param is used.
  • fun_name: the name of the kernel function. It is either a function in the recompiled kernels, or those added by calling add_kernels function before.

Refer to the global_work_size in OpenCL Document on clEnqueueNDRangeKernel

source code