/// <summary>
/// Gets the compile work-group size specified by the <c>__attribute__((reqd_work_group_size(X, Y, Z)))</c> qualifier.
/// </summary>
/// <param name="device"> One of the <see cref="OpenCLKernel.Program.Device"/>s. </param>
/// <returns> The compile work-group size specified by the <c>__attribute__((reqd_work_group_size(X, Y, Z)))</c> qualifier. If no such qualifier is specified, (0, 0, 0) is returned. </returns>
public long[] GetCompileWorkGroupSize(OpenCLDevice device)
{
return(OpenCLTools.ConvertArray(
GetArrayInfo <CLKernelHandle, CLDeviceHandle, OpenCLKernelWorkGroupInfo, IntPtr>(
Handle, device.Handle, OpenCLKernelWorkGroupInfo.CompileWorkGroupSize, CL10.GetKernelWorkGroupInfo)));
}
/// <summary>
/// Enqueues a command to execute a range of <see cref="OpenCLKernel"/>s in parallel.
/// </summary>
/// <param name="kernel"> The <see cref="OpenCLKernel"/> to execute. </param>
/// <param name="globalWorkOffset"> An array of values that describe the offset used to calculate the global ID of a work-item instead of having the global IDs always start at offset (0, 0,... 0). </param>
/// <param name="globalWorkSize"> An array of values that describe the number of global work-items in dimensions that will execute the kernel function. The total number of global work-items is computed as global_work_size[0] *...* global_work_size[work_dim - 1]. </param>
/// <param name="localWorkSize"> An array of values that describe the number of work-items that make up a work-group (also referred to as the size of the work-group) that will execute the <paramref name="kernel"/>. The total number of work-items in a work-group is computed as local_work_size[0] *... * local_work_size[work_dim - 1]. </param>
/// <param name="events"> A collection of events that need to complete before this particular command can be executed. If <paramref name="events"/> is not <c>null</c> or read-only a new <see cref="OpenCLEvent"/> identifying this command is created and attached to the end of the collection. </param>
public void Execute(OpenCLKernel kernel, long[] globalWorkOffset, long[] globalWorkSize, long[] localWorkSize, IReadOnlyList <OpenCLEventBase> events = null, IList <OpenCLEventBase> newEvents = null)
{
int eventWaitListSize;
CLEventHandle[] eventHandles = OpenCLTools.ExtractHandles(events, out eventWaitListSize);
CLEventHandle[] newEventHandle = (newEvents != null) ? new CLEventHandle[1] : null;
OpenCLErrorCode error = CL10.EnqueueNDRangeKernel(Handle, kernel.Handle, globalWorkSize.Length, OpenCLTools.ConvertArray(globalWorkOffset), OpenCLTools.ConvertArray(globalWorkSize), OpenCLTools.ConvertArray(localWorkSize), eventWaitListSize, eventHandles, newEventHandle);
OpenCLException.ThrowOnError(error);
if (newEvents != null)
{
lock (newEvents)
{
newEvents.Add(new OpenCLEvent(newEventHandle[0], this));
}
}
}
