/// <summary>
/// Compiles and launches an explicltly-grouped kernel.
/// </summary>
static void CompileAndLaunchKernel(Accelerator accelerator, int groupSize)
{
// Create a backend for this device
using (var backend = accelerator.CreateBackend())
{
// Create a new compile unit using the created backend
using (var compileUnit = accelerator.Context.CreateCompileUnit(backend))
{
// Resolve and compile method into a kernel
var method = typeof(Program).GetMethod(nameof(GroupedKernel), BindingFlags.NonPublic | BindingFlags.Static);
var compiledKernel = backend.Compile(compileUnit, method);
// Info: use compiledKernel.GetBuffer() to retrieve the compiled kernel program data
// -------------------------------------------------------------------------------
// Load the explicitly grouped kernel
// Note that the kernel has to be disposed manually.
var kernel = accelerator.LoadKernel(compiledKernel);
var launcher = kernel.CreateStreamLauncherDelegate <Action <GroupedIndex, ArrayView <int>, int> >();
// -------------------------------------------------------------------------------
using (var buffer = accelerator.Allocate <int>(1024))
{
// You can also use kernel.Launch; however, the generic launch method involves boxing.
launcher(
new GroupedIndex(
(buffer.Length + groupSize - 1) / groupSize, // Compute the number of groups (round up)
groupSize), // Use the given group size
buffer.View,
42);
accelerator.Synchronize();
// Resolve and verify data
var data = buffer.GetAsArray();
for (int i = 0, e = data.Length; i < e; ++i)
{
if (data[i] != 42 + i)
{
Console.WriteLine($"Error at element location {i}: {data[i]} found");
}
}
}
kernel.Dispose();
}
}
}
/// <summary>
/// Compiles and launches an auto-grouped implicitly-grouped kernel.
/// </summary>
static void CompileAndLaunchAutoGroupedKernel(Accelerator accelerator)
{
// Create a backend for this device
using (var backend = accelerator.CreateBackend())
{
// Create a new compile unit using the created backend
using (var compileUnit = accelerator.Context.CreateCompileUnit(backend))
{
// Resolve and compile method into a kernel
var method = typeof(Program).GetMethod(nameof(MyKernel), BindingFlags.NonPublic | BindingFlags.Static);
var compiledKernel = backend.Compile(compileUnit, method);
// Info: use compiledKernel.GetBuffer() to retrieve the compiled kernel program data
// -------------------------------------------------------------------------------
// Load the implicitly grouped kernel with an automatically determined group size.
// Note that the kernel has to be disposed manually.
var kernel = accelerator.LoadAutoGroupedKernel(compiledKernel);
var launcher = kernel.CreateStreamLauncherDelegate <Action <Index, ArrayView <int>, int> >();
// -------------------------------------------------------------------------------
using (var buffer = accelerator.Allocate <int>(1024))
{
// Launch buffer.Length many threads and pass a view to buffer.
// You can also use kernel.Launch; however, the generic launch method involves boxing.
launcher(buffer.Length, buffer.View, 42);
// Wait for the kernel to finish...
accelerator.Synchronize();
// Resolve and verify data
var data = buffer.GetAsArray();
for (int i = 0, e = data.Length; i < e; ++i)
{
if (data[i] != 42 + i)
{
Console.WriteLine($"Error at element location {i}: {data[i]} found");
}
}
}
accelerator.Synchronize();
kernel.Dispose();
}
}
}
public ParallelTranslator(Accelerator device)
{
_backend = device.CreateBackend();
_compileUnit = device.Context.CreateCompileUnit(_backend);
}
