/// <summary cref="MemoryBuffer{T, TIndex}.CopyFromView(
/// AcceleratorStream, ArrayView{T}, LongIndex1)"/>
protected internal unsafe override void CopyFromView(
AcceleratorStream stream,
ArrayView <T> source,
LongIndex1 targetOffset)
{
var binding = Accelerator.BindScoped();
var sourceAddress = new IntPtr(source.LoadEffectiveAddress());
var targetAddress = new IntPtr(ComputeEffectiveAddress(targetOffset));
var lengthInBytes = new IntPtr(source.LengthInBytes);
switch (source.AcceleratorType)
{
case AcceleratorType.CPU:
case AcceleratorType.Cuda:
CudaException.ThrowIfFailed(
CurrentAPI.MemcpyAsync(
targetAddress,
sourceAddress,
lengthInBytes,
stream));
break;
default:
throw new NotSupportedException(
RuntimeErrorMessages.NotSupportedTargetAccelerator);
}
binding.Recover();
}
/// <summary cref="MemoryBuffer{T, TIndex}.CopyToView(AcceleratorStream, ArrayView{T}, Index)"/>
protected internal unsafe override void CopyToView(
AcceleratorStream stream,
ArrayView <T> target,
Index sourceOffset)
{
var binding = Accelerator.BindScoped();
var targetBuffer = target.Source;
var sourceAddress = new IntPtr(ComputeEffectiveAddress(sourceOffset));
var targetAddress = new IntPtr(target.LoadEffectiveAddress());
switch (targetBuffer.AcceleratorType)
{
case AcceleratorType.CPU:
CudaException.ThrowIfFailed(CudaAPI.Current.MemcpyDeviceToHost(
targetAddress,
sourceAddress,
new IntPtr(target.LengthInBytes),
stream));
break;
case AcceleratorType.Cuda:
CudaException.ThrowIfFailed(CudaAPI.Current.MemcpyDeviceToDevice(
targetAddress,
sourceAddress,
new IntPtr(target.LengthInBytes),
stream));
break;
default:
throw new NotSupportedException(RuntimeErrorMessages.NotSupportedTargetAccelerator);
}
binding.Recover();
}
/// <summary cref="MemoryBuffer{T, TIndex}.CopyFromView(AcceleratorStream, ArrayView{T}, Index)"/>
protected internal unsafe override void CopyFromView(
AcceleratorStream stream,
ArrayView <T> source,
Index targetOffset)
{
var clStream = (CLStream)stream;
switch (source.AcceleratorType)
{
case AcceleratorType.CPU:
CLException.ThrowIfFailed(
CLAPI.WriteBuffer(
clStream.CommandQueue,
NativePtr,
false,
new IntPtr(targetOffset * ElementSize),
new IntPtr(source.LengthInBytes),
new IntPtr(source.LoadEffectiveAddress())));
break;
case AcceleratorType.OpenCL:
CLException.ThrowIfFailed(
CLAPI.CopyBuffer(
clStream.CommandQueue,
source.Source.NativePtr,
NativePtr,
new IntPtr(source.Index * ElementSize),
new IntPtr(targetOffset * ElementSize),
new IntPtr(source.LengthInBytes)));
break;
default:
throw new NotSupportedException(RuntimeErrorMessages.NotSupportedTargetAccelerator);
}
}
/// <summary>
/// Performs a reduction using a reduction logic.
/// </summary>
/// <typeparam name="T">The underlying type of the reduction.</typeparam>
/// <typeparam name="TShuffleDown">The type of the shuffle logic.</typeparam>
/// <typeparam name="TReduction">The type of the reduction logic.</typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="input">The input elements to reduce.</param>
/// <param name="output">The output view to store the reduced value.</param>
/// <param name="shuffleDown">The shuffle logic.</param>
/// <param name="reduction">The reduction logic.</param>
/// <remarks>Uses the internal cache to realize a temporary output buffer.</remarks>
public static void Reduce <T, TShuffleDown, TReduction>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView <T> input,
ArrayView <T> output,
TShuffleDown shuffleDown,
TReduction reduction)
where T : struct
where TShuffleDown : struct, IShuffleDown <T>
where TReduction : struct, IReduction <T>
{
if (!input.IsValid)
{
throw new ArgumentNullException(nameof(input));
}
if (input.Length < 1)
{
throw new ArgumentOutOfRangeException(nameof(input));
}
var tempStorageSize = accelerator.ComputeReductionTempStorageSize(input.Length);
var temp = accelerator.MemoryCache.Allocate <T>(tempStorageSize);
accelerator.CreateReduction <T, TShuffleDown, TReduction>()(
stream,
input,
output,
temp,
shuffleDown,
reduction);
}
/// <summary>
/// Calculates the histogram on the given view (without overflow checking).
/// </summary>
/// <typeparam name="T">The input view element type.</typeparam>
/// <typeparam name="TIndex">The input view index type.</typeparam>
/// <typeparam name="TBinType">The histogram bin type.</typeparam>
/// <typeparam name="TIncrementor">
/// The operation to increment the value of the bin.
/// </typeparam>
/// <typeparam name="TLocator">
/// The operation to compute the bin location.
/// </typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="view">The input view.</param>
/// <param name="histogram">The histogram view to update.</param>
public static void HistogramUnchecked <
T,
TIndex,
TBinType,
TIncrementor,
TLocator>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView <T, TIndex> view,
ArrayView <TBinType> histogram)
where T : unmanaged
where TIndex : unmanaged, IIndex, IGenericIndex <TIndex>
where TBinType : unmanaged
where TIncrementor : struct, IIncrementOperation <TBinType>
where TLocator : struct, IComputeMultiBinOperation <T, TBinType, TIncrementor>
{
accelerator.CreateHistogramUnchecked <
T,
TIndex,
TBinType,
TIncrementor,
TLocator>()(
stream,
view,
histogram);
}
/// <summary>
/// Performs a reduction using a reduction logic.
/// </summary>
/// <typeparam name="T">The underlying type of the reduction.</typeparam>
/// <typeparam name="TShuffleDown">The type of the shuffle logic.</typeparam>
/// <typeparam name="TReduction">The type of the reduction logic.</typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="input">The input elements to reduce.</param>
/// <param name="shuffleDown">The shuffle logic.</param>
/// <param name="reduction">The reduction logic.</param>
/// <remarks>Uses the internal cache to realize a temporary output buffer.</remarks>
/// <returns>The reduced value.</returns>
public static T Reduce <T, TShuffleDown, TReduction>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView <T> input,
TShuffleDown shuffleDown,
TReduction reduction)
where T : struct
where TShuffleDown : struct, IShuffleDown <T>
where TReduction : struct, IReduction <T>
{
if (!input.IsValid)
{
throw new ArgumentNullException(nameof(input));
}
if (input.Length < 1)
{
throw new ArgumentOutOfRangeException(nameof(input));
}
var tempStorageSize = accelerator.ComputeReductionTempStorageSize(input.Length);
var storage = accelerator.MemoryCache.Allocate <T>(tempStorageSize + 1);
var output = storage.GetSubView(0, 1);
var temp = storage.GetSubView(1);
accelerator.Reduce(
stream,
input,
output,
temp,
shuffleDown,
reduction);
stream.Synchronize();
accelerator.MemoryCache.CopyTo(out T result, 0);
return(result);
}
/// <summary>
/// Transforms elements in the source view into elements in the target view using
/// the given transformer.
/// </summary>
/// <typeparam name="TSource">
/// The source type of the elements to transform.
/// </typeparam>
/// <typeparam name="TSourceStride">The 1D stride of the source view.</typeparam>
/// <typeparam name="TTarget">
/// The target type of the elements that have been transformed.
/// </typeparam>
/// <typeparam name="TTargetStride">The 1D stride of the target view.</typeparam>
/// <typeparam name="TTransformer">
/// The transformer to transform elements from the source type to the target type.
/// </typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="source">The source elements to transform</param>
/// <param name="target">
/// The target elements that will contain the transformed values.
/// </param>
/// <param name="transformer">The used transformer.</param>
public static void Transform <
TSource,
TSourceStride,
TTarget,
TTargetStride,
TTransformer>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView1D <TSource, TSourceStride> source,
ArrayView1D <TTarget, TTargetStride> target,
TTransformer transformer)
where TSource : unmanaged
where TSourceStride : struct, IStride1D
where TTarget : unmanaged
where TTargetStride : struct, IStride1D
where TTransformer : struct, ITransformer <TSource, TTarget> =>
accelerator.CreateTransformer <
TSource,
TSourceStride,
TTarget,
TTargetStride,
TTransformer>()(
stream,
source,
target,
transformer);
/// <summary cref="MemoryBuffer{T, TIndex}.CopyFromViewInternal(ArrayView{T, Index}, AcceleratorType, TIndex, AcceleratorStream)"/>
protected internal override unsafe void CopyFromViewInternal(
ArrayView <T, Index> source,
AcceleratorType acceleratorType,
TIndex targetOffset,
AcceleratorStream stream)
{
switch (acceleratorType)
{
case AcceleratorType.CPU:
Buffer.MemoryCopy(
source.Pointer.ToPointer(),
GetSubView(targetOffset).Pointer.ToPointer(),
source.LengthInBytes,
source.LengthInBytes);
break;
case AcceleratorType.Cuda:
CudaException.ThrowIfFailed(CudaAPI.Current.MemcpyDeviceToHost(
GetSubView(targetOffset).Pointer,
source.Pointer,
new IntPtr(source.LengthInBytes),
stream));
break;
default:
throw new NotSupportedException(RuntimeErrorMessages.NotSupportedTargetAccelerator);
}
}
public ComputeSession(string sessionID, Accelerator accelerator, ComputeHost host)
{
SessionID = sessionID;
_stream = accelerator.CreateStream();
_host = host;
}
/// <summary>
/// Performs a reduction using a reduction logic.
/// </summary>
/// <typeparam name="T">The underlying type of the reduction.</typeparam>
/// <typeparam name="TReduction">The type of the reduction logic.</typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="input">The input elements to reduce.</param>
/// <remarks>
/// Uses the internal cache to realize a temporary output buffer.
/// </remarks>
/// <returns>The reduced value.</returns>
public static Task <T> ReduceAsync <T, TReduction>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView <T> input)
where T : unmanaged
where TReduction : struct, IScanReduceOperation <T> =>
Task.Run(() => accelerator.Reduce <T, TReduction>(stream, input));
/// <summary cref="MemoryBuffer{T, TIndex}.CopyToViewInternal(ArrayView{T, Index}, AcceleratorType, TIndex, AcceleratorStream)"/>
protected internal override void CopyToViewInternal(
ArrayView <T, Index> target,
AcceleratorType acceleratorType,
TIndex sourceOffset,
AcceleratorStream stream)
{
switch (acceleratorType)
{
case AcceleratorType.CPU:
CudaException.ThrowIfFailed(CudaAPI.Current.MemcpyDeviceToHost(
target.Pointer,
GetSubView(sourceOffset).Pointer,
new IntPtr(target.LengthInBytes),
stream));
break;
case AcceleratorType.Cuda:
CudaException.ThrowIfFailed(CudaAPI.Current.MemcpyDeviceToDevice(
target.Pointer,
GetSubView(sourceOffset).Pointer,
new IntPtr(target.LengthInBytes),
stream));
break;
default:
throw new NotSupportedException(RuntimeErrorMessages.NotSupportedTargetAccelerator);
}
}
/// <summary cref="MemoryBuffer{T, TIndex}.CopyFromView(AcceleratorStream, ArrayView{T}, Index)"/>
protected internal unsafe override void CopyFromView(
AcceleratorStream stream,
ArrayView <T> source,
Index targetOffset)
{
var binding = stream.BindScoped();
var targetAddress = ComputeEffectiveAddress(targetOffset);
switch (source.AcceleratorType)
{
case AcceleratorType.CPU:
Unsafe.CopyBlock(
targetAddress,
source.LoadEffectiveAddress(),
(uint)source.LengthInBytes);
break;
case AcceleratorType.Cuda:
CudaException.ThrowIfFailed(CudaAPI.Current.MemcpyDeviceToHost(
new IntPtr(targetAddress),
new IntPtr(source.LoadEffectiveAddress()),
new IntPtr(source.LengthInBytes),
stream));
break;
default:
throw new NotSupportedException(RuntimeErrorMessages.NotSupportedTargetAccelerator);
}
binding.Recover();
}
/// <summary>
/// Access a specific slice of either a column 'c' or row 'r' of this vector
/// </summary>
/// <param name="gpu"></param>
/// <param name="row_col_index"></param>
/// <param name="row_col"></param>
/// <returns></returns>
public Vector _AccessSlice(int row_col_index, char row_col)
{
if (this.Columns == 1)
{
throw new Exception("Input Vector cannot be 1D");
}
int[] ChangeSelectLength;
int OutPutVectorLength;
switch (row_col)
{
case 'r':
//ChangeSelectLength = new int[5] { 0, 1, row_col_index, 0, vector.Columns };
//OutPutVectorLength = vector.Columns;
return(this._AccessRow(this, row_col_index));
case 'c':
ChangeSelectLength = new int[5] {
1, 0, 0, row_col_index, this.Columns
};
OutPutVectorLength = this.Value.Length / this.Columns;
break;
default:
throw new Exception("Invalid slice char selector, choose 'r' for row or 'c' for column");
}
//this is bad and I should feel bad
Accelerator gpu = this.gpu.accelerator;
AcceleratorStream Stream = gpu.CreateStream();
var kernelWithStream = gpu.LoadAutoGroupedKernel <Index1, ArrayView <float>, ArrayView <float>, ArrayView <int> >(AccessSliceKernal);
var buffer = gpu.Allocate <float>(OutPutVectorLength);
var buffer2 = gpu.Allocate <float>(this.Value.Length);
var buffer3 = gpu.Allocate <int>(5);
buffer.MemSetToZero(Stream);
buffer2.MemSetToZero(Stream);
buffer3.MemSetToZero(Stream);
buffer2.CopyFrom(Stream, this.Value, 0, 0, this.Value.Length);
buffer3.CopyFrom(Stream, ChangeSelectLength, 0, 0, ChangeSelectLength.Length);
kernelWithStream(Stream, OutPutVectorLength, buffer.View, buffer2.View, buffer3.View);
Stream.Synchronize();
float[] Output = buffer.GetAsArray(Stream);
buffer.Dispose();
buffer2.Dispose();
buffer3.Dispose();
Stream.Dispose();
return(new Vector(this.gpu, Output));
}
/// <summary cref="MemoryBuffer.MemSetToZero(AcceleratorStream)"/>
public override void MemSetToZero(AcceleratorStream stream) =>
CLException.ThrowIfFailed(
CLAPI.FillBuffer <byte>(
((CLStream)stream).CommandQueue,
NativePtr,
0,
IntPtr.Zero,
new IntPtr(LengthInBytes)));
/// <inheritdoc/>
protected internal override unsafe void MemSetInternal(
AcceleratorStream stream,
byte value,
long offsetInBytes,
long lengthInBytes)
{
stream.Synchronize();
ref byte targetAddress = ref Unsafe.AsRef <byte>(NativePtr.ToPointer());
/// <summary cref="MemoryBuffer.MemSetToZero(AcceleratorStream)"/>
public override void MemSetToZero(AcceleratorStream stream)
{
var binding = Accelerator.BindScoped();
CudaAPI.Current.Memset(NativePtr, 0, new IntPtr(LengthInBytes), stream);
binding.Recover();
}
/// <summary>
/// Performs an initialization on the given view.
/// </summary>
/// <typeparam name="T">The element type.</typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="view">The element view.</param>
/// <param name="value">The target value.</param>
public static void Initialize <T>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView <T> view,
T value)
where T : unmanaged =>
accelerator.CreateInitializer <T, Stride1D.Dense>()(
stream,
view,
value);
/// <summary>
/// Performs a reduction using a reduction logic.
/// </summary>
/// <typeparam name="T">The underlying type of the reduction.</typeparam>
/// <typeparam name="TReduction">The type of the reduction logic.</typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="input">The input elements to reduce.</param>
/// <param name="output">The output view to store the reduced value.</param>
public static void Reduce <T, TReduction>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView <T> input,
ArrayView <T> output)
where T : unmanaged
where TReduction : struct, IScanReduceOperation <T> =>
accelerator.CreateReduction <T, Stride1D.Dense, TReduction>()(
stream,
input,
output);
/// <summary>
/// Performs an initialization on the given view.
/// </summary>
/// <typeparam name="T">The element type.</typeparam>
/// <typeparam name="TStride">The 1D stride of the target view.</typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="view">The element view.</param>
/// <param name="value">The target value.</param>
public static void Initialize <T, TStride>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView1D <T, TStride> view,
T value)
where T : unmanaged
where TStride : struct, IStride1D =>
accelerator.CreateInitializer <T, TStride>()(
stream,
view,
value);
/// <summary>
/// Computes a new sequence of values from 0 to view.Length - 1 and writes
/// the computed values to the given view.
/// </summary>
/// <typeparam name="T">The element type.</typeparam>
/// <typeparam name="TSequencer">The type of the sequencer to use.</typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="view">The target view.</param>
/// <param name="sequencer">The used sequencer.</param>
public static void Sequence <T, TSequencer>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView <T> view,
TSequencer sequencer)
where T : unmanaged
where TSequencer : struct, ISequencer <T> =>
accelerator.CreateSequencer <T, Stride1D.Dense, TSequencer>()(
stream,
view,
sequencer);
/// <summary>
/// Performs an initialization on the given view.
/// </summary>
/// <typeparam name="T">The element type.</typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="view">The element view.</param>
/// <param name="value">The target value.</param>
public static void Initialize <T>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView <T> view,
T value)
where T : struct
{
accelerator.CreateInitializer <T>()(
stream,
view,
value);
}
/// <summary>
/// Performs a reduction using a reduction logic.
/// </summary>
/// <typeparam name="T">The underlying type of the reduction.</typeparam>
/// <typeparam name="TReduction">The type of the reduction logic.</typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="input">The input elements to reduce.</param>
/// <remarks>
/// Uses the internal cache to realize a temporary output buffer.
/// </remarks>
/// <returns>The reduced value.</returns>
public static T Reduce <T, TReduction>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView <T> input)
where T : unmanaged
where TReduction : struct, IScanReduceOperation <T>
{
var output = accelerator.MemoryCache.Allocate <T>(1);
accelerator.Reduce <T, TReduction>(stream, input, output);
output.CopyToCPU(stream, out T result, 1);
return(result);
}
/// <summary>
/// Computes a new sequence of batched values of length sequenceBatchLength, and
/// writes the computed values to the given view. Afterwards, the target view will
/// contain the following values:
/// - [0, sequenceBatchLength - 1] = 0,,
/// - [sequenceBatchLength, sequenceBatchLength * 2 -1] = 1,
/// - ...
/// </summary>
/// <typeparam name="T">The element type.</typeparam>
/// <typeparam name="TSequencer">The type of the sequencer to use.</typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="view">The target view.</param>
/// <param name="sequenceBatchLength">The length of a single batch.</param>
/// <param name="sequencer">The used sequencer.</param>
public static void BatchedSequence <T, TSequencer>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView <T> view,
LongIndex1D sequenceBatchLength,
TSequencer sequencer)
where T : unmanaged
where TSequencer : struct, ISequencer <T> =>
accelerator.CreateBatchedSequencer <T, Stride1D.Dense, TSequencer>()(
stream,
view,
sequenceBatchLength,
sequencer);
/// <summary>
/// Transforms elements in the source view into elements in the target view using
/// the given transformer.
/// </summary>
/// <typeparam name="T">The type of the elements to transform.</typeparam>
/// <typeparam name="TTransformer">
/// The transformer to transform elements from the source type to the target type.
/// </typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="source">The source elements to transform</param>
/// <param name="target">
/// The target elements that will contain the transformed values.
/// </param>
/// <param name="transformer">The used transformer.</param>
public static void Transform <T, TTransformer>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView <T> source,
ArrayView <T> target,
TTransformer transformer)
where T : unmanaged
where TTransformer : struct, ITransformer <T, T> =>
accelerator.CreateTransformer <T, TTransformer>()(
stream,
source,
target,
transformer);
/// <summary cref="MemoryBuffer.MemSetToZero(AcceleratorStream)"/>
public override void MemSetToZero(AcceleratorStream stream)
{
var binding = Accelerator.BindScoped();
CudaException.ThrowIfFailed(
CurrentAPI.Memset(
NativePtr,
0,
new IntPtr(LengthInBytes),
stream));
binding.Recover();
}
/// <summary>
/// Computes a new sequence of values from 0 to view.Length - 1 and writes
/// the computed values to the given view.
/// </summary>
/// <typeparam name="T">The element type.</typeparam>
/// <typeparam name="TSequencer">The type of the sequencer to use.</typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="view">The target view.</param>
/// <param name="sequencer">The used sequencer.</param>
public static void Sequence <T, TSequencer>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView <T> view,
TSequencer sequencer)
where T : struct
where TSequencer : struct, ISequencer <T>
{
accelerator.CreateSequencer <T, TSequencer>()(
stream,
view,
sequencer);
}
public CudaError MemcpyAsync(
IntPtr destination,
IntPtr source,
IntPtr length,
AcceleratorStream stream)
{
var cudaStream = stream as CudaStream;
return(cuMemcpyAsync(
destination,
source,
length,
cudaStream?.StreamPtr ?? IntPtr.Zero));
}
public CudaError MemcpyDeviceToDevice(
IntPtr destinationDevice,
IntPtr sourceDevice,
IntPtr length,
AcceleratorStream stream)
{
CudaStream cudaStream = stream as CudaStream;
return(MemcpyDeviceToDevice(
destinationDevice,
sourceDevice,
length,
cudaStream?.StreamPtr ?? IntPtr.Zero));
}
public CudaError Memset(
IntPtr destinationDevice,
byte value,
IntPtr length,
AcceleratorStream stream)
{
var cudaStream = stream as CudaStream;
return(Memset(
destinationDevice,
value,
length,
cudaStream?.StreamPtr ?? IntPtr.Zero));
}
/// <summary>
/// Computes a new repeated sequence of values from 0 to sequenceLength, from 0 to
/// sequenceLength, ... and writes the computed values to the given view.
/// Afterwards, the target view will contain the following values:
/// - [0, sequenceLength - 1] = [0, sequenceLength]
/// - [sequenceLength, sequenceLength * 2 -1] = [0, sequenceLength]
/// - ...
/// </summary>
/// <typeparam name="T">The element type.</typeparam>
/// <typeparam name="TStride">The 1D stride of the view.</typeparam>
/// <typeparam name="TSequencer">The type of the sequencer to use.</typeparam>
/// <param name="accelerator">The accelerator.</param>
/// <param name="stream">The accelerator stream.</param>
/// <param name="view">The target view.</param>
/// <param name="sequenceLength">The length of a single sequence.</param>
/// <param name="sequencer">The used sequencer.</param>
public static void RepeatedSequence <T, TStride, TSequencer>(
this Accelerator accelerator,
AcceleratorStream stream,
ArrayView1D <T, TStride> view,
LongIndex1D sequenceLength,
TSequencer sequencer)
where T : unmanaged
where TStride : struct, IStride1D
where TSequencer : struct, ISequencer <T> =>
accelerator.CreateRepeatedSequencer <T, TStride, TSequencer>()(
stream,
view,
sequenceLength,
sequencer);
