/// <summary>
/// Invokes the specified kernels.
/// </summary>
/// <param name="kernels">The kernels.</param>
/// <param name="result">The result.</param>
/// <param name="srcA">The source a.</param>
/// <param name="srcB">The source b.</param>
/// <param name="weight">The weight.</param>
/// <returns>Tensor.</returns>
public static NDArray Invoke(ElementwiseKernels kernels, NDArray result, NDArray srcA, NDArray srcB, float weight)
{
var context = CudaHelpers.TSContextForTensor(srcA);
var cudaContext = context.CudaContextForTensor(srcA);
var writeTarget = TensorResultBuilder.GetWriteTarget(result, srcA, false, srcA.Shape);
var elementCount = writeTarget.ElementCount();
var ptx = kernels.GetPtx(context.Compiler);
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "lerp", writeTarget, srcA, srcB, weight, elementCount);
return(writeTarget);
}
/// <summary>
/// Invokes the specified kernels.
/// </summary>
/// <param name="kernels">The kernels.</param>
/// <param name="result">The result.</param>
/// <param name="srcY">The source y.</param>
/// <param name="srcX">The source x.</param>
/// <returns>Tensor.</returns>
public static Tensor Invoke(ElementwiseKernels kernels, Tensor result, Tensor srcY, Tensor srcX)
{
var context = CudaHelpers.TSContextForTensor(srcY);
var cudaContext = context.CudaContextForTensor(srcY);
var writeTarget = TensorResultBuilder.GetWriteTarget(result, srcY, false, srcY.Shape);
var elementCount = writeTarget.ElementCount();
var ptx = kernels.GetPtx(context.Compiler);
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "atan2", writeTarget, srcY, srcX, elementCount);
return(writeTarget);
}
public static Tensor Invoke(ElementwiseKernels kernels, Tensor result, Tensor srcA, Tensor srcB, float weight)
{
TSCudaContext context = CudaHelpers.TSContextForTensor(srcA);
CudaContext cudaContext = context.CudaContextForTensor(srcA);
cudaContext.SetCurrent();
Tensor writeTarget = TensorResultBuilder.GetWriteTarget(result, srcA, false, srcA.Sizes);
long elementCount = writeTarget.ElementCount();
byte[] ptx = kernels.GetPtx(context.Compiler);
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "lerp", writeTarget, srcA, srcB, weight, elementCount);
return(writeTarget);
}
/// <summary>
/// Invokes the specified kernels.
/// </summary>
/// <param name="kernels">The kernels.</param>
/// <param name="funcName">Name of the function.</param>
/// <param name="result">The result.</param>
/// <param name="lhs">The LHS.</param>
/// <param name="rhs">The RHS.</param>
/// <returns>Tensor.</returns>
public static NDArray Invoke(ElementwiseKernels kernels, string funcName, NDArray result, NDArray lhs, NDArray rhs)
{
var context = CudaHelpers.TSContextForTensor(lhs);
var cudaContext = context.CudaContextForTensor(lhs);
var writeTarget = TensorResultBuilder.GetWriteTarget(result, lhs, false, lhs.Shape);
var elementCount = writeTarget.ElementCount();
var ptx = kernels.GetPtx(context.Compiler);
if (result == lhs)
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t1_" + funcName, writeTarget, rhs, elementCount);
}
else
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t2_" + funcName, writeTarget, lhs, rhs, elementCount);
}
return(writeTarget);
}
/// <summary>
/// Invokes the specified kernels.
/// </summary>
/// <param name="kernels">The kernels.</param>
/// <param name="funcName">Name of the function.</param>
/// <param name="result">The result.</param>
/// <param name="src">The source.</param>
/// <param name="value">The value.</param>
/// <returns>Tensor.</returns>
public static Tensor Invoke(ElementwiseKernels kernels, string funcName, Tensor result, Tensor src, float value)
{
var context = CudaHelpers.TSContextForTensor(src);
var cudaContext = context.CudaContextForTensor(src);
var writeTarget = TensorResultBuilder.GetWriteTarget(result, src, false, src.Shape);
var elementCount = writeTarget.ElementCount();
var ptx = kernels.GetPtx(context.Compiler);
if (result == src)
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t1_" + funcName, writeTarget, value, elementCount);
}
else
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t2_" + funcName, writeTarget, src, value, elementCount);
}
return(writeTarget);
}
/// <summary>
/// Invokes the specified kernels.
/// </summary>
/// <param name="kernels">The kernels.</param>
/// <param name="result">The result.</param>
/// <param name="src">The source.</param>
/// <param name="min">The minimum.</param>
/// <param name="max">The maximum.</param>
/// <returns>Tensor.</returns>
public static NDArray Invoke(ElementwiseKernels kernels, NDArray result, NDArray src, float min, float max)
{
var funcName = "clamp";
var context = CudaHelpers.TSContextForTensor(src);
var cudaContext = context.CudaContextForTensor(src);
var writeTarget = TensorResultBuilder.GetWriteTarget(result, src, false, src.Shape);
var elementCount = writeTarget.ElementCount();
var ptx = kernels.GetPtx(context.Compiler);
if (result == src)
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t1_" + funcName, writeTarget, min, max, elementCount);
}
else
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t2_" + funcName, writeTarget, src, min, max, elementCount);
}
return(writeTarget);
}
public static Tensor Invoke(ElementwiseKernels kernels, string funcName, Tensor result, Tensor lhs, Tensor rhs, Tensor rhs2, Tensor rhs3)
{
TSCudaContext context = CudaHelpers.TSContextForTensor(lhs);
CudaContext cudaContext = context.CudaContextForTensor(lhs);
cudaContext.SetCurrent();
Tensor writeTarget = TensorResultBuilder.GetWriteTarget(result, lhs, false, lhs.Sizes);
long elementCount = writeTarget.ElementCount();
byte[] ptx = kernels.GetPtx(context.Compiler);
if (result == lhs)
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t1_" + funcName, writeTarget, rhs, rhs2, rhs3, elementCount);
}
else
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t2_" + funcName, writeTarget, lhs, rhs, rhs2, rhs3, elementCount);
}
return(writeTarget);
}
public static Tensor Invoke(ElementwiseKernels kernels, string funcName, Tensor result, Tensor src, Tensor src2, float step_size, float value)
{
TSCudaContext context = CudaHelpers.TSContextForTensor(src);
CudaContext cudaContext = context.CudaContextForTensor(src);
cudaContext.SetCurrent();
Tensor writeTarget = TensorResultBuilder.GetWriteTarget(result, src, false, src.Sizes);
long elementCount = writeTarget.ElementCount();
byte[] ptx = kernels.GetPtx(context.Compiler);
if (result == src)
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t1_" + funcName, writeTarget, src2, step_size, value, elementCount);
}
else
{
ApplyOpInvoke.Invoke(context, cudaContext, ptx, "t2_" + funcName, writeTarget, src, src2, step_size, value, elementCount);
}
return(writeTarget);
}