DML_FILL_VALUE_SEQUENCE_OPERATOR_DESC structure (directml.h)

Fills a tensor with a sequence. This operator performs the following pseudocode.

for each elementIndex in OutputTensor
    OutputTensor[elementIndex] = Value + elementIndex * Delta
endfor

Syntax

struct DML_FILL_VALUE_SEQUENCE_OPERATOR_DESC {
  const DML_TENSOR_DESC *OutputTensor;
  DML_TENSOR_DATA_TYPE  ValueDataType;
  DML_SCALAR_UNION      ValueStart;
  DML_SCALAR_UNION      ValueDelta;
};

Members

OutputTensor

Type: const DML_TENSOR_DESC*

The tensor to write the results to. This tensor may have any size.

ValueDataType

Type: DML_TENSOR_DATA_TYPE

The data type of Value field, which must match OutputTensor.DataType.

ValueStart

Type: DML_SCALAR_UNION

The initial value to fill the first element in the output, with ValueDataType determining how to interpret the field.

ValueDelta

Type: DML_SCALAR_UNION

A step to add to the value for each element written, with ValueDataType determining how to interpret the field.

Examples

Example 1. 1D ascending step

ValueStart = 3
ValueDelta = 2
ValueDataType = DML_TENSOR_DATA_TYPE_FLOAT32

OutputTensor: (Sizes:{1,1,1,3}, DataType:FLOAT32)
    [[[[3, 5, 7]]]]

Example 2. 2D ascending step

ValueStart = 10
ValueDelta = -2
ValueDataType = DML_TENSOR_DATA_TYPE_UINT8

OutputTensor: (Sizes:{1,1,2,2}, DataType:UINT8)
    [[[[10, 8],
       [ 6, 4]]]]

Availability

This operator was introduced in DML_FEATURE_LEVEL_2_1.

Tensor support

DML_FEATURE_LEVEL_5_0 and above

DML_FEATURE_LEVEL_4_0 and above

DML_FEATURE_LEVEL_2_1 and above

Requirements