Deep Neural Network Library (DNNL)  1.2.0
Performance library for Deep Learning
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Groups Pages

API Reference

The reorder primitive copies data between different memory formats but doesn't change the tensor from mathematical perspective:

\[ dst(\overline{x}) = src(\overline{x}) \]

As described in Basic Concepts in order to achieve the best performance some primitives (such as convolution) require special memory format which is typically referred as an optimized memory format. The optimized memory format may match or may not match memory format that data is currently kept in. In this case a user can use reorder primitive to copy (reorder) the data between the memory formats.

Using the attributes and post-ops users can also use reorder primitive to quantize the data (and if necessary change the memory format simultaneously).

Implementation Details

General Notes

  1. The reorder primitive requires the source and destination tensors to have the same shape. Implicit broadcasting is not supported.
  2. While in most of the cases the reorder should be able to handle arbitrary source and destination memory formats and data types it might happen than some combinations are not implemented. For instance:
    • Reorder implementations between weights in non-plain memory formats might be limited (but if encountered in real practice should be treated as a bug and reported to the DNNL team);
    • Weights in one Winograd format cannot be reordered to the weights of the other Winograd format;
    • Quantized weights for convolution with dnnl_s8 source data type cannot be dequantized back to the dnnl_f32 data type;
  3. To alleviate the problem a user may rely on fact that the reorder from original plain memory format and user's data type to the optimized format with chosen data type should be always implemented.

Data Types Support

The reorder primitive supports arbitrary data types for the source and destination according to the Data Types page.

When converting the data from one data type to a smaller one saturation is used. For instance:

reorder(src={1024, data_type=f32}, dst={, data_type=s8})
// dst == {127}
reorder(src={-124, data_type=f32}, dst={, data_type=u8})
// dst == {0}

Data Representation

The reorder primitive works with arbitrary data tensors. There is no special meaning associated with any logical dimensions.

Post-ops and Attributes

The reorder primitive support the following attributes and post-ops:

Attributes / Post-ops Meaning
Output scales Copy and scale the data according to the scaling factors
Sum post-op Instead of copy the data accumulate it to the previous data

For instance, the following pseudo-code

src = {dims={N, C, H, W}, data_type=dt_src, memory_format=fmt_src},
dst = {dims={N, C, H, W}, data_type=dt_dst, memory_format=fmt_dst},
attr ={
post-ops = { sum={scale=beta} },

would lead to the following operation:

\[ dst(\overline{x}) = \alpha \cdot src(\overline{x}) + \beta \cdot dst(\overline{x}) \]

The intermediate operations are being done using single precision floating point data type.

Implementation Limitations

  1. No primitive specific limitations. Refer to Data Types for limitations related to data types support.

Performance Tips