Reorder

Intel® oneAPI Deep Neural Network Developer Guide and Reference

Download PDF

ID 768875

Date 6/20/2024

Version

Public

A newer version of this document is available. Customers should click here to go to the newest version.

Visible to Intel only — GUID: GUID-B34F459F-5B63-4D01-BE18-BE9DF9D6B6C9

View Details

Reorder

API Reference

General

The reorder primitive copies data between different memory formats but doesn’t change the tensor from mathematical perspective (the variable names follow the standard Naming Conventions):

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 to 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).

Execution Arguments

When executed, the inputs and outputs should be mapped to an execution argument index as specified by the following table.

Primitive input/output	Execution argument index
	DNNL_ARG_FROM
	DNNL_ARG_TO
	DNNL_ARG_ATTR_SCALES \| DNNL_ARG_FROM
	DNNL_ARG_ATTR_SCALES \| DNNL_ARG_TO

Implementation Details

General Notes

The reorder primitive requires the source and destination tensors to have the same shape. Implicit broadcasting is not supported.
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 oneDNN 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;
- Only reference support is available for reorders to or from f8_e4m3.
- Optimized implementation of reorders to or from f8_e5m2 is available on Intel(R) Data Center GPU Max Series Only.
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
Scales	Scales the corresponding tensor by the given scale factor(s)
Zero points	Sets zero point(s) for the corresponding tensors
Sum post-op	Instead of copy the data accumulate it to the previous data

For instance, the following pseudo-code

reorder(
        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 ={
            scales={ src={mask=0} },
            zero_points= { src={mask=0}, dst={mask=0} },
            post-ops = { sum={scale=beta} },
        })

would lead to the following operation:

NOTE:

The intermediate operations are being done using single precision floating point data type.
and must be passed during execution runtime as a separate memory argument. Using argument will lead to multiplication of tensor values by a scale value. Using argument will lead to division of tensor values by a scale value.

Implementation Limitations

Refer to Data Types for limitations related to data types support.
CPU
- Reorders between bf16, f16 and s32 data types are not supported.
GPU
- Only tensors of 6 or fewer dimensions are supported.
- Runtime dimensions are not supported.

Performance Tips

N/A

Example

Reorder Primitive Example

This C++ API demonstrates how to create and execute a Reorder primitive.

Key optimizations included in this example:

Primitive attributes for output scaling.

Select Your Language

Using Intel.com Search

Quick Links

Recent Searches

Advanced Search

Only search in

Intel® oneAPI Deep Neural Network Developer Guide and Reference

Reorder

General

Execution Arguments

Implementation Details

Implementation Limitations

Performance Tips

Example